1996_TI_High Performance_FIFO_Memories_Designers_Handbook 1996 TI High Performance FIFO Memories Designers Handbook
User Manual: 1996_TI_High-Performance_FIFO_Memories_Designers_Handbook
Open the PDF directly: View PDF .
Page Count: 482
Download | |
Open PDF In Browser | View PDF |
• TEXAS INSTRUMENTS ======= High.Performance FIFO Memories 1996 1996 Advanced System Logic Products General Information • .. ~============~ Product OveNiew ... ~s=p=e=ci=fi=C=A=P=p=li=ca=t=lo=n=R=e=p=o=rt=s===========. Power Considerations .• Device Models .. .. ~========================= Mechanical and Thermal Information ~ ~==============~. Quality and Reliability Assurance . High-Performance FIFO Memories Designer's Handbook :'IlExAs INSTRUMENTS . Printad on Recycled " - IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or representthat any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated Printed in U.S.A. by Custom Printing Company Owensville, Missouri INTRODUCTION The Texas Instruments (TI) High-Performance FIFO Memories Designer's Handbook has been created to provide users and potential users of TI's FIFOs with a comprehensive collection of information and data used as a reference guide during the design-in process. In a much broader sense, the designer's handbook also is a means to further the overall understanding and awareness of TI's FIFOs and their many applications. The contents of the handbook complement the information contained in the 1996 High-Performance FIFO Memories Data Book, literature number SCAD003C. In addition to an expanded series of FIFO application notes, the handbook also contains information that is useful to the deSigner, such as sample power-dissipation calculations, mechanical packaging data, thermal-resistance data, and quality/reliability assurance information. Section 6, Device Models, includes lists of available VHDL [VHSIC (very high-speed integrated circuits) hardware-description language] models and logic-modeling behavior models. This deSigner's handbook is organized into seven major sections: General Information (section 1), Product Overview (section 2), Specific Application Reports (section 3), Power Considerations (section 4), Mechanical and Thermal Information (section 5), Device Models (section 6), and Quality and Reliability Assurance (section 7). Section 1 contains a glossary of symbols, terms, and definitions that are used throughout the handbook. These symbols, terms, and definitions are presented in accordance with those currently agreed upon by the JEDEC Council ofthe Electronic Industries Association. Section 2 provides an overview of TI's FIFO products. The summary charts in this section allow quick reference of key FIFO parameters, such as architecture, organization, speed sort, maximum clock frequency, and available packaging options. The features associated with TI's advanced application-specific FIFOs are summarized in this section. Section 2 also includes a flow chart that can be used as an aid in selecting a FIFO architecture based on a known bus width. Section 3 provides a comprehensive set of FIFO application reports. The application reports are organized into subsections based on the subject matter of each report. The subsections are FIFO Performance and Reliability, FIFO Features, and FIFO Applications. Section 4 provides typical power characteristics in the form of active supply current versus frequency for each of TI's advanced FIFOs. Plots of idle supply current versus frequency also are given for a select number of FIFOs. In addition to this data, sample power-clissipation calculations are performed for a representative set of FIFOs. The application report in this section is used as a guideline for these calculations. Calculations and equations are provided for CMOS FIFOs and Advanced BiCMOS Technology (ABT) FIFOs. Section 5 contains mechanical drawings for each FIFO packaging option. The official JEDEC descriptor is used to identify each package type. These drawings typically include the following dimensions: lead pitch (tip to tip), body width and length, shoulder-to-shoulder insertion width, lead width, thickness, and angles, and package maximum height, and stand-off clearances from seating plane to bottom of the package. Included with the mechanical data is thermal data for each FIFO packaging option. Thermal resistance values for varying conditions and power-dissipation derating curyes for varying air flows are v '\ presented for each package type. Several application reports and papers also are included in this section to provide further detail regarding overall thermal considerations and thermal-resistance measurements, including the design and development of the thermally enhanced thin quad flat package (TQFP). Section 6 contains a list ofthe behavioral models currently available from TI. All speed sorts of the devices under consideration are included in the models. VHDL models of FIFO devices included in this section may be obtained by calling the Advanced System LogiC hotline at 903-868-5202. Section 7 addresses the issue of quality and reliability assurance for TI's FIFO products. Concepts such as the qualification of products and processes, quality and reliability assurance in integrated-circuit design, and quality and reliability monitoring are discussed. For further information on TI's FIFO products or applications, please contact the Advanced System Logic hotline at 903-868-5202. For information on TI's military FIFO devices, contact military Advanced System Logic marketing at 915-561-7289. vi Contents Section 1 - General Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . •. 1-1 Glossary .•...••••••.•••••••.••••••.••••..•.••••••••••••.•.••••••••••••.•••••.•••••••• 1-3 Section 2 - Product Overview. • . . . • . . . . . . . . . . • . . . . . . . . • . . . . • . . . . . . . . .. 2-1 FIFO Product Offerings ••••••••••••••••••••••..•••••.•••.••••••••••••.•••••.•••••.•••• 2-3 Advanced Clocked FIFOs ......................................................... 2-5 Advanced Strobed FIFOs ......................................................... 2-5 Advanced Application-Specific Clocked FIFOs ....................................... 2-6 Advanced Application-Specific FIFO Features ........................................ 2-6 Synchronous Mature FIFOs ....................................................... 2-6 Asynchronous Mature FIFOs ...................................................... 2-7 FIFO Functionality ................................................................ 2-8 . FIFO Selection Flow Chart •••.••••••••••••••••••.•••••••••••••••.•••••••••••.•••••••• 2-11 Sect/on 3 - Specific Application Reports . .•...•.......•.............. " 3-1 FIFO Performance and Reliability ••••••••••••••••••.•••••••••••••••••.••••••••.••••••.• 3-3 Metastability Performance of Clocked FIFOs ......................................... 3-5 FIFO Memories: Solution to Reduce FIFO Metastability •............................. 3-17 Simultaneous-Switching Noise Analysis for Texas Instruments FIFO Products ........... 3-23 FIFO Solutions for Increasing Clock Rates and Data Widths .......................... 3-41 FIFO Features •••••••••••••.••••••••.•.•••••••••••••••••••••••••.•.••••••••••••••.••. 3-51 FIFO Patented Synchronous Retrans~it: Programmable DSP-Interface Application for FIR Filtering .............................................................. 3-53 FIFO Mailbox-Bypass Registers: Using Bypass Registers to Initialize DMA Control ....... 3-65 Advanced Bus-Matching/Byte-Swapping Features for Internetworking FIFO Applications .. 3-75 Parity-Generate and Parity-Check Features for High-Bandwidth-Computing FIFO Applications ....................•............ 3-67 FIFO Applications ••••••••••••••••••••••.•••••••••••••••••••••.•••••••••••••••••••••• 3-95 Multiple-Queue First-In, First-Out Memory SN74ACT53861 ...•....................... 3-97 Interfacing TI Clocked FIFOs With TI Floating-Point Digital Signal Processors ......... 3-121 FIFOsWith a Word Width of One Bit ............................................. 3-141 Internetworking the SN74ABT3614 ..........•................................... 3-161 High-Speed, High-Drive SN74ABT7819 FIFO ..................................... 3-181 SPARC MBus-to-Futurebus+ Bridge Using the Texas Instruments Futurebus+ Chipset .. 3-193 1K x 9 x 2 Asynchronous FIFOs SN74ACT2235 and SN74ACT2236 ............... 3-207 64-Byte FIFOs SN74ALS2232A and SN74ALS2233A .............................. 3-217 Section 4 - Power Considerations • •••.•...•••••.•..•....••.•••....•... 4-1 Introduction •••••••••••••.••••••••.•••••••.•••••.•••••••••••••••••••••••••••••••••••. 4-3 Key Power-Dissipation Equations for ACT FIFOs ...............•.........•........... 4-5 Key Power-Dissipation Equations for ABT FIFOs ...................•................. 4-6 Application Report .................................................................... 4-9 Power-Dissipation Calculations for TI FIFO Products ................................. 4-11 Single-Bit FIFOs •••••••••.•••••••••••••.••••••••••••.•••••••••••••••••••••••••••••..• 4-31 Sample Power-Dissipation Calculations for SN74ACT2229 ........................... 4-33 Power Characteristics ........................................................... 4-34 SN74ACT2226 and SN74ACT2228 Single FIFO Supply Current Versus Clock Frequency .................•.................................. 4-34 SN74ACT2227 and SN74ACT2229 Single FIFO Supply Current Versus Clock Frequency .................................................... 4-34 vii Section 4 - Power Considerations (Continued) 3~Blt Clocked FIFOs •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 4~ Sample Power-Dissipation Calculations for SN74ABT3613 ...............•....•...••. 4-37 Power Characteristics •.•.....••.•.•........•.....•.•........•................... 4-3.8 SN74ABT3611 Supply Current Versus Clock Frequency .....•.......•.•...•.•..•.. 4-38 SN74ABT3612 Supply Current Versus Clock Frequency ...........•....•...••..•.. 4-38 SN74ABT3613 Supply Current Versus Clock Frequency .•.•........•.............. 4-39 SN74ABT3614 Supply Current Versus Clock Frequency .......•...•............... 4-39 SN74ABT3614 Idle Current With ClKA Switching, Other Inputs at 0 or Vec - 0.2 V and Outputs Disconnected .................................................. 4-40 SN74ABT3614 Idle Current With ClKB Switching, Ottier Inputs at 0 or Vee - 0.2 V and Outputs Disconnected .................................................. 4-40 Sample Power-Dissipation Calculations for SN74ACT3641 •..•.................•....• 4-41 Power Characteristics ........................................................... 4-42 SN74ACT3632 Supply Current Versus Clock Frequency ........................... 4-42 SN74ACT3638 Supply Current Versus Clock Frequency ......•...... , ............. 4-42 SN74ACT3631 and SN74ACT3641 Supply Current Versus Clock Frequency ..•.•..•. 4-43 SN74ACT3641 Idle Current With ClKA Switching ................................. 4-43 SN74ACT3641 Idle Current With ClKB Switching' ................................. 4-44 SN74ACT3641 Active Current With ClKA and ClKB Switching, Simultaneous ReadIWrlte and ClKB as Data Output ...•••..•.....•.......•..... 4-44 18-Blt Clocked/Strobed FIFOs •••••••••••••••••••••••••••••••••••••••••••••••••••••••• 4-45 Sample Power-Dissipation Calculations for SN74ABT7819 ....•......•....•...•...... 4-47 Power Characteristics ...................•..•................•.........•....•.... 4-48 SN74ABT7819 Supply Current Versus Clock Frequency ....•.....................• 4-48 SN74ABT7820 Supply Current Versus Clock Frequency ••..•...........•....•....• 4-48 Sample Power-Dissipation Calculations for SN74ACT7803 ...••......•.............•• 4-49 Power Characteristics •..•.•....•..........•....................................• 4-50 SN74ACT7804 Supply Current Versus Clock Frequency .............•.....•....... 4-50 SN74ACT7806 Supply Current Versus Clock Frequency •...•...........•..•......• 4-50 SN74ACT7814 Supply Current Versus Clock Frequency ......•.•.•...•............ 4-51 SN74ACT7803 Supply Current Versus Clock Frequency ......................•..•. 4-51 SN74ACT7803 Idle Current With RDClK or WRTClK Switching .•.•.•.••........... 4-52 SN74ACT7805 Supply Current Versus Clock Frequency ..........•................ 4-52 SN74ACT7813 Supply Current Versus Clock Frequency ........•........••...••... 4-53 9-Bit Clocked/Strobed FIFOs •••••••••••••••••••••••••••••••••.••••••••••••••••••••••• 4~5 Sample Power-Dissipation Calculations for SN74ACT7807 ........................... 4-57 Power Characteristics .•.........•.................................•............. 4-58 SN74ACT7807 Supply Current Versus Frequency .....•..................•....•. .4-58 SN74ACT78071dle Current With WRTClK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected .....•...•......•.... 4-58 SN74ACT7807 Idle Current With RDClK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected .....•..•..........•. 4-59 SN74ACT7808 Supply Current Versus Frequency ................................ 4-59 Section 5 - Mechanical and Thermal Information ••••••••••••••••••••••• 5-1 Comparison Summary of Advanced Packaging Derating Curves (In Stili Air) •••••••••••••• 5-3 Introduction .•.........••..•....•...............•...........................•.... 5-5 Application Reports/Papers ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 5-7 FIFO Surface-Mount Package Information ...•...••.•.•..•....•.......•.•...•....•..• 5-9 FIFO Memories: Fine-Pitch Surface-Mount Manufacturability .•.•......•.•....••.....•. 5-19 Package Thermal Considerations ................................................. 5-29 K-Factor Test-Board Design Impact on Thermal-Impedance Measurements ....•......•. 5-41 More Power in less Space: A Thermal Enhancement Solution for Thin Packages ........ 5-55 FIFO Surface-Mount Packages for PCMCIA Applications .........•.......•........... 5-69 viii Section 5 - Mechanical and Thermal Information (Continued) Package Outlines and Thermal Data •..•••..•••.••...•••..••..•••..•••••••..•.•••••..• 5-81 Introduction .................................................................... 5-83 Plastic Dual-In-Line Package (PDIP) N/R-PDIP-T16 (16-Pin DIP) .................................................... 5-84 N/R-PDIP-T20 (20-Pin DIP) .................................................... 5-86 N/R-PDIP-T28 (28-Pin DIP-600 mil) ............................................. 5-88 N/R-PDIP-T40 (40-Pin DIP) .................................................... 5-90 NP/R-PDIP-T28 (28-Pin DIP-300 mil) ........................................... 5-92 NT/R-PDIP-T24 (24-Pin DIP) ................................................... 5-94 Plastic J-Leaded Chip Carrier (PQCC) FN/S-PQCC-J20 (20-Pin PLCC) ................................................ 5-96 FN/S-PQCC-J28 (28-Pin PLCC) ................................................ 5-9S FN/S-PQCC-J44 (44-Pin PLCC) .............................................. 5-100 FN/S-PQCC-J6S (6S-Pin PLCC) .............................................. 5-102 RJ/R-PQCC-J32 (32-Pin PLCC) .............................................. 5-104 Plastic Wide-Body Small-Outline Package (PDSO) DW/R-PDSO-G16 (16-Pin SOIC) ............................................. 5-106 DW/R-PDSO-G20 (20-Pin SOIC) ............................................. 5-108 DW/R-PDSO-G24 (24-Pin SOIC) .............................................. 5-110 DW/R-PDSO-G28 (28-Pin SOIC) .............................................. 5-112 Plastic Small-Outline Package (PDSO) DV/R-PDSO-G28 (28-Pin SOIC) .......................................•....... 5-114 Plastic Shrink Small-Outline Package (PDSO) DLIR-PDSO-G56 (56-Pin SSOP) .............................................. 5-116 Plastic Quad Flatpack (PQFP) PAGlS-PQFP-G64 (64-Pin TQFP) ............................................. 5-118 PCBlS-PQFP-G120 (120-Pin TQFP) .......................................... 5-120 PHlR-PQFP-GSO (SO-Pin TQFP) ....•........................................ 5-122 PM/S-PQFP-G64 (64-Pin TQFP) ............................................. 5-124 PN/S-PQFP-GSO (SO-Pin TQFP) .............................................. 5-126 PQ/S-PQFP-G132 (132-Pin TQFP) .................................... : ....... 5-12S PZlS-PQFP-G100 (100-Pin TQFP) ............................................ 5-130 Section 6 - Device Models . ••••••••..•••...••...••......••..•••.•.•••• 6-1 VHDL Models ••.•••••••••••••.•••••••.•••••••••••••••••••••••.••......•••.••••.•.•.•• 6-3 Logic-Modeling Behavioral Models ••••••••••••••••••••••••••••.••...•.•.•••••••••..••• 6-5 Section 7 - Quality and Reliability Assurance •....•.••...•.••..•......• 7-1 Qualification of Products and Processes •••••••••••••••••••••••••••••••••••••••••••••.• 7-3 Introduction •••••••••••••••••••••••••..••••••••••••.••••.••••.••••.•.•••• , ••••.•••••• 7-5 Quality and Reliability Assurance In Integrated-Circuit Design ••••••••..••••••••••••.•••• 7-7 General Quality and Reliability Rules ............................................... 7-9 Process-Specific Design Rules ..................................................... 7-9 Methodology-Oriented Design Rules ............................................... 7-9 Change Control ............•..................................................... 7-'iJ ix Section 7 - Quality and Reliability Assurance (Continued) .............. 7-1 Quality and Reliability Monitoring ••••••••••.•••••••••••••••.•••••••.••••••••••••••••• 7-11 Environmental Laboratory •............•.....••....... ·...•........................ 7-13 Methods of Measuring Component Reliability ...•................................... 7-13 Failure-Rate Calculations for FIFO Products ........................................ 7-13 Qualification Data for FIFO Products ............................................... 7-13 Life-Test and ESD-Characterization Data ................................ , .... : ...... 7-14 x 1-1 C) CD :::s CD --Dl:::s .. 03 ao_. :::s 1-2 GLOSSARY SYMBOLS, TERMS, AND DEFINITIONS INTRODUCTION These symbols, terms, and definitions are in accordance with those currently agreed upon by the JEDEC Council ofthe Electronic Industries Association (EIA) for use in the USA and by the International Electrotechnical Commission (IEC) for international use. operating conditions and characteristics (in sequence by letter symbols) Cj Input capacitance The internal capacitance at an input of the device Co Output capacitance The internal capacitance at an output of the device Cpd Power dissipation capacitance Used to determine the no-load dynamic power dissipation per logic function (see individual circuit pages): Po = Cpd Vee2 f + Icc Vee· f max Maximum clock frequency The highest rate at which the clock input of a bistable circuit can be driven through its required sequence while maintaining stable transitions of logic level at the output with input conditions established that should cause changes of output logic level in accordance with the specification Icc Supply current The current into· the Vee supply terminal of an integrated circuit ~ICC Supply current change The increase in supply current for each input that is at one of the specified TTL voltage levels rather than o or Vee 'CEX Output high-leakage current The maximum leakage current into the collector of the pulldown-output transistor when the output is high and the output forcing condition Vo = 5.5 V 'I (hold) Input-hold current Input current that holds the input at the previous state when the driving device goes to a high-impedance state High-level input current The current into· an input when a high-level voltage is applied to that input Low-level Input current The current into· an input when a low-level voltage is applied to that input loff Input/output power-off leakage current The maximum leakage current int%ut of the input/output transistors when forcing the input/output to 4.5 V and Vee = 0 IOH High-level output current The current into· an output with input conditions applied that, according to the product speCification, establishes a high level at the output IOL Low-level output current The current into· an output with input conditions applied that, according to the product speCification, establishes a low level at the output 'Current out of a terminal is given as a negative value. ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 1-3 GLOSSARY SYMBOLS, TERMS, AND DEFINITIONS IOZ RElJA Off-state .(hlgh-Impedance-state) output current (of a 3-state output) The current flowing into* an output having 3-state capability with input conditions established that, according to the product specification, establishes the high-impedance state at the output , Junctlon-ta-ambient thermal resistance The thermal resistance from the semiconductor junction(s) to the ambient RElJC Junctlon-to-case thermal resistance The thermal resistance from the semiconductor junction(s) to a stated location on the case ta Access time The time interval between the application of a specified input pulse and the availability of valid signals at an output tc Clock cycle time Clock cycle time is 1/fmax . tcils Disable time (of a 3-state or open-collector output) The propagation time between the specified reference points on the input and output voltage waveforms with the output changing from either of the defined active levels (high or low) to a high-impedance (off) state NOTE: For 3-state outputs, tdis = tpHZ or tpLZ. Open-collector outputs change only if they are low at the time of disabling, so ~is =tpLH. ten Enable time (of a 3-state or open-collector output) The propagation time between the specified reference pOints on the input and output voltage waveforms with the output changing from a high-impedance (off) state to either of the defined active levels (high or low) NOTE: In the case of memories, this is the access time from an enable input (e.g., OE). For 3-state outputs, ten = tpZH or tpZL. Open-collector outputs change only if they are responding to data that would cause the output to go low, so ten = tpHL. th Hold time The time interval during which a signal is retained at a specified input terminal after an active transition occurs at another specified input terminal NOTES: 1. The hold time is the actual time interval between two signal events and is determined by the system in which the digital circuit operates. A minimum value is specified that is the shortest interval for which correct operation of the digital circuit is to be expected. 2. The hold time may have a negative value, in which case, the minimum limit defines the longest interval (between the release of the signal and the active tranSition) for which correct operation of the digital circuit is to be expected. tpd Propagation delay time The time between the specified reference pOints on the input and output voltage waveforms with the output changing from one defined level (high or low) to the other defined level (tpd = tpHL or tpLH) tpHL Propagation delay time, hlgh-ta-Iow level output The time between the specified reference pOints on the input and output voltage waveforms with the output changing from the defined high level to the defined low level tpHZ Disable time (of a 3-state output) from high level The time interval between the specified reference points on the input and the output voltage waveforms with the 3-state output changing from the defined high level to the high-impedance (off) state ·Current out of a terminal is given as a negative value. ~TEXAS 1-4 INSTRUMENTS POST OFACE BOX 655303 • DALLAS, TEXAS 75265 GLOSSARY SYMBOLS, TERMS, AND DEFINITIONS tpLH Propagation delay time, low-to-hlgh level output The time between the specified reference points on the input and output voltage waveforms with the output changing from the defined low level to the defined high level tpLZ Disable time (of a a-state output) from low level The time interval between the specified reference points on the input and the output voltage waveforms with the 3-state output changing from the defined low level to the high-impedance (off) state tpzH Enable time (of a a-state output) to high level The time interval between the specified reference points on the input and output voltage waveforms with the 3-state output changing from the high-impedance (off) state to the defined high level tpzL Enable time (of a a-state output) to low level The time interval between the specified reference points on the input and output voltage waveforms with the 3-state output changing from the high-impedance (off) state to the defined low level tsu Setup time The time interval between the application of a signal at a specified input terminal and a subsequent active transition at another specified input terminal NOTES: 1. The setup time is the actual time interval between two signal events and is determined by the system in which the digital circuit operates. A minimum value is specified that is the shortest interval for which correct operation of the digital circuit is to be expected. 2. The setup time may have a negative value, in which case, the minimum limit defines the longest interval (between the active transition and the application of the other signal) for which correct operation of the digital circuit is to be expected. tw Pulse duration (width) The time interval between specified reference points on the leading and trailing edges of the pulse waveform VIH High-level Input voltage An input voltage within the more positive (less negative) of the two ranges of values used to represent the binary variables NOTE: A minimum is specified that is the least-positive value of high-level input voltage for which operation of the logic element within specification limits is to be expected. VIL Low-level input voltage An input voltage within the less positive (more negative) of the two ranges of values used to represent the binary variables NOTE: A maximum is specified that is the most-positive value of low-level input voltage for which operation of the logic element within specification limits is to be expected. VOH High-level output voltage The voltage at an output terminal with input conditions applied that, according to product specification, establishes a high level at the output VOL Low-level output voltage The voltage at an output terminal with input conditions applied that, according to product specification, establishes a low level at the butput VIT+ Positive-going Input threshold level The voltage level at a transition-operated input that causes operation of the logic element, according to specification, as the input voltage rises from a level below the negative-going threshold voltage, VIT- ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 1-5 GLOSSARY SYMBOLS, TERMS, AND DEFINITIONS VIT_ Negative-going Input threshold level The voltage level at a transition-operated input that causes operation of the logic element, according to specification, as the input voltage falls from a level above the positive-going threshold voltage, VIT+ VOHV High-level output voltage change during simultaneous switching The minimum (valley) voltage induced on a quiescent high-level output during switching of other outputs VOlP Low-level output voltage change during simultaneous switching The maximum (peak) voltage induced on a quiescent low-level output during switching of other outputs definitions asynchronous FIFO Data writes are initiated by a low-level pulse on the write-enable input when the full flag Is not asserted. Likewise, data reads are initiated by a low-level pulse on the read-enable input when the empty flag is not asserted. The empty and full flags are not synchronized to a particular clock and reflect the instantaneous comparison of the read and write pointers. clocked FIFO Data is written by a low-to-high transition of a write clock when write-enable inputs are asserted and the input-ready flag is not asserted. Likewise,· data is read by a low-to-high transit!on of a read clock when read-enable inputs are asserted and the output-ready flag is asserted. The input-ready flag is multistaged synchronized to the write clock and the ouput-ready flag is multistaged synchronized to the read clock, improving metastability. strobed FIFO Data is written on a low-to-high transition on the load-<:Iock input when the full flag is not asserted. Likewise, data is read on a low-to-high transition on the unload-<:Iock input when the empty-flag is not asserted. The empty and full flags are not synchronized to a particular clock and reflect the instantaneous comparison of the read and write painters. synchronous FIFO The term synchronous refers to a port-control method and does not imply that data writes and reads must be synchronous to one another. Data is written by a low-to-high transition of a write clock when write-enable inputs are asserted and the full flag is not asserted. Likewise, data is r~ad by a low-to-high transition of a read clock when read-enable inputs are asserted and the empty flag is not asserted. The empty flag is single-staged synchronized to the read clock and the full flag is single-staged synchronized to the write clock. ~TEXAS 1-6 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 2-1 ...""D ·0 Co c 2 o <' (I) _. ; ~ 2-2 FIFO Product Offerings Page Advanced Clocked FIFOs .•......................................................... Advanced Strobed FIFOs ........................................................... Advanced Application-Specific Clocked FIFOs ......................................... Advanced Application-Specific FIFO Features ......................................... Synchronous Mature FIFOs ......................................................... Asynchronous Mature FIFOs ...................................•.................... FIFO Functionality ................................................................. 2-5 2-5 2-6 2-6 2-6 2-7 2-8 2-3 2-4 PRODUCT OVERVIEW FIFO PRODUCT OFFERINGS Advanced Clocked FIFOs DEVICE ORGANIZATION SPEED SORT to (ns) MAXIMUM FREQUENCY ACCESS TIME (ns) (MHz) IOH/lOL (mA) PACKAGE PITCH (mm) AREA (mm2) SN74ACT7884 SN74ACT7882 SN74ACT7881 4K x 18 2K x 18 1K x 18 -15, -20,-30 67,50,33 11,13,18 8/16 8o-pin TQFP (PN) 68-pin PLCC (FN) 0.5 1.27 196 310 SN74ACT7811 1K x 18 -15, -18, -20, -25 40, 35, 29, 17 15,18, 20,25 8/16 8o-pin TQFP (PN) 68-pin PLCC (FN) 0.5 1.27 196 310 SN74ACT7803 SN74ACT7805 SN74ACT7813 512 x 18 256 x 18 64 x 18 -15,-20,-25, -40 67,50,40,30 12,13, 15,20 8116 56-pin SSOP (OL) 0.635 191 SN74ACT7807 2K x 9 -15, -20, -25, -40 67,50,40,25 12,13, 18,25 8/16 64-pin TQFP (PM/PAG) 44-pin PLCC (FN) 0.5 1.27 144 310 SN74ABT7819 512 x 18 x 2 -12, -15, -20, -30 80,67,50,33 9,10,12, 14 12124 8o-pin TQFP (PN) 80-pin PQFP (PH) 0.5 0.8 196 432 512 x 18 256 x 18 64 x 18 -20, -25,-40 25,40,50 13,15,20 8/16 56-pin SSOP (OL) 0.635 191 PITCH (mm) AREA (mm2) SN74ALVC7803 SN74ALVC7805 SN74ALVC7813 Advanced Strobed FIFOs MAXIMUM FREQUENCY ACCESS IOH/lOL TIME (ns) (mA) (MHz) ORGANIZATION SPEED SORT SN74ACT7802 1K x 18 -25, -40, -60 40,25,17 30,35,45 8116 80-pin TQFP (PN) 68-pin PLCC (FN) 0.5 1.27 196 635 SN74ACT2235 SN74ACT2236 1K x 9 x 2 -20, -30, -40, -60 50,33,25,17 25,25, 35,45 8/16 64-pin TQFP (PM/PAG) 44-pin PLCC (FN) 0.5 1.27 144 310 SN74ACT7804 SN74ACT7806 SN74ACT7814 512 x 18 256 x 18 64 x 18 -20, -25, -40 50,40,25 15,18,20 8/16 56-pin SSOP (OL) 0.635 191 SN74ACT7808 2K x 9 -20, -25, -30, -40 50,40,33,25 15,18, 20,22 8116 64-pin TOFP (PM/PAG) 44-pin PLCC (FN) 0.5 1.27 196 635 SN74ABT7820 512 x 18 x 2 -15,-20,-25, -40 67,50,40,33 12,14, 15, 17 12124 80-pin TQFP (PN) 8o-pin PQFP (PH) 0.5 0.8 196 432 512 x 18 256 x 18 64 x 18 -20, -25,-40 25,40,50 13,15,20 8/16 56-pin SSOP (OL) 0.635 191 DEVICE SN74ALVC7804 SN74ALVC7806 SN74ALVC7814 to (ns) PACKAGE ~ThXAS INSTRUMENTS POST OFFICE BOX 655303 • DALlAS. TEXAS 75265 2-5 PRODUCT OVERVIEW FIFO PRODUCT OFFERINGS Advanced Appllcatlon-8peclflc Clocked FIFOs DEVICE ORGANIZATION SPEED SORT (ns) to MAXIMUM FREQUENCY (MHz) ACCESS TIME (ns) IOH/IOL (mA) PACKAGE PITCH (mm) AREA (mrn2) 165 SN74ACT2226 SN74ACT2228 64 x 1 256 x 1 22 20 8/16 24-pin SOIC (OW) 1.27 SN74ACT2227 SN74ACT2229 64 x 1 256 x 1 60 9 8/16 28-pin SOIC (OW) 1.27 192 0.4 0.635 256 781 SN74ACT3638 512 x 32 x 2 -15.-20. -30 67.50.33 11.13.15 4/8 12O-pin TQFP (PCB) 132-p1n PQFP (PQ) SN74ACT3622 SN74ACT3632 SN74ACT3642 256x36x2 512 x 36 x 2 1K x 36 x 2 -15.-20.-30 67.50.33 11.13.15 4/8 12D-pin TQFP (PCB) 132-pin PQFP (PQ) 0.4 0.635 256 781 SN74ACT3631 SN74ACT3641 SN74ACT3651 512 x 36 1K x 36 2K x 36 -15.-20.-30 67.50.33 11.13.15 4/8 120-pin TQFP (PCB) 132-p1n PQFP (PQ) 0.4 0.635 256 781 SN74ABT3611 SN74ABT3613 64 x 36 64 x 36 -15. -20. -30 67.50.33 10.12.15 4/8 12D-pin TQFP (PCB) 132-pin PQFP (PQ) 0.4 0.635 256 781 SN74ABT3612 SN74ABT3614 64x36x2 64x36x2 -15.-20.-30 67.50.33 10.12.15 418 120-pin TQFP (PCB) 132-pin PQFP (PQ) 0.4 0.635 256 781 : Advllnced Application-Specific FIFO Features TELECOMMUNICATIONS FIFOs SN74ACT2226 SN74ACT2228 SN74ACT2227 SN74ACT2229 Ouallndependent FIFO with separate output enables. separate 1/0. separate resets. characterized to industrial temperature specification: -40·C to 85·C DIGITAL-SIGNAL-PROCESSING FIFOs SN74ACT3638 SN74ACT3631 SN74ACT3641 SN74ACT3651 Microprocessor interface-controllogic. synchronous retransmit capability. mailbox-bypass registers for each FIFO SN74ACT3622 SN74ACT3632 SN74ACT3642 Microprocessor interface-controllogic. mailbox-bypass registers for each FI FO HIGH-BANDWIDTH COMPUTING FIFOs SN74ABT3611 SN74ABT3612 Microprocessor interface-controllogic. parity generation and parity check. mailbox-bypass registers for each FIFO INTERNETWORKING FIFOs SN74ABT3613 SN74ABT3614 Microprocessor interface-controllogic. parity generation and parity check. bus matching and byte swapping. mallbox-bypass registers for each FIFO Synchronous Mature FIFOs DEVICE SN74ACT72211L SN74ACT72221L SN74ACT72231 L SN74ACT72241 L ORGANIZATION , 512 x 1K x 2K x 4K x 9 9 9 9 SPEED SORT (n8) MAXIMUM FREQUENCY (MHz) ACCESS TIME (n8) -15. -20. -25. -50 67. 50. 40. 20 10.12. 15.25 to ~1ExAs 2-6 INSTRUMENTS' POST OFFICE eox 665303 • DALlAS. TEXAS 75266 PACKAGE 32-pin PLCC (RJ) PRODUCT OVERVIEW FIFO PRODUCT OFFERINGS Asynchronous Mature FIFOs DEVICE SN74ACT7200L SN74ACT7201 LA SN74ACT7202LA SN74ACT7203L SN74ACT7204L ORGANIZATION SPEED SORT tc (ns) MAXIMUM FREQUENCY ACCESS TIME (ns) (MHz) 256 x 9 -15,-25,-50 67,40,20 10,12,20 512x9,1Kx9 -15,-25,-35, -50 67,50,40,20 10,12, 15,25 -15, -25, -50 67,40,20 10,12,20 2K x 9 4K x 9 PACKAGE 28-pin DIP (NP) 28-pin SOIC (DV) 32-pin PLCC (RJ) SN74ACT7205L 8K x 9 -15,-25,-50 67,40,20 10,12,20 28-pin DIP (NP) SN74ACT7206 16K x 9 -15,-25,-50 67,40,20 10,12,20 32-pin PLCC (RJ) SN74ALS2238 32x9x2 40 40-pin DIP (N) 44-pin PLCC (FN) SN74ALS2233A 64 x 9 40 28-pin DIP (N) 44-pin PLCC (FN) SN74ALS2232A 64 x 8 40 24-pin DIP (NT) 28-pin PLCC (FN) SN74ALS235 64 x 5 25 20-pin DIP (N) 20-pin SOIC (DW) SN74ALS233 16 x 5 30 20-pin DIP (N) 20-pin SOIC (DW) SN74ALS232B 16 x 4 30 16-pin DIP (N) 16-pin SOIC (DW) 20-pin PLCC (FN) SN74ALS229B 16 x 5 30 20-pin PLCC (FN) 20-pin DIP (N) 20-pin SOIC (DW) SN74S225 16 x 5 10 20-pin DIP (N) -!!1TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS, TEXAS 75265 2-7 FIFO Functionality ~ ~ 'il' s. I • ! ~f c. DEVICE ~ ~~ ~~ !fr !I'~ iii ~ ill8l II i r j B ~ f! .IJ II c !: iii I I I t ti3E cI a iD- i~ I Jl I r/ r/ r/ r/ r/ r/ r/ r/ :ra- 0 15 ID ... .5 !a ~ 8'" ~!! ;t ... i ! l Do I iE I! l Do Iiii 'C .z ""0 :::&:J j ~g Ii I Ii .!!II :1.5 i:I II I! •c «1 ~ :. .. E II Ia:: II ... a. ... C J... ia:: &.. P .cl!! J :. ol! IDa:: ID 0 i: 1 a:: S.1S l I .c »!l ID 1 ... J! •E ~ l Do oC c: ~ aI ~ m < !:i :::&:J ~ SN74ACT2226 64 1 20 r/ SW4ACT2227 64 1 9 til SN74ACT2226 256 1 20 r/ SN74ACT2229 256 1 9 r/ r/ SN74ALS232 16 4 23 r/ r/ SN74ALS234 64 4 17 r/ r/ r/ SN74ALS236 64 4 17 r/ r/ r/ SN74ALS229 16 5 30 r/ r/ SN74ALS233 16 5 30 r/ r/ SN74S225 16 5 75 r/ r/ r/ SN74ALS235 64 5 17 r/ r/ r/ SN74Al.S2232 64 8 26 r/ r/ SN74ALS2238 32 9 33 r/ r/ SN74ALS2233 64 9 26 r/ r/ SN74ACT7200. 256 9 15 r/ SN74ACT72211 512 9 10 r/ SN74ACT7201 512 9 15 r/ SN74ACT72221 1K 9 10 r/ SN74ACT2235 1K 9 25 r/ r/ r/ r/ i SN74ACT2236 lK 9 25 r/ r/ r/ r/ i SN74ACT7202 1K 9 15 SN74ACT7807 2K 9 12 SN74ACT72231 2K 9 10 SN74ACT7806 2K 9 15 SN74ACT7203 2K 9 SN74ACT72241 4K 9 SN74ACT7204 4K 9 15 r/ r/ r/ SN74ACT7205 8K 9 15 r/ r/ r/ SN74ACT7206 16K 9 15 SN74ACT7813 64 18 12 r/ r/ ! i r/ r/ I r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ r/ 15 r/ r/ 10 r/ r/ I r/ I r/ i r/ r/ r/ r/ r/ r/ r/ r/ * r/ r/ - - -~ FIFO Functionality (Continued) ... 'Ii" . s- .. I~g ~~~ ~{iJ ~ % ~~ .5 ... t! ::I ~ ..,. . C E ::& 'ii ~ ..,i l! t c ~ ... ...c E ~ I ! i C SN74ALVC7813 64 18 13 t/ t/ SN74ACT7814 64 18 15 t/ t/ SN74ALVC7814 64 18 18 t/ t/ SN74ACT7805 258 18 12 t/ t/ t/ SN74ALVC7805 258 18 13 t/ t/ t/ SN74ACT7806 258 18 15 t/ t/ SN74ALVC7806 258 18 18 t/ t/ SN74ACT7803 512 18 12 t/ t/ t/ SN74ALVC7803 512 18 13 t/ t/ t/ SN74ACT7804 512 18 15 t/ t/ SN74ALVC7804 512 18 18 t/ t/ SN74ABT7819 512 18 9 t/ t/ SN74ABT7820 512 18 12 t/ t/ SN74ACT7881 lK 18 11 t/ t/ t/ t/ SN74ACT7811 lK 18 15 t/ ,; t/ t/ SN74ACT7802 lK 18 30 t/ t/ SN74ACT7882 2K 18 11 t/ t/ t/ t/ SN74ACT7884 4K 18 11 t/ t/ t/ t/ SN74AC753881 4K 18 11 t/ ,; t/ SN74ACT3638 512 32 11 ,; ,; SN74ABT3611 64 36 10 t/ SN74ABT3613 64 36 10 t/ SN74ABT3612 64 36 10 t/ t/ SN74ABT3614 64 36 10 t/ SN74ACT3622 256 36 11 t/ t/ SN74ACT3631 512 36 11 t/ t/ SN74ACT3632 512 36 11 t/ t/ SN74ACT3641 lK 36 11 t/ t/ SN74ACT3642 lK 36 11 t/ t/ SN74ACT3651 2K 36 11 ,; t/ ~ ~~ It iL ie. DEVICE ~ E ;:: ~ 5! ~ 0 * .., 5 :I 80 111~!. go :z:~ Bidirectional configurable without additional logic 0 I .. -8 C 0 ~ ~Q. c'J'! t/ t/ i:i! ID * * * * * U 'ii .. c oh ~ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ .5 ::I .5 E E I! ~ IL ... 'ii aJ > ... t/ t/ t/ t/ ~ ;: ~ 0 ....I .. g ic . II ... B 'iii It U ID IE ::& U .., .. J!I I! c CI b 'C :. ... !u . a: c ::I II I! li a: b .., :. a: 'C :I 11 :I 0= CE e .. ~c ol! ~li ilia: ::& ID ID t . i ...C 'li. ... .i ...c c s E E 0 .!! :;: .. OJ ::I ~to I! : . ::I ::I ... ~ § IL ::& t/ ,; t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ '" t/ t/ t/ t/ t/ '" t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ t/ ""0 :J:J oC c: t/ ~ '" t/ m t/ t/ m t/ ~ ~ :e 2-10 FIFO Selection Flow Chart 2-11 2-12 PRODUCT OVERVIEW FIFO SELECTION FLOW CHART This chart can be used to select the appropriate strobed or clocked FI FO for the application based on the desired FI FO width. Which FIFO? Require More Than One 9-Blt FIFO (Cascading)? Y N Faster Than 33 MHz? N Y N Faster Than 33 MHz? N Y ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 2-13 2-14 3-1 tn "'C (1) (') _. _. --- (') l> "'C "'C -_. (') ...._. Q) o ::s ::0 CD "'C o S 3-2 FIFO Performance and Reliability Page Metastability Performance of Clocked FIFOs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-5 FIFO Memories: Solution to Reduce FIFO Metastability ................................ 3-17 Simultaneous-Switching Noise Analysis for Texas Instruments FIFO Products ............ 3-23 FIFO Solutions for Increasing Clock Rates and Data Widths ............................ 3-41 3-3 Metastability Performance of Clocked FIFOs First-In, First-Out Technology Chris Wellheuser Advanced System Logic - Semiconductor Group ~TEXAS INSTRUMENTS SCZAOO4A IMPORTANT NOTICE Texas Instruments (TIl reserves the right to make changes to its products or to discontinue any . semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications',). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant orrepresent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-6 Contents Title Page Introduction .............................................................................. 3-9 Metastability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-9 TI Clocked FIFOs ........................................................................ 3-11 Test Setup for Measuring FIFO Flag Metastability ................. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-12 Test Results .............................................................................. 3-14 MTBF Comparisons ...................................................................... 3-15 Conclusion .............................................................................. 3-16 References ............................................................................... 3-16 3-7 3-8 Introduction This report is intended to help the user understand more clearly the issues relating to the metastable performance of Texas Instruments (TI) clocked FIFOs in asynchronous-system applications. It discusses basic metastable-operation theory, shows the equations used to calculate metastable failure rates for one and two stages of synchronization, and describes the approach TI has used for synchronizing the status flags on its series of clocked FIFOs. Additionally, a test setup for measuring the failure rate of a device to determine its metastability parameters is shown and results are given for both an advanced BiCMOS (ABT) FIFO and an advanced CMOS (ACT) FIFO. Using these parameters, calculations of MTBF under varying conditions are performed. Metastability Metastability in digital systems occurs when two asynchronous signals combine in such a way that their resulting output goes to an indeterminate state. A common example is the case of data violating the setup and hold specifications of a latch or a flip-flop. In a synchronous system, the data always has a fixed relationship with respect to the clock. When that relationship obeys the setup and hold requirements for the device, the output goes to a valid state within its specified propagation delay time. However, in an asynchronous system, the relationship between data and clock is not fixed; therefore, occasional violations of setup and hold times can occur. When this happens, the output may go to an intermediate level between its two valid states and remain there for an indefinite amount of time before resolving itself or it may simply be delayed before making a normal transition 1. In either case, a metastable event has occurred. Metastable events can occur in a system without causing a problem, so it is necessary to define what constitutes a failure before attempting to calculate a failure rate. For a simple CMOS latch, as shown in Figure I, valid data must be present on the input for a specified period of time before the clock signal arrives (setup time) and must remain valid for a specified period of time after the clock transition (hold time) to assure that the output functions predictably. This leaves a small window of time with respect to the clock (to) during which the data is not allowed to change. If a data edge occurs within this aperture, the output may go to an intermediate level and remain there for an indefinite amount of time before resolving itself either high or low, as illustrated in Figure 2. This metastable event can cause a failure only if the output has not resolved itself by the time that it must be valid for use (for example, as an input to another stage); therefore, the amount of resolve time allowed a device plays a large role in calculating its failure rate. Data - - - - - - - 1 Input .... .x)---~I------ Clock - - - - - - - - ' Figure 1. A Simple CMOS Latch * 1,...---------j4-- tsu -+I Clock ~·tw-~ I j4- 1 Data Input Output 1 -----'. 1 I th--l I I1414---- tr I ~I I _______~I____~I____~/~--------~i(r------ Figure 2. Output at Intermediate Level Due to Data Edge Within to Aperture Output The probability of a metastable state persisting longer than a time, tr. decreases exponentially as 11: increases2. This relationship canbe characterized by equation 1: . f - (r) - e( - tr/t) (1) where the function f(r) is the probability of nonresolution as a function of resolve time allowed, tr. and the circuit time constant t (which has also been shown to be inversely proportional to the gain-bandwidth product of the circuit)3,4. For a single-stage synchronizer with a given clock frequency and an asynchronous data edge that has a uniform probability density within the clock period, the rate of generation of metastable events can be calculated by taking the ratio of the setup and hold time window previously described to the time between clock edges and multiplying by the data edge frequency. This generation rate of metastable events coupled with the probability of nonresolution of an event as a function of the time allowed for resolution gives the failure rate for that set of conditions. The inverse of the failure rate is the mean time between failure (MTBF) of the device and is calculated with the formula shown in equation 2: failure rate (2) Where: 11: t - resolve time allowed in excess of the normal propagation delay time of the device - metastability time constant for a fJip-flop to - a constant related to the width of the time window or aperture wherein a data edge triggers a metastable event fc - clock frequency fd asynchronous data edge frequency The parameters to and t are constants that are related to the electrical characteristics of the device in question. The simplest way to determine their values is to measure the failure rate of the device under specified conditions and solve for them directly. If the failure rate of a device is measured at different resolve times and plotted, the result is an exponentially decaying curve. When plotted on a semilogarithmic scale, this becomes a straight line the slope of which is equal to t; therefore, two data points on the line are sufficient to calculate the value of t using equation 3: tr2 - trl t = In(Nl/N2) (3) Where: 11:1 - resolve time 1 11:2 - resolve time 2 Nl - number of failures relative to 11:1 N2 - number of failures relative to 11:2 Mter determining the value for t, to may be solved for directly. The formula for calculating the MTBF of a two-stage synchronizer, equation 4, is merely an extension of equation 2: (4) Where: 11:1 - resolve time allowed for the fIrst stage of the synchronizer 11:2 - resolve time allowed in excess of the normal propagation delay fco fd, t, and to are as previously defmed, with t and to assumed to be the same for both stages. 3--10 The first tenn calculates the MTBF of the fust stage of the synchronizer, which in effect becomes the generation rate of metastable events for the next stage. The second tenn then calculates the probability that the metastable event will be resolved based on the value ofta, the resolve time allowed external to the synchronizer. The product of the two tenns gives the overall MTBF for the two-stage synchronizer. TI Clocked FIFOs The TI clocked FIFOs are designed to reduce the occurrence of metastable errors due to asynchronous operation. This is achieved through the use of two- and three-stage synchronizing circuits that generate the status-flag outputs input ready (IR) and output ready (OR). In a typical application, words may be written to and then read from the FIFO at varying rates independent of one another, resulting in asynchronous flag-signal generation (internally) at the boundary conditions of full and empty; for example, the operation when the FIFO is at the full boundary condition with writes taking place faster than and asynchronous to reads. The IR flag is low, signifying that the FIFO is full and can accept no more words. When a read occurs, the FIFO is no longer completely full. This causes an internal flag signal to go high, allowing another write to take place. Since the exit from the full state happens asynchronously to the write clock (WRTCLK) of the FIFO, this flag is not useful as a system write-enable signal. The solution is to synchronize this internal flag to the write clock through two D-type flip-flop stages and output this synchronized signal as the IR flag (see Figure 3). The OR status flag is generated in a similar manner at the empty boundary condition and is synchronized to the read clock through a three-stage synchronizing circuit. Internal Asynchronous Flag Signal IR CLK Internal Logic Delay CLK WRTCLK----~--------------------------------------~ Figure 3. IR-Flag Synchronizer The remainder of this report pertains to the metastability perfonnance of the two-stage IR synchronizer, which is the limiting case of the two in tenns of MTBF characteristics. The internal flag signal that goes high on a read and low on a write is synchronized to the write clock through two D-type flip-flop stages. Since this results in the IR flag status of the FIFO being delayed for two clock cycles, a predictive circuit is used to clock the status into the synchronizer at (full minus two) words so that the action of the IR flag going low coincides with the actual full status of the FIFO. However, once the FIFO is full and IR is low, a read that causes the internal flag to go high is not reflected in the status of the IR flag until two write clocks occur. With the FIFO full and the IR flag low, a read causes the internal flag signal to go high. This signal is clocked into the ftrSt stage of the two-stage synchronizer on the next write clock. Because these two signals are asynchronous to one another, the potential for the output of the first stage of the synchronizer to go to a metastable state exists. If this condition persists until the next write clock rising edge, a metastable condition could be generated in the second stage and reflected on the IR flag output. This metastable condition manifests itself as a delay in propagation time and is considered a failure only if it exceeds the maximum delay allowed in a design. The effectiveness of the two-stage synchronizer becomes apparent when attempting to generate failures at a rate high enough to count in a reasonable period of time. A metastable event generated in the ftrSt stage must persist until the next write clock, i.e., when that data is transferred to the second stage. The resolve time for the first stage is governed by the frequency or period of the write clock. At slower frequencies, the failure rate of the first stage is very low, resulting in a low metastable generation rate to the second stage. The second stage of the synchronizer further reduces the probability of a metastable failure based on the resolve time allowed at the output. The overall failure rate of the device may be affected by increasing the initial asynchronous data generation rate (adding jitter to the data centered about the setup and hold window), by decreasing the resolve time of the first stage (increasing the write clock frequency), and by reducing the external resolve time at the output. 3-11 Test Setup for Measuring FIFO Flag Metastability The failure rate of a device is measured on a test fixture as shown in Figure 4. The input waveforms used on this setup are also shown in Figure 4. Rising data is jittered asyilchronously about the setup and hold aperture of the device under test (DUT) in a ±400-ps window with respect to the device clock (CLK). The output of the OUT is then clocked into two separate flip-flops, FFI and FF2, by two different clock signals, CLKI and CLK2. The resolve time, tp is set by the relationship between CLKI and CLK and is measured as the delta between the normal output transition time and the rising edge ofCLKl minus the setup time required for FFl. CLK2 occurs long enough afterCLKl to allow sufficient time for the OUT to have resolved itself to a valid state. The outputs ofFFI and FF2 are compared by the exclusive OR gate, the output state of which is latched intoFF3 by CLK3. When a metastable failure occurs, the output ofthe exclusive OR gate goes high caused by FFl and FF2 having opposite data due to the OUT not having resolved itself by time tr. On the next cycle, low data is clocked into the OUT and FFl and FF2 in order to reset the status latch, FF3. Failures are counted for different resolve times, and t is then calculated using equation 3. Using the test setup in Figure 4, failure rates are measured for both an SN74ABT7819, 512x 18 x 2 clocked FIFO, and an SN74ACT7807, 2K x 9 clOCked FIFO. The device is initially written full to set IR low at the boundary condition. A read clock is generated to send the internal flag high, and ajitter signal is superimposed on it to sweep asynchronously with respect to the write clock in an 800-ps-wide envelope and centered such that the IR flag goes high alternately on the second and third write clocks. The nominal write-clock frequency of the test setup is 40 MHz, but to increase the failure rate to an observable level, a pulse is injected into the write-clock stream just after the read clock occurs such that the first and second write clocks (the ones that clock the status through the synchronizer) are only 5.24 ns apart. This increases the effective write clock frequency to 191 MHz, reducing the resolve time allowed in the first stage and increasing the failure rate. This test setup and these actions together create the.necessary conditions to generate a metastable occurrence on the IR output that is seen after the second write clock and manifests itself as a delay in propagation time. In this instance, the write clock is the syilchronizing clock and·the read clock generates the asyilchronous internal data signal. CLKI is adjusted to vary the external resolve time, ta, and the resulting failure rates are recorded (see Table 1). 3-12 RClK DUT (FIFO) ClK D Q ClK ClK FF1 FF2 ClK1 ClK2 ClK3 ...-----Jltter RDClK (data) WRTClK (clock) Metastable Event IR (out) //II 7 X' __\..10--.__ i I ClK1 ClK2 load FF1 Reset FF1 ___________.~_~~~~~ _____~r_l.- _______ _ ____'r_l.- load FF2 Reset FF2 ---'nL.-~ __________________ __ load FF3 ClK3 ~n~~ Reset FF3 ~~ Figure 4. Metastable Event Counter and Input Waveforms 3-13 Test Results Table 1. SN74ABT7819 Failure Ratest RESOLVE TIME, tr2 (ns) NUMBER OF FAILURESIHOUR NUMBER OF FAILURES/SECOND MTBF (seconds) 0.27 890 0.2472 4.04 0.39 609 0.1692 S.91 0.53 396 0.1101 9.08 t VCC = 4.S V. TA = 2SoC After measuring the metastable performance of the SN74ABTI819, some assumptions must be made to calculate the parameters't and to. Because the individual flip-flops comprising the two-stage synchronizer cannot be measured separately, it is first assumed that the values fon and to are the same for both. This is a safe assumption, as these constants are driven by the process technology and because the schematics are identical. The other assumption made involves determining the resolve time allowed in the first stage of the synchronizer. The clock period is set at 5.24 ns, but the delay through the flip-flop and the setup time to the next stage must be subtracted from the clock period to arrive at the true resolve time (tel)' These values could not be measured directly and were, therefore, estimated from SPICE analysis to be 1.3 ns. Using equation 4 and the measured failure rates to calculate't results in a value of 0.33 ns for the conditions given. The following values from the test setup must be used to solve for to: Where: tel ta - 3.94 ns (5.24-ns clock period - 1.3-ns setup and delay time) = 0.27 ns (set externally at IR output by CLK1) fc fd MTBFl - 40MHz 125 MHz (4-MHz input adjusted by 25/0.8 jitter ratio) 4.04 s Substituting these values into equation 4 and solving for to yields a value of 16.9 ps. Table 2 summarizes the results for the SN74ABTI819 and SN74ACTI807 clocked FIFOs. An internal setup and delay time of 1.8 ns was assumed for the SN7 4ACTI807. Table 2. Values of 't and to for SN74ABT7819 and SN74ACT7807 TA 2SoC vcc SN74ABT7819 ~ (ns) to (ps) SN74ACT7807 ~ (ns) to (ps) 4.SV 0.33 16.9 O.SO 1.13 SV 0.30 7 0.40 2.0S S.SV 0.23 28.8 0.30 9.40 These numbers indicate the performance of only a few devices and are not intended to represent a fully characterized parameter. However, they should be valid for the purpose of relative performance comparisons, and the values do fall within the expected range given the circuit configuration and process technology in which the devices are fabricated. 3-14 MTBF Comparisons With the constants t and to now known, calculations of the MTBF of the device under different operating conditions may be performed. First, however, consider an example of the metastability performance of a single-stage synchronizer using equation 1 and the circuit constants t and to from Table 2. Assume an application running with a 33-MHz write clock, an S-MHz read clock, a 9-ns maximum propagation delay time for the IR path, and a 5-ns setup time for IR to the next device. Therefore: tr = 16 ns (30-ns clock period - 9-ns propagation delay - 5-ns tsu) fc = 33 MHz fd - SMHz Using equation 2 to calculate the MTBF gives 2.55 y 1017 seconds or a little bit more than S billion years. The reliability of a one-stage synchronizer degrades as operating frequency increases. With a 50-MHz write clock, a 12-MHz read clock, a 9-ns maximum delay, and a 5-ns setup time: tr - 6 ns (20-ns clock period - 9-ns propagation delay - 5-ns tsu) fc - 50MHz fd - 12MHz Substituting these values into equation 2 yields an MTBF of about 2 hours. This performance is unacceptable, even with a device fabricated in the O.S-mm BiCMOS process, which is more resistant to metastability than other processes. The benefits of two-stage synchronization become evident with the next example. Using the conditions stated in the last example: tr1 IS.7 ns (20-ns clock period - 1.3-ns setup and delay time) 6 ns (20-ns clock period - 9-ns propagation delay - 5-ns tsu) fc - 50MHz fd - 12MHz tr2 Using equation 4 to calculate the MTBF gives 3.16 y 1028 seconds or 1.00 y 1021 years. Table 3 gives a performance summary of both one- and two-stage synchronizing solutions under different conditions. Table 3. MTBF Comparlsonst CONDITIONS fe=33 MHz. fd=8MHz fe=40 MHz. fd-10MHz ACT 1 STAGE 8400 years ABT1 STAGE 8.1 x 109 years 92 days 1400 years ACT 2 STAGE 2.62 x 1028 years 3.S6 x 1019 years 2 hours 4.90 x 1010 years fe= SO MHz. fd=12MHz fe= 67 MHz. fd = 16 MHz fe= 80 MHz. fd=20 MHz 417 years ABT 2 STAGE 4.77 x 1047 years 2.18 x 1034 years 1.00 x 1021 years 1.28 x 109 years 2900 years t Assumptions for the MTSF comparisons: - The values for to and t are those given previously for both the AST and ACT devices with VCC- 4.S V. TA _ 2SoC. - Flag propagation delay time (IR or OR) is assumed to be 9 ns. -Setup times to the next device are S ns (up to SD-MHz operation). 4 ns (up to 67-MHz operation). and 3 ns (up to 80-MHz operation). 3-1S Conclusion Metastability failures must be accounted for in the design of asynchronous digital circuits. These failures become increasingly prevalent at higher operating frequencies. When higher frequencies are used, extreme care must be taken to ensure that system reliability is not adversely affected due to inadequate synchronization methods. ClockedFlFOs from 11 provide a solution to this problem by synchronizing the boundary flags with at least two flip-flop stages to improve the metastable MTBF over one-stage synchronization. This architecture allows designers to utilize the high-throughput performance of the memory without endangering the reliability of their end products. References 1. J. Horstmann, H. Eichel, and R. Coates, "Metastability Behavior of CMOS ASIC Flip-Flops in Theory and Test," p. 146, IEEE Journal of Solid State Circuits, February 1989. 2. H. Veendrick, "The Behavior of Flip-Flops Used as Synchronizers and Prediction of Their Failure Rate," p. 169, IEEE Journal of Solid State Circuits, April 1980. 3. S. T. Flannagan, "Synchronization Reliability in CMOS Technology," p. 880, IEEE Journal of Solid State Circuits, August 1985. 4. T. Kacprzak and A. Albicki, "Analysis of Metastable Operation in RS CMOS Flip-Flops," p. 59, IEEE Journal of Solid State Circuits, February 1987. 5. L. Kleeman and A. Cantoni, "Metastable Behavior in Digital Systems," p. 4, IEEE Design and Test of Computers, December 1987. 3-16 FIFO Memories: Solution to Reduce FIFO Metastability First-In, First-Out Technology Tom Jackson Advanced System Logic - Semiconductor Group ~TEXAS .INSTRUMENTS SCAA011A 3-17 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is uriderstood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance; customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-18 As system operating frequencies continue to increase in excess of 33 MHz, designers must begin to address the issues of overall system reliability due to increased chance of a metastable event occurring. A metastable event is defined as the time period when the output of a logic device is neither at a logic high nor at a logic low but rather in an indeterminate level. The chancel of a metastable occurrence is exponentially increased if single-stage synchronization is employed, as in the case of the '722xx synchronous-style devices versus the two-stage synchronization that is implemented by Texas InstrumentS (11) (see Figure 1). The following information assists designers in understanding and improving upon the metastable characteristics of '722xx synchronous-style devices and their reliability. - 1.00E+67 1.00E+52 1.00E+47 ,- •• 1.00E+42 1.00E+37 II) 1.00E+32 ... 1~OOE+27 :Ii 1.00E+22 I CD .... .'" 0 1.00E+17 1.00E+12 0 1.00E+07 1.00E+02 o .I' o 10 20 30 40 60 Two-Stage Synchronization On.Stage Synchronization I u 1.00E-43 1.00E-48 • 080 70 Frequency - MHz Figure 1. MTBF for Metastability as a Function of Frequency Metastability may occur when using a FIFO to synchronize two digital signals operating at different frequencies. This type of application is a familiar one to many design engineers. Triggering a metastable event is common in single-stage (single flip-flop) synchronized FIFOs that are used to synchronize different clock signals (see Figure 2). With this method, the asynchronous input might change states too close to the clock transition, violating the flip-flop's setup and hold times. This causes an increase in resolve time (tc) which then results in an overall increase in propagation delay (tpd)' Once a metastable event is triggered, the probability of the output recovering to a high or low level increases exponentially with the increased resolve time. The expected time until the output of a single flip-flop with asynchronous data has a metastable event is described by the mean time between failure (MTBF) equation (see equation 1). The first term of the equation is the probability that the asynchronous data will trigger a metastable event. The second term is the data rate. The third and fmal term is the probability of the metastable event recovering given the resolve time. A linear increase in resolve time exponentially increases the MTBF of a metastable event. 3-19 Clock Asynchronoua Flag Clock D / tau I I I t :::;!~ f/) Asynchronous Flag Synchronized Flag Synchronized Flag Q ~ tpd \ r { ~tr--t/ Figure 2. Single-Stage Synchronizer MTBFl = _1_ x .1 x exp(!t) tofe fd 't (1) Where: to tr - flip-flop constant representing the time window during which changing data invokes a failure resolve time allowed in excess of the normal propagation delay 't - flip-flop constant related to the settling time of a metastable event fc clock frequency fd - asynchronous data frequency (for OR-flag analysis, it is the frequency at which data is written to empty memory; for IR-flag analysis, it is the frequency at which data is read from full memory). 11 has increased the metastable MTBF by several orders of magnitude over single-stage synchronization with its advanced FIFO family by employing two-stage synchronization (see Figure 3). The output of the first flip-flop is clocked into the second flip-flop on the next clock cycle. For the output of the second stage to become metastable, the first stage must have a metastable event that lasts long enough to encroach upon the setup time of the second stage. The addition of the second flip-flop to the single-stage synchronizer allows the flip-flops more time to resolve any metastable output. This is statistically equivalent to increasing its resolve time by the clock period minus its propagation delay. MTBF for a two-stage synchronizer is given in equation 2. All terms, except for the third one, are the same as in equation 1. The third term represents the additional propagation delay through the added flip-flop. [t -~] 1 1 (tr) MTBF . 2 = -x-xexp - - xexptofe fd 't 't tpd MTBF2 Where: 3-20 propagation delay through the first flip-flop MTBFI (2) Q Clock Aaynchronous Flag Clock Asynchronous Flag First Stage ;:@~ ----~£ I Synchronized Flag Synchronized Flag D D I ---I Q ''--___r ''--___I { / ~------~ I -. tpd If- tr --It f7///////////$d ! High or Low (does not mattar) Figure 3. Two-Stage Synchronizer The functional block diagram in Figure 4 illustrates the connections necessary to add the second-stage synchronization to the '72211 synchronous FIFO. A quick and inexpensive schematic to resolve metastability of a synchronous FIFO is shown in Figure 5. In this case, the FIFO is the '72211U and, by implementing a single TI SN74F74 D-type positive-edge-triggered flip-flop and a TI SN74F08 two-input positive AND gate, the metastability characteristics of this circuit can be dramatically improved. The TI SN74F74 acts as the second stage for this circuit, increasing the resolve time as described in the previous paragraphs. The TI SN74F08 is implemented to act as the control-empty and control-full flags to the receiving device. These control lines of the first-stage and second-stage synchronized flags are then ANDed together to create the control flags (control empty and control full). The control lines are essentially read enables that ensure the synchronization of the device. As is shown in the logic diagram and truth table, synchronization is complete only when the empty flags (EF) of both the second stage (truth table input A) and the device (truth table input B) are high. The empty flag is used for read control and the full flag (FF) is used for write control. If either flag from the synchronizer or the device is held low or becomes metastable, a read is not permitted (truth table output Y) until the write flag is synchronized. As can be seen in today's digital systems, synchronous and asynchronous operations can and will produce random errors due to metastability in single-stage FIFO designs like those of the '722xx synchronous FIFO family. The described method of implementing a second stage for flag synchronization is extremely useful for clock speeds that are either approaching or exceeding 33 MHz. Metastability can be virtually eliminated in the '722xx synchronous FIFO family by the simple addition of a second flip-flop. The second-stage synchronizer greatly reduces metastability, thereby increasing the MTBF and allowing designers to use faster microprocessors and higher data-transfer rates for greater overall system performance and reliability. To reduce metastability and improve system reliability, TI offers a complete line of high-performance FIFO memory devices. TI's FIFOs have dual-stage synchronization designed onto each chip. This eliminates the need for any external discrete solution and reduces critical board space by fully utilizing TI's family of fine-pitch surface-mount packaging. 3-21 72211 WCLK Second Stage EF I FF Control Logic RCLK I - - - - - - - E > Control Empty EF Q Control Full FF ~----------~~~D Figure 4. Connecting the Second-Stage Synchronizer to the '72211 Synchronous FIFO TISN74F74 D-Type Positive-Edge-Triggered Flip-Flop 72211 TISN74F08 2-lnput Positive AND Gate Control Empty EF , To Receiving Device 11 . ......'" Control FuliFF F = D - Control Empty Two Stage E _ EF INPUTS F = 'D - Control Full Two Stage F _ FF OUTPUT Y A B H H X H L X L L L Figure 5. Resolving Metastability of a Synchronous FIFO 3-22 Simultaneous-Switching Noise Analysis for Texas Instruments FIFO Products Navid Madani Advanced System Logic - Semiconductor Group ~1ExAs INSTRUMENTS SCAAOOSA 3-23 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthetime of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applicatlons'1. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product deSign, software performance, or Infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-24 Contents Title Page Abstract. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• 3-27 Introduction ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 3-27 TI Solution for Simultaneous-Switching Noise • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• Reducing Package Inductance ....................•...................................... Multiple GND and Vee Pins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Output Edge Control (OECTM) Method .................................................... Dirty and Clean Grounds in 36-Bit FIFO Families ........................................... 3-29 3-29 3-29 3-31 3-32 Simultaneous-Switching Tests Performed to Ensure Reliability of FIFO Products ••••••••••••••••••• SPICE Simulation .................................................................... GroundIPower-Noise Measurements (VOLP and VORY) ...................................... Special Test Performed on 36-Bit FIFOs (VIH and VIL Testing) ..............................•. 3-32 3-32 3-32 3-33 Summary ................................................................................ 3-33 References ............................................................................... 3-33 Acknowledgements ....................................................................... 3-33 Appendix A .............................................................................. 3-34 Appendix B .............................................................................. 3-39 3-25 List of mustratioDS FigUre 1 Title Page Power/Ground Noise Coupling Through a dc-on Driver to External Circuitry ................... 3-27 2 Simultaneous Switching ................•............................•................ 3-28 3 Example of Ground-Bounce Waveform ...•............................................. 3-28 4 SN74ACTI814 FIFO With Multiple GND Pins . . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . • . . . .. 3-30 5 CMOS-Transistor-Gate Layout ....................................................... 3-31 6 TI's Patented OEC Circuitry . . . . • . . . . . . . . • • • . . . . • . . . . . . • . . . • . . . • . . . . . . . . . . . . . . . . . . . . .. 3-32 A-I SN74ACT3638 VOLP, Vee - 5.5 V ................................................... 3-34 A-2 SN74ACT3638VOHV,Vee-5.5V .........•.......................................•. 3-34 A-3 SN74ABT3611 VOHV. Vee - 5 V •............•........•............................. 3-35 A-4 SN74ABT3611 VOHV. Vee-5.5V ..••........•..........•...................•....... 3-35 A-5 SN74ABT3611 VOLP, Vee-5V ..........•.......•.................................. 3-36 A-6 SN74ABT3611 VOLP. Vee - 5 V ..................................................... 3-36 A-7 SN74ABT3613VOHV. Vee-5V .................................................... 3-37 A-8 SN74ABT3613 VOHV , Vee - 5.5 V ............•...•.................................. 3-37 A-9 SN74ABT36I3 VOLP. Vee - 5 V ..............•................•.......••......•..... 3-38 A-I0 SN74ABT3613 VOLP. Vee - 5.5 V .................•...................•............. 3-38 B-1 Surface-Mount Package Options . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . • . • . . . . . . . . . • . . . . . • . . . .. 3-39 B-2 Surface-Mount Package Area by Package Type ....•...................................... 3-40 List of Thbles Table 3-26 Title Page 1 Inductance Value per Pin for Most Package Types Used for FIFO Products ..................... 3-29 2 Number of Data Output Pins per Ground Pin for Different FIFO Products ...................... 3-30 3 SPICE Simulations for the SN74ACT3632 Device When 18 or 36 Outputs Switch Simultaneously .. 3-32 4 Sample VOLP and VOHV Test Results for 36-Bit FIFO Products. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-33 5 Thst Required for the 36-Bit SN74ACT3632 ............•........•....................... 3-33 Abstract Analysis of circuit-noise immunity during simultaneous switching of multiple outputs is crucial in the high-speed advanced logic families including ACT and ABT FIFO products. Consequently, reduction of simultaneous-switching noise is of the utmost concern to the FIFO design team at Texas Instruments (T!). TI offers reliable FIFO products that meet the fast-speed requirement of today's technology. In this application report, a thorough explanation of noise-reduction techniques for TI's FIFO devices is provided. This report assists component and system design engineers in their evaluations of simultaneous-switching noise for TI's ACT and ABT FIFO products. Introduction One concern in advanced integrated circuit (lC) design is the challenge of minimizing simultaneous-switching noise while increasing switching speed of the device. This application report presents the achievements TI has made in providing very high-speed FIFO products with minimum simultaneous-switching noise. This report provides an introduction to advanced CMOS simultaneous-switching noise and the approaches taken by the FIFO design group to effectively reduce the noise. lest procedures for measuring noise during simultaneous switching of multiple outputs are also presented. Test results provide the data necessary to ensure proper operation of the FIFO during simultaneous switching of multiple outputs. In high-speed, high-density CMOS VLSI devices, many output drivers may switch simultaneously. During the transition, the excessive current drawn from the power supply can produce a significant amount of power/ground noise called simultaneous-switching noise (SS noise). The noise may be generated in the package VOOIVss planes and also in the internal (on-chip) VoolVSS buses. Figure 1 shows the power/ground noise coupling through a dc-on driver to external circuitry. Vee VI (VSS) Vss J~ +- }y~ Offj ..,.;v v2 J } j J v3 (Vee) '\Jw---+ 00 vee }y~ jt vss Figure 1. Power/Ground Noise Coupling Through a dc-on Driver to External Circultry1 A simultaneous-switching scenario where several drivers switch at the same time is shown in Figure 2. The electrical path from the ground/power IC pads to the package terminals is inductive as shown in Figure 2. The IC pads are connected through bonding wires to the package pads that are connected through a multilayer package to the package terminals, which, in tum, are connected to ground/power planes on the printed circuit board (PCB). All of these different elements in the packaging of the IC behave as inductances with negligible resistive components. 2 Switching Drivers PCPVCC Driver I Quiescent Driver -=Ole GND Vgl PCP GND Ole GND Vgl PCP GND CL Die GND I Vgl PCP GND Figure 2. Simultaneous Swltchlng2 The physics of the device package plays a fundamental role in the voltage-noise spike. The major effect on a high-speed device is the induced voltage on the GND and Vee terminals caused by the transient currents from switching capacitive loads. 3 If only one output is switched, the ground noise is calculated by equation 1 where Lg is the inductance of the ground terminal due to the bond wire, lead, and via, and dildt is the time rate of change in transient current driving the capacitive load. Equation 2 illustrates transient current where dVoidt is the change of output voltage in time. VONO--Lg x dildt (1) i(t) = CL x dVoidt (2) The induced ground bounce appears on the quiescent output as shown in Figure 3. o-to·1 Trsnsltlon of Switching Drivers o Vn (ground noise) J.- _______________ _ 1·to·0 Trsnsltlon of Switching Drivers Figure 3. Example of Ground·Bounce Waveform 2 When the number of simultaneously switchingputputs increases, ground noise increases. For large ICs, the relationship between ground-bounce amplitude and the number of switching drivers is no longer linear. 1,2, 4 Unless these power/ground noise fluctuations are controlled, simultaneous-switching noise can degrade or even limit system performance. Uncontrolled noise spikes can lead to loss of stored data, severe speed degradation, output glitches, and reduction in system-noise immunity.3 From a functional perspective, ground bounce reduces noise margins of the gate and may cause false switching of quiet gates. Noise margins for the low state are usually smaller than noise margins for the high state; therefore, noise of the ground bus in the IC concerns designers the most. 2, 5 Several techniques have been proposed for reducing simultaneous-switching noise. At the package level, one approach is to reduce the inductance by improved packaging techniques, such as decreasing the various inductive contributions 3-28 to ground bounce. 2 Surface-mount packages, such as PQFPs, are a better package option than through-hole packages, such as DIP or PGA, because the former have shortened pins or a lower-profile package. Another approach is to decrease the inductance of the ground pins by placing as many ground! power pins in the package as possible. I , 2 At the design level, some designers have proposed output edge control (OEe™) as a solution to reduce noise. 4 At the circuit level, simultaneous-switching noise can be reduced by skewing the output drivers and! or by damping out power and ground noise with additional damping resistors at the source end of both p- andn-channel transistors of output drivers,3 and/or by adding bypass capacitors that reduce the current noise associated with output buffers driving off-chip loads. 4 This application report concentrates only on TI's approaches to reduce the simultaneous-switching noise in high-performance advanced FIFO products. TI Solution for Simultaneous-Switching Noise TI's solutions to minimize noise caused by simultaneous switching of outputs include reducing package inductance by using multiple ground pins, controlling the output edge, and separating the ground pins. Measurements used by TI in evaluating the chip's performance are included in the following discussion. Reducing Package Inductance To reduce voltage spikes, the value of lead inductance (Lg) in equation 1 should be lowered. Lead inductance is dependent upon lead lengths as well as the location of GNDIVcc pins in the package. Decreasing the overall size of the FIFO package lowers the package inductance. The inductance value per pin for most of the package types used for FIFO products is shown in Table 1. TI's current technology has provided high-performance 9-, 18-, and 36-bit FIFO products with less inductance per pin, giving TI a performance edge in the FIFO market. More information on TI package types is provided in Appendix B. Table 1. Inductance Value per Pin for Most Package Types Used for FIFO Products PITCH Standard-Pitch Option Fine-Pitch Option PACKAGE TYPE FIFO TYPE INDUCTANCE PER PIN (nH) 24-pin DIP 4-,5-,8-, and 9-bit FIFO 3-15 28-pin DIP 9-bit (lOT) FIFO 2-15 44-pin PLCC 9-bitFIFO 6-8 28-pinSOIC 1-bit FIFO 3-8 12o-pin TQFP 36-bitFIFO 4-5 ao-pinTQFP 18-bltFIFO 64-pinTQFP 9-bitFIFO 5 3-4 Multiple GND and Vee Pins By adding more GND and Vee pins on a chip, TI offers advanced FIFO products with lower noise compared to other products with only one GND and one Vee corner pin. The previous section discussed simultaneous-switching noise being directly proportional to the inductance of the ground! power leads. Multiple ground! power pins improve the noise immunity of the chip by reducing the total ground! power lead inductance because the total inductance is a parallel combination of the lead inductances of the ground!power pins. For example, SN74ACTI814 FIFO memory in Figure 4 has four GND pins distributed among the outputs. The total ground-lead inductance of this chip is approximately one-fourth of that of a similar chip with only one GND pin. Assuming Ll, L2. L3, and L4 are the lead iilductances of the four ground pins on the chip, and assuming these inductances are equal, the combination of the parallel inductances is 114 of the inductance when only one GND pin is on the chip (see equation 3). (3) OEC is a trademark of Texas Instruments Incorporated. 3-29 Multiple ground! power pins are used on all TI FIFO products. Table 2 shows the nlllilber of data output pins per ground pin for different FIFO products. DLPACKAGE (TOP VIEW) RESET 017 016 015 014 013 012 011 010 VCC 09 08 GNO 07 06 05 04 03 02 01 00 HF PEN AF/AE LOCK NC NC FULL 1 V56 2 55 3 54 4 53 52 5 51 6 7 50 8 49 9 48 10 47 11 48 12 45 44 13 14 43 15 42 16 41 17 40 18 39 19 38 20 37 21 36 22 35 23 34 24 33 25 32 26 31 27 30 28 29 OE Q17 Q16 Q15 GNO Q14 VCC Q13 Q12 Qll Ql0 Q9 GNO as Q7 as as VCC Q4 Q3 Q2 GNO Ql QO UNCK NC NC EMPTY Figure 4. SN74ACT7814 FIFO With Multiple GND Pins Table 2. Number of Data Output Pins per Ground Pin for Different FIFO Products FIFO 9 Bits 18 Bits FIFO PRODUCT SN74ACT2335 102492 SN74ACT2236 102492 SN74ACT7807 20489 SN74ACT7803 51218 SN74ACT7811 SN74ABT7819 32 Bits 36 Bits SN74ACT3638 102418 512182 512322 SN74ABT3614 64362 SN74ACT3632 512362 SN74ACT3641 3-30 SIZE 102436 PACKAGE TYPE GND PINS DATA OUTPUTS PERGNDPIN 4.5 44-pin PLCC (FN) 4 64-pin TQFP. (PM) 12 1.5 44-pin PLCC (FN) 4 4.5 44-pin PLCC (FN) 6 1.5 64-pin TQFP (PM) 12 0.75 4 4.5 68-pin PLCC (FN) (OL) 10 1.8 8O-pin TQFP (PN) 14 1.3 8O-pin QFP (PH) 14 2.6 SO-pin TQFP (PN) 14 2.6 120-pinTQFP (PCB) 14 4.6 132-pin PQFP (PQ) 15 4.3 120-pln TQFP (PCB) 10 7.2 132-pin PQFP (PQ) 18 4.0 120-pin TQFP (PCB) 14 5.1 132-pin PQFP (PQ) 14 5.1 120-pin TQFP (PQ) 15 2.4 132-pin PQFP (PCB) 15 2.4 Output Edge Control (OECTM) Method OECTM is another method for controlling simultaneous-switching noise. This is a circuit method that reduces the dildt portion of equation 1. The output transistor is split into many small subtransistors with sequential turnon of each subtransistor. By splitting the total current into a series of smaller currents distributed over time, the effective diJdt is reduced. The delay in turnon of the successive subtransistors reduces the maximum peak diJdt for the entire output transistor. Drains \ 0 0 0 0 ~ Drains / Polyslllcon Gate \ I~ 0 0 0 0 I~ 0 0 0 0 ~ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 / (a) TYPICAL ARRANGEMENT ~ 0 0 0 0 0 C 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Polyslllcon Gate ~ 0 0 0 0 0 0 0 0 0 0 ~ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ---- 0 0 0 0 0 (b) SERPENTINE ARRANGEMENT Figure 5. CMOS-Transistor·Gate Layout At the output, the structure of the polysilicon gate is modified to grade the tumon by removing portions of the polysilicon gate to form a serpentine arrangement and by driving the gate from one end (see Figure 5). The resistance of the polysilicon and the capacitance of each gate segment form a distributed RC network that slows the turnon of each succeeding segment. Figure 6 shows the equivalent-circuit schematic for the distributed output transistor. 4 3-31 Out In Figure 6. Tl's Patented OEC Circuitry The OEC circuitry implemented in output structures reduces simultaneous-switching noise by reducing the edge rate. The distributed output transistor with pull-down transistors evenly added gives a fast-turnoff feature to the circuit and minimizes the through current as well. The undesirable slow turnoff is resolved by evenly adding pull-down transistors to the distributed output transistor. Thmoff transistors minimize through current by rapidly turning off all the segments of the output-transistor circuit; therefore, the OEC method not only provides an effective means for controlling dildt noise in high-speed CMOS FIFO products, but adds a fast-turnoff feature to the output circuit. Dirty and Clean Grounds In 36-81t FIFO Families To reduce effects of simultaneous-switching noise on 36-bit FIFOs, TI divides the ground pins into dirty and clean ground terminals. A dirty ground is used only for device outputs and a clean ground is used for inputs and other internal circuit connections. A dirty ground is isolated from a clean ground on the chip, but users can connect them to the same external ground. Isolating the. two grounds benefits the FIFO chips because output-switching noise does not affect the rest of the chip, thereby reducing the possibility of false clocks and intermittent data errors. \ Simultaneous-Switching Tests Performed to Ensure Reliability of FIFO Products SPICE Simulation Simultaneous-switching SPICE simulations were performed for the SN74ACT3632 device during the design process for 36-bit FIFO products. Simulation results for SN74ACT3632 indicate that Vee droop and ground bounce when 18 bits are switching simultaneously are only 4.39 and 0.67 V, respectively. When 36 bits are switching simultaneously, Vee droop and ground bounce only change to 4 and 0.9 V, respectively. The results of this SPICE simulation illustrate the reliability of TI's high-performance FIFO products against ground noise6 (see Table 3). Table 3. SPICE Simulations for the SN74ACT3632 Device When 18 or 36 Outputs Switch Simultaneously SN74ACT3632 SWITCHING SIMULTANEOUSLY 18 bits 36 bits Vcc DROOP (V) 4.39 4.0 GROUND BOUNCE.(V) 0.67 0.9 Ground/Power-Noise Measurements (VOlP and VOHV) Noise measurements evaluate the performance of FIFO products while simultaneously switching the outputs. A typical simultaneous-switching test is performed to determine the magnitude of the disturbance on the output that is not being switched, as well as stored data integrity for devices with multiple outputs. The voltage induced on a quiescent output during simultaneous switching is referred to as VOLPandVOHV. ForVOLP (VOHV) measurements ,theoutputundertest is held low (high) while the rest of the outputs are switching from high to low (low to high). VOLP, the maximum (peak) . voltage induced on a quiescent low-level output during switching of other outputs, is measured with respect to a ground reference near the output under test. VOHV is the minimum (valley) voltage induced on a quiescent high-level output during switching of other outputs. Table 4 summarizes the VOLP and VOHV results for 36-bit FIFO products. Data was taken on an automatic test machine (lIP 82000) at room temperature (25°C). 3-32 Table 4. Sample VOLP and VOHV Test Results for 36·Blt FIFO Products DEVICE SN74ABT3614 VOLPM 0.75 SN74ACT3632 1.0 VOHV(V) VCc(V) 0.2 5.5 1.4 5.0 The results ofVOLP and VOHVmeasurements perfonned on n's 36-bitFIFO products indicate the reliability and noise immunity of n's high-perfonnance FIFO products. Sample wavefonns for SN74ACT3638, SN74ACT361I, and SN74ACT3613 are presented in Appendix A. Special Test Performed on 36-Bit FIFOs (VIH and VIL Testing) VIH and VIL values for 36-bit FIFO families are tested while outputs are simultaneously switching. For example, the SN74ACT3632 continues to function properly with VIH and VIL values shown in Table 5. These results show that separating the output ground from the ground used for the rest of the chip results in excellent input-noise margins for the 36- and 32-bit-wide FlFOs. Table 5. Test Required for the 36-Blt SN74ACT3632 Vee 4.5 V 5.5 V VIH 1.8V 1.9V VIL 1.3 V 1.3 V Summary Fast switching speeds in today's technology require solutions to problems such as simultaneous-switching noise. The fast switching of drivers can cause uncontrolled noise spikes on the chip's ground bus, which lead to false clocks or incorrect data and control signals on the device. As more outputs of an IC switch simultaneously, noise effects increase and limit the usefulness of the device. Better packaging options, multiple ground/ power pins, output edge control, arid separating the ground pins as clean and dirty ground pins reduce the simultaneous-switching noise. Finally, results obtained from simultaneous-switching tests are provided to illustrate the noise immunity of n's high-perfonnance FIFO devices. References 1. R. Senthinathan, J. L. Prince and S. Nimmagadda. "Noise Immunity Characteristics of CMOS Receivers and Effects of SkewinglDamping CMOS Output Driver Switching Wavefonn on the Simultaneous Switching Noise," 23 pp. 29-36, Microelectronics Journal, 1992. 2. Lilianan Diaz-Olavarrieta, "Ground Bounce in ASIC's: Model and Thst Results," IEEE International Symposium on Electromagnetic Compatibility, 1991, Cherry Hill, NJ. 3. Texas Instruments Advanced CMOS Logic Designer's Handbook, 1988. 4. Craig Spurlin and Dale Stein, "EPIC Advance CMOS Logic Output Edge Control," Texas Instruments TechnicalJournal, March-April 1989,. 5. H. Hashemi, U. Ghoshal, K. Ziai, and P. Sandborn, "Analytical and Simultaneous Study of Switching Noise in CMOS Circuits," Proceedings of the Technical Conference, September 1990, pp. 762-773. 6. Design Review Reports for 74ACT3632 FIFO. Acknowledgements The VOLP and VOHV graphs in Appendix A are provided with the assistance of Al Sawyer. Packaging infonnation in Appendix B is provided by n packaging engineers and organized by Tom Jackson. 3-33 Appendix A Appendix A shows VOLP and VOHV measurements during simultaneous switching. Measurements are made with respect to ground at 25°C with an HP8200 automatic test machine. Viis the quiescent voltage on the output being tested prior to switching other outputs. V2 is the peak voltage on the output being tested while switching other outputs. Vn is the difference in V I and V2. 70 60 > E I 50 & 40 i .& .. 'S 0 I .p 30 20 10 0 -10 0 5 10 15 20 25 30 35 40 45 50 t-Tlme-na VI-25mV, V2-14.7mV, Vn-12.2mV Figure A-1. SN74ACT3638 VOLP. Vee =5.5 V 7 > 6 i 'S .& . 5 I ~ 0 4 3 2 I .p 01 -1 0 5 10 15 20 25 30 35 40 45 t-Tlme-na VI - 3.63 V, V2 - 5.31 V, Vn -1.69V Figure A-2. SN74ACT3638 VOHV. Vee =5.5 V 50 7 > I i I 8 5 4 ~ 3 J 2 I ~ 0 -1 0 5 10 15 20 25 30 35 40 45 50 t-Tlme-na VI - 3.25 V, V2 - 2.93 V, VD - -312.5 mV Figure A-3. SN74ABT3611 VOHY. Vee = 5 V 7 8 > I 5 j 4 ~ 3 ! 2 'S "" I ~ -- 0 -1 0 5 10 15 - AI""'\.. J 20 25 30 35 40 t-Tlme-na VI - 3.71mV, V2 - 3.21 mY, VD - -500 mV Figure A-4. SN74ABT3611 VOHY. Vee =5.5 V 45 50 7 >I I I.p 6 5 4 \ 3 2 I uJ~ o ~ I -1 o 5 10 15 20 25 30 35 40 45 50 45 50 t-Tline-ns VI - 218.75 mY, V2 - 843.75 mY, Vo - 625 mV Figure A-5. SN74ABT3611 VOLP, Vee =5 V 7 > I 6 t 5 I.p 3 ~ 4 h 2 I ...1\ o -1 W' o 5 10 15 20 25 30 35 40 t-Tlme-ns VI - 218.75 mY, V2 - 843.75 mY, Vo - 625 mV Figure A-6. SN74ABT3611 VOLP, Vee 3-36 =5 V 7 ,.. >I III ~ ::I .& ::I 6 5 ~ "",...., 4 3 j 2 0 I .p 0 -1 0 5 10 15 20 25 30 35 40 45 50 45 50 t-Tlme-ns Vl- 3.18 V, V2 -2.71 V, VD=-468.75mV Figure A-7. SN74ABT3613 VOHV. Vee = 5 V ,.. > I GI ~ ::I .& ::I 0 7 6 5 4 3 2 I .p 0 -1 0 5 10 15 20 25 30 35 40 t-Tlme-ns Vl-3.18V, V2 = 2.71 V. VD--468.75mV Figure A-S. SN74ABT3613 VOHV. Vee =5.5 V 3-37 7 6 > I 5 i 4 3 I.p 2 I -1 " .f'\. o MAo ~ o 5 10 15 20 25 30 35 40 45 50 t-Tlme-ne Vl-250mV, V2-1.12V, VD-87SmV Figure A-9. SN74ABT3613 VOLP, Vee =5 V 7 >I i I.p 6 5 4 \ 3 2 I .I \ o -1 .-. ~ o 5 10 15 20 25 30 35 40 t-Tlme-ne Vl-218.7SmV, V2-1.l8 V, VD-968.7SmV Figure A-10. SN74ABT3613 VOU), Vee = 5.5 V 45 50 Appendix B FIFO Package Types 1 Bit 9 Bit 18 Bit 36 Bit 27,94 It-- 17,6 ---+I 44-Pln PLCC(FN) 17,6 = = Area 192,0 mm 2 Height 2,65 mm Lead Pitch 1,27 mm = rr~~~ 25,2 Area 635,04 mm 2 Height 4,37 mm Lead Pitch 1,27 mm = 132·Pln PQFP(PQ) 2794 PLCC (FN) = = Area =309,78 mm 2 Height =4,37 mm Lead Pitch 1,27 mm ~ = = Area 780,64 mm 2 Height 4,32 mm Lead Pitch =0,635 mm n = wuu:i l. JWIWIIIJWIIII16 1j I 12()"Pln TQFP (PCB) U°'----' = Area 165,0 mm 2 Height 2,65 mm Lead Pitch 1,27 mm = = = = Area 224,5 mm 2 Height 3,05 mm Lead Pitch 1,27 mm = Area 415,36 mm 2 Height 2,95 mm Lead Pitch 0,8 mm = = = = = Area 258 mm 2 Height 1,58 mm Lead Pitch 0,4 mm = 14--14,95~ IT II ~18,42~ 32.Pln ",-co (RJ) T 56-Pin SSOP (OL) 10,35 JJmmmmmummrnm' = Q = Area 186,62 mm 2 Height 3,30 mm Lead Pitch 1,27 mm = = Area 190,647 mm 2 Height =2,6 mm Lead Pitch 0,635 mm = 1--12~ n 64-Pln .l.!~QFP (PM) = = Area 144 mm 2 Height 1,5 mm Lead Pitch 0,5 mm = = = Area 196 mm 2 Height 1,5 mm Lead Pitch 0,5 mm = Figure B-1. Surface-Mount Package Options 3-39 SOO *7S1 700 "635 600 500 415 N E E 400 310 I ~ 300 cr: 200 100 0 24-Pln SOIC 28-Pln SOIC '---v---' 1 Bit 64-Pln TQFP 32·Pln PLCC 28-Pln SOIC V 9 Bit 44-Pln PLCC / 8O-Pln TQFP \ 58-Pln SSOP SD-Pln PQFP 68-Pln PLCC V 1SBIt "Competitors' Best Solution Figure 8-:2. Surface-Mount Package Area by Package Type 3-40 120-Pln TQFP 132·Pln SOIC '---v---' 36 Bit FIFO Solutions for Increasing Clock Rates and Data Widths First-In, First-Out Technology Kam Kittrell Advanced System Logic - Semiconductor Group ~1ExAs INSTRUMENTS SZZA001A 3-41 IMPORTANT NOTICE Texas Instruments (TIl reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is curr~nt. TI warrants performance of Its semiconductor products and related software to the specifications applicable at the time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage C'Critical Applications'1. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products In such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of petents or services described herein. Nor does TI warrant or represent that any license, ' either express or implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-42 Contents Title Page Introduction ............................................................................. 3-45 Clocked FIFOs ........................................................................... 3-45 Flag Synchronization ...................................................................... 3-45 . Compact Packaging ....................................................................... 3-47 New Clocked FIFOs ....................................................................... 3-48 Conclusion .............................................................................. 3-49 List of mustrations Figure I 2 3 4 5 Title Page Triggering a Metastable Event With a One-Stage Synchronizer ............................... 3-46 Two-Stage Synchronizer ............................................................. 3-46 Storage Oscilloscope Plots Taken Over a I5-Hour Duration ................................. 3-47 Surface-Mount Package Area Comparison ............................................... 3-48 Bidirectional Configuration for the SN74ACT7803 ................. : ...................... 3-49 3--43 3-44 Introduction Steady increases in microprocessor operating frequencies and bus widths over recent years have challenged system designers to find FIFO memories that meet their needs. To assist the designer, new FIFOs from Texas Instruments (TI) are available with features that complement these microprocessor trends. Higher data-transfer rates have dictated the need for FIFOs to evolve into clocked architecture wherein data is moved in and out of the device with synchronous controls. Each synchronous control of the clocked FIFO uses enable signals that synchronize the data exchange to afree-running (continuous) clock. Since the continuous clocks on each port of a clocked FIFO can operate asynchronously to each other, internal status signals indicating when the FIFO is empty or full can change with respect to either clock. To use a status signal for port control, it is synchronized to the port's clock on a clocked FIFO. Synchronization of these signals with flip-flops introduces metastability failures that increase with clock frequency. TI uses two-stage flag synchronization to greatly improve reliability. Higher clock frequencies augment raw speed, but greater bandwidth is also achieved by increasing the data width. Wider datapaths can have the associated cost of large board area due to increased package sizes. New compact packages for TI's FIFOs reduce this cost. Clocked FIFOs Clocked FIFOs have become popular for relieving bottlenecks in high-speed data traffic. Data transfers for many systems are synchronized to a central clock with read and write enables. These free-running clocks can be input directly to a clocked FIFO with the same enables controlling its data transfer on the low-to-high transition of the clock. Reducing the number of clocks keeps the interface simple and easy to manage. Extra logic is needed to produce a gated pulse when using a FIFO that accepts a clock only for a data transfer request. The generated clock signal is a derivative of the master clock with a margin of timing uncertainty. At high clock frequencies, this timing uncertainty is not tolerable and costly adjustments are needed. Additional logic also is conserved by implementing flag synchronization on the clocked FIFO. Tracking is done to generate flags that indicate when the memory is empty or full. In many applications, the input and output to the FIFO are asynchronous and the flag signals must be synchronized for use as control. A read is not completed on the FIFO if no data is ready, so the EMPTY signal is synchronized to the read clock. This synchronous output-ready (OR) flag is useful for controlling read operations. Likewise, the FULL signal is synchronized to the write clock, producing the input-ready (IR) flag. Flag Synchronization As previously explained, one of the advantages of the clocked FIFO is the on-board synchronization of the EMPTY and FULL status flags when the input and output are asynchronous. In one method of synchronization, a single flip-flop captures the asynchronous flag's value (see Figure 1). With this method, the rising transition of data can violate the flip-flop's setup time and produce a metastable event (metastability is a malfunction of a flip-flop wherein the latch hangs between high and low states for an indefinite period of time). 3-45 Clock Q Asynchronous Flag Clock D I llf ~ ~tsu Flag I I I OR Synchronized Flag { ( tpd~ r \ ~tr4 Figure 1. Triggering a Metastable Event With a One-Stage Synchronizer Once a metastable event is triggered, the probability of the output recovering to a high or low level. increases exponentially with increased resolve time (tr). The expected time until the output of a single flip-flop with asynchronous data has a metastable event that lasts tr or longer is characterized by the following mean time between failures (MTBF) equation: exp(¥) MTBF, =of t f ed Where: to - flip-flop constant representing the time window during which changing data invokes a failure tr - resolve time allowed in excess of the normal propagation delay t - flip-flop constant related to the settling time of a metastable event fc - clock frequency fd - asynchronous data frequency. For OR-flag analysis, it is the frequency at which data is written to empty memory. For IR-flag analysis, it is the frequency at which data is read from full memory. The MTBF decreases as clock and data frequency increase and as the time allowed for a metastable event to settle (tr) decreases. Metastability failures are a formidable issue for short-clock cycle times. Increasing the clock frequency linearly increases the number of metastable events triggered, but the shortened available resolve time exponentially increases the failure rate. It is impossible to eliminate the possibility of a metastable event under these conditions, but solutions exist to reliably increase the expected time between failures. Clock Aaynchronous Flag Q D Q D Figure 2. Two-Stage Synchronizer 3-46 Synchronized Flag 11 increases the metastable MTBF by several orders of magnitude for m and OR flags by employing two-stage synchronization (see Figure 2). For the output of the second stage to be metastable, the first stage must have a metastable event that lingers until it encroaches upon the setup time of the second stage. Adding another stage to a single flip-flop synchronizer is statistically equivalent to increasing its resolve time by the clock period minus its propagation delay. The mean time between failures for a two-stage synchronizer is given by: exp[ t, +t-tpj t C Where: tp - propagation delay of the fJrst flip-flop fo= 50 MHz, fd = 5 MHZ, VCC =5 V (8) ONE-STAGE SYNCHRONIZATION fo = 55.7 MHz, fd =6.7 MHz, VCC =5 V (b) TWO-STAGE SYNCHRONIZATION Figure 3. Storage OSCilloscope Plots Taken Over a 15-Hour Duration Figure 3 compares the two synchronization methods previously discussed. Both plots were taken at room temperature and nominal Vee while each data transition violated setup time. Figure 3(a) shows the performance of an EMPTY flag synchronizer using only one flip-flop, while Figure 3(b) is the m flag of an SN74ACI'7S07 with the write clock operating at maximum frequency. Compact Packaging Microprocessor bus widths have continuously doubled every few years to maximize their performance. Bus widths of 32 and 64 bits are commonplace today, whereas they were almost unheard of a few years ago. The downside to the increased bit count is that each subordinate device in the system must match this width with corresponding increases in board size. New shrink packages forTI's clocked FIFOs provide a solution to this problem. Multiple-byte datapaths can be buffered while covering only a fraction of the area of conventional packages. These new FIFO packages are presently available in 56-, 64-, and SO-pin configurations. Dubbed shrink quad flat package (SQFP), the 64-pin package is used for 9-bit-wide FIFOs, and the SO-pin package is used for IS-bit-wide FIFOs. Both SQFP packages have a lead pitch of 0.5 mm. The 56-pin shrink small-outline package has a O.025-inch lead pitch and also houses IS-bit-wide FIFOs. A variety ofTI's FIFOs are offered in these new packages (see Table 1). 3-47 Table 1. FIFOs Available In Space-Efficient Packages DEVICE CLOCKED ORGANIZATION CLOCK CYCLE TIME (n8) PACKAGES SN74ACT2235 No 1Kx9x2 20,3040,50 64 TQFP 44 PLCC SN74ACT7802 No 1Kx18 25,40,60 60TQFP 68 PLCC SN74ACT7811 Yes 1Kx18 15,18,20,25 80TQFP 68PLCC SN74ACT7803 SN74ACT7805 SN74ACT7813 Yes 512 x 18 256 x 18 64x 18 15,20,25,40 56SS0P SN74ACT7804 SN74ACT7806 SN74ACT7814 No 512x 18 256 x 18 64x 18 20,25,40 56SS0P SN74ACT7807 Yes 2Kx9 15, 20, 25, 40 64 TQFP 44 PLCC SN74ACT7808 No 2Kx9 20,25,30,40 64 TQFP 44 PLCC Figure 4 compares the space savings of the new compact packages compared to competitive surface-mount solutions. A 4-byte path constructed with four clocked FIFOs in 32-pin PLCC packages occupies 1.16 in2, while two 56-pin SSOP packages occupy only 0.59 in2. 1 0.9 0.8 - IN 0.7 0.6 C :::. I CC 0.5 0.4 0.3 0.2 0.1 0 64 TQFP 56 SSOP 80 SQFP. 32 PLCC 44 PLCC 68 PLCC Figure 4. Surface-Mount Package Area Comparison New Clocked FIFOs Four new CMOS clocked FIFOs from 11 offer a variety of memory depths. All four can match applications that require maximum clock frequencies of 67 MHz and access times of 12 ns. Suited for buffering long packets, the 2K x 9 SN74ACT7807 is the deepest of the four and is available in the 44-pin PLCC or 64-pin TQFP. The SN74ACT7803, SN74ACT7805, and SN74ACT7813 are organized as 512 x 18,256 x 18, and 64 x 18, respectively, and have the same pin arrangement in the 56-pin SSOP. Every 11 clocked FIFO is easily expanded in word width, and the SN7 4ACT7803/05/13 can also be arranged to form a bidirectional FIFO. With the two FIFOs connected as in Figure 5, no extra logic is needed for bidirectional operation. . 3-48 'ACT7803 ClKA WRTCLK RDCLK ClKB WiRA WRTENl OEl W/RB A CS WRTEN2 RDEN 012 cst .J 18 DO-D17 QO-Q17 BO-B17 'ACT7803 - L AO-A17 18 RDCLK WRTCLK r-- OEl WRTENl r-- RDEN WFi'i'EN2 OE2 QO-Q17 DO-D17 Figure 5. Bidirectional Configuration for the SN74ACT7803 Silicon is currently available for a bidirectional clocked FIFO fabricated in TI's Advanced BiCMOS (ABT) process. The SN74ABT7819 is organized as 512 x 18 x 2 with two internal independent FIFOs. Each port has a continuous free-running clock, a chip select (CS), a read/write select (WW), and two separate read and write enables for control. It supports clock frequencies in excess of 80 MHz and a maximum access time below 10 ns. This device is packaged in the 80-pin QFP and 80-pin SQFP. Conclusion Several semiconductor manufacturers, including TI, have responded to customer needs by providing clocked FIFOs whose synchronous interfaces conform to the requirements of many high-performance systems. Capitalizing on the available continuous system clocks, this architecture limits the amount of necessary glue logic and the number of timing constraints. Flag synchronization is important for clocked FIFOs buffering between asynchronous systems. Flip-flop synchronizers used for this task have a metastable failure rate that grows exponentially with clock frequency. TI employs two stages of synchronization that improve the flags' reliability significantly. Finally, providing a FIFO buffer for wide buses has historically consumed large amounts of board area. Designers seeking relief from this problem can fmd it in the packaging options offered for TI's FIFOs. Used to house 9- and I8-bit devices, these packages require only about 50% of the space required for conventional surface-mount packages. 3-49 3-50 FIFO Features Page FIFO Patented Synchronous Retransmit: Programmable DSP-Interface Application for FIR Filtering ............................................................. FIFO Mailbox-Bypass Registers: Using Bypass Registers to Initialize DMA Control ........ Advanced Bus-Matching/Byte-Swapping Features for Internetworking FIFO Applications ... Parity-Generate and Parity-Check Features for High-Bandwidth-Computing FIFO Applications ............................... 3-53 3-65 3-75 3-87 3-51 FIFO Patented Synchronous Retransmit: Programmable DSP-Interface Application for FIR Filtering Steve Strom and Kam Kittrell Advanced System Logic - Semiconductor Group ~lExAs INSTRUMENTS SCAAOO9A 3-53 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthetime of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage (''Critical Appllcations',. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license. either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright@ 1996, Texas Instruments Incorporated 3--'54 Contents ntle Page Introduction .........................•...•.............•................................. 3-57 Description of Synchronous Retransmit ......•............................................... 3-57 Example of Retransmit for FIR Filtering .....•............................................... 3-59 Modified Code for TMS320C3x FIR Filtering .•.....••....•.•...••.............•.......•...... 3-61 Conclusion ............................................................................... 3-63 Figure List of mustrations ntle Page I SN74ACf3638 Functional Block Diagram ...............................•.............. 3-58 2 FIFOI Retransmit Timing Diagram Showing Minimum Retransmit Length ......••..•••...•.. " 3-59 3 Using a FIFO for Coefficient Storage in Multiply/Accumulate Operations ...................... 3-59 4 5 6 Bidirectional FIFO Interface ...•.........•............•...••.......................... 3-60 Interconnection Example . . . . . . . . . . . . . . . • • . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . .. 3-60 Retransmit Timing for Interconnection Example .......................................... 3-61 7 IOF Register Bit Summary .•..•......................•.........................•..... 3-61 8 FIFO Retransmit Control for FIR Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . .. 3-62 9 Control Timing for FIFO Retransmitfor the FIR Filter ..........•.......................... 3-63 3-55 3-56 Introduction TIris application report presents one example of the many uses of the synchronous-retransmit feature of Texas Instruments (TI) digital signal processing (DSP) application-specific FIFOs. This report describes TI's patented synchronous-retransmit feature and shows how this feature can be used in conjunction with a DSP for fmite-Iength impulse-response (FIR) filtering. The TMS32OC31 floating-point DSP and the SN74ACT3638 bidirectional clocked FIFO are the examples for this discussion. Description of Synchronous Retransmit An SN74ACT3638 functional block diagram with the synchronous-retransmit logic block highlighted is shown in Figure 1. The synchronous-retransmit feature of the SN7 4ACT3638 allows data stored within the FIFO to be reread starting at a selected position. FIFO 1, one of two 512 x 32 dual-port SRAM FIFOs on board the SN74ACT3638 device, buffers data from port A to port B. FIF01 is placed in the retransmit mode to select a beginning word and to prevent ongoing FIFO write operations from destroying data to be retransmitted. Data vectors with a minimum length of three words can be retransmitted repeatedly starting at the selected word. The FIFO can be taken out of the retransmit mode at any time, allowing normal operation to resume. Figure 2 shows the FIF01 retransmit timing and minimum retransmit length. FIFOl is placed in the retransmit mode by a low-to-high transition on CLKB when the retransmit mode (RTM) input is high and the port-B output-ready (ORB) flag is high. This rising clock edge marks the data present in the FIFO 1 output register as the first retransmit word. FIFO 1 remains in the retransmit mode until a low-to-high transition of CLKB occurs while RTM is low. When two or more reads have been performed past the initial retransmit word, a retransmit is initiated by a low-to-high transition on CLKB when the read-from-mark (RFM) input is high. This rising CLKB edge shifts the first retransmit word to the FIF01 output register and subsequent reads begin immediately. While FIFOl is in the retransmit mode, retransmit loops can be performed repeatedly with each pulse of the RFM terminal. When FIF01 is in the retransmit mode, it operates with two read pointers. The current-read pointer operates normally, incrementing each time anew word is shifted to the FIFO 1 output register. TIris pointer is used as a reference by the ORB and port-B almost-empty (AEB) flags. The shadow-read pointer stores the SRAM location at the time FIFOl is placed in the retransmit mode and does not change until FIF01 is taken out of the retransmit mode. TIris pointer is used as a reference by the port-A input-ready (IRA) and almost-full (AFA) flags. While the FIFO is in the retransmit mode, data writes to the FIFO may continue. AFA is setlow by the write that stores (512 - Y1) words after the frrst retransmit word, where 512 is the FIFO depth and Y1 is the almost-full-flag offset value. The IRA flag is set low following 512 writes after the first retransmit word. When FIF01 is in retransmit mode and RFM is high, a rising CLKB edge loads the current pointer with the shadow-read-pointervalue. The ORB flag immediately reflects the new level offill. If the retransmit changes the status of FIFO 1 such that it is no longer within the almost-empty range, up to two CLKB rising edges after the retransmit cycle are required before the AEB flag is asserted. The rising CLKB edge that takes FIFO lout of the retransmit mode shifts the read pointer used by the IRA and AFA flags from the shadow-read pointer to the current-read pointer. 3-57 MBF1 I CLKA CSA WiRA ENA MBA R::~r - - Port·A Control 1 - Logic 1 - - l- l ... Jr~ - r--- ~ I 512x32 SRAM I'S I+- FIF01, Mal11 Reaat Logic C2. .5 ""'-- 1.32 ~ RTM f-+--4 I Write Pointer II RFM Reed Pointer T IRA AFA I . 1f-' I ORA AEA I I I .. .,. Logic -.. Reed Pointer T~ II .,. I I Write Pointer IRB AFB II .. 512x32 SRAM ~~ l32 -. I I'S C2. --.5 I ROYB BO-B 31 ._- ...1 [FIF02 , ORB AEB I . ProgrammableFlag Offset ", ROYA I ~ FSO FS1 AO-A31 T Logic -.. I R::~~r I FIF02, Mall2 Reaat -+-'RST2 Logic f-4-< .... I-- Port·B -4-CLKB '--- I-- Control .... r-- I Figure 1. SN74ACT3638 Functional Block Diagram 3-58 Logic ....::=i~:B ..... ·MDD eLKB I I \ \ ~ Ih(EN) I??j I 'W tau(EN) ENB tau(RM) I I I I I I I I I I I ~ "'(AM) IW I~ RTM I I I I I I RFM ORB II High I4-ta-tl wo BO-B31 Initiate Retransmit Mode WIth WO .. First Word * I \ tau(EN) ~ I I I I I I I I I ~ta...., I4-ta""" * W2 Retransmit From Selected Position "'(AM) '&~ I~ #» -!Ut§\) I \ I I I tau(AM) '--ta""" W1 I \ ~ Ih(EN) * wo End Retransmit Mode tsu(EN) Setup time, CSA, WIRA, ENA, end MBA before CLKAt; ese, WIRB, ENB, MBB, ATM, end RPM slter CLKBt tsU(EN) Hold time, CSA, WiRA, ENA, end MBA slter CLKAt; ese, WIRB, ENB, MBB, ATM, end RFM after CLKBt fa Access time, CLKAttoAO-A31 end CLKBtto 80- 831 * W1 Figure 2. FIF01 Retransmit nmlng Diagram Showing Minimum Retransmit Length Example of Retransmit for FIR Filtering In addition to the typical interface functions, such as rate matching and clock partitioning, FIFOs with retransmit capabilities can provide arepeated sequence of data to a processing element such as a DSP. This sequence ofinfonnation may take the form of coefficients for use in a DSP multiply/accumulate operations as shown in Figure 3. Retransmit Loop Initializing Informetion forDSP Data Block 1 Coefffcients for FFTofFIR Data WrIte! Reed Order ---+ Figure 3. Using a FIFO for Coefficient Storage In Multiply/Accumulate Operations 3-69 \ Many DSP applications require filtering. The FIR fIlter is a type of digital filter that is implemented very efficiently by the TMS32OC31. The FIR filter in the time domain takes the general form of: N-I y(n) = I h(i) x x(n - I) i=O Where: y(n) is the output sample at time n, h(i) is the ith coefficient or impulse response, and x(n - i) is the (n - i)th input sample. The capability for parallel multiply/add operations and circular addressing permits easy implementation of the FIR filter with the TMS32OC31 DSP. The former allows a multiplication and addition operation to execute in one machine cycle; the latter generates a finite buffer of length N for the data x(n). When used for coefficient storage, the FIFO serves as a zero-wait-state SRAM. Applications in which coefficients or other data are stored in external SRAM or EPROMs can be greatly simplified, thereby reducing cost, space requirements, and overall device count. In other instances where DSP internal RAM is used to store the coefficients, a penalty is often paid in the form of overhead time for transferring the coefficients from the buffering FIFO to RAM. This overhead penalty and inefficient use of RAM can be eliminated by the use of the patented synchronous-retransmit feature of the 11 FIFO. 'I\vo TMS32OC31 external input/output (110) flags (XFO and XFI) can be configured as input or output terminals under software control. In the example of FIR filtering, I/O flags can be implemented to control the retransmit function of the FIFO, providing a programmable DSP interface; Figure 4 shows a block-diagram representation of the bidirectional interface to the programmable DSP. rG ....._H_o_st__ 32 4/~ 32 4/~ SN74ACT3638 I"-~ Point . DSP Figure 4. Bidirectional FIFO Interface Figure 5 shows an interconnection example for the SN74ACT3638-30 FIFO and TMS32OC31-40 DSP. The DSP XFO and XFI terminals are configured for general-purpose output and are directly connected to the RTM and RPM terminals of the FIFO, respectively. The retransmit timing associated with this interface is shown in Figure 6. The I/O flag register (lOp), which is one of28 registers in the TMS32OC3 I CPU register file, controls the external pins XFO and XFI. Figure 7 shows a summary of 10F register bit assignments. Additional information on the 10F register may be obtained by consulting the TMS32OC3x User's Guide (literature number SPRU031C). SN74ACT3638-30 TMS320C31-40 ClKS 0 css Decode ENS W/RS ~ A23,A20 W/R XFO RFM XF1 33 Figure 5. Interconnection Example 3-60 ~ STRS RTM SO-S31 H3 DO-D31 ~ Fetch/Load + Decode H3 OUTFxBlt Oar 1 14-- 13ns~ I XFxPln ________________________________________________--J~ Oor1 Figure 6. Retransmit Timing for Interconnection Example 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 Ixx Ixxlxxl xxlxxlxxl xxlxxl xx xx xx I xxi xx xx xx I xxi 151413121110 9 8 7 6 5 4 3 2 0 Ixxlxxlxxlxxlxxlxxlxxlxxl INXF1 I OUTXF11 T/OXF1 Ixxl INXFO IOUTXFOI T/OXFO Ixxl R R R/W R/W R/W R/W NOTES: A. xx =reserved bit, read as ,0 B. R =read, W =write BIT o NAME RESET VALUE Reserved o Read as 0 i/OXFO o o o o o If ItOXFO '" 0, XFO is configured as a general-purpose input terminal. If I/OXFO '" 1, XFO is configured as a general-purpose output terminal. 2 OUTXFO 3 INXFO 4 Reserved 5 i/OXF1 6 OUTXF1 7 INXF1 o o Reserved 0-0 31-8 FUNCTION Data output on XFO Data Input on XFO. A write has no effect. Read as 0 If ItOXF1 '" 0, XF1 is configured as a general-purpose input terminal. If I/OXF1 '" 1, XF1 is configured as a general-purpose output terminal. Data output on XF1 Data input on XF1. A write has no effect. Read as 0 Figure 7. IOF Register Bit Summary Modified Code for TMS320C3x FIR Filtering The FIFO retransmit control for FIR filtering can be structured as in the following modified code fragment from the TMS32OC3x User's Guide (see Figure 8). The values loaded into the IOF register are chosen to set and reset the RTM and RFM terminals of the FIFO as appropriate, providing retransmit control. Figure 9 shows the control timing associated with the FIFO retransmit for the FIR filter. * TITLE FIR FILTER (I! denotes changes from code example in * the TMS320C3x User's Guide) * SUBROUTINE FIR * EQUATION: yen) = h(O) * x(n) + hell * x(n-1) + * ... + h(N-l) * X(n-(N-l» TYPICAL CALLING SEQUENCE * LOAD ARO * LOAD AR1 * LOAD RC * LOAD BK * LOAD IOF * II * CALL FIR * ARGUMENT ASSIGNMENTS: * ARGUMENT I " FUNCTION * * --------+------------------------------------------------ * II ARO ADDRESS OF FIFO where h vector is stored starting with h(N-1) * ADDRESS OF x(n-(N-1» * AR1 LENGTH OF FILTER - 2 (N-2) * RC LENGTH OF FILTER (N) * BK * I I IOF XFO, XF1 configured as outputs. XFO is high, XF1 is low. * Initial register content is 026h. FIFO in retransmit mode. * ! I REGISTERS USED AS INPUT: AR1, RC, BK, IOF * I I REGISTERS MODIFIED: RO, R2, ARO, AR1, RC, IOF * REGISTER CONTAINING RESULT: RO .global FIR * Initialize RO FIR MPYF3 *ARO,*AR1++(l)%,RO I I ARO not incremented h(N-1) * x(n-(N-1» -> RO * LDF O.O,R2 Initialize R2 FILTER (1 <= i < N) * RPTS RC Setup the repeat cycle I I ARO not incremented * *ARO,*AR1++(1)%,RO MPYF3 h(N-1-i)*x(n-(N-1-i»->RO RO,R2,R2 ADDF3 Multiply and add operation II * * ADDF RO,R2,RO Add last product LDI 066h,IOF Retransmit FIFO data starting with h(N-1) by asserting RFM (XF1) high LDI 026h,IOF I! * * * II * * * 3-62 RETURN SEQUENCE RETS end .end End RFM (XF1) high pulse to begin normal data reads Return Figure 8. FIFO Retransmit Control for FIR Filtering CLKA (H3) RTM (XFO) RFM (XF1)' _ _ _+-.....JI\-....J 80-831 h(N-1) h(O) Set FIFO In Retransmit Mode \ (LDI O2Oh.IOF) h (N_-_1_)1-__--'-_""""I__ Retransmit CoefIIclent V8CICr (LDI 066h.IOF) End RFM Pulse (LDI 028h.IOF) / h(N-1) h(O) Retransmit CoefIIclent Vector (LDI 088h.lOF) h(N-1) End RFM Pulse (LDI028h.IOF) End Retransmit Mode on FIFO (LDI 022h.IOF) v~-----_--J Generate yin) Cell FIR M Times 10 Generate y Vector ~----------~v~--------------J Generate y Vector of Length M + 1 Figure 9. Control Timing for FIFO Retransmit for the FIR Filter Conclusion Unlike conventional retransmit, 11's patented synchronous-retransmit feature allows the user to select or mark the FIFO data to be retransmitted. Synchronous retransmit is easily controlled by two FIFO terminals: R1M and RPM. As previously discussed in this application report, synchronous retransmit provides a very efficient method for transferring a series of FIR filter coefficients to a DSP without storing the coefficients in a standard SRAM or EPROM. By interfacing the DSP external 110 terminals to the FIFO retransmit tenninals, the DSP can effectively request the FIR filter coefficients on demand. The following FIFOs belong to the DSP application-specific family featuring synchronous retransmit. Table 1. DSP Applicatlon·Speclflc Family SPEED SORTS tc (n8) MAXIMUM FREQUENCY (MHz) MAXIMUM ACCESS 512x32x2 -15,-20,-30 67 11 SN74ACT3631 512 x 36 -15,-20,-30 67 11 SN74ACT3641 1Kx36 -15,-20,-30 67 11 SN74ACT3651 2Kx36 -15,-20,-30 67 11 DEVICE ORGANIZATION SN74ACT3638 (n8) 3-63 3-64 FIFO Mailbox-Bypass Registers: Using Bypass R~gisters to Initialize DMA Control Kam Kittrell and Steve Strom Advanced System LogIc - Semiconductor Group :illExAs INSTRUMENTS SCAA007A 3-65 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the Information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may Involve potential risks of death, personal injury, or severe property or environmental damage C'Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any licanse, either express or Implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, TE!xas Instruments Incorporated Contents Title Page Introduction ............................................................................. 3-69 Mailbox-Bypass Register Operation ......................................................... 3-69 DMA Controller Description ............................................................... 3-71 Initializing DMA Controllers With Mailbox-Bypass Registers .................................... 3-72 Conclusion .............................................................................. 3-74 List of IDustrations Figure Title Page SN74ACT3641 FIFO Functional Block Diagram .......................................... 3-69 2 Timing Diagram for Maill Register and MBFl Status Flag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-70 3 TMS32OC31 Memory Map in Microprocessor Mode Showing Mailbox-Bypass Register and FIFO SRAM Addresses .......................................................... 3-71 4 TMS320C31 DSP Block Diagram ..................................................... 3-72 5 Using DMA Controlto Transfer Large Data Blocks ........................................ 3-72 6 Using Mailbox-Bypass Registers to Initialize DMA Control ................................. 3-73 3-68 Introduction This application report describes the operation of FIFO mailbox-bypass registers and shows their application in an example that implements the direct memory access (DMA) control of a digital signal processor (DSP). All Texas Instruments (TI) 32- and 36-bit FIFOs contain mailbox-bypass registers that transmit priority data from one FIFO port to the other, either from port A to port B or port B to port A, without storing the data in the FIFO SRAM buffer. The SN74ACT3641, a unidirectional, clocked-DSp, application-specific FIFO (lK x 36), is used as the example for our discussion. In the following scenario, this device connects to a DMA controller that may be viewed as an integral part of any generic processing element. This discussion focuses on the DMA controller that is resident on board the TMS320C3! DSP. Mailbox-Bypass Register Operation The operation of the FIFO mailbox-bypass register is straightforward. Each FIFO that features the mailbox option has two 32- or 36-bit bypass registers (Maill and Mai12) to pass command and control information between both ports without queuing the information in the FIFO SRAM buffer. The functional block diagram of the SN74ACT3641 is shown in Figure 1. The associated timing of the Mail! register in the FIFO functional block diagram is shown in Figure 2. + CLKA, - CSA,- Port-A W/RA, - Control ENA, MBA,- I Mall1 Reglstar Logic - ~ - -@] ......... - I 'S Go 512 x 36, 1024 x 36, or 2048x36 SRAM ~ .s - LogiC I T~ r.- ! ! .......-::J T := RTM .cEo 3~ 0 I I Write Pointer AO-A35 I IR I AF I II Read Pointer T ~ 0 - ~iS II: -44 RFM T BO-B35 Status-Flag Logic OR AE + Flag-offset Reglatar FSo/SD FS1/SEN '"""- 10 '---- 0( .... I Mal12 Register I I Port-B Control Logic ~ ~ f+- CLKB CSB W/RB ENB MBB T Figure 1. SN74ACT3641 FIFO Functional Block Diagram 3-69 ClKA I \ , ________~~~(~EN~l~ r-__ \\-__--J/ \\-_ _.....J/ t_h(_EN_l___________________________________ H : W' W/RA ??ZlIT!ZZT4 MBA ?2lZZZl/ZZZZ? I I I I I : I W' ENA AO-A35 ClKB I W/RB o/ZIflf: II ENB ten BO-B35 II. I I MBB II i"--t ! 1..1.1 I" ~, : tsu(ENl~ tpd(M-DV) 1 ,I mdJ i__. ~......, k~w 1h(ENl fu\'\ tpd(C-MRl I I I -tI ~ tdls ------~~::.~:::::.:::).~ W1(remalns valid In Mall1 register after readl j..-. FIFO Output Register Figure 2. Timing Diagram for Mall1 Register and MBF1 Status Flag Operation of the SN74ACT3641 mailbox~bypass registers is summarized below and shown in Figure 2. The mailbox-select inputs (MBA and MBB) are used to choose between a mail register and the FIFO SRAM for a port data transfer operation. A low-to-high transition on CLKA writes the data on AO-A35 to the Maill register when a port-A write is selected by CSA, WiRA, and ENA while MBA is high. Likewise, a low-to-high transition on CLKB writes the data on BO-B35 to the Mail2 register when a port-B write is selected by CSB, W/RB, and ENB while MBB is high. Writing data to a mail register sets its corresponding status flag (MBFl or MBF2) low. Attempted writes to a mail register are ignored while the mail flag is low. When the port-B data outputs (BO-B35) are active and the mailbox-select input (MBB) is low, data on the bus comes from the FIFO output register. When the port-B data outputs are active and the port-B mailbox-select input is high, data on the bus comes from the mailbox register (Maill). The MaiU register data is always present on the port-A data outputs (AO-A35) when they are active. The Maill register flag (MBFl) is set high by a low-to-high transition on CLKB when MBB is high and a port-B read is selected by CSB, W/RB, and ENB. The Mai12 register flag (MBF2) is set high by a low-to-high transition on CLKA when MBA is high and a port-A read is selected by CSA, wiRA, and ENA. Data in a mail register remains intact after it is read and changes only when new data is written to the register. The mailbox-bypass registers can be easily mapped to a location within aDSPthat is separate from the location assigned to the FIFO SRAM within the same memory map. This is made possible by virtue of the mailbox-register control terminals and timing similarities between the mailbox-register read/write operation and the FIFO SRAM read/write operation. An example of assigning the mailbox register and FIFO to the memory map of the TMS32OC31 is shown in Figure 3. 3-70 Oh 03Fh 040h Interrupt Vectors and Reserved External STRB Active (NSMwords) 7FFFFFh SOOOOOh S07FFFh 808000h S097FFh S09800h S09BFFh 809COOh S09FFFh 32KWords Reserved Peripheral eBus Memory-Mapped Registers IntemalRAM Block 0 Internal RAM Block 1 SOAOOOh 9OO000h AOOOOOh I FIFO Read I FIFO Write External STRB Active FFFFFFh Figure 3. TMS320C31 Memory Map in Microprocessor Mode Showing Mailbox-Bypass Register and FIFO SRAM Addresses DMA Controller Description The TMS32OC3 I DSP contains an on-board direct-memory-access (DMA) controller that minimizes the requirement for the CPU to perform input/output operations. The TMS320C31 DSP block diagram is shown in Figure 4. Because of the DMA controller, the TMS320C31 can operate with slow external memories, peripherals, or analog-to-digital converters, for example, without slowing the computational throughput of the CPU. Address and data buses are specifically dedicated to the DMA. As a result, there exists minimal conflict between the CPU and DMA controller. Using its address generators, source and destination registers, and transfer counter, the DMA controller can react to interrupts much like the CPU. Performing data transfers based upon interrupts received enables the DMA to execute input/output transfers that are typically the task of the CPU. While the DMA is receiving and transmitting data, the CPU is permitted to continue processing data. A DMA transfer consists of two operations: reading from a memory location and writing to a memory location. These read and write operations can be as a block or a single word. The DMA controller can read from and write to any location in the TMS32OC31 memory map, including all memory-mapped peripherals. 3-71 Program cache (64 x 32) ROY, HOlOA, STRB, RJW, 031-00, A23-AO RESET INT3-0 4( - +- ClKIN VOO VSS - SHZ - X2! l! ~ 8 RAM Block 1 (1K x 32) OateBuaes 1 1: - lACK XF1-0 MCBl! MP X1 RAM Block 0 (1K x 32) 'I' ..j CPU OMA Integer! Integer! Floating-Point Floating-Point Multiplier AlU Address Generators Generator 0 ......... Serial PortO II) .' :::I III i!III Control Registers 8 Extended-Precision Registers Address D=l .c i ......... Timer ......... Timer 0 Address Generator 1 8 Auxllllary Registers 1 12 Control Reglstera "'''--- ~ Figure 4. TMS320C31 DSP ,Block Diagram Initializing DMA Controllers With Mailbox-Bypass Registers FIFO memories, such as the SN74ACT3641, typically channel data between a generic local or backplane bus and a TMS32OC3l DSP. The FIFO collects incoming information from the bus and develops packets or vectors of data for transfer to the DSP via the DMA controller (see Figure 5). The packet size of the stored data is easily defined by using the FIFO programmable almost-empty or almost-full flags and, if necessary, its empty and full flags. Data transfers via the DMA controller are performed block by block from the FIFO to the DSP or to off-chip memory such as RAM. The DMA-controlled transfer is preferred for moving large blocks of data. Instead of using the DSP's CPU for each single-word transfer, investing in a small amount of setup overhead allows the DMA to initialize the transfer of several words. In this case, the DMA controller becomes the bus master and performs the block transfer while the CPU is not using the external bus. The CPU is free to accomplish its primary task of performing mathematical operations. Host Processor 36 FIFO SRAM r------, I I I I 36 I OMA I 1-.....-t-......~l-7y...-+I~ Control I I I IL _ _ _ DSP _ _ _ .JI SN74ACT3641 Unidirectional FIFO Figure 5. Using DMA Control to Transfer Large Data Blocks 3-72 Generally, the DMA controller requires the following information for data-block transfers: location of the data, destination of the data, and size of the data block to be transferred. The DMA controller is initialized by using the FIFO mailbox-bypass registers. This concept is shown in Figure 6. AJ; previously, mailbox-bypass registers are useful in separating a control word from the data in a FIFO queue. In the example shown in Figure 6, the bypass registers of the SN74ACT3641 FIFO provide the DMA controller with the block-length initialization before performing the block transfer. At the same time, the mailbox-register status flags alert the DMA controller that the FIFO data is ready for transfer. In other instances, the mailbox register can also store a destination address in the DSP memory for incoming data. In the earlier discussion, data flow through the mailbox register has been assumed to be in the bus-to-DSP direction. However, because of the bidirectional nature of the mailbox feature, the Mai12 register (see Figure 5) can also be used for transferring data in the DSP-to-bus direction. Many bus architectures support burst writes or have virtual addresses that can use the mailbox-bypass register for initialization. FIFO Memory FIFO Write { LocaUon Block 1 Date ••• Bypass Register ••• Date Write Order Block 1 Length -+ Block 2 Length (a) BU8-To-FIFO DATA FLOW FIFO Read { Location FIFO Memory ••• Bypass Register ••• Initialize DMA With Block Length 1tl1 Date Read Order -+ DMA Transfer (b) FIFo-To-BUS DATA FLOW Figure 6. Using Mailbox-Bypass Registers to Initialize DMA Control 3-73 Conclusion Mailbox-bypass registers are very useful in performing block-data transfers from a bus to a processing element or vice versa. Integrating the 32- or 36-bit mailbox registers on board the FIFO chip significantly reduces the device count per system. Likewise, implementing on-board mailbox registers with access timing similar to the FIFO SRAM also reduces the requirement for control logic. TI's DSP application-specific FIFOs, in addition to the other application-specific FIFOs with mailbox-bypass registers, are summarized in Table 1. Table 1. FIFOs Featuring Mailbox-Bypass Registers 3-74 DEVICE ORGANIZATION SPEED SORTS MAXFREQ MAX ACCESS APPLICATION SN74ACT3622 256x36x2 -15,-20,-30 67 11 DSP SN74ACT3621 512x36 -15, -20,-30 67 11 DSP SN74ACT3632 512x36x2 -15,-20,-30 67 11 DSP SN74ACT3638 512x32x2 -15,-20,-30 67 11 DSP SN74ACT3641 1Kx36 -15, -20, -30 67 11 DSP SN74ACT3642 1Kx36x2 -15,-20,-30 67 11 DSP SN74ACT3651 2Kx36 -15, -20, -30 67 11 DSP SN74ABT3611 64x36 -15,-20,-30 67 10 High Bandwidth SN74ABT3612 64x36x2 -15, -20, -30 67 10 High Bandwidth SN74ABT3613 64x36 -15, -20, -30 67 10 Intemetworking SN74ABT3614 64x38x2 -15, -20, -30 67 10 Intemetworking Advanced Bus-Matching/ Byte-Swapping Features for Internetworking FIFO Applications Tom Jackson Advanced System Logic - Semiconductor Group :II TEXAS INSTRUMENTS SCAA014A 3-75 IMPORTANT NOTICE Texas Instruments (TIl reserves the right to make changes to Its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtaln the latest version of relevant information to verify, before placing orders, that the information being relied on Is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications',). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. " TI assumes no liability for applications assistance, customer product design, software performance, or Infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or Implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-76 Contents Title Page Introduction ............................................................................. 3-79 Conventional Bus-Matching Data Reads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-80 Tl's Bus-Matching Data Reads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-81 Conventional Bus-Matching Data Writes ..................................................... 3-84 TI's Bus-Matching Data Writes ............................................................. 3-85 TI's Byte-Swapping Feature .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-86 Conclusion .............................................................................. 3-86 List of illustrations Figure Title Page Big-Endian and Little-Endian Fonnats ...... ~ ........................................... 3-79 2 Data-Read Cycle for 36-Bit to 9-Bit Bus Matching ........................................ 3-80 3 Bus Sizer ......................................................................... 3-81 4 Port-B Data-Read Cycle for 36-Bit to 9-Bit Bus Matching .................................. 3-82 5 Little-Endian Data-Output Structure .................................................... 3-83 6 Big-Endian Data Output ............................................................. 3-83 7 Big-Endian Data-Output Structure ..................................................... 3-83 8 Data-Write Cycle for 9-Bit to 36-Bit Bus Matching ........................................ 3-84 9 Port-B Data-Write Cycle for 9-Bit to 36-Bit Bus Matching .................................. 3-85 10 Big-Endian Fonnat of Byte Write ...................................................... 3-85 11 Byte Swapping ..................................................................... 3-86 3-77 3-78 Introduction Bus matching and byte swapping are features that Texas Instruments (TI) has added to their intemetworking family of application-specific, flrst-in flrst-out memories (FIFOs). The first two FIFOs available are the SN74ABT3614 and SN74ABT3613. The bus-matching feature allows the user to dynamically select the desired bandwidth, either long-word format (36 bits), word format (18 bits), or byte format (9 bits) for mixed-bus systems. Byte swapping allows the user to reconfigure protocols for different microprocessor-based systems, such as big-endian format (i.e., RISe-based microprocessors such as the Me68000, IBM370) where the most signiflcant bit (MSB) is 0 or little-endian format (i.e., elSe-based microprocessors such as iPAX, x86, DEe VAX) where the least significant bit (LSB) is 0 (see Figure 1). The bus-matching and byte-swapping features can be used independently or in conjunction with one another, which allows the user a wide range of possible solutions. Motorola Microprocessor (big endlan) Intel r " Microprocessor (little end Ian) • XOOO8 X0004 ~------:------- I-t-----------31 o • :~~:t1--------------:--------------o 31 (MSB) (LSB) Figure 1. Big-Endian and Little-Endian Formats Intel is a trademark of Intel Corp. 3-79 Conventional Bus-Matching Data Reads With the evolution of 32-bit microprocessors and digital signal processors (DSPs), designers must add large complex discrete circuits to provide data continuity between mixed data-bus systems. A typical solution requires four 9-bitFIFOs and considerable board space (s~ Figure 2). ROCLK Status Flags ROCLK FIFO 1 9 00-08 Q1 ROEN1,OE1 Q1 09-017 FIFO 2 9 Q2 J I \ ~ 3B~e1_;) I i~-f ROEN2,OE2 018-026 FIFO 3 9 Q3 Q3,Z 027-0~5 FIFO 4 Status Flags 9 Q4,Z Q4 9 I L.-t I I I I I I I I I I I I i Q2,Z 1 0 1 0 10to-15 ns ~ KB~2_1>B-- ~I I QO-Qa Figure 2. Data-Read Cycle for 36-Blt to 9-Blt Bus Matching To provide bus matching from a 36-bit bus to a 9-bit bus, access time and flag synchronization are critical issues due to the combination of separate components. Figure 2 shows the required circuitry and associated timing diagram for a conventional bus-matching operation. The fIrSt byte of the long word is written into FIFOl on 00-08. On the rising edge of the read clock with read enable-l (RDEN1) and output enable-l (OEl) held high, the first byte of FIFO 1 (byte 1_1) is read out. During that first read cycle, the remaining three FIFOs must be disabled. Before reading the second byte from FIF02 (byte 2_1), the read enables and output enables of FIFO 1 are asserted low, putting the FIFO outputs into 3-state, which prevents any bus-arbitration problems. FIF01 must be disabled before FIF02 is enabled. In addition to the associated propagation delays of enabling and disabling the FIFOs, the typical access time of a single FIFO can range from 10 to 20 ns. To ensure proper device synchronization, the FIFO access time must be increased to allow for the propagation delays of the control signals. The increased propagation time for data reads can cause a bottleneck for data that is waiting to be written into the other FIFOs. This delay dramatically impacts system performanCe and data throughput. If the timing parameters are violated, the result is bus contention and lost data. To complete the long-word read, the cycle is repeated three more times. For subsequent long-word data reads, the process begins with the second byte of FIFO 1 (byte 1_2). This data ping-ponging is prolonged and requires excessive access times and data setUp and hold times to perform a single long-word read cycle. These factors contribute to reducing the effective maximum operating frequency of the system. 3-80 Tl's Bus-Matching Data Reads 11 has designed the intemetworking FIFOs to provide the user a single-chip solution for dynamic bus matching, in addition to fast data access times (ta - 10 ns). The bus-matching feature offers a flow-through architecture that maintains port-to-port transparency and eliminates the need for any bus-arbitration control logic. Bus matching is performed with the FIFO on the port B bus and can be configured in long-word format (36 bits), word format (18 bits), or byte format (9 bits) for data reads from FIFOI or written to FIF02, in the case of the SN74ABT3614 bidirectional FIFO. Port-B bus size can be changed dynamically and synchronous to CLKB to communicate with peripherals of various bus widths. The bus-matching feature is implemented using the big-endian (BE) format and the port-B bus size-select (SIZO, SIZ1) terminals (see Table 1). Table 1. Bus Size-Select Terminals TERMINAL NUMBER PACKAGE BE SIZ1 12O-pin TOFP (PCB) 47 50 SIZO 51 132-pin POFP (PO) 132 129 128 The sizing function is performed on the output of port B after a long word has been written into the FIFO on port A (see Figure 3). If an output register is not used (e.g., 9- or 18-bit data reads), no line terminations such as pullup resistors are required due to the bipolar output structures of 11's advanced BiCMOS technology. 36 Sizer Port B Figure 3. Bus Sizer By varying the_ assertion levels of the three control terminals, five different bus-format configurations can be selected (see Table 2). Table 2. Bus Configurations BE SIZO SIZ1 X L L L H H L H H L L L H L H BUS CONFIGURATION Long-word size Word size - big endlan Word size -little endian Byte size - big Bndlan Byte size -1!We endlan 3-81 ----------- The byte-order arrangement of data that is read from or written to the FIFO can be changed synchronous to the clock. The bytes are rearranged within the long word, but the bit order within the bytes remains constant. The byte-swapping feature is implemented by asSerting port-B byte-swap select (SWO, SWl) terminals (see Table 3). Table 3. Byte-Swap Select Terminals TERMINAL NUMBER PACKAGE SW1 SWO 12o-pln TOFP (PCB) 48 49 132-pin POFP (PO) 131 130 The example as shown in Figure 4 takes the conventional 36-bit to 9-bit bus-matching example as shown in Figure 2 one step further by incorporating the byte-swapping feature. A timing diagram of a little-endian, byte-size, byte-swap data read from port B using the SN74ABT3613 unidirectional FIFO is shown in Figure 4. With a 36-bit-long word written into memory from port A, data read can be performed. On the rising edge of CLKB with the port-B chip select (CSB) asserted low, the size and swap functions can be selected. The littIe-endian format is chosen by asserting BE high. Byte size is selected by asserting SIZI high and SIZO low. On the second clock cycle, the byte swap is performed by asserting SWI low and SWO high for one clock cycle. ClKB ENB ~~~ ~~~~~~I I I I I: taU(SWl ~ lh(sZ)~ ~. I I I l.-.:- lh(8W) ~ ~ $ No Operation I ~ ij I . I I tau(EN) --*-t! I I I I I ~i;....:hI~,-h.~~~~A !I ~ ! I I lh(SZ,~ I ---l4--tI BE_ I I ,.,.. !I ~ I I !I I ~ ~ I I Not tau(PO) ~ tan BO-B8 t ---t-+I ~ j4- ta ~ * Previous Data SIZO = Hand SIZ1 • H selects the mail1 register for output on BO-B35. Figure 4. Port-B Data-Read Cycle for 36-Blt to 9-Blt Bus Matching I IcIIS~ Read 4 j-- During the second clock cycle, the first byte appears on the output bus BO-B8. After four successive read cycles are completed, the entire long word is parsed out onto the bus in four 9-bit data packets (see Figure 5). In byte-size or word-size data reads, the unused bytes hold the last FIFO output values. After the four bytes are read, the configuration can be dynamically changed. Example: Little Endlan, Byte Size, Byte Swap BE = H, SIZ1 =H, SIZO= L,SW1 =L,SWO= H Firat Read A •• • B c Fourth Read D Figure 5. Little-Endian Data-Output Structure If the example shown in Figure 4 is configured for big-endian format, the data is output on bus B27-B35 (see Figure 6 and Figure 7). No line termination is required for the unused data outputs. This is a dramatic improvement, not only in design ease, but in performance over the conventional data ping-ponging technique shown in Figure 2. B27-B35 ------«PrevloU8 Dats X Read 1 X Read 2 X Read 3 Figure 6. Blg-Endlan Data Output Example: Big Endlan, Byte SIze, Byte Swap BE = L, SIZ1 = H, SIZO = L, SW1 = L, SWO = H I I I I A Firat Read B c ••• D Fourth Read Figure 7. Blg-Endlan Data-Output Structure X Read 4 r- Conventional Bus-Matching Data Writes Conventional bus-matching data writes experience the same timing restrictions as data reads. The example in Figure 8 shows the circuitry required for 9-bit to 36-bit data writes. Just as in the example in Figure 2. four 9-bit FIFOs are required in addition to an extra control-logic block to synchronize all the write enables. WRTCLK Status Flags QO-Qa Q9-Q17 9 FIFO 1 1-+-----..., 9 Q1 FIFO 2 I-+----i-..., 9 W3 Q18-Q26 FIFO 3 1-1-----1---l Q3 9 W1 W2 W3 W4 W4 WEN Control Logic Q27-Q35 Status Flags FIFO 4 1------+----' 9 Q4 9 Figure 8. Data-Write Cycle for 9-Bit to 36-Blt Bus Matching The write-control logic controls each data write in a round-robin style. The control logic consists of a bank of flip-flops that generate the appropriate write-enable signal. After four successful data writes. a long word can be read from the FIFO bank. However. the FIFO bank must have its full- and empty-status flags monitored to ensure data integrity. The empty status is monitored from FIF04. the last FIFO in the chain. Upon an empty signal. additional data reads are immediately disabled. The full status is monitored from FIFO!. the tIrst FIFO in the chain. When a full status is indicated. further data writes are disabled. To ensure maximum performance. the status flags require fast propagation delays for proper data synchronization. Otherwise. data overwrites can occur. corrupting the FIFO data. or additional wait states must be introduced into the system. Due to the synchronization issues. additional bus control or interrupts become extremely difficult by using the half-full. almost-full or almost-empty flags. This also limits FIFO operations. The data setup and hold times also must be increased to ensure there are no bus contentions during a write operation. TI's Bus-Matching Data Writes The timing diagram for perfonning a little-endian, byte-size, byte-swap data write to port B of FIF02 using the SN74ABT3614 is shown in Figure 9. By implementing TI's SN74ABT3614 bidirectional FIFO in a design, bus matching can be performed in either direction without the need for additional glue logic or loss of system performance. elKB FFB (HIgh) I I I I I I I I thlEN) -+I 14- IsUIEN)=ri 1. : lj"--- ~~I--------------------------------------------------------~: I W/RB ENB SW1, SVo/O mmllmm II SIZ1,SIZO Write 1 t SIZO =Hand SIZ1 Write 2 Write 3 Write 4 =H wrHes data to the mail2 register. Figure 9. Port-B Data-Write Cycle for 9-Bit to 3~Bit Bus Matching On the rising edgeofCLKB with the portB selected (CSB), the size and swap functions can be selected. The little-endian format is chosen by asserting BE high. Byte swap is selected by asserting SIZO high and SIZllow. These assertion levels are maintained for the entire write cycle. On the second clock cycle, the byte swap is performed by asserting SWllow and SWO high for one clock cycle. The data is then written into BO-B8, since the little-endian format has been selected. If Figure 9 is configured for big-endian format, the data is written into B27-B35 (see Figure 10). No line tennination in the form of pullup resistors is required for the unused data inputs. B27-B35 WrIte 1 Write 2 WrIte 3 WrIte 4 Figure 10. Blg-Endian Format of Byte Write 3-85 TI's Byte-Swapping Feature TI has designed the intemetworking FIFOs to provide designers maximum flexibility and ease of use. In addition to the bus-matching feature, a byte-swapping option has been added. The byte-swapping feature allows communication between systems with mixed bus protocols such as those using by RISe and else microprocessors. The previous examples of TI's bus matching (Figure 2 through Figure 10) have included the byte-swapping function to demonstrate the true power and flexibility these features provide when implemented together. Byte swappin:g is performed on port B of the FIFO (see Figure 11) after the bus-matching function has been executed. Either feature can be implemented separately depending on the system requirements. PortA A B c o PortB 9 9.... , 9, 9 FIFO SN74ABT3614 9, 9 , o c 9 , B 9 , A i i Byte Data Byte Swap Figure 11. Byte Swapping As with the bus-matching function, there are several variations of byte swapping, depending upon the assertion levels of the port-B byte-swap select terminals (see Table 4). Table 4. Byte--Swapping Option SWO SW1 L L H H L H L H BUS CONFIGURATION No swap Byte swap Word swap Byte-word swap Bus matching and byte swapping are performed in the following sequence for all data reads; the 36-bit word is first read, the swap is performed, followed by the bus-size function. The converse is true for data writes. Conclusion The ability to dynamically select the desired bus configuration and format is a very useful feature for today's designs. Many systems, such as networlc switches and routers, implement high-speed backplanes that are typically 32 to 36 bits to ensure maximum bandwidth for data; however, there are many 8-bit and 16-bit controllers and buses in existence. TI's bus-matching feature ensures a flow-through, high-speed architecture that permits multiple logical permutations; There is no longer the need for rerouting bytes on buses and manually controlling bus arbitration through a large and costly discrete solution. The SN74ABT3613 FIFO provides a unidirectional datapath with bus matching and byte swapping on port B. The SN74ABT3614 provides a full bidirectional datapath and supports bus matching and byte swapping in either direction. Both of these FIFOs feature TI advanced-clocked architecture ina space-saving single-chip solution that offers a maximum clock speed of 67 MHz with 10-ns access time. Parity-Generate and Parity-Check Features for High-Bandwidth-Computing FIFO Applications Tom Jackson Advanced System Logic - Semiconductor Group ~1ExAs INSTRUMENTS SCAA015A 3-87 IMPORTANT NOTICE Texas Instruments' (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant Information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or Implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated Contents Title Page Introduction ............................................................................. 3-91 Parity-Generate Feature ..................... : ............................................. 3-91 Using Parity Error to Force an Exception ..................................................... 3-93 Conclusion .............................................................................. 3-94 3-89 3-90 Introduction Parity-generate and parity-check features are available in both the intemetworking family of first-in, first-out memories (FIFOs), SN74ABT3614 and SN74ABT3613, and the high-bandwidth-computing family, SN74ABT3612 and SN7 4ABT3611. Parity generate and parity check are needed in high-bandwidth and high-speed computing applications where demanding data integrity levels are required. All of TI's 36-bit FIFOs have bidirectional mailbox registers that allow quick access to data by bypassing the FIFO static random-access memory (SRAM) core. The same input/output (I/O) is shared between the mailbox registers and the FIFO data registers that allow parity generate and parity check on both the FIFO memory data and the mailbox-register data. The parity-generate and parity-check features are designed for fault-tolerant systems, such as those in computing and telecom that require error-detection techniques, in addition to many of today's microprocessors that have provisions for parity detection. Parity-Generate Feature The parity-generate feature enables the user to generate odd or even parity in the most significant bit (MSB) of each byte on either port A or port B of the FIFO. Odd and even parity are defmed as follows: Odd parity: The parity bit is set to one for an even number of ones, including the parity bit. Even parity: The parity bit is set to zero for an odd number of ones, including the parity bit. Parity is generated for data reads from either port A or port B of a bidirectional FIFO by asserting parity generate A (PGA) and parity generate B (pGB), respectively ~CSAIB low, PGAIB high, ENA/B high). In Figure I, a write to a FIFO or mailbox register stores the levels applied to all 36 inputs, regardless of the state of the parity-generate inputPGB. When data is written to a port, the lower eight bits of each byte are copied to the parity-generate tree. These bits are then used to generate a parity bit according to the level of the ODDIEVEN select pin (see Table 1). When PGB is asserted, the results of the parity generator are written to the MSB of the output register. If PGB is low, the original contents of the MSB of the byte under consideration are transferred to the corresponding location in the output register. Memory Odd/Even PGB P 8 Data Input Output Register BO-B8 Parity Generator Memory Out Figure 1. Parity-Generate Circuit 3-91 Table 1. Parity-Generate Input Pins PACKAGE PEFA PIN NUMBER PEFB ODD/EVEN 120-pin TOFP (PCBl 39 53 44 132-pin POFP (POl 9 125 3 Each mailbox register has an associated parity-generate and check (gen/check) circuit (see Figure 2) that enables parity to be generated and checked on either port of the FIFO. The circuit that generates parity for the Maill mailbox register is shared by the port-B bus (BO-B35) and generates and checks parity for the FIFO data bus, as well as to check parity for the Mail2 mailbox register. The circuit that generates parity for the Mail2 mailbox register is shared by the port A bus (AO-A35) to check parity for FIFO data and mailbox register. The shared parity trees of a port generate parity bits for the data in a mailbox register when the port write/read select input is low, port-mail select input is high, and port-parity-generate select is high. Generating parity for mail register data does not change the contents of the register. cLKA-----I CSA - - - - - I Port-A r---r=:::::,========= I wiifA ENA - - - - - I Control Lagle MBA - - - - - I Parity ,J-I....- - - - , ,..-_-, ~++=++:""'__...r-"jM~a;nIli"1-1+'1 ~Gen/Chec..:J+, RSf - ~:E~ - C 1===;~IL!R~eg~l~ster~..J-+9;:::==~_I::;;==~~.......__ PGB Devlce r-~ ---------~-=r~~~ Control ~ ~ I I i i _ 64 x 36 a: rr.. SRAM IS ... ~e I"" III c! I "-.. II I Q) a.,c 36t++-~11+f j AO-A35--....- :M:B:F:1 PEFB I Write II ""-- - I Read I Pointer Pointer I "t"t I.- Hf~t, I I .. I ! r:I 36 I II ...........-BO-B35 I FF---+--~t+~~I~---~iSte~N;.~F~la~g.,-----~I~-~++~---AEFE AF---4---rI1--~'~~-IIL-~LO~)g~llc~~r------r-'-ttl1l1---- L~~-----4---------J I FSO I FS1----~~.-------I ~ I ProgrammableFlag Offset Register J PGA-----tt====~~~~~b,~r_~~~_,I-~--~ ~ .,~I Mall2 : ~ Parity _~ PEFA _ _ _ _~~~~.-_-_-_-_·_... 1IcGe=nJ::C:.:hec:::Jkl I MBF2--------.-------~ Register ........ ..---CLKB ..---CSB Port-B ...... Control "'---W/RB Lagle "'---ENB "'---MBB ...... --~ Figure 2. SN74ABT3611 Mailbox Registers and Associated Parity Gen/Check Circuits Functional Block Diagram The parity-generate and parity-check features allow the user to select odd or even parity and to passively check the results of all incoming data to either port A or port B of the FIFO without disrupting normal operations. Both port A (AO-A35) inputs and port B (BO-B35) inputs have four 9-bit parity trees to check the parity of incoming or outgoing data (see Figure 3). Parity is checked on the ninth MSB of each byte. 3-92 Odd/Even - - - - , BE----, Odd/Even SIZ1 SIZO A35 9 B35 ER ER ER ER ER ER ER ER 9 9 9 9 9 9 9 AO BO Figure 3. Parity Trees A parity failure on one or more bytes of the input bus is indicated by a low level on the port parity-error flag (PEFA, PEFB) (see Table 2). The parity-error flags can be ignored if this feature is not desired. Table 2. Parity-Error Input Pins PACKAGE PIN NUMBER PGA PGB ODD/EVEN 120-piri TOFP (PCB) 38 54 44 132-pin POFP (PO) 10 124 3 The user can choose odd or even parity by asserting the ODDIEVEN input or allow the FIFO to default to even parity. In this manner, the user can select the parity format that best fits the application requirements. Since foUr 9-bit parity trees are used, it is possible to implement the parity-check function on the bus-configuration port in conjunction with the bus-matching feature of the intemetworking FIFOs. In this manner, any bus width that has been selected, 9-bit through 36-bit, can have parity checked. The parity-checking circuit is designed to ignore all error flags that may be generated on unused bytes. As in parity generate, the four parity trees used to check the port-A inputs are shared by the Mail2 mailbox register. Port-B inputs on bidirectional FIFOs are shared by the Maill mailbox register (see Figure 2); therefore, parity errors are detected before the data is entered into the FIFO SRAM core. Using Parity Error to Force an Exception Although the parity-check feature is passive, it permits the designer to disregard data before it is written into the FIFO SRAM core. This type of functionality is easily implemented by the circuit shown in Figure 4. Since parity is checked on the inputs before being written to the FIFO, it is possible to capture the erred data and force an exception. Figure 4 shows a data error that has been detected by a low onPEFA. The associated propagation delay, [tpd(D-PE)] of a valid error flag is IOns for the -15 speed sort. This allows adequate time for a fast programmable logic device (tpd PLD) to disallow a data write prior to the data becoming valid. If no parity error is detected, the data write is performed. If FIFOs with slower speed sorts (-20, -30) are used, the associated propagation delay is increased. This method eliminates the need for external counters to track the erroneous data through the FIFO to the output. By forcing an exception, the parity is captured and the clock cycle passes without writing the datil: to the FIFO memory core. 3-93 CLK WRTCLK T WEN PLD PEFA ~ ~: WRTCLK I I \ --./ tau j---- II W E N _ " " T ' " :_ __ tpd(pLD) PEFA ~I I "'''''''''''''''''''~'''''''''''''''''''''''''''''''''''''''''''~'''''''''7''7\ : ~----':----. DATA ~ . tpd(D-PE) ~.I ~~,.,.,..,..,..,..,.~""""""..,..,..,..,..,.,.,..,..,..,..,.~ i 14- I : tpd(D-CLK) ~ Figure 4. Parity-Error Exception Circuit Conclusion As systems become more integrated and bus speeds increase, there is a growing need to ensure data integrity. When parity generate or parity check is required by dynamic random-access memory (DRAM) refresh cycles, bus noise, or other card-to-card performance issues, U's parity-generate and parity-check features provide a high-speed, space-saving alternative. 3-94 FIFO Applications Page Multiple-Queue First-In, First-Out Memory SN74ACT53861 ............................. 3-97 Interfacing TI Clocked FIFOs With TI Floating-Point Digital Signal Processors ............ 3-121 FIFOs With a Word Width of One Bit ................................................ 3-141 Internetworking the SN74ABT3614 ................................................. 3-161 High-SpeEld, High-Drive SN74ABT7819 FIFO ..................•..................... 3-181 SPARC MBus-to-Futurebus+ Bridge Using the Texas Instruments Futurebus+ Chipset .... 3-193 1K x 9 x 2 Asynchronous FIFOs SN74ACT2235 and SN74ACT2236 .................. 3-207 64-Byte FIFOs SN74ALS2232A and SN74ALS2233A ......•.......................... 3-217 3-95 3-96 Multiple-Queue First-In, First-Out Memory SN74ACT53861 Peter Forstner Semiconductor Group SCAA026A ~ThxAs INSTRUMENTS 3-97 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of Its semiconductor products and related software to the specifications applicable at the time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications'1. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products In such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. . TI assumes no liability for applications assistance, customer product design, software performance, or infringenient of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3--98 \ Contents TItle Page Introduction ............................................................................ 3-101 Main Areas of Application . ................................................................ 3-102 The Multi-Q FIFO ....................................................................... 3-103 Construction of the Multi-Q FIFO ......................................................... 3-103 Configuration Registers ............................................................... 3-105 Allocation of Queues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-105 Cells Instead of Words of Data .......................................................... 3-106 Flags ......................................... ; .................................... 3-108 Programming ....................................................................... 3-109 Extension of Word Width ................................................................ 3-111 Progranuning Examples ................................................................. 3-115 Applications ............................................................................ 3-117 Summary ............................................................................... 3-120 3-99 List of IDustrations Figure 1 TItle Page FIFO Data Flow ......................•...............•..•......................... 3-101 2 A1M Te1eco=unications Exchange System Block Diagram . .. • . . . . . . . . . . . . . . . . . • . . . . . . . .. 3-102 3 A1M-Header Structure ....................................••....................... 3-103 4 Multi-Q FIFO Functional Block Diagram ......•.......•..•....•.............•.......... 3-104 5 Data Stream With Odd Cell Size ..............................•....................... 3-106 6 Writing Cells Into the FIFO ..........................•.. '.' . . . . . • . . . . . . . . . . . • . . . . . • . . . 3-107 7 Faulty Writing of Cells Into the FIFO: ISOC Comes Too Soon ... .. . . . . . .. . . . . . .. . . . .. .. .. .. 3-107 8 Faulty Writing of Cells Into the FIFO: ISOC Comes Too Late or Not at All .................... 3-107 9 Reading Cells Out of the FIFO ....................................................... 3-108 10 Hysteresis of the PFI Flags With Configuration Registers PFl_W and PFl..R, .................. 3-109 11 Connection of a Microcontroller to the Auxiliary Bus ..................................... 3..:.109 12 Extension of Word Width With 18-Bit or 36-Bit Input and/or Output Data .............•....... 3-112 13 Extension of Word With 9-Bit Input Data ............................................... 3-113 14 Data Flow of an A1M Data Stream in Two Multi-Q FIFOs ........... ,...................... 3-114 15 A1M-ExchangeReceivingUnit ...................................................... 3-117 16 A1M-Exchange Transmitting Unit .................................................... 3-117 17 Connection of a Multi-Q FIFO to a Receiving Unit Using an 8-Bit or 16-Bit UTOPIA Interface With One Queue ............................... 3-118 18 Priority-Controlled Connection of a Multi-Q FIFO to a Receiving Unit Using an 8-Bit or 16-Bit UTOPIA Interface ............................................. 3-118 19 Connection of a Multi-Q FIFO to a Transmitting Unit Using an 8-Bit or 16~Bit UTOPIA Interface With One Queue .............. " ...... , •....... 3-119 20 Priority-Controlled Connection of a Multi-Q FIFO to a Transmitting Unit Using an 8-Bit or 16-Bit UTOPIA Interface ............................................. 3-119 21 Switching Matrix With Bottleneck Between Two Switching Elements ........................ 3-120 List of Tables Table 1 2 3 4 5 6 7 8 3-100 ,TItle Selecting the Queue When Reading the FIFO.. . . . .. . .. .. .. . . .. .. .. . . .. . .. . .. .. . . .. . . . ... Configuration Registers . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Port-Control Register PORT ..................•....•..................••........•.... Multi-Q FIFO Flags ...................•............................................ Configuration-Registers Access Order ................................................. Example of Configuration Registers Programmib.g: 18-Bit W:rite, 18-Bit Read ................. Example of Configuration Registers Programming: 9-Bit Write, 18-Bit Read .................. Example of Configuration Registers Programming: 18-Bit Write, 9-Bit Read .................. Page 3-104 3-105 3-106 3-108 3-11 0 3-115 3-115 3-116 Introduction This application report presents a detailed description of the versatile functions of the SN74ACT53861 multiple-queue (Multi-QTM) first-in, first-out (FIFO) memory. Examples of circuits show how the device can be controlled and cascaded. Typical application examples show how the device can be used in asynchronous transfer mode (ATM) telecommunications exchange systems. Memories are indispensable circuit components of digital-system subassemblies. There are a large number of memories in various configurations for many application requirements. Each memory is suited for specific and specialized applications. One of these specialized memories is the FIFO memory, which provides intermediate storage of data being transferred between two electronic systems. The designation FIFO indicates how the data flows. A FIFO has separate data input and data output; however, the first word of data written into the memory is the first to leave when it is read (see Figure 1). Within the FIFO, words of data wait in a data queue. If a FIFO is configured between two systems that are working asynchronously, the FIFO must be able to manage the synchronization of the data flow to both systems to prevent metastable situations. I Input Data I II FIFO "'V I Stored Data I t I Stored Data I 't I I Stored Data • •• Stored Data I I t I Stored Data I 't I Stored Data L Output Data I U J Figure 1. FIFO Data Flow FIFOs differ from one another iIJ. their word widths, memory capacity, and in the way they are controlled. Texas Instruments (TI) offers FIFOs with word widths from 1 to 36 bits and memory capacities from 64 to 4096 words. Because FIFOs have alternative methods of control, TI offers strobed FIFOs and clocked FIFOs. A detailed description of the various methods of controlling FIFOs can be found in other TI application reports. The various word widths and memory capacities available are described in the 1996 High-Performance FIFO Memories Data Book, literature number SCADOO3C. In addition to standard FIFOs, versions for special purposes have been designed for specific applications. The TI Multi-Q FIFO is an application-specific FIFO designed for ATM telecommunications exchange systems. Multi-Q is a trademark of Texas Instruments Incorporated. 3-101 Main Areas of Application The SN74ACT53861 Multi-Q FIFO is designed specifically for ATM telecommunications exchange systems. As shown in Figure 2, ATM telecommunications exchange systems can have three functional parts: • Receiving unit (one per channel) • Switching matrix • Transmitting unit (one per channel) The ATM used for data transmission supplies the receiving unit with digital information, which is usually apportioned in cells having a length of 53 bytes. Each cell consists of a 5-byte cell header and a 48-byte payload. The cell header ! includes: • Ultimate destination: virtual channel identifier (VCI) • Immediate next destination, i.e., the next ATM exchange installation through which the ultimate destination is reached: virtual path identifier (VPI) • The type of information contained in the cell: payload type (PI') • The importance, or priority, of the cell: cell-loss-priority (CLP) bit • Error-correction controller: header error control (HEC) In certain applications, extending the cell header by one to two bytes provides the ATM exchange installation with internal information (tagged cells) (see Figure 3). Interfaces • Interfaces Switching Mstrlx • • • • • Controller Figure 2. ATM Telecommunications Exchange System Block Diagram 3-102 Bit: 8 7 6 4 5 2 3 User Definable VPI VCI VPI 2 3 VCI PT VCI HEC I CLP 4 5 Figure 3. ATM-Header Structure If delays occur because transmission channels in the ATM exchange are not available, the Multi-Q FIFO allocates cell priorities, known as quality of service (QOS), by interpreting the PT infonnation and the CLP bit in the cell header. Data that is critical as to the time taken for transmission, such as audio or video signals, is swept more rapidly through the ATM exchange than, for example,less critical computer data. If the CLP bit is set to 0, the cell contains important data that must reach its destination; whereas cells with the CLP bit set to 1 can be deleted. In a digital ATM exchange system, a priority control for cell transmission must be implemented. The cells being received arrive asynchronously to the clock signal of the exchange system; therefore, synchronization of the input data stream to the system clock is necessary. The Multi-Q FIFO solves synchronization problems and controls transmission priority with minimal complexity. The architecture of this FIFO, unlike conventional FIFOs, is not based on words of data but on cells. This device can control up to three priorities. The writing of the input data can be perfonned completely asynchronously with respect to the reading of the output data. The Multi-Q FIFO The most remarkable feature of the Multi-Q FIFO is that memory can be allocated to three independent queues. These queues allow the implementation of three QOS priorities. . Construction of the Multi-Q FIFO Figure 4 shows the functional block diagram of the Multi-Q FIFO, which is clocked; i.e., it has inputs for free-running write ~d read clocks. Write accesses occur at the rising edges of the write clock when one of the three write-enable-x, (WRTENx) (x - 1,2, or 3) lines is set. Read accesses are implemented at the rising edges of the read clock by setting the read-enable (RDEN) line. Reading or writing stops when a low level is applied to WRTENx or RDEN. For writing operations, the three control lines, WRTENx per queue, are individually brought out. The control lines for write accesses are operated by a multiplexer. The desired queue is chosen with MUXO and MUXI selecting access to the chosen queue using RDEN (see Table 1). Before use, this device must be reset by four rising edges of the write clock (WRTCLK) and four rising edges of the read clock (RDCLK) while the reset input (RST) is high. 3-103 Table 1. Selecting the Queue When Reading the FIFO MUX1 MUXO 0 0 SELECTED QUEUE Queue 1 0 1 Queue 1 1 0 Queue 2 1 1 Queue 3 POE DS RiW -BREQ P0-P7 DWRDY ... WRTCLK WRTEN1 WRTEN2 WRTEN3 -ABRT Auxiliary-Bus Control 8 + 8 I I ..... ..... ..... ...... FF2 PF3 FF3 --------------- ... Queue2 ... b r-"I :; Q, - .5 1 2 4096 x 18 Dual-Port SRAM In 16 256 x 18 Divisions 15 16 ,.... CR3 iH &01 ~ .. 4- &~I r-+- ~ ~ iii! Read-Address Generstlon ~ 'a:;! it; &0 ~ Figure 4. Multl·Q FIFO Functional Block Diagram 3-104 CR1 .... I Write-Address jleneratlon ....., T- MUXO MUX1 RDEN CR2 r-+- ~ RDCLK OSOC . Queue3 I +f-+__ f-+- Flags Queue1 DO-D17 RST T PF1 FF1 I Configuration Registers ISOC ..... - r - -ALER ""'""- PF2 Reset Logic - ~ OE QO-Q17 Configuration Registers Eleven configuration registers allow matching the FIFO to requirements of a particular application (see Table 2). These configuration registers can be written to and read from using a microcontroller through the auxiliary-bus control interface. Table 2. Configuration Registers REGISTER SYMBOL REGISTER NAME NO. OF DEFAULT PROGRAMMABLE BITS VALUE RANGE FUNCTION Port control 5 0 Bit-slice control Chooses the data input and output bus size and format. Controls output byte destufling. QU Queue 1 length 5 a 0-16 Defines the number of256 x 18 memory blocks for Queue 1 QL2 Queue 2 length 4 6 0-15 Defines the number of 256 x 18 memory blocks for Queue 2 QL3 Queue 3 length 4 2 0-15 Defines the number of 256 x 18 memory blocks for Queue 3 Cell size 6 27 10-32 Defines the cell size in 18-bit words PF1_W Programmable flag 1, write threshold 9 71 0-409 Defines the number of cells in Queue 1 to set PFllow PF1_R Progremmable flag 1, read threshold 9 70 1-408 Defines the number of cell~ in Queue 1 to set PFl high PF2_W Programmable flag 2, write threshold 9 51 0-383 Defines the number of cells in Queue 2 to set PF210w PF2_R Programmable flag 2, read threshold 9 50 1-382 Defines the number of cells in Queue 2 to set PF2 high PF3_W Programmable flag 3, write threshold a 13 1-383 Defines the number of cells in Queue 3 to set PF310w 8 12 0-382 Defines the n':lmber of cells in Queue 3 to set PF3 high PORT CLSZ PF3_R Programmable flag 3, read threshold Allocation of Queues The Multi-Q FIFO memory consists of 4096 IS-bit words that have a maximum of three independent queues. These queues can be called up to control up to three QOS priorities of ATM cells. Using configuration registers QL1, QL2, and QL3, the sizes of the individual queues can be allocated in steps of 256 IS-bit words. The initial value of QL1 = S if Queue 1 has a size ofS x 256 - 2048 IS-bit words. The development engineer has access only to configuration registers QLI and QL2 and can only determine the size of the fIrst two queues; after that, the Multi-Q FIFO automatically reserves the part of the memory that is still available for the third queue. Programming queue lengths of zero allocates the memory to one or two queues. The word width of the memory is IS bits; however, the development engineer can choose between 9-bit and I8-bit access when reading and writing. In these cases, the bus widths for reading and writing operations can be different. For example, it is possible to write with 9-bit access but implement the reading cycle with a word width of 18 bits. If the 9-bit access is chosen, the FIFO can write the fll'St 9-bit word to the lower signifIcant half of the IS-bit memory and the second 9-bit word to the higher signifIcant half (little endian). Alternatively, this order can be reversed (big endian). The programming for write accesses is performed in the configuration register PORT using bits INSIZ, OUTSIZ, and INBE (see Table 3). With read accesses, the 9-bit data word is output on bits Q8-QO in little-endian data format and on the bits Q17-Q9 in big-endian format. In this case, the hardware wiring determines the data format; whereas with the input data, the software programming determines the data format. 3-105 Table 3. Port-Control Register PORT OUTSTF OUTSIZ Blta INST BIt2 INBE Bit 1 INSIZ X X X 0 18-blt input bus X X 0 0 1 9-bit input bus with an even number of bytes per cell in little-endian data format X X 0 1 1 9-blt input bus with an even number of bytes per cell in big-endian data format X X 1 0 1 9-bit input bus with an odd number of bytes per cell In little-endian data format X X 1 1 1 X 0 X X X 18-bit output bus Bit 4 X FUNCTION Bit 0 9-bit input bus with an odd number of bytes per cell in big-endian data format 0 1 X X X 9-bit output bus with an even number of bytes percell 1 1 X X X 9-blt output bus with an odd number of bytes percell Cells Instead of Words of Data The Multi-Q FIFO flags (e.g., empty, full, etc.) indicate the presence or the absence of complete cells. The cell size can be set with the configuration register CLSZ in the range of 10 to 3218-bit words to allow a cell size of 20 to 64 bytes. The Multi-Q FIFO can also be programmed to odd cell sizes (e.g., S3 bytes) with 9-bit writing access by byte stuffing and with 9-bit reading access by removing the stuffmg bytes (see Figure S). This property can be chosen in the configuration register PORT with the help of bits INST and OUTSTF (see Table 3). 8-Blt Input Data Stream 53-Blt Cell 53-Bit Cell ~ 8-Blt Output Data Stream 18-Blt Data Stream In FIFO Multl·Q FIFO 53-Bit Cell 53-Blt Cell ~ Stuffing Bytes Figure 5. Data Stream With Odd Cell Size When writing into acell, the Multi-Q FIFO must be informed of the beginning of a cell with the input start-of-cell (ISOC) signal, as shown in Figure 6. At the rising clock-pulse edge when the first data word of a cell is written into the FIFO, both ISOC and the valid data word must be set high. If a cell has been written completely into the FIFO, ISOC must again be set with the beginning of the next cell. The FIFO compares the beginning of a cell, which has been indicated, with the expected cell beginning in accordance with the previously implemented programming of the cell size and indicates any fault at the alarm (ALER) output (see Figure 7 and Figure 8). If a fault of this kind occurs and ALER is low, the fault must be reset with the abort (ABRT) input signal before further cells can be written into the FIFO. When reading from cells, the output start-of-cell (OSOC) signal indicates the beginning of a cell. OSOC can be used to control subsequent parts of the circuit (see Figure 9). 3-106 " II ~' Figure 6. Writing Cells Into the FIFO Do-D18 (f ALER ABRT II If II I I I I II \. I I I I \..LI Figure 7. Faulty Writing of Cells Into the FIFO: ISOC Comes Too Soon Do-D18 I II { ,f I I I (f ALER ABRT II I ,tI \..lJ Figure 8. Faulty Writing of Cells Into the FIFO: ISOC Comes Too Late or Not at All 3-107 Figure 9. Reading Cells Out of the FIFO Flags Table 4 defmes the functions of flags that indicate the extent to which the memory is filled in the Multi-Q FIFO. A fonn of hysteresis is implemented with the programmable flags PFl, PF2, and PF3. The number of requir~ cells in Queue 1 to set PFI low is determined using the configuration register. At subsequent readout, PFI is reset to high as soon as the number of the cells still remaining in the memory reaches the value PF1_R in the configuration register. The extent to which the FIFO is filled can be set with configuration register PFl_W. From that point, ATM cells whose CLP bit is set to 1 are erased and no longer written into the FIFO. Only when the FIFO is again filled below the value in configuration register PFCR does an external cell-priority logic accept the writing in of cells whose CLP bit has a value of 1. The purpose of adjustable hysteresis is explained using as an example a standard FIFO having only one simply programmable almost-full (AF) flag without hysteresis. H the FIFO is filled to the predetermined value, the FIFO displays this at the AF flag output This process is repeated when the FIFO again exceeds the predetermined value and ignores CLP - 1 cells. As a result of the reading out of a cell, the AF flag is reset and the external cell-priority logic immediately allows the storage of CLP - 1 cells. At this point, the external cell-priority logic switches between acceptance and rejection of CLP - 1 cells. The implementation of hysteresis in the Multi-Q FIFO allows the user to suppress continuous switching between acceptance and rejection of CLP - 1 cells (see Figure 10). Hysteresis can be suppressed by an appropriate choice of threshold values for PFl_W and PFl_R. Table 4. Multi·Q FIFO Flags FLAG SYNCHRONIZED TO DWRDY WRTCLK Data write ready. DWRDY must be high before data can be written into the FIFO. FF1 WRTCLK Full flag, Queue 1. When FF1 is low, there is no more room for an additional cell in Queue 1. PF1 WRTCLK FF2 WRTCLK PF2 WRTCLK Programmable flag, Queue 1. Indicates the extent to which Queue 1 is occupied, as previously defined with configuration registers PF1 Wand PF1 R Full nag, Queue 2. When FF2 is low, there Is no more room for an additional cell in Queue 2. Programmable flag, Queue 2. Indicates the extent to which Queue 2 is occupied, as previously defined with configuration registers PF2 Wand PF2 R FF3 WRTCLK 3-'108 FUNCTION PF3 WRTCLK CR1 RDCLK Full flag, Queue 3. When FF3 is low, there is no more room for an additional.cell in Queue 3. Programmable nag, Queue 3. Indicates the extent to which Queue 3 is occupied, as previously defined with configuration registers PF3_W and PF3_R Cell ready, Queue 1. If there is at least a complete cell In Queue 1, CR1 is high. CR2 RDCLK Cell ready, Queue 2. If there is at least a complete cell in Queue 2, CR2 is high. CR3 RDCLK Cell ready, Queue 3. If there Is at least a complete cell in Queue 3, CR3 is high. Cell 74 Cell 70 Cell 65 Cell 60 t. oaf Cell 55 iii fIl'~ 11 ..... = PF1 W From this point, CLP 1 cells are no longer accepted. Cell 50 13a Cell 45 PF1_R From this point, CLP are again accepted. Cell 40 =1 cells Cell 35 ::I» :tt c8 !!o Cell 25 cZ Cell 20 Cell 30 Cell 15 Cell 10 Cell 5 Celli Figure 10. Hysteresis of the PF1 Flags With Configuration Registers PF1_W and PF1_R Programming The Multi-Q FIFO can be set up to meet the requirements of a particular application after resetting and before writing in the first word of data with the. configuration registers. These registers are written to and read from using a microcontroller via the auxiliary-bus control interface (see Figure 11). MC66302 Multl-Q FIFO A23-A0 AS 24 I Address Decode r-;1 I I I DTACK I....- DS D7-00 RiW 8 I 1 BREQ DWRDY DS P7-P0 RiW ..... POE Figure 11. Connection of a Mlcrocontroller to the Auxiliary Bus 3-109 Table 5. Configuration-Registers Access Order ACCESS ORDER REGISTER SYMBOL 1 PORT 2 REGISTER NAME PROGRAM BUS BIT WIDTH 5 MSB P4 " LSB Port control QL1" Queue 1 length 5 P4 PO 3 QL2 Queue 2 length 4 P3 PO 4 CLSZ P5 PO 5 Cell size Programmable flag 1, write threshold P7 PO PO 6 PFCW PF1 R Programmable flag 1, read threshold 6 9 9 P7 PO 7 PF2_W Programmable flag 2, write threshold 9 P7 PO 8 PF2 R Programmable flag 2, read threshold P7 PO 9 10 PF3 W PF3_1'1 Programmable flag 3, write threshold Programmable flag 3, read threshold 9 8 8 P7 P7 PO PO The writing into the configuration registers is perfonned sequentially (see Table 5). Access to register QL3 is unnecessary because the content of this register always consists of the memory size of the Multi-Q FIFO of 4096 words of data minus the values of registers QLl and QL2. To open access to the configuration registers, the bus request (BREQ) signal must be low. As a result, the data write ready (DWRDY) output replies with a low level after two rising edges of the write clock (WRTCLK). DWRDY indicates an active data access. When DWRDY is high, access to the FIFO is through the D inputs. When DWRDY is low, access is through the P tenninals to the configuration registers (see Figure 4). At every falling edge of the data strobe (DS) signal, the FIFO writes an 8-bit data word from the P tenninals in sequence to the configuration registers. If all ten configuration registers from Table 5 are filled with values, the FIFO ignores all further write accesses. Only after a renewed reset of the device are write accesses to the configuration registers again possible. The following rules apply for the values that are permitted to be written into the configuration registers. Rules for the length of the queues QLl, QL2, QL3 are: • The minimum value is O. • For QLl, the maximum value is 16. • For QL2 or QL3, the maximum value is 15. • The sum of QLl and QL2 must not exceed a value of 16; it can be less than 16. • Only QLl and QL2 can be programmed by the user. The value of QL3 is determined by the Multi-Q FIFO in that it is infonned of the length of the memory that is still available. Rules for the cell-size (CLSZ) register are: • The minimum value is 10. • The maximum value is 32. Rules for programmable flag values PFI_W, PF2_W, and PF3_Ware: • The minimum value is 1. • The value may not be larger than the number of whole cells for which there is room in the queue. • The PFCW, PF2_W, and PF3_W registers are nine bits. The higher-valued eight bits are programmable by the development engineer. The least significant bit (LSB) is always 1. Accordingly, all PFx_W values are odd numbers. Rules for programmable flag values PFCR, PF2_R, and PF3_R are: • The minimum value is 1. • The value must be smaller than the value of the corresponding PFx_W register. • The PFl_R, PF2_R. and PF3_R registers each consist of nine bits. The higher-valued eight bits are programmable by the development engineer. The LSB is always O. Accordingly, all PFx_R values are even numbers. 3-110 Extension of Word Width An extension of word width is possible with a 36-bit access. As shown in Figure 12 (36-bit access), all input control lines must be switched in parallel while the flag outputs are connected together with AND or OR gates. In theory, both FIFOs must have the same internal state and, accordingly, signal-identical flags; however, when there is unfavorable ovedapping, the flag of one device can change one clock cycle later than the other device. This does not cause differences in the contents of memory or loss of data. The flag synchronization can decide on a clock-pulse edge sooner or later, resulting in differences in the display. In this case, the connection with AND or OR gates ensures reliable results. If an IS-bit access is desired with an extension of word width, this can be achieved as shown in Figure 12. The only difference, in this case, is that both FIFOs are programmed for 9-bit access and only nine data lines per FIFO (DS-DO and Q8--QO) are used. 3-111 Multl-QFIFO RST RST WRTCLK WRTCLK WRTEN1 WRTEN1 ROEN ROEN WRTEN2 WRTEN2 MUXO MUXO WRTEN3 WRTEN3 MUX1 FF1 - PF2 - m - PF1 PF3 FF3 ISOC ALER ABRT - PF1 & ~ ~ ROCLK MUX1 T CR1 CR1 FF1 CR2 r====ID- CR2 CR3 ==ill- CR3 ISOC OSOC ALER OE rffi- PF2 = <1 FF2 .=. & PF3 ~ ----0= FF3 OSOC OE ABRT 035-00 017-00 36 ROCLK Q17-QO 18 18 36 Q35-QO Multl-Q FIFO RST '--- WRTCLK ROCLK ' - - - WRTEN1 ROEN WRTEN2 MUXO WRTEN3 MUX1 '--- PF1 CR1 FF1 CR2 PF2 CR3 - FF2 PF3 FF3 ISOC OSOC ALER OE ABRT 017-00 18 Q17-QO 18 Figure 12. Extension of Word Width With 18-Blt or 38-Bit Input and/or Output Data 3-112 FIF02 RST WRTCLK WRTCLK I-- WRTEN RST ISOC - ~ J ISOC ALER Q ABRT C ~ K R "7" QI- 0 - FIF01 Q RST C R WRTCLK --r& WRTEN ISOC - In-- ALER ABRT 08-00 o Flip-Flop I"- 08-D0 J.K"~"opr WRTEN ALER ABRT 08-D0 9 9 Figure 13. Extension of Word With 9-Blt Input Data If a 9-bit access to two Multi-Q FIFOs having extended word width is desired, these devices must be provided with external logic to control them in accordance with the ping-pong principle. InFigure 13, WRTEN and ISOC control lines demonstrate the ping-pong principle; i.e., the first 9-bit word is read into FIFO 1 and the second 9-bit word is read into FIF02. In this case, ISOC must also be generated for the second 9-bit data word, because this data word represents the beginning of a cell of FIF02. The order in which the 9-bit words are read into the two FIFOs is shown in Figure 14. 3-113 Ef !(~ Wl Wl W64 ECIl W53 W53 ef W53 liS; W52 W52 ~o W52 ~l W52 W51 S! ::len W2 W2 Wl j~ I·· ell W3 j~ W51 ~ :: Wl " ~l' "'~:.l W3 W2 W4 liS: W51 ~.~~4 h~ ~~ :: ::len W2 W64 W51 ;:I; ~j~ !(~ W3· Wl Wl ef W55 ::len !(~ W3 W2 Wl Wl " " -I1lli Wl W3 W2 W4 Wl W3 W2 W4 Wl W3 W2 W4 W5 W7 W6 W8 W5 Wl W6 W8 W5 W7 W6 W8 W5 Wl W6 W8 W9 Wll Wl0 W12 W9 Wll Wl0 W12 W9 Wll Wl0 W12 W9 Wll Wl0 W12 W49 W61 W50 W52 W49W51 W50 W52 W49 W51 W50 W52 W3 W2 W4 W53l1lll1tflW2 W4 . W53llllih9JIIW64 ill~iiillil Wl W3 W6 W8 Wl Wl W9 Wll Wl0 W12 W51Wl Wl0 W12 W51W7 W6 W8 W5 W7 FIFO 1 FIFO 2 FIFO 1 FIFO 2 FIFO 1 FIFO 2 FIFO 1 FIFO 2 Cell Size: 13 18-Blt Words Cell Size: 13 18-Blt Words Cell Size: Cell Size: 13 18-Blt Words Cell Size: Cell Size: Cell Size: Cell Size: 14 18-Blt Words W3 W2 W4 W53 W56 W64 W5 Wl W6 W8 Wl W3 14 14 14 14 18-Blt Words 18-Blt Words 18-Blt Words 18-Blt Words ~ =9-Blt Data Word, Third Word of ATM Data Stream • =Stuffing Byte NOTE A: Two Multi-Q FIFOs are connected as a 36-bit-wide FIFO with 9-bit data access. Figure 14. Data Flow of an ATM Data Stream In Two Multi·Q FIFOs 3-114 Programming Examples Before use, the Multi-Q FIFO must be reset and programmed to perform the desired function using the configuration registers (see Table 2). Table 6, Table 7, and Table 8 show examples of register programming. Table 6. Example of Configuration Registers Programming: 18-Bit Write, 18-Blt Read Function: REGISTER 53 bytes ~ 27 18-bit words 18bH 18bH 75 ATM cells ~ 2048 18·bit words 56 ATM cells ~ 1536 18·bit words 18 ATM cells ~ 512 18-bit words 65 ATM cells 55 ATM cells 50ATMcelis 40ATMcelis 15ATMcelis 10ATMcelis Cell size: Write access: Read access: Size of Queue 1: Size of Queue 2: Size of Queue 3: PF'_W: PF1_R: PF2_W: PF2_R: PF3_W: PF3_R: P7 P6 P5 P4 PORT 0 0 0 0 QL1 0 0 0 QL2 0 0 0 P3 P2 P1 0 0 0 0 1 0 0 0 0 1 1 PO HEX DESCRIPTION 00 PO = 0 ~ 18-bit input bus P3 = 0 ~ 18-bit output bus 0 08 8 x 256 0 06 6 x 256 53 cells ~ 27 18-bit words 0 = 2048 18-bit words = 1536 18-bit words CLSZ 0 0 0 1 1 0 1 1 18 PF1_W 0 1 0 0 0 0 0 1 41 65ATM cells PF1 R 0 0 1 1 0 1 1 1 37 55ATMcelis PF2 W 0 0 1 1 0 0 1 0 32 50ATMcelis PF2 R 0 0 1 0 1 0 0 0 28 40ATMcelis PF3 W 0 0 0 0 1 1 1 1 OF 15ATMcelis PF3 R 0 0 0 0 1 0 1 0 A 10ATMcelis Table 7. Example of Configuration Registers Programming: 9-Bit Write, 18-Blt Read Function: REGISTER 53 bytes ~ 27 18-bit words 9 bit, little endian 18bit 66 ATM celis ~ 1792 18-bit words 56 ATM celis ~ 1536 18-bit words 28 ATM celis ~ 768 18-bit words 60ATMcelis 50ATMcelis 50ATMcelis 40ATMcelis 24ATMcelis 16ATMcelis Cell size: Write access: Read access: Size of Queue 1: Size of Queue 2: Size of Queue 3: PF1 W: PFCR: PF2_W: PF2 R: PF3::::W: PF3 R: P7 P6 P5 P4 P3 P2 P1 PO HEX DESCRIPTION 1 00 PO = 1 ~ 9-bit input bus P1 = 0 ~ little endian P2 = 1 ~ odd-numbered celi size P3 = 0 ~ 18-bit output bus 1 1 07 7 x 256 1 0 06 6 x 256 53 cells ~ 27 18-bit words PORT 0 0 0 0 0 1 0 QL1 0 0 0 0 0 1 QL2 0 0 0 0 0 1 =1792 1a-bit words =1536 1a-bit words CLSZ 0 0 0 1 1 0 1 1 18 PFCW 0 0 1 1 1 1 0 0 3C 60ATM celis PF1 R 0 0 1 1 0 0 1 0 32 50ATM celis PF2_W 0 0 1 1 0 0 1 0 32 50ATMcelis PF2_R 0 0 1 0 1 0 0 0 28 40ATMcelis PF3 W 0 0 0 1 1 0 0 0 18 24ATMcelis PF3 R 0 0 0 1 0 0 0 0 10 16ATM cells 3-115 Table 8. Example of Configuration Registers Programming: 18-Blt Write, 9-Blt Read Function: Register Cell size: Write access: Read access: Size of Queue 1: Size of Queue 2: Size of Queue 3: PF1_W: PFCR: PF2_W: PF2_R: PF3_W: PF3_R: 54 bytes -7 27 18-bit words 18 bit 9 bit 56 ATM cells -7 153618-bit words 56 ATM cells -7 153618-bit words 37 ATM cells -7102418-bit words 50 ATM cells 40 ATM cells 50 ATM cells 40ATMcelis 30ATMcelis 20ATMcelis P7 P3 P6 P5 P4 P2 PI PO , HEX Description 0 1 0 0 0 00 PO =0 -7 18-bit Input bus P3 =1 -7 9-bit output bus P4 = 0 -7 even-numbered cell size 0 0 0 1 1 0 07 6 x 256 0 0 0 1 1 0 06 6 x 256 = 1536 18-bit words PORT 0 0 0 QU 0 0 Ql2 0 0 =1536 18-bit words CLSZ 0 0 0 1 1 0 1 1 1B 54 cells -7 27 18-bit words PF1 W 0 0 1 1 0 a 1 0 32 50ATMcelis PF1 R 0 0 1 0 1 0 0 0 28 40 ATM cells PF2_W 0 0 1 1 0 0 1 0 32 50 ATM cells PF2_R 0 0 1 0 1 0 0 0 28 40 ATM cells PF3_W 0 0 Q 1 1 1 1 0 1E 30 ATM cells PF3_R 0 0 0 1 0 1 0 0 14 20 ATM cells 3-116 Applications The Multi-Q FIFO provides several alternatives for arranging the priority control of various QOS classes. A common implementation is the priority control in the receiving unit (see Figure 15) and transmitting unit (see Figure 16) of an ATM exchange. If the content of the transmitted ATM cells in the receiving unit is larger than the capacity of the switching matrix, a priority control must be installed and cells of less importance put in a waiting queue or eliminated completely. The same phenomenon can arise with the transmitting unit when the capacity of the outgoing line cannot accept the cells received from the switching matrix. In both cases, use of a Multi-Q FIFO is recommended. Flaga J ATM Data Stream -+- PHY Priority Control 8 t Control Linea Multl-QFIFO 9 9,18,36 To Switching Matrix - PHY = Physlcallntarfaca Figure 15. ATM-Exchange Receiving Unit Flaga I From Switching Matrix Control Lines - Multl·Q FIFO Priority Control 9,18, 38 t 9,18, 38 PHY 8 r--+ ATM Data Stream ~ PHY =Physlcallntarfaca Figure 16. ATM-Exchange Transmitting Unit 3-117 The universal test and operations physical interface to ATM (UTOPIA) in 8-bit and 16-bit bus widths has become the preferred interface between the physical interface (pHY) and the subsequent or preceding stages. Figure 17 shows the connection of the Multi-Q FIFO on the receiving side to a PHY with a UTOPIA interface when one queue is used. When priority control of the ATM cells is implemented, an arrangement as shown in Figure 18 can be used. Similarly, the connection on the transmitting side to a PHY with a UTOPIA interface can be implemented as shown in Figure 19 and Figure 20. UTOPIA Multl-Q FIFO 07-00,015-00 8, 16 08,016-017 ~ 1,2 Q. CJI ATM OataStream .s:c: Start of Cell ISOC RxEmpty 'a:1 WRTEN1 FF WRTCLK PHY =Physlcallntarfaca Figure 17. . Conne~lon of a Multl-Q FIFO to a Receiving Unit Using an 8-Bit or 16-Bit UTOPIA Interface With One Queue UTOPIA 8,16 08,016-017 1,2 ISOC >- iECJI c: j WRTEN1 RxEmpty WRTEN2 Priority Logic WRTEN3 PF1 RxEnable PF2 PF3 WRTCLK PHY =Physical interfaca ' - - - - - - - - I Oscillator Figure 18. Priority-Controlled Connection of a Multl-Q FIFO to a Receiving Unit Using an 8-Blt or 16-Bit UTOPIA Interface 3-118 UTOPIA Multl·Q FIFO Q7-QO, Q15-Q01--..........----~------~----...... 8,16 Q8, Q16-Q171--.......... ---"""'1--------Pr----...... OSOC~----1,-2--------r_~s~m~rt~o~f~Ce~I~I--;_------~~ TxFull RDEN1 ~r------------+------------~--------~ ATM Dam Stream CR1 MUXO MUX1 RDCLK Oscillator 1-----------' PHY = Physical Interface Figure 19. Connection of a Multl·Q FIFO to a Transmitting Unit Using an 8-Blt or 16-Blt UTOPIA Interface With One Queue Multl·QFIFO Q7~,Q16-QOI------~~---~~-----~--~~ OSOCr-------------------~~~~~~--~--~~ RDEN MUXO MUX1 CR1 CR2 CR3 RDCLK ~.-~----~~----------~--~~~----~~ Oscillator 1-----------' PHY =Physical Interface Figure 20. Priority-Controlled Connection of a Multl-Q FIFO to a Transmitting Unit Using an 8-Blt or 16-Bit UTOPIA Interface 3-119 To Output 000-+ To Output 001 - + 11 000 12 001 13 010 14 011 15 100 16 101 17 110 18 111 Figure 21. Switching Matrix With Bottleneck Between Two Switching Elements There are different versions of the switching matrix. A simple example is shown in Figure 21. In this case, a bottleneck arises between the next-to-last and the last switching elements. This problem can be solved by increasing the transmission bandwidth of this part of the transmission path to double that of an input channel or by installing a priority control for the ATM cells to be transmitted. A Multi-Q FIFO is a suitable device for implementing this priority control. In view of the many ways in which an ATM exchange system can be implemented, there are certainly a large number of potential applications for the Multi-Q FIFO. When the priority control of up to three QOS classes is required, the Multi-Q FIFO is the logical choice. Summary The Multi-Q FIFO is designed to fulfill the particular requirements of ATM telecommunications exchange systems by: • Buffering ATM cells until they are passed on to the switching matrix • Matching asynchronous rates of data flow between a transmission line and the switching matrix • Managing up to three different priorities (QOS classes) of ATM cells • Matching the bus width (for example, from a 9-bit input bus to a 36-bit output bus, or vice versa) Programming the device by using ten configuration registers allows it to be used in a variety of applications. The 11 SN74ACT53861 Multi-Q FIFO is an outstanding component that fulfills the requirements of telecommunications applications. 3-120 Interfacing TI Clocked FIFOs With TI Floating-Paint Digital Signal Processors First-In, First-Out Technology ~1ExAs INSTRUMENTS SCAA005A 3-121 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas InstrumentS Incorporated 3-122 Contents TItle Page Introduction ............................................................................ 3-125 DSP Applications Using FIFOs ............................................................ 3-125 Communication Between a TI Bidirectional Clocked FIFO and a TI Floating-Point DSP ............ 3-127 DMA Considerations ................................................................... 3-130 Example Control of the SN74ACT3632 Using the DMA ....................................... 3-131 Programming the FIFO Almost-Full-Flag and Almost-Empty-Flag Offsets ........................ 3-135 Calculating FIFO Flag-Offset Values ....................................................... 3-135 Hardware Interface ...................................................................... 3-136 Read and Write Cycles .................................................................. 3-137 Interrupt Generation .................................................................... 3-138 Conclusion ............................................................................. 3-139 List ormustrations Figure ntle Page Clocked FIFOs Used for High-Speed Data Acquisition .................................... 3-126 2 Clocked FIFOs Used in Pipelined Image-Processing Systems ............................... 3-127 3 Bidirectional Clocked FIFO Used for Bus-Speed Matching ................................. 3-127 4 TMS320C31 DSP ................................................................. 3-128 5 SN74ACT3632512 x 36x 2 Bidirectional FIFO ......................................... 3-129 6 TMS320C31 Memory Map in Microprocessor Mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-130 7 Routine for Almost-Empty Flag CPU Interrupt to Schedule FIFO Reads ...................... 3-132 8 Scheduling FIFO Writes ..................................... .' ...................... 3-133 9 Routine for DMA Interrupt to CPU When Transfer Counter Reaches Zero ..................... 3-134 10 Almost-Full-Flag and A1most-Empty-Flag Offset Selection ................................ 3-136 11 TMS32OC31-40InterfacetoanSN74ACT3632-30FIFO .................................. 3-136 12 TMS32OC31-40 Read-Read-Write Timing Diagram With an SN74ACT3632 512 x 36 x 2 FIFO ... 3-137 13 TMS320C31 Interrupt Generation by FIF02 Almost-Empty Flag . . . • . . . . . . . . . . . . . . . . . . . . . . .. 3-138 14 TMS32OC31 Interrupt Generation by FIFO 1 Almost-Full Flag . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-138 3-123 3-124 Introduction Digital signal processors (DSPs) are used in a variety of applications to analyze real-time data or speed computationally intensive tasks. A DSP is a microprocessor tuned to the task of number crunching by an instruction set that conveniently ties together special hardware components needed for fast floating-point and fixed-point math and by powerful input/output (110) functions that keep data flowing quickly. Design of the 110 for a digital-signal-processing system is one of the major factors that dictates the machine's performance. First-in, flfSt-out (FIFO) memories often are used as data rate buffers to optimize the throughput of digital-signal-processing systems and increase overall performance. A FIFO is a dual-port memory with built-in write and read addressing to pass out data in the same order it is written. Data reads and writes can be done asynchronous to one another. Flag circuitry indicates when the queue is empty or full, preventing simultaneous read/write access to the same memory location. Advanced FIFO memories from Texas Instruments (11) produced in CMOS or BiCMOS technology also have user-programmable almost-empty and almost-full flags to measure the number of words in memory. FIFOs provide a seamless bridge between two buses operating at different clock speeds and acting as temporary data bins to exchange information between two systems without handshaking delay. 11's TMS32OC3x and TMS320C4x processors are popular DSPs that include a 4O-/32-bit floating-Iflxed-point math unit, one or two 32-bit external buses, and an on-board direct-memory-access (DMA) controller. Unidirectional and bidirectional clocked FIFO devices from 11 frequently are used to support systems built around these processors. Attractive features offered by 11 clocked FIFOs are synchronous (clocked) interface on each port, asynchronous I/O capability, programmable flags, maximum write/read frequencies up to 80 MHz, maximum read access times as low as 9 ns, and fine-pitch surface-mount packaging. DSP Applications Using FIFOs DSP systems doing real-time data analysis or control functions use analog-to-digital (AID) converters to translate continuous-time, real-valued signals into discrete-time, integer-valued sequences. The rate used to sample the analog signal is chosen based on the frequency bandwidth of the signal. This sample rate is independent of the microprocessor-bus rate, and asynchronous buffering is required to pass the information to the DSP. Serial ports on the TMS32OC3x1C4x processors provide an asynchronous interface with AID converters and are adequate when the incoming data traffic has a relatively low bit rate. For higher bit rates, unidirectional clocked FIFOs provide a parallel buffer between the converters and the DSP bus. Figure 1 shows several digitized signals, each using a FIFO for rate buffering to the processor bus. An example of this application is multiplexing several analog telephone lines for compression or symbol detection. An input signal packet is gathered in the FIFO and burst into memory by the DMA unit on the TMS32OC3x/C4x. This method also is useful when the analog data is sampled at a high rate for short duration, as in some medical-imaging equipment. Each FIFO holds its AID samples in queue until the processor retrieves the information that must be completed before the next sampling period. The block labeled FIFO Enable can have a single-memory-space address and control the FIFOs in round-robin fashion as the DMA fills the random-access memory (RAM) with digitized signals. 3-125 FIFO Enable -J EOC --+ Data AID r--- 1WEN OEIRDEN f-- I-- H3 D Clocked FIFO Q ~ - c Address, STRB, W f c( ~ 'iii i 'V --+ AID 112;<18 II. (/) 12-18 ~ RCLK WCLK Clocked FIFO I & I j::: - DO-D31 Q •• • ••• c c( --+ AID 112;<18 Clocked FIFO Q D A RAM Figure 1. Clocked FIFOs Used for High-Speed Data Acquisition DSP algorithms are drawn as functional boxes with interconnecting lines representing data streams. This concept can translate directly to a hardware organization as shown in the pipelined image-processing system (see Figure 2), wherein the unidirectional TI clocked FIFOs provide the data connection between the TI floating-point DSPs. The FIFO that connects the bus to the first processing element (PE) performs the task of rate matching, as the bus generally operates at a slower rate than the DSP bus. FIFOs that connect adjacent PEs are used as packet builders; that is, a packet of data is stored and then detected with the use of the almost-full/almost-empty or half-full flags and read by the next processor. Transferring a known packet size simplifies DMA control. The FIFO interconnect between PEs eliminates the need for processor interlock protocols and reduces clock-distribution requirements by allowing each PE to utilize its own independent clock. 3-126 VME, SBus, etc. '--C-I-oC-ked--' r------------------, I I I PE1 FIFO - -.... Addr o --+ Q I-i--+---+_ .....----' I I I I I I I Addr Clocked FIFO TI OSP Data 0 --+ PE2 Q I I ~------------------~ .. o Clocked FIFO Q Output Figure 2. Clocked FIFOs Used In Plpellned Image-Processing Systems Software applications often are written for a general-purpose workstation platfonn, incorporating signal-processing functions that are not efficiently perfonned by the workstation. A solution to this problem is to use a special-purpose DSP system as a coprocessor for the application and communicate with the host computer via a local or backplane bus (see Figure 3). The bidirectional clocked FIFO is most useful when data traffic is heavy both to and from the host computer, such as when the host provides the input data and receives the processed results. Bidirectional FIFOs also can be used as instruction queues between a host processor and the DSP. The FIFO in the datapath provides clock partitioning so each bus can operate at its maximum rate; it also eliminates transfer delay required for a bus request to be granted to either the host or the DSP. VME, SBus, etc. ... ["'.,.... Clock-0ff-set-----"'l"'O~1 No. of Words In FIF02 S Offset -t\ INT1t __________________ t-- tpd ----~~=~~~S1~ -tI t-- tpd ~~:2~2~?~------- t Low level on INT1 Indicates an available FIF02 packet space. Figure 13. TMS320C31 Interrupt Generation by FIF02 Almost-Empty Flag H3 H1 I4-tI tpd(C-AF) I4-tt tpd(C-AF) _ _-"t""Z",,,..~-r:N:":'0';'.-of~E~m""pty~":::s""pa"ces~':"ln-:F:::IF::O::":1:->~Off==set"""!"'--"11:\::~i No. of Empty Spaces In FIF01 S Offset t-- INTot -t\ tpd --~\~~~~~~ __________________ t-- -tI tpd ~e:2~2~?~------- t Low level on INTO indicates an available FIF01 packet space. Figure 14. TMS320C31 Interrupt Generation by FIF01 Almost-Full Flag 3-138 Conclusion FIFO memories are used in DSP systems for matching two datapaths with asynchronous clock or data rates. The SN74ACT3632 512 x 36 x 2 clocked FIFO provides a single-chip bidirectional buffering solution that interfaces nicely with TI floating-point DSPs. Programmable FIFO flags enable a variety of DMA control techniques to be used in handling data flow, and the FIFO control signals are easily derived from TMS32OC3x outputs. Available in a variety of speed options, the SN74ACT3632 can interface a DSP to buses operating up to 67 MHz. 3-139 3-140 FIFOs With a Word Width of One Bit First-In, First-Out Technology Peter Forstner Mixed Signal Logic Products ~TEXAS INSlRUMENTS SCAAOO6A 3-141 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to Its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable althe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Appllcations'1. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product deSign, software performance, or infringement of patents or services descrlbed herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or procElss in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-142 Contents Title Page Introduction ............................................................................ 3-145 FIFO Basics ............................................................................ 3-145 Telecommunications . ..................................................................... 3-147 Digital-Transmission Methods ............................................................ 3-148 A PDH Application Example ............................................................. 3-149 Frame Structure of the First Hierarchy Level .............................................. 3-149 Frame Structures of the Second to Fifth Hierarchy Levels .................................... 3-150 Clock Adjustment With FlFOs .......................................................... 3-154 Modems With Data Compression . .......................................................... 3-156 Signal-Processor Interfaces . ............................................................... 3-158 Teletext Decoders ........................................................................ 3-159 Summary ............................................................................... 3-160 3-143 List of Illustrations Figure I Title Page First-In, First-Out Data Flow ......................................................... 3-145 2 Connections of a Strobed FIFO . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-146 3 Connections of a Clocked FIFO ...................................................... 3-146 4 Time-Division Multiplex of Several Channels ........................................... 3-147 5 Logarithmic 13-Segment Characteristics for the Coding of Telephone Signals .................. 3-148 6 Frame Str\Jcture of the 2048-kbitls Multiplex Signal (First Hierarchy Level) ..•................ 3-150 7 Frame Structure of the 8448-kbit/s Multiplex Signal (Second Hierarchy Level) ................. 3-151 8 Frame Structure of the 34368-kbit/s Multiplex Signal (Third Hierarchy Level) ................. 3-151 9 Frame Structure of the 139264-kbit/s Multiplex Signal (Fourth Hierarchy Level) ............... 3-151 10 Frame Structure of the 564992-kbit/s Multiplex Signal (Fifth Hierarchy Level) . . . . . . . . . . . . . . . .. 3-152 11 Clock Adjustment at the Transmitting End With Positive Pulse-Stuffing Techniques Block Diagram ................................... 3-154 12 Bit Stream at the FIFO of the Transmitter-Clock Adjustment ............................... 3-154 13 Clock Adjustment at the Receiving End With Positive Pulse-Stuffing Techniques Block Diagram ................................... 3-155 14 Bit Stream at the FIFO of the Receiver-Clock Adjustment .................................. 3-155 15 Clock Signals at the Receiver ........................................................ 3-156 16 Data Transmission by Modem With Data Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-157 17 Serial-Port Data-Transmission Protocol of a TMS320CXX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-158 18 Serial-Port Data-Transmission Protocol With Inverted Signals .............................. 3-158 19 Connection of 1\vo Signal Processors via a Serial Port With the Help of the FIFO SN74ACT2229 .............................................. 3-159 20 Video Signal ..................................................................... 3-159 21 Teletext Decoder With 1 x 512 FIFO Block Diagram ....•................................ 3-160 22 Extending MemOry Depth of a SN74ACT2229 FIFO to 512 x 1 Bit ....... , ...... , ........... 3-160 List of Tables Table Title Page 1 Plesiochronous Digital Hierarchies .. • . . .. . . .. .. .. .. .. . .. .. . . . . .. .. . . . . . .. .. . .. . . . . . . .. 3-149 2 3 Synchronous Digital Hierarchy and SONET ............................................ 3-149 4 Spread of the Transmission Capacity of a 34368-kbit/s Signal Consisting of Four Multiplexed 8448-kbitls Signals With a Net Nominal Transmission Speed of 4 x 8448 kbit/s - 33792 kbit/s .... 3-153 5 Spread of the Transmission Capacity of a 139264 kbit/s Signal Consisting of Four Multiplexed 34368-kbit/s Signals With a Net Nominal Transmission Speed of 4 x 34368 kbit/s - 137472 kbit/s ..................................................... 3-153 6 Spread of the Transmission Capacity of a 564992-kbit/s Signal Consisting of Four Multiplexed . 139264-kbit/s Signals With a Net Nominal Transmission Speed of 4 x 139264 kbit/s - 557056 kbit/s .................................................... 3-153 7 One-BitFIFOs From TI .............•............................................... 3-156 3-144 Spread of the Transmission Capacity of a 8448-kbit/s Signal ConSisting of Four Multiplexed 2048-kbit/s Signals With a Net Nominal Transmission Speed of 4 x 2048 kbit/s - 8192 kbit/s ..... 3-153 Introduction In every digital system, data is continually being exchanged between various subsystems. Intennediate storage is always necessary if data arrives at the receiving subsystem at a high rate or in batches but can then only be processed slowly or irregularly. Such intermediate stores are familiar to us in our daily lives, for example, as queues of customers at the checkout of a supennarket or cars waiting at traffic lights. The checkout of goods. at the supennarket point of sale proceeds slowly and regularly, while customers arrive there unpredictably; if many customers all want to pay at the same time, a queue builds up that the cashier processes on the principle of first come, first served. Queues of cars at traffic lights result from the sporadic arrival of traffic, which the lights allow to proceed only in regular batches. An intennediate store or memory that operates on the above principle is known as a fIrst-in, first-out (FIFO) memory. The fIrst data written into a FIFO is also the first to leave it at readout. Texas Instruments (TI) offers a variety ofFIFOs. These are available with word widths from 1 bit to 36 bits, storage densities from 64 to 2048 words of data, and clock speeds of up to 80 MHz. This application report is concerned exclusively with FIFOs having a word width of one bit and it suggests various possible applications for them. Whenever a buffer memory is needed for serial-data transmission, there is a requirement for I-bit-wide FIFOs. Digital telecommunications, local-area networks (LANs), serial transmission of data with the help of data compression, and communication between signal processors are all examples of serial data-transfer applications that require I-bit-wide FIFOs. In some applications, the FIFOs are already integrated into the application-specifIc integrated circuit (ASIC) or the chip set with LANs. However, very often discrete FIFO components are required. FIFO Basics Every memory component for which the fIrst word of data written to the memory is also the flfSt to leave it when the memory is read out can be classified as a FIFO (see Figure 1). In practice, a further characteristic often required from a FIFO is asynchronism between the writing and reading processes. This kind of FIFO is known as an asynchronous FIFO. • •• Figure 1. First-In, First-Out Data Flow 3-145 A FIFO has an input to which data words are written and a separate output from which data words are read. Since writing can take place completely asynchronously to reading, it is permissible for the writing and reading pulses to have completely different clock speeds, chosen at will. Control signals such as full, empty, half full, and almost full allow the controlling circuitry to monitor the internal state of the FIFO before every writing or reading process. According to the control signals to write and read, asynchronous FIFOs can be classified into two groups; strobed FIFOs (see Figure 2) and clocked FIFOs (see Figure 3). WRTCLK RDCLK Strobed FIFO Input Data Output Data Figure 2. Connections of a Strobed FIFO Free-Running WRTCLK Free-Running RDCLK WRTEN IR Clocked FIFO Input Data RDEN OR Output Data Figure 3. Connections of a Clocked FIFO The strobed FIFO enters a word of data into its internal memory at every rising (or every falling) edge of the write clock (WRTCLK). FULL shows whether there is room in the memory for a data word. Reading adata word takes place at every rising (or falling) edge of the read clock (RDCLK). If there is no data word awaiting readout, this is indicated by the status signal EMPTY. The disadvantage of this kind of FIFO is that the status signals cannot be fully synchronized with the corresponding clock signals. Clocked FIFOs require a free-running write clock (WRTCLK) and read clock (RDCLK). The writing and reading processes are controlled by the control signals write enable (WRTEN) and read enable (RDEN). The status signals input ready (IR) and output ready (OR) indicate the internal state of the FIFO. As a result of the two free-running clock signals, all status signals can be synchronized within the FIFO. The IR signal changes its level exclusively in synchronism with the writing pulse, while OR switches synchronously with the reading pulse. The I-bit FIFOs in this application report (SN74ACT2226, SN7 4ACT2227, SN74ACT2228, and SN74ACT2229) are, without exception, clocked FIFOs with complete built-in synchronization of all available status signals, including: Output ready (OR) synchronized with read clock (RDCLK) Input ready (IR), half full (HF), and almost fulValmost empty (APIAE) synchronized with write clock (WRTCLK) 3-146 Telecommunications The rapidly increasing need for telecommunication installations cannot, in the long run, be met by providing a separate line for every telephone connection; the simultaneous use of one line for several channels is a requirement. Digital transmission via pulse-code modulation (PCM) techniques enables the cost-effective use of single lines for multichannel transmission. Using these techniques, digitized telephone signals are switched successively onto a connecting line with the help of a multiplexer and separated from one another at the end of the line with a demultiplexer (see Figure 4). 81 83 1841 1 84 1 1 1 1 1 f4 ta t2 t1 81 1 1 1 1 1.1 1 1 1 1 1 1 1 1 1 f4 ta t2 t1 83 1841 1 84 1 1 1 1 1 f4 t3 t2 t1 Figure 4. Time-Division Multiplex of Several Channels With the 3.4-kHz upper bandwidth limit of a telephone channel and the internationally standardized sampling frequency for digitizing the signal (fo = 8 kHz), there remains enough space in the frequency band to insert the edge of the necessary bandwidth-limiting low-pass filter. Although extensive tests of syllable intelligibility have shown that 7-bit quantization with 128 quantization intervals is adequate even with successive analog-to-digital-to-analog conversion, an 8-bit quantization with 256 intervals has been made the standard. For the compression of the instantaneous value of the signal, the logarithmic 13-segment characteristic shown in Figure 5 is used. For the transmission of a channel, a bit rate of 8 kHz x 8 bit = 64 kbit/s is necessary and, correspondingly, a line for 32 multiplexed channels must attain a transmission rate of64 kbit/s x 32 = 2048 kbitls (CCITIrecommenc4ttions G.732 and G.704). 3-147 Input Level [dB] -20 -15 -10 If -5 -4-3 -2 -1 0 2 • +3 01111111 01110000 01100000 01010000 01000000 00110000 00100000 1 718 6/8 5/8 00000000 0 418 3/8 218 11100000 11010000 11000000 10110000 10100000 10010000 10000000 -218 -3/8 -418 -5/8 -618 -7/8 -1 1 -1 • Coder Input 0 Figure 5. Logarithmic 13-Segment Characteristics for the Coding of Telephone Signals Digital-Transmission Methods At present, four different digital-transmission methods are used for telecommunications: European plesiochronous digital hierarchy (pOH, see Table 1) American plesiochronous digital hierarchy (POH, see Table 1) Japilnese plesiochronous digital hierarchy (pOH, see Table 1) Synchronous digital hierarchy (SOH, see Table 2) Signals coming from various clock generators should have the same bit speeds but, in practice, the bit speed may deviate by a certain tolerance from the nominal value. These signals are referred to as plesiochronous signals. The lack of worldwide standardization of the three POH transmission methods makes world networking much more difficult, and the use of equipment from various manufacturers is limited to the networks of individual national telecommunications organizations. The fact that the standard for synchronous digital hierarchy (SOH, see Table 2) has worldwide validity does, however, offer the promise of assistance. SOH evolved from the North American synchronous optical network (SONET) specifications but is based, (as described in the CCITTrecommendations 0.707, 0708, and 0709) on a bit rate of 155520 kbit/s (see Table 2); that is, exactly three times the SONET basic bit rate of 51840 kbitls. The SOH basic signal is designated as synchronous transport module level one (STM-I); higher hierarchy levels are whole integer multiples of the level-one bit rate. 3-148 Table 1. Pleslochronous DIgItal HIerarchIes HIERARCHIES BASED ON 2 Mbltls HIERARCHY LEVEL HIERARCHIES BASED ON 1.5 Mbltls EUROPE, SOUTH AMERICA USA JAPAN 1 2048 kbills 1544 kbit/s 1544 kbit/s 2 8448 kbills 6312 kbit/s 6312 kbit/s 3 34368 kbit/s 44736 kbit/s 4 139264 kbit/s 32064 kbills 97728 kbit/s Table 2. Synchronous DIgItal HIerarchy and SONET SONET SOH BIT RATE LEVEL SIGNAL IDENTIFICATION 1 STM-1 51840 kbit/s 155520kbit/s 466560 kbit/s LEVEL SIGNAL IDENTIFICATION ST5-1 OC-1 ST5-3 0C-3 ST5-9 00-9 ST5-12 00-12 933120 kbit/s ST5-18 00-18 1244160 kbit/s ST5-24 00-24 1866240 kbit/s ST5-36 0C-36 STS-48 OC-48 622080 kbit/s 2488320 kbills 4 16 STM-4 STM-16 A PDH ApplicatIon Example The plesiochronous digital hierarchy and the application of FIFOs for the synchronization of the PDH signals are demonstrated using as an example European transmissions based on a bit speed of 2048 kbit/s. Frame Structure of the First Hierarchy Level The bit speed of the first hierarchy level (2048 kbit/s, see Table 1) allows the transmission of 32 telephone channels, each of 64 kbitls, over a normal telephone line. In this case, only 30 telephone conversations are transmitted, since two channels are required for the following additional information (see Figure 6): Frame recognition word for the synchronization of the receiver Cyclic-redundancy-check (CRC4) bits for the recognition of bit faults during the transmission Service bits for initiating alarms Registration bits for national and international telecommunication traffic Telephone exchange technical identification (signalization) Each of the eight bits of the 32 channels is multiplexed bit by bit; that is, bit 0 of the 32 channels is first sent serially over the line followed by 32 times bit 1, etc. These 8 x 32 bits - 256 bits are consolidated in a frame (see Figure 6). Channels 0 and 16 contain the necessary control information, while the remaining channels can be used for the transmission of 30 telephone connections. The transmission of a 256-bit frame of this kind at 2048 kbit/s requires a time period of 125 /LS. Sixteen frames together make up a 16 x 256 - 4096-bit multiple frame with a transmission time period of 2 ms. The 256 control bits in channels 0 and 16 can be seen in Figure 6. 3-149 --- ---- --- --- FRAME BITS FRAME NO. 1 2 3 4 5 6 7 0 Cl 0 0 1 1 0 1 ---- - FRAME BITS 8 FRAME NO. 1 2 3 4 5 6 7 1 0 0 0 0 0 X Y X °1 02 dl a17 b17 °17 d17 dz a18 b18 °18 d18 1 0 1 0 N Sn Sn Sn Sn 1 al bl 2 C2 0 0 1 1 0 1 1 2 a2 bz 8 X 3 0 1 0 N Sn Sn Sn Sn 3 as bs 03 d3 a19 b19 °19 d19 4 C3 0 0 1 1 0 1 1 4 84 b4 °4 d4 a20 bzo 020 d20 b5 °5 d5 a21 b21 021 d21 be OS d6 a22 bz2 022 dzz °7 08 d] a23 b23 023 d23 d8 a24 bz4 024 d24 5 0 1 0 N Sn Sn 5 C4 0 0 1 Sn 1 Sn 6 0 1 1 6 as as 7 0 1 0 N Sn Sn Sn Sn 7 a7 b] 8 Cl 0 0 1 1 0 1 1 8 a8 b8 9 0 1 0 N Sn Sn Sn Sn 9 ag bg Cg d9 a25 bz5 025 d25 10 C2 0 0 1 1 0 1 1 10 al0 bl0 010 dl0 a26 bz6 026 d26 11 0 1 0 N Sn Sn Sn Sn 11 all bll °11 dll a27 bz7 027 d27 12 C3 0 0 1 1 0 1 1 12 a12 b12 012 d12 a28 bz8 028 d28 13 0 1 0 N Sn Sn Sn Sn 13 a13 b13 013 d13 a29 bz9 029 dz9 14 C4 0 0 1 1 0 1 1 14 a14 b14 c14 d14 aso b30 030 d30 15 0 1 0 N Sn Sn Sn Sn 15 a15 b15 c15 d15 asl b31 c31 d31 0011011 C4··· C1 D,N SI,Sn Frame Recognition Word CRC4 Control Bits Bits for Alarm Initiation Registration Bits OOOOXYXX an ••• d n Multiple Frame Recognition Word Signalization for Channel n Figure 6. Frame Structure of the 2048-kblt/s Multiplex Signal (First Hierarchy Level) Frame Structures of the Second to Fifth Hierarchy Levels For further sections, four 2048-kbit/s signals are transmitted with successive bit-by-bit time-division multiplexing combined with the pulse-stuffing procedure at bit speeds of 8448 kbit/s, 34368 kbit/s, 139264 kbit/s, and 564992 kbit/s. If several plesiochronous signals are multiplexed, they must be synchronized before the multiplexing process. Plesiochronous signals have nominally the same bit speeds; but, in practice, the following kinds of asynchronism can arise: The bit rates deviate from one another within the specified tolerance (drift). As a result of long transmission distances and significant differences of temperature, etc., the bit speeds fluctuate for short periods (jitter). For the synchronization of many plesiochronous 2048 kbit/s signals, positive pulse-stuffing techniques are used when multiplexing these signals into an 8-mbit/s signal. The principle of this technique is based on the fact that, in the multiplexed signal, a bandwidth is made available that is wider than the nominal bit rate requires. If at particular points in the transmission information bits or empty bits (so-called stuffing bits) are sent out, the bit speed can be reduced and thus adjusted to suit the input signal. This technique also compensates for driftand jitter of the input signal. 3-150 4 x ~I 212 Bit = 848 Bit I Block II Block I I Block III NI RKW: S: NI: St: B: Block IV NI Frame-Recognition Word SeNlceBlts Useful Information Bits Stuffing Information Bits Stuffing or Useful Bits Bit Position per Block Figure 7. Frame Structure of the 8448-kblt/s Multiplex Signal (Second Hierarchy Level) '4~------------1 4 X 348 Block II Block I BI~= 1536 Bit I - - - - - - - - - - - - -.. ~ Block III Block IV 1 5 NI RKW: S: NI: St: B: NI Frame-Recognition Word SeNlceBlts Useful Information Bits Stuffing Information Bits Stuffing or Useful Bits Bit Position per Block Figure 8. Frame Structure of the 34368-kbit/s Multiplex Signal (Third Hierarchy Level) 1'4'<11-------------1 Block I Block II 4 X 488 BI~= 2928 Bit - -.........- - - - - - - - - -....... ~ Block III I RKW: S: NI: St: B: Block IV Block V Frame-Recognition Word Service Bits Useful Information Bits Stuffing Information Bits Stuffing or Useful Bits Block VI 1 Bit Position per Block Figure 9. Frame Structure of the 139264-kblt/s Multiplex Signal (Fourth Hierarchy Level) 3-151 ... ~------------ I Block I Block II Block III 71384BIt=2688~1t I Block IV I - - - - - - - - - - - -...... ~ Block V Block VI Block VII I RKW: Fram.Recognltlon Word s: HI: St: B: Service Blta Useful Information Blta Stumng Infonnatlon Blta Stumng or Useful Blta Bit Position per Block Figure 10. Frame Structure of the 564992·kblt/s Multiplex Signal (Fifth Hierarchy Level) When multiplexing with positive pulse-stuffing techniques, a frame that is constructed with a 8448-kbitls signal is partitioned into four blocks (see Figure 7). This frame structure envisages four stuffmg bits in block N in bit positions 5 to 8. These stuffing bits can either contain useful information or they can be empty bits. The stuffing information in bit positions 1 to 4 in blocks II, ill, and N indicates whether empty bits or useful bits are present in block N. This 4-bit stuffmg information is transmitted three times (blocks II, ill, and N) to assure a correct decision about the information content of the stuffing bits in the case of bit faults within the stuffing information. If there is a conflict between the individual bits of the three transmissions of stuffing bits, a majority decision can be used to avoid a false conclusion that would result in a bit slip and, consequently, a loss of synchronization of the 2-mbit/s systems. If, for example, one of the bit combinations 0-0-0, 0-0-1,0-1-0, or 1-0-0 is received as stuffing information in blocks II, ill, and N for the ftrSt stuffmg bit, a useful bit follows at bit position 5 in block N; the reception of 1-1-1, 1-1-0, 1-0-1, or 0-1-1 indicates an empty bit. A 2-bit fault in the stuffmg information results in the loss or gain of a bit (bit slip) and, consequently, in loss of the frame synchronism of the multiplexed signals. Techniques similar to multiplexing with positive pulse stuffing with the 8448-kbit/s signal are also performed with the 34368-kbit/s, 139264-kbit/s, and 564992-kbit/s signals (see Figures 8, 9, and 10). As a result of these stuffing techniques, a 8448-kbit/s frame has a transmission speed in the range of 8169-kbitls to 8209-kbit/s useful bits. With the nominal transmission speed for four multiplexed 2048-kbit/s signals of 8192-kbit/s, fluctuations in the transmission speed in the range of about ±0.2% can be compensated for (see Tables 3, 4,5, and 6). 3-152 Table 3. Spread of the Transmission Capacity of an 8448-kblt/s Signal Consisting of Four Multiplexed 2048-kblt/s Signals With a Net Nominal Transmission Speed of 4 x 2048 kblt/s 8192 kblt/s = USEFUL BITS FRAME oSTUFFING BITS Frame Capacity Transmission Speed 2 STUFFING BITS 4 STUFFING BITS 848 bits 820 bits 822 bits 824 bHs 8448 kbit/s 8169 kbit/s 8189 kbH/s 8209 kbit/s 8192 kbit/s 8192kbit/s 8192 kbit/s -0.28% -0.04% +0.21% Nominal Value Deviation from Nominal Value Table 4. Spread of the Transmission Capacity of a 34368-kblt/s Signal Consisting of Four Multiplexed 8448-kblt/s Signals With a Net Nominal Transmission Speed of 4 x 8448 kblt/s 33792 kblt/s = USEFUL BITS FRAME oSTUFFING BITS Frame Capacity Transmission Speed 2 STUFFING BITS 4 STUFFING BITS 1536 bits 1508 bits 1510 bits 1512 bits 34368 kbH/s 33742 kbit/s 33786 kbH/s 33831 kbH/s 33792 kbit/s 33792 kbH/s 33792 kbH/s -0.15% -0.02% +0.12% Nominal Value Deviation from Nominal Value Table 5. Spread of the Transmission Capacity of a 139264·kblt/s Signal Consisting of Four Multiplexed 34368-kblt/s Signals With a Net Nominal Transmission Speed of , 4 x 34368 kbit/s = 137472 kblt/s USEFUL BITS FRAME oSTUFFING BITS Frame Capacity Transmission Speed 2 STUFFING BITS 4 STUFFING Bits 2928 bits 2888 bits 2890 bits 2892 bits 139264 kbit/s 137361 kbit/s 137457 kbit/s 137552 kbit/s 137472 kbH/s 137472 kbH/s 137472 kbit/s -0.08% -0.01% +0.06% Nominal Value Deviation from Nominal Value Table 6. Spread of the Transmission Capacity of a 564992·kblt/s Signal Consisting of Four Multiplexed 139264-kblt/s Signals With a Net Nominal Transmission Speed of 4 x 139264 kblt/s 557056 kbit/s = USEFUL BITS FRAME Frame Capacity Transmission Speed Nominal Value Deviation from Nominal Value o STUFFING BITS 2 STUFFING BITS 4 STUFFING BITS 2688 bits 2648 bits 2650 bits 2652 bits 584992 kblt/s 556584 kblt/s 557005 kbit/s 557425 kbit/s 557056 kbit/s 557056 kbit/s 557056 kbit/s -0.08% -0.01% +0.07% 3-153 Clock Adjustment With FIFOs Clock Adjustment at the Transmitting End A block diagram showing the principle of clock adjustment at the transmitting end with positive pulse-stuffmg teclmiques is shown in Figure 11. In this case, each channel is provided with an elastic memory in the form of a FIFO. RDCLK RDEN WRCLK FIFO Input Slgnal---4._------...., 1D 2048kblt/a Control Logic HF HF I----+l~ 3 Mora Telephone Channela M With 2048 kblta/a Each, -:--+-1 U X Frame Recognition Word,. Service Blta, • Stuffing Information Blta, • Stuffing or Useful Blta - - + I Output Signal 8448kblt/a Figure 11. Clock Adjustment at the Transmitting End With Positive Pulse-Stuffing Techniques Block Diagram The input data is written into this FIFO with the help of a circuit for clock recovery. The FIFO takes on the buffering of the input data while the frame and stuffmg information is being transmitted. If information bits are to be transmitted. the control logic of the transmission path extracts the data from the FIFO. With positive pulse-stuffmg teclmiques, the net bit speed of the transmission path is slightly higher than the bit speed of the incoming signal. As a result, the transmission-path controller reads the data from the FIFO more quickly than it can deliver it to the input channel. Whenever the FIFO contains less than a certain minimum filled level (e.g., half full), the transmission path sends at the next possible moment a stuffmg bit instead of a data bit. As a result, the input channel has enough time to raise the filled level of the FIFO above the specified minimum levei by writing in further data (see Figure 12). NI: Uaefullnformatlon blta St: Stuffing Information blta \~~ I Frame I NI rJ.¢I' ~to NI NI NI Time Figure 12. Bit Stream at the FIFO of the Transmitter-Clock Adjustment If the minimum level of the FIFO when sending block II (see Figure 7) is not reached, the stuffmg information in block II can no longer be changed. Accordingly, a wait must be made until the next frame when the necessary stuffmg-information bits and the associated stnffmg bits can be transmitted. The maximum number of data words that can be stored in the FIFO should be chosen such that the FIFO will not become empty during this time period. In addition, the FIFO must be in a position to buffer the arriving data during the transmission of the frame bits. 3-154 Clock Adjustment at the Receiving End There is also an elastic memory (FIFO) at the receiving end of each channel. Figure 15 shows that the information is written into the FIFO with the multiplex clock pulse divided by n. As a result of the now well-known frame structure, writing must be inhibited while the additional information is being received. The writing process also must be interrupted when stuffing bits are received (see Figure 13). Consequently, received data is written into the FIFO block by block (see Figure 14). --.J Clock Recovary ... Control Logic . WRCLK WREN 1D l' ••• r+ RDCLK PLL FIFO Output Signal 8448kblt/s 1Q f---+ Input Signal 2048kblt/s ~ •• • Three More Telephone Channels ~ Figure 13. Clock Adjustment at the Receiving End With Positive Pulse-Stuffing Techniques Block Diagram NI: Useful Information bits St: Stuffing Information bits J :t: III , Frame' NI NI NI Figure 14. Bit Stream at the FIFO of the Receiver-Clock Adjustment The write clock of the FIFO has as a nominal clock frequency (the multiplex clock divided by n); however, during the reception of the frame and the stuffmg bits, several clock periods are omitted. Over a long period of time, the bit speed is identical with that of the original signal at the transmitter end (see Figure 15). As a read pulse for the FIFO, a regular clock without gaps is needed so that a continual bit stream conforming to the original signal is supplied. A PLL circuit reconstitutes this continuous clock signal from the clock signal containing gaps, although there is a small amount of jitter. 3-155 Input Clock Dlvldedbyn Clock With Gap. Regular Clock Without Gap. Figure 15. Clock Signals at the Receiver Types of FIFOs Suitable for Clock Adjustment The width of a FIFO data word for clock adjustment at the transmitting or receiving end is merely one bit and, consequently, the FIFOs listed in Table 7 can be considered as candidates for this application. Table 7. One-Bit FIFOs From TI SN74ACT2229 FIFO TYPE SN74ACT2226 SN74ACT2227 SN74ACT2228 Word Width 1 bit 1 bit 1 bit 1 bit Memory Capacity 64 words 64 words 256 words 256 words FIFOs per Package 2 2 2 2 Clocked FIFO ..J ..J fmax 22 MHz 60 MHz ..J 22 MHz 60 MHz Totem-Pole Q Output ..J 3-State Q Output ..J ..J ..J ..J Half-Full Flag ..J ..J ..J ..J Almost-Full Flag ..J ..J ..J ..J Modems With Data Compression Modems are now widely used for transmitting data over telephone lines. The telephone network was, however, originally developed for speech communications and for the transmission of analog audio signals. The result is that only alternating-current signals having an upper bandwidth limit of 3.4 kHz can be transmitted. Binary-digital information must be modulated, or converted, into another kind of signal. With acoustic couplers, frequency modulation is used such that a 0 is audible as a high note and a 1 as a lowernote. This frequency-modulated signal is analog, with 2100 Hz used for 0 and 1700 Hz for 1. These frequencies lie within the frequency band that can be transmitted over a telephone line. The maximum transmission rate is only 600 baud. Since significantly higher frequencies cannot be transmitted by a telephone network, a trick must be used to attain higher transmission speeds. If the number of possible states (e.g., frequencies) is created from two to four, two bits can be transmitted simultaneously without exceeding the upper bandwidth limit of 3.4 kHz. A further sophistication of this multistage modulation process to 16 or even 32 states (4 or 5 bits can be simultaneously transmitted) resulted in modems having atransmission capacity of up to 9600 bitls but at the same time a transmission system that was more susceptible to interference. A further increase of transmission speed by means of yet more sophisticated modulation methods would have been difficult; therefore, data compression has been used to improve performance. This involves examining the bit stream for redundant information, then compressing it. The receiver recognizes the parts of the signal that have been compressed and expands them in order to reconstitute the original signal. In a typical case, redundancy of the transmitted bit stream allows a 50% reduction of the original data, whereby the possibility for compression can typically range from 0% to 75%. 3-156 If, for example, a computer sends data via synchronous serial interface to a modem having a data rate of 4800 baud, the modem uses data compression to reduce the information to a transmission speed of 2400 baud and subsequently sends it without problems over a telephone line (see Figure 16). Variations in the compressibility of the signal are, in this case, buffered by a FIFO. If the transmitted data is not compressible, the data received from the interface line is temporarily stored by a FIFO in the modem. When the potential for data compression increases to over 50%, the modem again accepts data stored in the FIFO. Only if the compressibility of the transmitted data deviates significantly and for a long time from the average value (50%) must the arriving data stream be halted or the data stream that is leaving be interrupted. The same speed variations arise with data expansion at the receiver as with compression at the transmitter. A FIFO also is used here to buffer the data and to ensure a constant flow of data to the receiver. FIFO FIFO Interface Cable 4800 Baud Figure 16. Data Transmission by Modem With Data Compression Since in this application a serial stream needs to be buffered by the FIFO, the FIFOs having a word width of one bit shown in Table 7 are suitable. The two FIFOs needed for duplex operation (for transmitter and receiver) have already been integrated with these FIFO types into a single package. 3-157 Signal.Processor Interfaces The signal processors from TI's TMS320CXX family have one or more serial ports to allow them to communicate with other signal processors or for data exchange with peripheral equipment such as the analog interface circuit (AIC). For data transmission, the signal processors make use of the following signals: Transmit clock - clock transmit (CLKX) Transmitter control- frame sync transmit (FSX) Transmit data - data transmit (DX) Receiver clock - clock receive (CLKR) Receiver control- frame sync receive (FSR) Receive data - data receive (DR) The protocol for the transmission of data is shown in Figure 17. The fact that data is to be transmitted is signaled by FSX, which occurs on the falling edge of the clock pulse CLKX. To make the waveform of the signal processor in Figure 17 compatible with that required by the FIFO, both the clock signal CLKX and the control signal FSX must be programmed to give an inverted output. The TMS320C30 offers the possibility of programming both the polarity of the clock signal and the control signal. The resulting signals shown in Figure 18 are directly compatible with the FIFO. CLKX FSX------~,~ ____________________________________JI DX ________~--------------__ Figure 17. Serial·Port Data·Transmission Protocol of a TMS320CXX ..I' FSX ______ , DX ______~~--------------__ Figure 18. Serial-Port Data-Transmission Protocol With Inverted Signals , With data transmission via a serial port, both the transmitter and the receiver must normally be ready to operate simultaneously since the TMS320CXX has only a single word of internal buffer memory apart from the transmit and receive buffers. If a'SN74ACT2229 FIFO is switched into the communication channel, both transmitter and receiver do not need to transfer data simultaneously. Each participant can complete the data transfer when time allows. The time that is saved is available for processing other jobs. Figure 19 shows the connection of two TMS32OC30 devices. The connection of an analog interface circuit (AlC) to a TMS320C30 is made similarly. Since two independentFIFOs are integrated into a single SN74ACT2229, full-duplex operation is possible with only one package. 3-158 TMS320C30 TMS320C30 SN74ACT2229 IR WRTCLK WRTEN XFO CLKX FSX DX RESET 256 x 1 FIFO D RESET OR RDCLK RDEN XF1 CLKR FSR DR r--- - r--256 x 1 FIFO Q - r+ OR RDCLK RDEN XF1 CLKR FSR DR g IS WRTCLK WRTEN D RESET r XFO CLKX FSX DX RESET RESET RESET Figure 19. Connection of Two Signal Processors via a Serial Port With the Help of the FIFO SN74ACT2229 Teletext Decoders With teletext, pages of text are transmitted as digital infonnation in addition to the nonnal television signal. To be compatible with existing TV receivers, this digital infonnation is transmitted in the picture-frequency blanking interval. The invisible picture lines, sent during beam flyback but after those for picture synchronization, contain the digital teletex data instead of picture infonnation (see Figure 20). With D2-MAC, 360 bits with a bit rate of 20.25 Mbit/s are transmitted per TV line; therefore, the teletext infonnation occupies 17.8 JlS of the 64 JlS for which the TV line lasts. In this example, a D2-MAC decoder extracts the digital teletext infonnation from the television sigual and conducts it to a 512 x 1 FIFO (see Figure 21). The D2-MAC decoder writes the data block by block at a rate of 20.25 Mbit/s into the FIFO. The teletext module is now able to read out and process the 360-bit digital infonnation within 64 JlS at a significantly slower rate of up to 5.625 Mbit/s. In this example, the FIFO undertakes the adjustment and synchronization of the two different rates. ,..-_-===============::::::::::::==_ +. Picture =-_=======::=:::::;::::::::::::__ ~~ I I I ~~ I Horizontal Synchronous Pulses ~I~I I I Vertical Synchronous Pulses I I I Horizontal Synchronous Pulses Figure 20. Video Signal 3-159 Teletext Teletext Data Data 22.25MH D2-MAC Decoder C B8-BO _ _ _.JX,-_D_..JX,-_C_..JX,-_B_..JX,-_A __ so S1 S2 S3 S4 - Select byte-sIze JltUe endlan Write byte 0 at poslUon 1; wrItes to FIFO dIsabled Write byte C at poslUon 2; wrItes to FIFO dIsabled WrIte byte B st posItIon 3; wrItes to FIFO dIsabled WrIte byte A at posItIon 4; load fuJI 3&-blt long word to memory Figure 8. Input Registers and Timing for Write Operation During Byte-Size, Little-Endian Configuration 3-171 Effect on Status Flags Bus sizing affects the operation of the almost-empty (AEB) and almost-full (AFB) flags associated with port B and, to some degree, the empty (EFB) and full (FFB) flags. The almost-empty and almost-full flags can be programmed to one of the preset values of 4,8,12, or 16 during device reset. These depths are based on full 36-bit words. When used with bus sizing, the internal flag still reacts to the programmed depth to a multiple of the byte or word size selected. For example, if a depth of eight long words is programmed and byte sizing is selected, the AEB flag indicates when there are 32 bytes remaining to be read from FIFO 1 before it goes empty. This results from the internal flag reacting to eight long words remaining in the memory. Due to the byte size being selected, it takes 32 read-clock'pulses to unload them. Similarly, if a depth of four long words is programmed and a bus size of 18-bit words is selected, the AFB flag reacts when there are eight more valid CLKB pulses before FIF02 is written full. This is based on the need for two write-clock pulses to load each long word into memory and the internal flag reacting when there are four available locations remaining in the memory. The empty and full flags for port B still indicate when the,associated boundary condition is met during sizing operations because the internal flags for FlFOl and FIF02 are based on 36-bit words only. The write to FIF02 memory during bus-size mode occurs on the last byte or word of the full 36-bit word being loaded. This means FFB does not indicate a full condition until all four 9-bit bytes (or both 18-bit words) of the last full long word are clocked into FIF02. If, for instance, only three of the bytes are loaded during byte sizing, the internal write to memory has not occurred yet, and the status flag does not indicate full until the last byte is clocked in and the internal write to FIF02 memory takes place. Likewise, due to an internal long-word read access being performed at the beginning of a bus-sized byte or word transfer, the internal empty-flag status updates immediately, since FIF01 has effectively been read empty. However, in the instance of byte sizing, there are still three bytes remaining in the pipeline to be shifted out before the user considers the EFB flag to be valid. To provide a proper empty-flag indication, the same signal that disables the internal memory accesses during remaining byte (or word) transfer is combined with the synchronized empty-flag~ut signal and prevents it from switching until the final byte (or word) is clocked to the outputs. This results in the EFB flag correctly indicating when the empty-boundary condition occurs, regardless of the bus size chosen. Dynamically Changing the Bus Size Many applications select a bus size at system initialization, which remains fixed due to the architectural design of the system. However, applications that require dynamic bus sizing to be performed can use the SN74ABT3614 to provide an effective interface. To avoid data loss when making the transition, the following explanation is provided. Mailbox operations, which are selected with bus-size control terminals, also are discussed. Bus-size selections must be made one cycle before they are to take effect. The values present on SIZO and SIZ1 are stored in registers XFF1 and XFF2 on the rising edge of CLKB as shown in Fignre 9. This latency allows the control logic t<;> be set up for the next transfer, resulting in short setup and hold times for these inputs and allows comparisons to the previous state of the inputs to determine if an actual change in bus size is being requested. This is important when using the size inputs to enter mailbox mode, then returning to the previous size for FIFO operations. In Figure 9, exclusive NOR (XNOR1 and XNOR2) gates are used to compare the current state of the size inputs to the,ir previous states. If either input is changed from its previous state, the signal COUNT_RES is driven low. This signal is a synchronous reset to the counters that keep track of the byte or word count. When a size change is requested, the rising clock edge that loads the new size value into the registers also resets the counters, ensuring proper byte tracking of the new size beginning with the next clock pulse. If a size change is requested before the last byte/word is transferred from/to the FIFO, the remaining data is lost or the entire word is not written (see Figure 10). 3-172 SWO~~------------------------------~~ SW1~~------------------------------~~ XFFO FF RES ..... ~ Clock Reset D XFF1 FF_RES Clock SIZO XNOR1 Reset XFF2 FF RES Clock SIZ1 XNOR2 Reset .>o--+t-l D XNOR3 Clock Q Reset Figure 9. Size-Control Block In the special case of mailbox operations, which use the size inputs to select mailbox mode on port B, it is essential not to interpret this as a change in bus size. When SIZI and SIZO inputs are driven high (mailbox mode), COUNT_RES is disabled (along with the counters) and the size registers are not updated (see XNOR2 in Figure 9). This allows mail operations to take place without disturbing any data transfers to/from the FIFO. 3-173 RCLK 0_,_1_____X'--________O_,O_ _ _ _ _ _ __ SIZO, SIZ1 _ _ _ _ _ W2A X W2B X W2C X W3 X W4 X W5 X~__ (a) TIMING FOR CHANGE FROM BYTE SIZE TO LONG WORD, LOST LAST BYTE OF WORD 2 0,1 SIZO,SIZ1 Data !!V< 0,0 X W1D X W2A X W2B X W2C X W2D X W3 X W4 X W5 (b) TIMING FOR CHANGE FROM BYTE SIZE TO LONG WORD, NO DATA LOSS WCLK 1_,_0___.IX'--________O,_O_ _ _ _ _ _ _ _ __ SIZO, SIZ1 _ _ _ _ Input 3< W2A X W2B X W3A X W4 X W5 X LO~:: --W-1-A~+-W-1B-""'Xr. . ===W=2A=+=W=2=B==:Xr-W-4~Xr-W5~X we X....__C we X. . ____ (e) TIMING FOR CHANGE FROM WORD SIZE TO LONG WORD, WORD W3A IS NOT WRITTEN 1,0 SIZO,SIZl Input Data Loaded X 0,0 3< W2A X W2B X W3A X W3B X W4 X W5 X W1A+W1B X W2A+W2B >E3A +W3BX W4 X W5 X C (d) TIMING FOR CHANGE FROM WORD SIZE TO LONG WORD, NO DATA LOSS Figure 10, Dynamically Changing Byte-/Word-Slze Transfers to Long-Word Data Transfers 3-174 Byte Swapping Four different modes of byte-order arrangement (byte swap, word swap, byte-word swap, and no swap) can be performed with any port-B size selection. When byte swap is performed, the order of the bytes are rearranged within the long word but the bit order remains the same. The port-B swap-select (SWI and SWO) inputs are used to achieve these byte-order arrangements. Table 2 lists the levels on the SWI and SWO input terminals and the respective size implemented. Figure II shows the different schemes of byte swap. Table 2. Control of Byte-Swap Operation on Port BUsing SW1 and SWO Port A PortB PortA SW1 SWO L L BUS CONFIGURATION No swap L H Byte swap H L Word swap H H Byte-word Swap PortB PortA Port B PortA Port B ~ cifB ~ ~ (a) No Swap (b) Byte Swap (e) Word SWap (d) Byte-Word Swap Figure 11. Byte-Swap for Long-Word-Size Data Transfers The byte swap is performed with any port-B size selection. Byte-order arrangement is implemented by the levels on the port-B swap select inputs on a CLKB rising edge that reads a new long word from FIFOI or writes a new long word to FIF02. When long-word-size (36-bit) transfers are selected on port B, the byte-order arrangement can be changed on each clock cycle. On the other hand, when byte (9-bit) or word-size (I8-bit) transfers are selected on port B, the byte order chosen on the first byte or word of a new long-word read from FIFO I or written to FIF02 is maintained until the entire long word is transferred, regardless of the swap-select input states during subsequent reads or writes. This implies that for byte- or word-sized data transfers, the byte-order arrangement can be changed only on the first byte or word data transfer of a new long-word read. Simultaneously performing a byte swap and bus size on port B results in the sequence of events shown in Figure 12. 3-175 FIF01 Reads FIF01 Writes No Yea 36-Blt Long Word Storad In FIF02 SRAM In ~pped Order Figure 12. FIF01 Data-Read and FIF02 Data-Write Sequence During Simultaneous Bus-Sizing and Byte-Swapping Operations 3-176 Parity Generation and Checking Parity Checking The odd or even parity-checking function can be selected on both port A and port B using the ODDIEVEN input. Four parity trees examine the parity of incoming or outgoing data bytes on each port as shown in Figure 13. Port-A bytes are arranged as AO-A8, A9-Al7, A18-A26, and A27-A35, with the most significant bit used as the parity bit. A parity error on each of these four bytes is indicated internally by a low signal on the individual-error (ER) flag. These four ER outputs are combined to output a single parity-error (PEFA) flag that indicates an error on one or more bytes on portA. Similarly, port-B bytes are arranged as BO-B8, B9-B17, B18-B26, and B27-B35, with the most significant bit used as the parity bit. Again, a single parity-error (PEFB) flag indicates an error on one or more bytes on port B. During the port-B sizing operation, the internal flag output for invalid data bytes is disabled; the parity-error flag indicates an error only on the bytes that are valid for the particular size selected. Parity checking on both ports is a passive operation; the port parity-error flags can be ignored if this feature is not desired. ODD/EVEN - - - - , BE---, ODD/EVEN SIZ1 SIZO B35 A35 9 9 ER ER· ER ER ER ER ER ER 9 9 9 9 9 9 BO AO Figure 13. Parity-Checking Block Diagram Parity Generation Parity generation for port reads from the FIFO or mailbox can be selected using the parity-generate select-input terminal for that port (pGA or PGB). ODDIEVEN selects the type of parity generated. Port-A bytes are arranged as AD-A8, A9-A17, A18-A26, and A27-A35, with the most significant bit used as the parity bit. Port-B bytes are arranged as BO-B8, B9-B17, B18-B26, and B27-B35, with the most significant bit used as the parity bit. A write to the FIFO or mailbox stores all 36 bits regardless of the state of the parity generate select (PGA) input. When data is read from the FIFO, the lower eight bits of each byte are used to generate the parity bit as shown in Figure 14. If parity generation is selected (PGA - 1), the levels originally written to the most significant bits of each byte are substituted by the generated parity bit as the word is read to the outputs. Otherwise, the levels originally written to most significant bit of each byte are output. 3-177 ODD/EVEN MemorY PGB P Output Register 8 Date Input BO-B8 Parity Generator Memory Out Figure 14. Parity-Generation Block Diagram Parity generation on mailbox reads for a port is performed by the four parity trees used to check the parity on the inputs to the port. When odd- or even-parity generation is selected on a port-A or port-B read from the mailbox, the port parity-error (pEFA or PEFB) flag is held high regardless of the state of the inputs AO-A35 or BO-B35. The generated parity does not change the contents of the mailbox register. Internetworking Intemetworking is the process of connecting a variety of local-area-network (LAN) and wide-area-network (WAN) computer systems and related devices to build the communications infrastructure required to satisfy the needs of an organization. Bridges and routers are two key devices that provide intemetworking capability (see Figure 15). Layer 7 - Application Layer 6 - Presentation Layer 5 - Session Layer 4 - Transport Layer 3 - Network Layer 2 - Data Link Layer 1 - Physical t EtherNet (a) OSI REFERENCE MODEL t Token Bus t Token Bus (b) BRIDGE t X.25 (c) ROUTER Figure 15. Bridge and Router Devices Bridge and router designs are critical in meeting the increasing bandwidth requirements and volume of data transfer in existing network elements. The use of FlFOs in these designs provides a viable means for improving the system performance. The FlFOs provide a link between the communications processors and buses operating at different speeds and data widths. The handshaking signals for control can provide operating speeds that match the bus-communication speeds. 3-178 A bridge operates at the data-link layer (layer 2) in the OS1 model to connect two similar LANs. A bridge reads the destination address contained on each incoming packet and uses the address to transmit the packets or to ignore them. This process is known as address filtering. An example of a bridge design using multiple FIFOs is shown in Figure 16. -.- - ~ Serlal-ta-Parallel Conversion Data ---4 Control Block Address Generation Data Multiplexing Read/Wrlta Control FIFO Flags Data Data Control FIFO Parallel Data to FIFOs Data Flags f-+- To Host Bus r+r+- .. I r+- ~ I Microprocessor I Figure 16. Implementation of a Bridge Using FIFOs A router, on the other hand, operates at the network layer (layer 3) in the OS1 model and can be used to connect two different networks; therefore, a router has the added complexity to perform frame conversion in addition to address filtering. For example, in the situation where devices on a WAN communicate with the devices on a token-ring LAN, the interconnecting device is required to convert HDLC frames to token-ring frames and vice versa as shown in Figure 17. ~r----------------~~----------------, I Layer 2 T1 36 36 36-Blt Router Bus Layer 2 1-4..,16"+0-1 Token-Ring HI-?''+-I Controller L ___________________________ _ Figure 17. WAN to Token-Ring Router Using SN74ABT3614 FIFO 3-179 The layer-2 device provides the functions of flag generation and detection, zero-bit insertion and deletion, cyclic-redundancy-check (eRC) generation and detection, and abort generation and detection. The layer-2 device interfaces with the T1 transceiver on the line side. The SN74ABT36l4 FIFO provides a bidirectional buffer between the layer-2 device and the high-speed router bus. In l¥idition, if the layer-2 device has a 36-bit interface on the bus side, the bus-matching function on the SN74ABT36I4 device is used to interface to a high-speed 36-bit bus. Similarly, the layer-2 device provides the necessary token-ring LAN protocols and links the token-ring LAN controller to the router bus. The SN74ABT36I4 is used to match the data width and rates between the token-ring controller and the 36-bitrouter bus. A FIFO solution implemented in the hardware offers a speed advantage over the latency for a software setup performing the same functions. Conclusion The SN74ABT3614 is a clocked, bidirectional 64 x 36-bit member of TI intemetworking family ofFIFOs. This FIFO integrates the glue logic necessary to simplify design of communications networks. In addition to providing the traditional FIFO function of removing input/output bottlenecks, this FIFO provides full functionality for bus-sizing, byte-swapping, and parity-generation checking logic, and mailbox operations. These functions are necessary for effective communication between microprocessors, communicatious processors, and buses. The SN7 4ABT3614 FIFO can operate at frequencies up to 66 MHz and can provide data access times as fast as lins. The FIFO also is available in speed sorts to provide 33-MHz and 50-MHz operation. 3-180 High-Speed, High-Drive SN74ABT7819 FIFO Teruo Ishii and Navid Madani Advanced System Logic - Semiconductor Group ~TEXAS INSTRUMENTS SCAA016A 3-181 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable althe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications''). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-182 Contents TItle Page Introduction ............................................................................ 3-185 Structure ..... .......................................................................... 3-186 High-Speed Performance ................................................................. 3-186 High-Drive Capability .................................................................... 3-186 Incident-Wave Switching Capability ........................................................ 3-187 VME Backplane Drive . .•................................................................. 3-190 Summary .................................... ............................................ 3-191 Reference .............................................................................. 3-191 Figure List of lliustrations Title Page 1 SN74ABT7819 Block Diagram ...................................................... 3-185 2 Output Wavefonn at 80-MHz Clock (CL - 50 pF) ........................................ 3-186 3 Output Characteristics of the SN74ABT7819 (Vee - 5 V, TA - 25°C) ........................ 3-187 4 5 6 7 Shelf Voltage When Switching a Transmission Line ...................................... 3-188 Backplane Driven From End . . . . . • . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-190 Backplane Driven From Center. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-190 Number of Slots Versus Shelf Voltage (Center Drive) ..................................... 3-191 ~183 3-184 Introduction First-in, fIrst-out (FIFO) memories are used as high-speed data-stream buffers for maximizing throughput between buses having different data-transfer speeds, for example, between multiprocessors or a microprocessor and peripheral circuits. Recently, the processing speed of microprocessors has become so fast that the gap between the speed of the microprocessor and peripheral circuits has widened. As a result, the demand for faster FIFOs that can maximize the use of the microprocessor capabilities has increased. Usually, when a memory integrated circuit drives a heavy-load bus line, external bus buffers are required. FIFO memories, however, help to minimize the design complexity and cycle time by providing the capability to buffer data while at the same time directly driving the bus line. The Texas Instruments SN74ABT7819 is a high-speed, high-drive, advanced BiCMOS FIFO with operating frequencies up to 80 MHz and a drive capability ofl()H/IoL - -12124 rnA (see Figure 1). PENA RENA WENA CSA Port-A Control logIc WiRA t I Read I Polntar ClKA RSTA I t j, 18 : Raglstar: 512 x 18 Dual·Port SRAM HRaglstar FIFOB-A 18 + 18 -+---+-I PoInter WrIte I Flag logIc FIFOB-A ORA AO-A17 - IRB AF/AEB HFB 8 ~ BO-B17 8 IRA AF/AEA HFA I Write "I Flag logIc FIFOA-B ORB I PoInter t 18 I Raglster r.- 512x 18 Dual·Port SRAM ~ HRaglstar FIFOA-B I ..." t Read I Polntar I RsrB ClKB Port·B Control logIc CSB WiRB WENB RENB PENB Figure 1. SN74ABT7819 Block Diagram 3-185 Structure The SN74ABT7819is a512x 18 dual-port bidirectional clocked FIFO (see Figure 1). It comprises two SRAMs (FIFOA, FIFOB), specific circuits for controlling these two SRAMs, and II 0 registers. Read/write operations are carried out on the SRAMs at the low-to-high transition of their respective free-running clocks, CLKB and CLKA. Read/write operations are synchronized by independent clocks from the system. Likewise, EMPTY and FULL flags are synchronized with independent clocks from the system. Neither an external clock-enable circuit nor an external circuit for synchronizing output flags to the clocks are required. Output data is buffered by output registers and can be enabled at alL times. These features facilitate design by eliminating the need for detailed timing considerations, especially in applications requiring high-speed operation. High-Speed Performance The SN7 4ABT7819-12 is a high-speed FIFO memory with an access time of9 ns (at CL - 50 pF). Operation is specified for free-running clock (CLKA, CLKB) inputs of up to 80 MHz. These high-speed characteristics allow microprocessor time management to be reduced and allow an efficient system to be configured. Figure 2 shows the waveform of output data read by an 80-MHz clock. 7 6 5 >I ; ~ 'S S- 4 11 3 I 2 J J I II a I ~ ,/' 11 / r- 0 -1 ~ lr ,..,J ~ lr L ,.,1 -2 -3 O. 10 20 30 40 50 60 70 60 90 100 t-Tlme-nl Figure 2. Output Waveform at SO-MHz Clock (CL = 50 pF) High-Drive Capability The advanced BiCMOS SN74ABT7819 comprises a bipolar-output circuit that achieves a high-drive capability. Figure 3 shows the output characteristics for the SN7 4ABT7819. Output impedance is equivalent to about 30 0 at high output and 50 0 at low output. Output impedance values for the SN74ABT7819 are equivalent to the output impedance for FAST and BCT bus-interface logic; therefore, a designer can effectively implement the SN74ABT7819 as an interface to a bus line. 3-186 1.1 5.5 5 > > 4.5 I I ! 0.9 CD G) al I 4 ~ 3.5 '$ 3 :::I ,..- V .s- 0 ./ 1.... 2.52 .J:. al :E 1.5 ~ I J: .p 0.5 / ~ ~ '$ .s-:::I V" 0.8 0.7 0.6 0 ] ~ ....~ V ""'" !oJ .p 0.5 0.4 0.3 0.2 0.1 o -100 -90 -80 -70 -80 -50 -40 -30 -20 -10 0 / o o IOH - High-Level Output Current - mA ........ 10 20 ~~ 30 40 - 50 ........ ~ ~- 60 70 80 90 100 IOL - Low-Level Output Current - mA Figure 3. Output Characteristics of the SN74ABT7819 (Vee =5 V, TA =25°C) Incident-Wave Switching Capability When specific data (rectangular wave) is carried over the transmission line, data is influenced by the impedance of the transmission line, introducing distortion in the data waveforms. This distortion can cause a mismatch in required data-transfer speed; therefore, the characteristic impedance of the transmission line must be considered in the system design. The following describes the relationship between the output characteristics of the SN74ABT7819 and the transmission line. The impedance of the transmission line and the drive capability of the device can introduce distortion in the waveforms at the low-to-high and high-ta-Iow transition referred to as shelf voltage. The shelf voltage lowers as the impedance of the transmission line and the drive capability of the device decrease. The shelf voltage is expressed by the following equations and is shown in Figure 4. Zo V OHS = ZONH X Zo X V OH (a low-to-high transition) (1) ZONL xZ XVOH ONL (a high-to-Iow transition) (2) VOLS=Z o Where: V OHS VOLS V OH ZoNH ZONL = Zo = Shelf voltage at the low-to-high transition (V) Shelf voltage at the high-to-Iow transition (V) High output voltage of device (V) High output on resistance of device (0) Low output on resistance of device (0) Impedance of transmission line (0) 3-187 v SN74ABT7819 ------, VOH I l ! l' VOH ~~--11 "'""" .. 0 Pol" ","L. 50 VOHS VOLS - ------1 ------ Time Figure 4. Shelf Voltage When Switching a Transmission Line When a low-impedance line is driven by a device with high output resistance, the sbelfvoltage occurs in the threshold region, causing delay in establishing the transmitted logic. In order to drive a low-impedance transmission line without any performance degradation, a high-drive capability is required. On the bus lines of backplanes or memories, the impedance of the transmission line drops as load capacitance is distributed over the transmission line. Generally, the characteristic impedance and the propagation delay time of the transmission line are shown by the following equations: (3) (4) Where: Zo = LO Co - tpd - Impedance of transmission line (0) Impedance per unit length (H) Capacitance per unit length (F) Propagation delay time of transmission line (s/m) When the load capacitance is applied on the transmission line, the characteristic impedance and the propagation delay time of the transmission line are changed as follows: (5) (6) Where: Z' - tpd' = Cl - Impedance after application of capacitance (0) Propagation delay time after application of capacitance (s/m) Applied capacitance (F) Example The effect of loading a Zo - 100 0 and tpd - 7 ns transmission line with 8-pf loads equally spaced at 3-cm intervals is calculated in equations 7 and 8. The high- and low-output impedance necessary to drive the loaded transmission line with no settling time delay is calculated in equations 9 and 10. 3-188 From equations 3 and 4: 1.0 - Zo x tpd Co - 100 x 7 x 10--9 - 700 nHlm - 21 nHl3cm tpdZo -7 x 10--9/100 -70 pF/m - 2.1 pF/3cm Substituting for Co in equations 5 and 6: Zo' = Zo ~ 1+ = -;:.=~100~:::;:=;:~ = 45.6 V ic; tpd' r.! (7) 8x10-12 1 + 2.1 X 10-12 = tpd V~ 1 + Co = 7 x -9 10 8 X 10-12 1 + 2.1 X 10-12 = 15.4 ns / m (8) The values calculated by equations 7 and 8 are Zo' - 45.6 a and tpd' - 15.4 nslm, respectively. The propagation delay time of the transmission line is about doubled, and the impedance of the transmission line is about halved. The output impedance required of a device to drive this transmission line without a settling time delay is calculated by the following equations: (high-output impedance) Z' 2SZ0NH~ZOXVOH (9) 45.6 x 35 2 < - ZONH + 45.6 . ZONH S 34.2 r.! (low-output impedance) 0.8 ~ Zo' ZONL + ZONL X VOH (10) ZONL S 13.5 r.! The transmission line must be driven by a device having an output impedance of 34 a or less at high output and 13 a at low output; therefore, this transmission line can be driven with the SN74ABT7819. 3-189 VME Backplane Drive Since the drive perfonnance of almost all memory integrated circuits is low, it has been difficult to drive memory lines or backplane buses directly; therefore, external bus buffers have been required to interface a memory to a bus driver. Now the SN74ABT7819 allows direct driving of medium-scale bus lines. Figure 5 shows the wavefonn of a 12-s10t VME backplane driven by an SN74ABT7819 FIFO from the line end. Although wavefonn distortion caused by multiple reflection due to the 3-cm stub length and DIN connector occurs, there is no influence of reflection in the threshold region; therefore, when the SN74ABT7819 drives the VME backplane from the end, direct driving is possible. However, when the 12-slot VME bus is driven from the center, the backplane line is regarded as a branch pattern and the impedance becomes one-half the impedance at that point, resulting in generation of a step in the threshold region (see Figure 6). For this reason, a bus buffer having a higher drive capability is required. 7 7 6 6 5 5 4 4 rvf !.oj"'" 3 > I 2 :I!! ~ >I :I!! 812 81 ~~ - -1 -2 o 20 30 VCC=5V 40 50 60 70 t-Tlme-ns 80 rt sa I 812 " .... I£. o ::r-- 1\1 -1 -2 90 100 VCC=5V o ro ~ ~ ~ ~ 80 t-Tlme-ns VCC=5V ~ 80 80 100 VCC=5V 3300 mo 3~0 4700 4700 4700 Pulse Generator Figure 5.. Backplane Driven From End 3-190 2 -3 10 Il ~ 3 ~ J 0 -3 "- Pulse Generator Figure 6. Backplane Driven From Center Figure 7 shows the stepped levels when the VME bus is driven from the center slot by the SN74ABT7819. The influence of multiple reflections increases and the level of the shelf approaches the threshold region as the number of slots increases (the transmission line of the backplane is optimized corresponding to the number of slots). Based on these results, a design of five or fewer slots is preferable for driving with sufficient margin,to preclude unwanted delay and possible data errors. 2.5...-----------------..., 1.5 6 8 10 12 Number of Slots Figure 7. Number of Slots Versus Shelf Voltage (Center Drive) Summary The SN74ABT7819 is a high-speed, high-drive capability FIFO memory that meets the high-speed data-transfer rate requirements and drives the bus lines directly. This application report presents the essential points that a designer should consider in using a SN74ABT7819 for high-speed data transfer. The importance of drive capability and the influence of distortion is explained. Comprehending the relationship between drive capability and transmission line characteristics is essential to obtain the best performance of the SN74ABT7819. Reference Bus-Interface Circuits Application and Data Book, Texas Instruments Incorporated, 1990 3-191 3-192 SPARe MBus-to-Futurebus+ Bridge Using the Texas Instruments Futurebus+ Chipset Robert Gugel Mixed Signal Product Group ~TEXAS INSTRUMENTS SCAA019A 3-193 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale in accordance with TI's standard warranty. Testing and other quality control techniques are utiiized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessariiy performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-194 Contents Title Page Introduction ••••••••••••••••••.•••••••••••••..••••.••••••••••••.•••••••••••••••••••••••• 3-197 Transaction Support •••••••••••••••.•••••••••.••••••••••••••••••.•••••••••••••••••••••••• 3-197 Bridge Architecture ••••••••••••••••••••••••••.••••••••••••••••••.•••••••••••••••••••••••• 3-199 Command •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.•• 3-200 Address ................................................................................ 3-200 Direct-Memory Model .................................................................. 3-200 Page-Memory Model ................................................................... 3-200 Datapatb ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••. 3-201 Considerations ......................................................................... 3-201 Byte-Lane Mapping .................................................................... 3-201 Control Logic ........................................................................... 3-203 Summary ............................................................................... 3-206 List of IUustrations Figure 2 3 4 5 6 7 8 Title Page SPARCFB+ ...................................................................... MBus-to-HIP Bridge ............................................................... MBus/FB+ Memory Map ........................................................... Byte-Lane Mapping for a 64-Bit-Only System ........................................... Byte-Lane Mapping for a 64-1 32-Bit System ............................................ Byte-Lane Mapping Showing Path Taken by Address Quadlet on HD ........................ MBus-Word Read to HIP Single-Read Transaction ....................................... MBus-Word Write to HIP Single-Write Transaction ...................................... 3-197 3-199 3-201 3-202 3-202 3-203 3-204 3-205 3-195 3-196 Introduction This application report describes the logic necessary to connect the SPARC MBus to the Texas Instruments (TI) Futurebus+ (FB+) chipset host interface (HIP). This logic is a translator of MBus transactions to HIP transactions and vice versa. The MBus-to-HIP bridge described is MBus level-I compliant. Level-I transactions are the noncache-coherent subset of MBus level-2 transactionS. Even though level-2 transactions are occurring on the MBus, they cannot cross the L1 bridge. Figure 1 shows a block diagram of a FB+ module featuring a SPARC processor that can access FB+. The MBus-to-HIP bridge block is highlighted. Since knowledge of FB+, the HIP, and MBus is necessary to understand this defmition, references to the applicable specifications also are included in Figure I. Memory SPARC L2MBu8 SPARC MBu8 Interface Specification, Rev 1.2 TI Futurebu8+ Interface Family Data Manual, Rev 1.2 Unceched FUtUrebU8+ NOTE: Reference IEEE P896.1 and P896.2 Figure 1. SPARe FB+ Transaction Support A transaction originating on the MBus must be transported to the HIP and then to the FB+. This requires the L1 bridge to be a slave to MBus transactions and a master to HIP transactions. Likewise, a transaction originating from FB+ is transported to HIP and then to the MBus. In this case, the LI bridge is a slave to HIP transactions and a master of the MBus transactions. Each of the three buses has a set of transactions that must be mapped to each other when crossing bridges. 3-197 Table 1 shows the translation of MBus transactions to HIP transaction that is perfonned by the L1 bridge when it is a slave to the MBus and a master of the HIP.. Table 1. MBus-to-HIF Transaction Mapping MBUS TRANSACTION TYPE Byte read fwrite Half-word (2 bytes) readfwrite RESULTING HIF TRANSACTION TYPE DW32 single l-byte readfwrite partial \ DW32 single 2-byte readfwrite partial Word (4 bytes) readfwrite DW32 single readfwrite Double word (8 bytes) readfwrite DW64 B-byte burst readfwrite 16-byte burst read fwrite DW64 16-byte burst readfwrite 32-byte burst readfwrite DW64 32-byte burst readfwrite 64-byte burst readfwrite DW64 64-byte burst readfwrite 12B-byte burst readfwrite Two DW64 64-byte burst readsfwrites chained together with HIF MORE signal Table 2 shows the translation of HIP transactions to MBus transactions that is perfonned by the L1 bridge when it is a slave to the HIP and a master of the MBus. The data-width 32 (DW32) HIP bursts always cause 4-byte-wide transactions on MBus, and data-width 64 (DW64) HIP bursts always cause 8-byte-wide MBus transactions. This means that dynamic bus sizing from HIP to MBus is not supported by the architecture suggested. Table 2. HIF-to-MBus Transaction Mapping HIF TRANSACTION TYPE 3-198 RESULTING MBUS TRANSACTION TYPE DW32 single l-byte readfwrite partial Byte readfwrite DW32 single 2-byte readfwrite partial Half-word (2 bytes) readfwrite DW32 single 3-byte readfwrite partial Half-word (2 bytes) readfwrite and a byte readfwrite DW32 single readfwrite Word (4 bytes) readfwrite DW32 8-byte burst readfwrite Two individual word (4 byte) readsfwrites DW32 l6-byte burst readfwrite Four individual word (4 byte) readsfwrites DW32 32-byte burst read fwrite Eight individual word (4 byte) readsfwrites DW32 64-byte burst readfwrite Sixteen individual word (4 byte) readsfwrites DW64 8-byte burst read fwrite Double-word (8 bytes) readfwrite DW64 16-byte burst readfwrite 16-byte burst readfwrite DW64 32-byte burst readfwrite 32-byte burst readfwrite DW64 64-byte burst readfwrite 64-byte burst readfwrite Bridge Architecture Figure 2 shows suggested implementation of the MBus to HIP bridge. -- ----------------------------------- COMMAN D I MBus-Attrlbute I 63-36 T ~- -> ADDRESS FB+ Page Register L MATTR TvDe Size Lock Encoder J TRW,DL,DW ---a4, TBST, TSIZE, K L HIF-Attrlbute Encoder HATTR-J -- ----------------------------------4 ~ -r+-,- 36 , Selected t-- ~ 32 , > 35-0 MBus Page Register 4 r-- - HA (31-0) -- ----------- ~------------------- -,--- DATA PATH MAD (63-56) MAD (63 -0)- HA (31-24) FIFO 64x36x2 MAD (31-24) r--.. HD (31-24) MAD (47-40) HA (15-8) MAD (15-8) HD (15-8) SN74ABT3614 - - I> r--.. < - MAD (55-48) HA(23-16) FIFO 64x36x2 MAD (23-16) HD (23-16) MAD (39-32) r--.. HA(7-0) MAD (7-0) HD (7-0) HD (31-0) SN74ABT3614 - -I-' "CONTROL HCLK ------------ -F- r----------------------r V MATTRY, MRTY, MERR MBC - - I-+- HATTR I-- HFC 4 BR,MBG,MBB I I MORE, HIP, HAS, HDS, DSACK (1 -0), ---Y, BSTRD BSTAT(1-0) MS (1-0) BSTRDY L---~ HBR, HBG, GACK Figure 2. MBus-to-HIF Bridge 3-199 A typical bus-bridge implementation consists of command, address, datapath, and control logic. The operations of these sections in this design are as follows: Command The host interface defines a set of discrete signals that indicate the attributes (i.e., type, size, etc.) of the transaction taking place. The MBus does the same thing; however, these signals are multiplexed onto signals in the field MAD (63:0) that are not used to carry the address during the address phase of transactions. These command attributes also are logically encoded differently by the two buses. When a slave to MBus, the command section of the L1 bridge must latch the transaction-specific information from MAD (63:36) during the address phase and encode it into host-interface attributes that correspond to the resulting transaction to be mastered on the host interface. When a slave to the host interface, the command logic encodes the HIP's attribute signals into MBus attributes that correspond to the resulting transaction to be mastered on the MBus. These encoded MBus attributes must be multiplexed onto MAD (63:36) along with theMBus address during the MBus-address phase. Address The host interface has a 32-bit physical address space with a 36-bit extension. The MBus defmes a 36-bit physical-address space. The address portion of the L1 bridge logic must dO three things: Recognize the address region that it must respond to as an MBus slave and as an HIP slave Transport the physical address from one protocol to the other Relate one bus-memory region to the other Direct-Memory Model The simplest memory model would be a logically direct connection between the MBus 36-bit address and the HIP 36-bit address. This would mean that all MBus addresses not within FB+ MEM_BASE and MEM_BOUND or UNIT_BASE and UNIT_BOUND would be mapped into the lower 64 Gbytes of the 64-bit FB+ address region. It would also mean . that all MBus memory would be accessible from FB+ (between MEM_BASE and MEM_BOUND) and no MBus memory would be private. Page-Memory Model A slightly more complex memory model is defined here that allows for private memory on the MBus. When a slave to the MBus, a 4-bit FB+ page register is used to map one of 16 4-Gbyte MBus memory regions into the FB+ 32-bit address space. Within the 4-Gbyte FB+ page, the MEM_BASE and MEM_BOUND registers contained in the TI FB+ chipset point to local public memory on the MBus. Other addresses within the page but outside of MEM_BASE and MEM_BOUND are remote addresses and are transported to FB+ via the HIP. When a slave to the host interface, the MBus-page register is used to map the incoming 32-bit FB+ address to one of 16 4-Gbyte MBus memory regions. Within the 4-Gbyte MBus page, the ME~BASE and MEM_BOUND registers contained in the TI FB+ chipset point to memory on the MBus, which can be accessed from FB+. If the FB+ and MBus page are kept the same, the addresses outside the page are private (he., not accessible by the other bus). Figure 3 shows the memory mapping between MBus and HIP for the page-memory model. 3-200 16 x 232 Page 15 - 15 x 232 FB+ Page Register Private MBuI Memory FB+ Page Register ~ Page 2 2 '" '" _----+------1 MEM_BOUND - --- l,-'" 4 x 232 x 232 Page 1 i'~--, ,, ,, 232 Page 0 0 ,. " ' ..."-_ _ _ _ _...... 0 MBuI FB+ Figure 3. MBus/FB+ Memory Map Datapath Considerations There are several considerations when designing the datapath interface between the MBus and the host interface. • • • • • Since MBus runs at 40 MHz nominally and the HIP at 20-2S MHz, a FIFO is needed to synchronize the two different time domains. MBus has a big-endian datapath and requires that words and half words be word and byte aligned. The HIP has no endian preference with the exception of big-endian access to FB+ CSR space. MBus always multiplexes address and data on 64 signals; the HIP has demultiplexed address and data when the data width is 32 bits, and multiplexed address and data when the data width is 64 bits. MBus does 1-,2-,4-, and 8-byte nonburst (word) and 16-, 32-, 64-..and 128-byte burst transactions while the HIP does 1-,2-,3-, and 4-byte nonburst (single) and 8-, 16-,32-, and 64-byte burst transactions. MBus bursts are always of data width 64. HIP bursts can be of data width 32 and 64. MBus has no parity protection on its address/data lines. The HIP address/data lines do have parity; however, the TI chipset can generate parity internally and pass it on to FB+ when sourcing data to FB+. Taking the above considerations into account requires the use of two bidirectional FIPOs capable of being clocked at 40 MHz. These FIPOs need to be 32 bits wide and 64 words deep. They need empty/full flags, clock enables, and port-direction control. A byte-swap function within a 16-bit word also is required. The TI SN74ABT3614 is an ideal candidate. Byte-Lane Mapping FB+ systems require address invariant byte-lane mapping. This means that data byte 0 (the byte pointed to by byte address 0) always appears onFB+ AD (7:0). MBus systems require that data transfers ofless than a double word (8 bytes) be aligned. Figure 4 shows byte-lane mapping between MBus, the FIFOs, the HIP, and FB+ for a 64-bit implementation only. 3-201 HA HA (31) Byte 3 MAD (63) AD (63) Byte 0 Byte 7 Byte 6 Byte 1 FIFO Byte 2 Byte 5 Byte 3 FB+ Chipset Byte 4 Byte 4 Byte 3 Byte 2 Byte 5 FIFO Byte 1 Byte 6 Byte 7 Byte 0 MAD (0) AD (0) Byte 0 MAD FB+ HD (0) HIF Figure 4. Byte-Lane Mapping for a 64-Blt-Only System Since 32-bit data transfers can take place on FB+ and the HIF, quadlet (4-byte word) steering is required to correctly align 32-bit data quadlets from the HIF to those of the MBus. Quadlet steering of this type require a cross-point switch with two 32-bit ports on each end. This could be implemented with four pairs of 16-bit Widebus™ transceivers; however, this is expensive in terms of board space and datapath performance. A FIFO with the ability to swap bytes within 16-bit words performs the quadlet-steering function with the byte lanes wired as shown in Figure 5. HA FIFO AD (63) MAD (63) Byte 0 Byte 7 Byte 1 Byte 6 Byte 2 Byte 5 HA(O) Byte 3 FB+ Chipset Byte 4 HD Byte 4 Byte 3 HD (31) Byte 5 Byte 2 Byte 6 Byte 1 Byte 7 Byte 0 MAD FB+ AD (0) MAD (0) HIF Figure 5. Byte-Lane Mapping for a 64-/32-Blt System Wldebus is a trademark of Texas Instruments Incorporated. 3-202 An example of this quadlet steering in action is the following scenario: the TI FB+ chipset is a slave to a single-beat DW64 write transaction on FB+ (8 bytes). The chipset's 64-bit HIP enable bit is not set; therefore, it perfonns 4-byte-wide, 8-byte burst writes onto the HIP. This HIP burst write contains two 4-byte data phases. The first data is an even-address quadlet and needs to be steered to MAD (63:32). The second data is an odd-address quadlet and needs to be steered to MAD (31 :0). Figure 6 shows the datapath for the odd-address quadlet. HA FIFO ~--'HA(31) r----, AD (53) MAD (53) r----, Byte 0 Byte 7 Byte 1 Byte 5 Byte 2 Byte 5 Byte 3 FB+ Chipset Byte 4 Byte 4 Byte 3 Byte 5 Byte 2 Byte 5 Byte 1 Byte 7 Byte 0 '-_..I AD (0) MAD (0) MAD FB+ HIF Figure 6. Byte-Lane Mapping Showing Path Taken by Address Quadlet on HD Control Logic The MBus controller (MBC) and host interface controller (HIPC) are synchronous state machines that operate in the two clock domains. They handle the protocols of their respective buses. They also handle arbitration protocols when bus mastership is required. These controllers drive the latch enables, 3-state controls, and FIFO control signals of the command, address, and datapath sections of the Ll bridge. They are responsible for coordinating data flow between the two buses that operate at different data rates by using the master hold-off and slave-wait capabilities of their respective buses. They also are responsible for manipulating arbitration protocols in response to locked-transaction requests. Details of the implementation of these bus controllers is outside the scope of this application report. An example of an MBus-initiated word-read transaction resulting in an HIP single read is shown in Figure 7. The L1 bridge is a slave to MBus and a master on the host interface. It responds to an MBus address as a selected slave and then arbitrates for HIP mastership. Likewise, an example of an MBus-initiated word-write transaction resulting in an HIP single write is shown in Figure 8. In this case, the MBus write is acknowledged before the HIP write is complete. If another write occurs to the L1 bridge before the HIP transaction is complete, an ~Bus relinquish and retry operation must be perfonned to back off the MBus master until the HIP transaction is finished. 3-203 I I IOn8 l250n8 I I MCLK MAS MAD (83-0) ----~~~------------------------------------~~~--------- \.J MRDY Miffi MEiiR Selected HBR HBG HBGACK _ _ _ _J~~_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ____ -----------\~~----~I ________ \~ J/~---------------------- \~--------------------~I HCLK HA (31-0) HD(31-0) HAS MS(1-0) HDS HIP TR_W TSIZ(1-0) DW84 TBST DSACK1 DSACKO i'STRi5V Figure 7. MBus-Word Read to HIF Single-Read Transaction 3-204 I 1250 ns I MCLK \J \J MAD (83-0) ~ ~ MRDY \J MAS \.J MRTY MERR Selected HBR· HBG HBGACK 1\ I \ I \ I \ HCLK HA (31-0) ( ( HD(31-0) HAS MS(1-0) \ TSIZ(1-0) ( DW84 I TBST I BSTRDY I I \ TR_W DSACKO X 01 \ HIP ) dO I \ X HDS DSACK1 ) Addr I 00 \ \ \ ) \ \ I I I FigureS. MBus-Word Write to HIF Single-Write Transaction 3-205 Summary Methods of connecting the SPARe MBus to the TIFB+ chipset's host interface are explored. Levell MBus transactions are mapped to HIF transaction by the bridge logic. The HIF transactions are then mapped to FB+ I/O transaction by the chipset. Direct- and paged-memory mapping are described. Techniques used to implement a 64-bit-only datapath and a datapath that has a dynamically configurable 32-bit MBus/32-bit HIF or 64-bit MBus/64-bit HIF also are described. Finally, the state-machine controller's task is summarized and example transactions directed by the controllers are shown. 3-206 1K x 9 x 2 Asynchronous FIFOs SN74ACT2235 and SN74ACT2236 First-In, First-Out Technology . Kam Kittrell Advanced System Logic - Semiconductor Group SCAA010A ~TEXAS INSTRUMENTS 3-207 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the Information being relied on is current. TI warrants performance of Its semiconductor products and related software to the specifications applicable at the time of sale In accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage C'Crltlcal Applications'1. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procadural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or othl\!r intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or l;Iervices might be or are used. Copyright © 1995, Texas Instruments Incorporated 3-208 Contents Title Page Introduction ............................................................................ 3-211 FIFO Control ........................................................................... 3-212 High-Frequency Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-212 Programmable Flags ................. ,................................................... 3-213 Output Drive ........................................................................... 3-214 Conclusion ............................................................................. 3-215 List ofmustrations Figure 1 Title SN74ACT2235 Block Diagram Page 3-211 2 Controlling the SN74ACT2235 Using a Clock, Write Enable, and Read Enable per System ....... 3-212 3 Read Operation When Cycle Time Is Less Than Access Time ............................... 3-213 4 AF/AEAFlag-ProgrammingLogicforFIFOA ........................................... 3-213 5 Programming AF/AEB Flag ofFIFOB From PortA ...................................... 3-214 6 SN74ACT2235 YOLP Measurement. .................................................. 3-215 3-209 3-210 Introduction Texas Instruments (11) designed the SN74ACf2235 to meet a variety of synchronous or asynchronous bidirectional applications. Two lK x 9 first-in, first-out (FIFO) memories are arranged in parallel to buffer data in opposite directions. Data ports also may exchange real-time data. Three-state control (GAB, GBA) and real-time/stored data select (SAB, SBA) match the popular '652 transceiver logic. Produced in 11's EPICTM CMOS process, the inputs accept TTI.-voltage levels. An option to the SN74ACT2235 is the SN74ACT2236, which has '646 transceiver control (DIR, G). The functional block diagram for the SN7 4ACT2235 is shown in Figure 1. 0(.7 ;- I UNCKA - - ' RSTA SBA,- - LDCKA, - GBA AO-AS - - ~ ~ =c Read Pointer FULLA EMPTYA AF/AEA HFA Flag Logic I I I Location 1 I 1024x9RAM Write Pointer Location 1024 G1 EN2 I 2V I ~ '-.- ~ 2:1 MUX I G1 I--EN2 I--2V I ~ - SAB GAB BO-BS 2:1 MUX Write Pointer h - UNCKB Read POinter ~~ - LDCKB Location 1024 1024x9RAM Location 1 HFB FULLB EMPTYB AF/AEB I I I Flag Logic I I - (')Figure 1. SN74ACT2235 Block Diagram EPIC is a trademark of Texas Instruments Incorporated. 3-211 FIFO. Control The SN74ACT2235 consists of two FIFO memories, FIFOA and FIFOB. Both FIFOs can be accessed from either port A or port B. Four control signal lines (GAB, GBA, SAB, and SBA) control the eight possible data flow paths through the device (these datapaths are illustrated in the device data sheet). Each FIFO has a load clock (LDCK) that writes data into memory and an unload clock (UNCK) that reads the data in the same order it was written. Both clocks are positive-edge-triggered and may operate asynchronously to one another. The first word loaded into an empty FIFO propagates directly to the outputs and the EMPTY flag switches high. EMPTY represents the valid state of data on the outputs (data is valid when EMPTY is high and invalid when EMPTY is low). EMPTY may be used to enable an UNCK pulse when it is synchronized with the bus that reads the data. FULL can qualify a LDCK pulse in the same way. Figure 2 is an example of an SN74ACf2235 interfacing two asynchronous systems. Each system provides a read enable, write enable, and free-running clock. Synchronization of a flag to the system clock is needed to use it as device-clock control. Although the flag's high-to-low transition is synchronous to the clock it enables, the low-to-high transition is asynchronous. The output of the latch qualifying this transition has the possibility of going metastable when bistable (setup and hold) conditions are not met. An output is metastable ifit lingers between the specified VOH and VOL levels. Two-stage synchronization of the flags reduces the probability of a metastable-induced failure. r------------.---r-- CLK2 Sye SYL--r--~----------~ CLKl SN74ACT2235 )-------+------i>LDCKA LDCKB LDCKA LDCKB11-_-_-_t1~+ ______...r"1:::;:::;__=~ ,_ UNCKB UNCKA .... ·--,~~;;;::==t----<~;::1GBA H SAB OAF 18-Blt Data ....' GAB t-;-::-~t===F;=~4 SBAI- FULLA DBF FULLB AO-A8 B~8 ~--~~~~ EM~~~~~--~ NOTE: Two devices are used for 18-bit width expansion. Figure 2. Controlling the SN74ACT2235 USing a Clock, Write Enable, and Read Enable per System High-Frequency Applications A unique feature of the SN74ACT2235 is that the UNCK cycle time may be less than the device access time. The SN74ACT2235-20 has a maximum LDCK and UNCK frequency of 50 MHz (20-ns cycle time) and a 25-ns maximum access time (lpd UNCKA or UNCKB to B bus or A bus). In a series of FIFO reads, the next access may be initiated before the present one is complete. The largest concern associated with this technique is the length of time data is assured as valid. Minimum access time from the rising edge of UNCK also may be viewed as minimum data hold time. Timing for this relationship is shown in Figure 3. Valid data time from the SN74ACT2235 over the commercial temperature range and ± 10% Vee is given by equation 1: tv 3-212 to + lpdmin -lpdmax (1) Data from an SN7 4ACT2235 operating at a 50-MHz clock frequency is valid for at least 7 ns. This allows a 4-ns setup and I-ns hold time with a 2-ns tolerance to the next device in the datapath. BO-BS W1 ,tpdmln-1 14 tpdmax For SN74ACT2235-20: tpd min = 12 ns, Ipd max" 25 ns, tv" 7 ns Figure 3. Read Operation When Cycle Time Is Less Than Access Time Programmable Flags Data is often transmitted in packets, where each packet is a specific number of bytes and must be delivered in an unbroken stream. A FIFO transmitting packeted data needs a flag that shows the number of bytes stored. This keeps from breaking the transmission of a packet due to an empty or full condition. The SN74ACT2235 has a programmable almost-full/almost-empty (AF/AB) flag for this application. The AFIABA offset value (X) and the AFIABB offset value (Y) are programmed separately. AFIABA is high when FIFOA contains X or fewer words or (1024 - X) or more words. It is low when FIFOA contains between (X + 1) and (1023 -X) words. AF/ABB functions in the same manner with its programmed value Y. The programmed or default value of 256 is chosen during a reset of each FIFO. Flag-programming logic is illustrated in Fi~ ~g the AFIAB flag value for each FIFO is done with the define-flag (OAF, DBF) inputs and resets (RSTA, RSTB). Define-flag inputs are negative-edge-triggered clocks that store input data to a register. IfDAF or DBF is low when the rising edge of RSTA or RSTB occurs, the registered value is used for the FIFO AF/AB flag. The flag uses the default value of 256 ifDAF orDBF is high during the rising edge of RSTA or RSTB. x>---t> RSTA---..., C MUX AO-AS 9 D 9 AF/AE Offset Default Value of 256 Value (X) Figure 4. AF/AEA Flag-Programming Logic for FIFOA 3-213 Programming both flag offset values from either port is possible using real-time select. Figure 5 is a timing example of programming APIAEB from port A. To program the APIAEB offset value (Y) from port A, the binary value for Y is on AO-A8, SAB is low, and GAB is high. With this COnitguration, the port-A aata appears on the inputs ofFlFOB and a falling edge of DBF stores the Y value. GBA ------------------------------------------------------ 0 SAB GAB AO-A8 "'---- / _ __ J ~~_________B_I_na_~_~_a_IU_e_Y________~·~ BO-B8 _______~_____B_I_na_~_~_a_IU_e_Y_____J~ Figure 5. Programming AF I AEB Flag of FIFOB From Port A Output Drive Charging and discharging the load of a bus with acceptable speed requires high device-output drive. The I/O ports of the SN74ACT2235 provide 16-mA IOL and 8-mA IoH for this task. Most memory devices have low drive capability and require buffers to interface a bus. They do not use larger transistors that support high current because the rate of change of current with respect to time (di/dt) increases. When several transistors switch simultaneously, the rate of change of current through ground and Vee lines multiplies. Voltage transients on the power lines are given by equation 2: V--L(dildt) (2) Where: L - inductance of the bond wire and package lead The SN7 4ACT2235 provides a two-fold solution to allow high-output current capability with low noise. One solution is to reduce inductance of ground and Vee lines. The SN74ACT2235 has four GND and two Vee pins in parallel. The resulting ground inductance is about 1/4 that of a single connection and divides Vee inductance in half. 3-214 Reducing di/dt per output transistor is another way to minimize voltage transients. TI's patented output-edge-control (OECTM) design divides a large transistor into smaller segments that turn on in series and turn off simultaneously. OECTM lowers di/dt, maintains a quick voltage transition through threshold, and avoids the high power consumed when gradually turned off .1 The result of a VOlP test on the SN74ACT2235 is shown in Figure 6. VOlP is a measurement of ground-voltage noise when all outputs of a bus are switched from high to low. Eight of nine outputs of a bus are switched, and the peak-voltage rise of the steady-state low output is measured. Maximum ground-voltage rise is only 700 mV. The output fall time is less than 3 ns with a 50-pF load. 5 r- - BO-B7 I ....... VCC=5V TA=25OC CL=5OpF RL=5000 \ 4 \ 3 ~ I 2 \ !!J ~ 1 0.7 B8 o -1 Ij' ,\ ~ \J( ;' o 5 10 15 20 25 1'\ - ;" 30 35 40 45 50 SwItching Time - n8 NOTE: Eight bus outputs switching, one remains low Figure 6. SN74ACT2235 VOlP Measurement Conclusion The SN74ACT2235 and SN74ACT2236 provide several advantages for high-speed asynchronous bus interface. Simple control logic offers great design flexibility. Programmable flags may be used for data flow optimization. High-output drive for bus leading is balanced with noise reduction through package and circuit design. 1 Advanced CMOS Logic Designer's Handbook, pages 3-1 through 3-12. OEC is a trademark of Texas Instruments Incorporated. 3-215 3-216 64-Byte FIFOs SN~ALS2232AandSN~ALS2233A First-In, First-Out Technology Kam Kittrell Advanced System Logic - Semiconductor Group ~1ExAs INS1RUMENTS SCAA023A 3-217 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to Its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale In accordance with TI's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device Is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applicatlons'? TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such epplications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or Infringement of patents or services described herein. Nor does TI warrant or represent that any license, either exprass or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 3-218 Contents Title Page Introduction ............................................................................ 3-221 Clocks .................................................................................. 3-221 Flags .................................................................................. 3-222 Noise Control ........................................................................... 3-222 Applications ............................................................................ Using the FULL and EMPTY Flags ........................................................ Width Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Bus Conversion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-224 3-224 3-224 3-225 List of mustrations Figure 1 2 3 4 5 6 7 Title Page FIFO Read Control With READ/OE Logic ............................................. 'ALS2232A and 'ALS2233A UNCK Control ............................................ Noise From a GND Pad to a 0-V Plane ................................................. 'ALS2232A and 'ALS2233A Clock Input Circuit ........................................ Clock Generation With Two-Stage Synchronization of FULL and EMPTY .................... Width Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Bus-Folding Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3-221 3-222 3-223 3-223 3-224 3-224 3-225 3-219 3-220 Introduction First-in, first-out (FIFO) memories are irreplaceable bus logic when interfacing two asynchronous systems. The Texas Instruments SN74ALS2232A 64 x 18 and SN74ALS2233A 64 x 9 FIFOs are ideal solutions for many high-speed buffering needs. These bipolar devices, produced in IMPACT-X technology, come in a 28-pin PLCC and a 24-pin DIP for the 'ALS2232A and 28-pin PLCC and DIP for the 'ALS2233A. Data is stored in a dual-port SRAM that supports transfer rates up to 40 MHz and maximum access times of 27 ns. Reads are accomplished independent of writes with separate internal addressing. Clocks The read enables of many FIFOs also control the active/high~impedance state of the data outputs. FIFOs using this logic must have a read-enable pulse long enough to include an access and hold time before it disables the outputs, which makes high-frequency clock design difficult (see Figures 1 and 2). READ/OE Data Out I I '------' I ~ta~ FIFO Data Qual _______+i__ tau~ ta tsu th a I J~~-----.~~th access time maximum setup time minimum hold time minimum Figure 1. FIFO Read Control With READ/OE Logic 3-221 UNCK OE Data Out (QO-Q7,Qa) FIFO Data Qual twl 12 ns tw2 13n8 _______...Ix\,,______ x W_O_rd_X _ _ _ _ _..J _ _----..J1 WordX+ 1 \_---- Figure 2. 'ALS2232A and 'ALS2233A UNCK Control Texas Instruments has allowed clock generation to be simple for its FIFOs. Load-clock (LOCK) and unload-clock (UNCK) inputs are edge triggered, which makes the device more suited for use as a buffer in a data-transmission path. Fewer constraints are placed on a design with edge-triggered clocks since duty cycles are permitted to vary greatly (see Figure 2). A separate output-enable (OE) input is provided for applications requiring 3-state buses. The LOCK and UNCK independently control all data transfers into and out of memory and can be synchronous or asynchronous. The first word loaded into an empty FIFO propagates directly to the data outputs. Any UNCK pulses that occur during an empty condition are ignored, while any LOCK pulses that occur during a full condition are ignored. Flags The'ALS2232A has two flags to indicate boundary conditions of the memory: EMPTY and FULL. In addition to these, the 'ALS2233A has the aImost-full/aImost-empty (AF/AE) and half-full (HF) flags. AF/AE is high when memory contains less than nine words or more than 55 words. To distinguish between an almost-full and an almost-empty state, HF is high when memory contains more than 31 words. The extra flags are provided for applications wherein full and empty conditions should be avoided. Noise Control Ground bounce is a result of current surges produced by output switching. Bond wire, lead, and board inductance cause .intemal ground levels to fluctuate from the current surge (rise above, then dip below 0 V). Extreme ground noise that causes input levels to cross the transition threshold may be detected as clock pulses by high-speed devices. Worst-case conditions for ground bounce are high Vee and outputs switching simultaneously from high to low. The'ALS2232A and 'ALS2233A have package-centered Vee and GND pins to combat ground bounce. The shortened bond wire and lead distance reduce package inductance from conventional comer-pin configurations. Figure 3 shows a large voltage transient that might be measured on the ground pad of any device referenced to a 0-V plane. An input at a steady high or steady low level is likely to cross the transition threshold as a result. 3-222 1.5 > 0.5 I S c "0 ID 0 'Q C e" -4.5 CJ " - 1\ \ V \ -1 -1.5 Time -10 na/Dlv Figure 3. Noise From a GND Pad to a O-V Plane Figure 4 is the equivalent circuit of the clock inputs for the 'ALS2232A and 'ALS2233A. This modified RS flip-flop is more likely to pass a very quick high pulse (0 to 5 ns) caused by noise when it is in the steady low state than it is to pass a very quick low pulse when in the steady high state. The clock input one-shot structure is immune to a very quick (0 to 5 ns) low pulse when in the steady high state. For improved noise protection. LDCK and UNCK signals may be generated as inactive high with a low pulse generated for clocking. UNCKor LDCK CLKPulse Figure 4. 'ALS2232A and 'ALS2233A Clock Input Circuit 3-223 AppliCations Using the FULL and EMPTY Flags The FULL and EMPTY flags are provided to indicate that the FIFO is at one of its boundaries. An example of how to qualify these flags as enables for the device clocks is shown in Figure 5. Without the flip-flop qualification, a flag can cause the asynchronous generation of a clock. The two-stage synchronization alternative shown reduces the chances of a metastable output from one-stage synchronization. SYS1 Clock - - - - - - . . - - - - , ,--------<.-----SYS2 Clock SN74ALS2232A or SN74ALS2233A Write Enable - - - - - , ~---Resd Enable LOCK UNCK o o 0 ~ 00-07,08 QO-Q7,Q8 Figure 5. Clock Generation With lWo-Stage Synchronization of FULL and EMPTY Width Expansion Several'ALS2232A devices can be used in width expansion to handle datapaths with several bytes. The ' ALS2232A can likewise be expanded and also pass parity for each byte. No special control logic is needed to implement this application (see Figure 6). SN74ALS2232A or SN74ALS2233A LOCK 2 Bytes Data In - 1 Byte , FULL UNCK EMPTY - 1 Byte , SN74ALS2232A or SN74ALS2233A LOCK .-0= FULL UNCK EMPTY 1 Byte 1 Byte Figure 6. Width Expansion 3-224 D- 2 Bytes Bus Conversion Systems frequently require that data be converted from l-byte buses to multiple-byte buses operating asynchronously. Figure 7 shows an IS-bit bus folded into a 9-bit bus using the I ALS2233A. The control logic can be implemented with a TIBPAL20R4. SYSCLK ,~-------------- EMPTYA _ _ _..£.../.,-""2",2,,,2.,,7 '......_---- EMPTYB _ _ _..//",,,,%,,,%,,,,% .../',,,,1 UNCKA ''---_/ ,'----.,,/ ' .... __J/ UNCKB ' ......_...J/ OEA _ _ _ _ _ _ _......,-'~...._ _---J~'-_ _ _ _ _ _ _ __ _ OEB ____________________--'~'--___J~...._ _ _ _ __ SN74ALS2232A SYS CLK or -+-----------------4....... SN74ALS2233A .--------------+-------r-------------i>UNCK .------~f-----------___I LDCK~----.-- OE .-------1 EMPTY SN74ALS2232A or SN74ALS2233A L---------------------------------------+-~--t>UNCK LDCK~--+_~ .---------------+--1---1 OE L-+_--1 EMPTY 1 Byte - ...~~-I Figure 7. Bus-Folding Logic 3-225 Table 1. Terminal Functions FUNCTION Control inputs TERMINAL NAME DEFINITION LOCK Load clock; rising-edge clock. Writes data into the FIFO; updates the flags. UNCK Unload clock; rislng-edge clock. Reads data out of FIFO; updates the flags. OE Output enable. Controls the actlvelhlgh-Impedance state of the data outputs. A high level on OE selects the active state; low selects high impedance. RESET Reset. Low level resets the read and write pointers to the first location and sets the flag status to empty. The FIFO must be reset after power up. EMPTY Empty flag. tpLH transitions are controlled by LOCK. tpHL transitions are controlled by UNCK or RESET. FIFO read pointers are unaffected by UNCK when EMPTY is low. FULL Full flag. tpHL transitions are controlled by LOCK. tpLH trensltions are controlled by UNCK or RESEi. FIFO memory and write pointers are unaffected by LOCK when FULL is low. . AF/AE Aimost-fuil/almost-empty flag: high level when FIFO is eight locations from a full or empty condition (FIFO contains less than nine words or more than 55 words) HF Half-full flag. tPLH transitions are controlled by LOCK. tpHL transitions are controlled by UNCK or RESET. HF Is at a high level when the FIFO contains more than 31 words. Status-flag outputs Data 3-,-226 00-07 (SN74ALS2232A) 00-08 (SN74ALS2233A) Data Inputs: data latched by LOCK Into marnory QO-Q7 (SN74ALS2232A) QO-OO (SN74ALS2233A) Data outputs: data read from FIFO 4-1 "i ... oo ::s en _. Q. CD ...o ;_. ::s en 4-2 Introduction Page Key Power-Dissipation Equations for ACT FIFOs . . . . . . . . . . . . . . . . . . • . . • . . • . . . . . . . . . . . . .. 4-5 Key Power-Dissipation Equations for ABT FIFOs ....................................... 4-6 Introduction This section of the FIFO Designer's Handbook shows typical power characteristics in the form of active supply current versus frequency plots for each of Texas Instruments (TI's) advanced FIFOs. In some cases, plots of idle supply current (IeeI) versus frequency are also included. In addition to the current versus frequency data, sample calculations of power dissipation are included in this section for a representative set of FIFOs. The application report entitled Power Dissipation Calculations for TI FIFO Products serves as a guideline for the example calculations. Equations for dynamic, quiescent, and total power dissipation for both ACT and ABT FIFOs are given, along with the idle lee versus frequency plots required for the sample calculations in this section. These and other key equations that are used in the calculation of power dissipation are repeated below and numbered for convenience. Key Power-Dissipation Equations for ACT FIFOs Quiescent Pq - vee x [leeI + (NTIL x Alec x DCYIH)] (1) Where: - fclock x pF(clock) - supply current when FIFO is idle - clock switching frequency fclock clock switching power factor (slope of lee versus fclock curve) pF(clock) = number of inputs driven by TIL levels NTIL - increase in supply current for each input at a TIL high level (see data sheet) Alec - percent of TIL signals at a high level of 3.4 Y DCVIH leeI Dynamic Power Pd - Vee x [leef + (NTIL x Alee x DCvrn)] + l:(CL x Vee2 x fo) (2) Where: Vee leef NTIL Alee DCVIH CL fo = supply voltage -, supply current when FIFO is transferring data - number of inputs driven by TIL levels - increase in supply current for each input at a TIL high level (see data sheet) = percent of TIL signals at a high level of 3.4 Y - load capacitance - output switching frequency - [112 (since maximum data rate is 112 clock frequency) x (fraction of outputs switching at a given time) x (frequency of the slowest of the port clocks)] Tota/Power Pt - Pd(DCd) + Pq(1-DCd) Where: Pd Pq DCd dynamic power dissipation quiescent power dissipation - duty cycle = (3) Key Power-Dissipation Equations for ABT FIFOs Quiescent Power Pq - Vcc x [DCENX (NHx ICCHINT+ NLx IcCrfNT) + (l-DCEN) x Iccz + Iccd (4) Where: vCC IcCI fclock pF(clock) DCEN IcCH ICCL ICCZ NL NH NT - supply voltage fclock x pF(clock) - supply current when FIFO is idle clock switching frequency clock switching power factor (slope of ICC versus fclock curve) percent duty cycle enabled power supply current when outputs are in the high state (see data sheet) power supply current when outputs are in the low state (see data sheet) = power supply current when outputs are in the high-impedance state (see data sheet) number of outputs in low state - number of outputs in high state - total number of outputs Dynamic Power Pq - VCC XIcCf+ I: [VCC xCLx(VOH- Vou xfo] (5) Where: vCC ICCf fo VOH VOL CL supply voltage supply current when FIFO is transferring data, active current output switching frequency - [112 (since maximum data rate is 112 clock frequency) x (fraction of outputs switching at a given time) x (frequency of the slowest of the port clocks)] - output voltage in high state output voltage in low state = load capacitance = = = Total Power Pt = Pd(DCd) + Pq(1-DC~ (6) Where: Pd Pq DCd - dynamic power dissipation quiescent power dissipation duty cycle In all of the following power calculation examples, the FIFO inputs are being driven by a 1TL device. In the case where the inputs are driven by a CMOS device, the MCC term in the power-dissipation equation equals zero and can be ignored. In each calculation, the number of inputs (or outputs) equal to the width of the FIFO are assumed to be switching in order to provide a worst-case solution for the conditions being as~umed. Figure 1 contains the results of the sample power-dissipation calculations. The plotis divided up into results for the ACf FIFOs and results for the ABT FIFOs. Within each of these technologies, the conditions for each sample calculation have been chosen to provide (as much as possible) an apples-to-apples comparison (see individual calculations in this section). As previously noted, the number of bits switching at a given time is assumed to be equal to the width of the FIFO under consideration. In addition to the results of the calculations, the maximum power-dissipation capability of the packages associated with these devices at 70°C with no air flow is also plotted. In every case, the dissipation capability of the package significantly exceeds the power being dissipated by the device, even under the rather severe conditions stated above. The specific packages and their associated dissipation capabilities are listed in Figure 1. 2000 .. c 0 ! ACT ~I ABT - 1500 r- ICL E • :::I I/) c ~ r1000 iICL j c - - 0 • 500 I ICL 0_ ...L JI Ju ACT3641 (36 Bits) ACT7803 (18 Bits) D - ACT7807 (9 Bits) .1 i. ACT2229 (1 Bit) PACKAGE A ...L - ABT3613 (36 Bits) I...L Device Dlsslpatlonl Consumption PackageA Dissipation Capability PackageS Dissipation Capability '-- ABT7819 (18 Bits) PACKAGE B NOMENCLATURE DISSIPATION CAPABILITY (mW)t NOMENCLATURE DISSIPATION CAPABILITY (mW)t ACT3641 120-pin TOFP (PCB) 1606 132-pin POFP (PO) 1610 ACT7803 56-pln SSOP (OL) 851 ACT7807 64-pin TOFP (PAG) 1121 44-pin PLCC (FN) 1244 ACT2229 24-pin SOIC (OW) 909 28-pin SOIC (OW) 978 ABT3613 12o-pin TOFP (PCB) 1606 132-pin POFP (PO) 1610 ABT7819 8O-pin TOFP (PN) 911 8o-pin POFP (PH) 954 DEVICE t The conditions of 70·C and no air flow are assumed. Figure 1. Comparison of Device and Package Power-Dissipation/Consumption capabilities 4-7 4-8 Application Report 4-10 Power-Dissipation Calculations for TI FIFO Products Navid Madani Advanced System Logic - Semiconductor Group ~TEXAS INSTRUMENTS SCAA013A 4-11 · IMPORTANT NOTICE Texas Instruments (TIl reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises Its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthetime of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications',. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or Implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright + Pq (1- DCd) - 1047.1625 mW x 1/3 + 202.436 mW x 213 - 484.0115 mW Therefore, total power is approximately: IT - 484mW The SN74ACT3632 is available in 120-pin TQFP and 132-pin PQFP packages (refer to Appendix B for the maximum power curve calculated for reliability purposes). At maximum ambient temperature (70°C) and no air flow for 132-pin PQFP and 120-pin TQFP packages, the maximum power that the packages can dissipate to free air is 1610 mW and 1606 mW, respectively. The SN74ACT3632, in this example, meets the reliability requirement since 484 mW is much less than 1610 mWor 1606 mW. 4-19 Quiescent or Nonswltchlng Power Dissipation In ABT FIFO Devices (BICMOS FIFO products) Unlike CMOS devices that have a single value for lec, BiCMOS devices have varying static current levels depending on the state of the output (ICCL, IeCR, ICcz). Quiescent power includes the power consumed while outputs are active, the power consumed when outputs are disabled, and the power consumed by the switching clocks. The design of the BiCMOS inputs is such that when a 1TL high level is applied at the input, it does not increase the current; therefore, the alec term (NTIL x alec x DCd> is excluded from the following equation (from Texas Instruments ABTAdvanced BieMOS Technology Data Book, 1993, literature # SCBDOO2A) in calculating power for ABT FIFO products. Pq - VCC x [DCEN x (NR x ICCRINT + NL x IeCLINT) + (1 - DCEN)IeCZ + lecd Where: VCC x [DCEN x (NR x IecwNT + NL x IeCIJNT)] - power consumed while outputs are active VCC x (1 - DCEN)IeCZ - power consumed when outputs are disabled VccXICCI - power consumed by switching clocks VCC - supply voltage IeCI - fclock x pF - supply current when FIFO is idle (the clocks are running but no data is written to or read from the FIFO) (not in the data sheet) fclock - clock switching frequency pF - clock switching power factor, the slope oflec versus fclock curve (see data sheet) % duty cycle enabled DCEN power-supply current when outputs are in high state (see data sheet) IeCR power-supply current when outputs are in low state (see data sheet) ICCL power-supply current when outputs are in high-impedance state (see data sheet) lecz number of outputs in low state NL number of outputs in high state NR total number of outputs NT Dynamic or SWitching Power Dissipation In ABT FIFO Devices (BICMOS FIFO products) For most applications, dynamic power accounts for most of the total power consumption of a BiCMOS device. Dynamic power consumption includes the device switching power consumed without the load, as well as the power consumed due to the capacitive load. Pd - VCC x IeCf+ r[VCC x CL x (VOR- VOL) x fol Where: VCC x ICCf - device switching power without the load r[Vcc x CL x (VOR - VOL) x fo 1- power consumed due to the output switching frequency and the load capacitance VCC - supply voltage IeCf - supply current when FIFO is transferring data, active current (see active ICC versus frequency plot in data sheet) fo - output switching frequency VOR output voltage in high state VOL output voltage in low state CL load capacitance ' 4-20 Example 2 This example shows how to calculate the power dissipation for an SN7 4ABT3614 bidirectional FIFO used in a system under the following conditions: Data input l~ (AO-A35) and the control signals (CLKA, CLKB, CSA, CSB, WiRA, W/RB. ENA, ENB, MBA, BE, RST, SIZO, SIZl, ODDIEVEN, SWO, SWl, PGA, PGB) are driven by a CMOS device. Only 213 of the inputs (or outputs) are switching at a given time. The output of the FIFO is fed to a memory device. The port-B clock rate is 33.3 MHz and the port-A clock rate is 40 MHz. The SN74ABT3614 is only used 60% of the time by the system. The load capacitance of each output is about 50 pF, and the supply voltage is set to 5 V. During the FIFO active period, the bus is enabled 75% of the time. When the bus is enabled, the output is in the high state 80% of the time. The SN74ABT3614 is used equally in both directions. Figures 5 through 7 and included information are needed to calculate power dissipation: 400 , ,!. , -' ' fdata = 1/2 fclock -TA=25°C CL=OpF 350 VCC=5.5V / V 300 VCC=5Z 0( E I 8 I I ~ 250 / 200 / :/ V / ./ . / :Y V~ ~ ~ /VCC=4.5V 150 100 ,~ 50 o o 10 ~ '? V 20 30 40 50 60 70 60 fclock - Clock Frequency - MHz Figure 5. SN74ABT3614 Active Icc With eLKA and eLKB Switching, Simultaneous Read/Write and eLKB as Data Output 4-21 40 c I _ ~ Vee=4.5V 36 . . . Vee=5V -'- ...... Vee=5.5V 30 TA = 250 ~r- E I 8 25 i 20 ~ 15 .....-' 10 ---' ~ ~ ~ ----- ------ ~ ~~~ ~~~ r~ I ---- ~ ~ L Slope = 0.25 5r-----~------_+------_r------4_------+_------~----~ o~----~------~------~------~------~------~----~ o 10 20 30 40 50 60 70 f - Frequency - MHz Figure 6. SN74ABT3614 Idle ICC With ClKA Switching, Other Inputs at 0 or VCC - 0.2 V and Outputs Disconnected 40r-----~-------.-------.------.-------.-------.------, 1 I = 35 ~ vec 4.5 V --+_------~----~f-----__+----::::;;oo.....,."-------__i . . . Vee=5V 30 ...... Vee = 5.5 V TA =250 --+-------,-+_----___l6......,e:~__+--__=_....~----__i 25r-------+------_+----~~"-----~~~----+_~~~~----~ 8 i I ~ 15. 10r-----~------_+---'----_r------4_------+_------~----~ 5r-----~------_+------_r------4_------+_------+_----___l O~----~------~------~------~------~------~----~ o 10 20 30 40 50 f - Frequency - MHz Figure 7. SN74ABT3614 Idle ICC With elKB Switching, Other Inputs at 0 or VCC - 0.2 V and Outputs Disconnected 4-22 60 70 The following parametric values are needed to calculate power dissipation: Vcc VOH VOL DCd 1 - DCd fclockA fclockB pF(A) pF(B) DCEN ICC! ICCf ICCH = ICCL Iccz NL/NT NWNT CL fo - supply voltage - 5 V VCC - 1.3 V 0.3V % time FIFO is switching - 0.6 % time FIFO is not switching - 0.4 clock switching frequency of port A - 40 MHz clock switching frequency of port B - 33.3 MHz clock-A switching power factor, the slope of Icc versus fclock curve (see data sheet) = 0.25 clock-B switching power factor, the slope of Icc versus fclock curve (see data sheet) - 0.28 % duty cycle enabled - 0.75 [fclockA x pF(Al + [fclockB x pF(B) - (40 x 0.25) + (33.3 x 0.28) - 19.32 mA idle supply current when FIFO is transferring data - 136.26 mA active supply current when outputs are in high state (see data sheet) = 30 mA power supply current when outputs are in low state (see data sheet) = 130 mA power supply current when outputs are in high-impedance state (see data sheet) - 30 mA ratio of number of outputs in low state to total number of outputs - 0.2 ratio of number of outputs in high state to total number of outputs - 0.8 load capacitance - 50 pF 112 (since maximum data rate is 112 clock frequency) x 2/3 (since 2/3 of the outputs are switching at a given time) x 33.3 MHz (slowest of the the two clock frequencies, fclockA or fclockB) - 11.1 MHz Solution PT - PT (from A to B) + PT (from B to A) - Pq (from A to B) + Pq (from B to A) + Pd (from A to B) + Pd (from B to A) - Pq +Pd Where Pq and Pd include power from A to B and from B to A directions. Quiescent Power Pq - Vee x [DCENx (NHx IcewNT+NLx IceIlNT) + (I-DCEN)Icez+ Icc!l - 5 V x [0.75 x (0.8 x 30 mA + 0.2 x 130 mA) + (1 - 0.75) 30 mA + 19.32 mAl - 5 V x [37.5 mA + 7.5 mA + 19.32 mAl- 321.6 mW Dynamic Power Pd - Vec x Ieef+ 'L[Vcc xCLx (VOH- VOL) xfol - 5 V x 136.25 mA + 'L[5 x 50 pF x (5 V - 1.3 - 0.3 V) x (11.1 MHz)l - 681.25 mW + (36 x 9.44 mW) - 681.25 mW + 339.66 mW = 1020.91 mW Total Power PT - Pd (DCd) +Pq (l-DCd) - 1020.91 mW x 0.6 + 321.6 x 0.4 -741.19 mW Therefore, total power is approximately: PT - 741.19 mW The SN74ABT3614 is available in 120-pin TQFP and 132-pin PQFP packages (refer to Appendix B for the maximum power curve calculated for reliability purposes). At maximum ambient temperature (70°C) and no air flow for the 132-pin PQFP and 120-pin TQFP packages, the maximum power that the package can dissipate to free air is 1610 mW and 1606 mW, respectively. The SN74ABT3614, in this example, meets the reliability requirement since 741.19 mW is much less than 1610 mW or 1606 mW. 4-23 Summary Power-dissipation calculations are essential to meet the design requirements related to the chip temperature and the system power. In this application report, a simple method of calculating power is provided to assist the design engineer with power-dissipation calculations for 11 CMOS and BiCMOS FIFO products. Total power includes quiescent power and dynamic power. For most applications using CMOS and BiCMOS FIFOs, dynamic power accounts for most of the power requirement. Examples of power-dissipation calculations are provided to show the practical use of this application report. In each example, the reliability of the chip was tested against the absolute maximum power dissipation in free air. For example, the total calculated power consumption for the SN74ACT3632 and SN74ABT3614 examples resulted in 484 mW and 741 mW, respectively. These values are much less than the maximum power dissipation of the 120-pin TQFP (1606 mW) or 132-pinPQFP (1610 mW) packages in still air. Icc versus frequency curves are provided in Appendix A. These graphs assist the design engineer in the search for the FIFO device with the minimum power consumption. After total power is calculated for a system, the design engineer can ensure that this value does not exceed the maximum power capability of the package type. The table of maximum power versus ambient temperature for different package options are included in Appendix B. Acknowledgements The author would like to express her appreciation to Kam Kittrell for his editorial comments and AI Sawyer for his assistance in providing the Icc versus frequency curves in Appendix A. 4-24 Appendix A Graphs of ICC Versus Frequency The following information is provided to assist the designer with the power-consumption calculations. Graphs of Icc versus frequency are shown for the SN74ACT7803, SN74ACT7811, SN74ACT3641, SN74ACT7807, and SN74ABT7819. While the FIFOs were idle, data was taken on five units on an automatic test machine (lIP 82(00). Five readings were taken for each frequency and the average was used to plot the graphs. The tests were done by setting VIL and VIH as shown below: VIL - OV VIH - Vee- 0.2V For each of the FIFOs, two graphs are provided for idle Icc. One graph shows the Icc versus frequency when WRTCLK is running, whereas the other graph shows the Icc versus frequency when RDCLK. is running. The slope of the 5-V supply voltage curve is calculated for both graphs and the largest of the two slopes is used as the power factor for power calculations. The slopes of the lceI versus frequency graphs in the tests performed were 0.09 (SN74ACT7807), 0.12 (SN74ACT7803, SN74ACT7811), 0.2 (SN74ACT3641), and 0.28 (SN74ABT3614); therefore, if the slope of the lceI versus frequency plot is not readily available, it is appropriate to estimate the slope as 0.2. 6 ....... VCC=4.6V ..... Vce=6V -IP- Vee 6.5 V TA=2S0 6 1 I = 4 0 E .!! 3! 3 I B 2 o~----~------~----~------~----~------~----~----~ o 5 10 15 20 25 30 35 40 f - Frequency - MHz Figure A-1. SN74ACT7811 Idle Icc With RDCLK or WRTCLK Switching 4-25 9 ..... VCC=4.6V ..... VCC=5V ...... VCC=5.5V TA =25· 8 7 1 8 8 5 i 4 B 3 I I 2 f - Frequency - MHz Figure A-2. SN74ACT7803 Idle Icc With RDCLK or WRTCLK Switching 18 ..... VCC=4.5V ..... VCC=5V ...... VCC=5.5V TA = 25· 18 14 1 12 8 10 I i 8 B 8 I 4 2 O~~ o ____L -______L -______L -______L -______L -______L -____ ro m ~ ~ ~ f - Frequency - MHz Figure A-3. SN74ACT3641 Idle Icc With CLKA Switching 4-26 ~ ~ ~ 14.----,----,----,----,----,----,-----. ...... Vee=4.5V ~ Vee = 15 V ~ Vee=5.5V TA = 25° 12 1 10 --+----+----+----+-----bJfl""""'----:I. I 8 i I ~ 8~---~---~---~--~~~~~F-~~-~--~ 8~---~---~--~~~~~~~--~---~--~ 4~---~--_2~~~~r----~---~---~--~ 2~--~~~~--~------r_----_+-------r------+_--~~ 10 20 30 40 eo 50 70 f - Frequency - MHz Figure A-4. SN74ACT3641 Idle Icc With ClKB Switching 2eo.----,----,----,----,----,----,---~ ...... Vcc =4.5 V ~ Vee=5V Vee = 5.5 V --+----+----~--_i----:::;JfI""""_+--:::;o,,-;;_i TA=25° 200 ~ 1I B 1eor------r----~r_----;_----~~--~~~~-r----__i I 100r----+----+--~~~~~~+----+----+---__i I = Slope 2.94 g 50r----~~~JfI""""T----+----+----+--~-+---__i o~----~------~------~----~------~------~----~ o 10 20 30 40 50 eo 70 f - Frequency - MHz Figure A-5. SN74ACT3641 Active Icc With ClKA and ClKB Switching, Simultaneous ReadlWrite and ClKB as Data Output 4-27 6 ...... VCC=4.5V ~ VCC=5V -lAP- Vee = 5.5 V TA = 26° 6 ~ 4 I 8 J! 3! 3 I ~ 2 O~----~------~------~------~------~------L-----~ o 10 20 30 40 50 60 70 f - Frequency - MHz Figure A-B. SN74ACT7807IdleJce With WRTCLK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected 7 ...... VCC=4.5V ~ Vee=5V -lAP- Vee = 6.5 V TA = 25° 6 ~ 5 8 4 I .!! 3! I 3 ~ 2 O~----~~----~------~------~------~------~----~ o ro ~ 30 40 50 f - Frequency - MHz Figure A-7. SN74ACT7807 Idle lee With RDCLK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnectea 4-28 60 N 200 I 180 - _ ./ Vee=4.5V V".....-I ~ Vee=5V 180 1 I ~ - -A- Vee = 5.5 V i l5 80 ~ ~~ /"~ ~" f-~~ ....d ~ ~ 80 I 40 20 o ....04 ./.~ ~ 120 100 2 ./".J~ 140 f- TA=25° o 10 20 30 40 Slope = 2.40 50 70 80 f - Frequency - MHz Figure A-8. SN74ACT7807 Active Icc With Simultaneous ReadIWrlte 200 I I I I Vee = 4.15 V 180 I- _ ........... r ~ Vee=5V 180 1 140 t- -A- Vee = 5.5 V t- TA = 260 ./V k-"""'" ./ I ~ Ic 120 100 I 80 B 80 ..-I~ ~ ~ ...-! ~ ..... ~ f' ...... ~ 40 .... ~ r--- ,. I"""'" ~~ ........ ......... r'"' ,... ~ ~ ......... ........ V V ........~ ~ ~ t-::: ~ ..... Ir---"" ~ ~ ~ ,- 20 o 10 15 20 25 30 315 40 45 50 55 80 65 70 75 80 65 90 95 100 f - Frequency - MHz Figure A-9. SN74ABT7819 Active Icc Versus Frequency 4-29 Appendix B Maximum Power Dissipation for Different Package Types For reliability purposes, maximum power is calculated for each package option using the following equation: Chip temperature - Power x 9lA + TA Where: Chip temperature - lS00C (absolute maximum chip temperature) 9lA - thermal characteristics of a package (known) TA - ambient temperature (known) Table 1 lists maximum power dissipation by package type for ambient temperature from 25°C to 90°C. Table B-1. Maximum Power Dissipation (mW) for Packaged FIFOs AMBIENT TEMPERATURE (OC) 25 30 35 40 45 50 65 60 1,911 1,811 65 70 75 60 65 90 PQl32PQFP 2,515 2,414 2,314 2,213 2,113 2,012 PCBl20TQFP 2,510 2,410 2,309 2,209 2,108 2,008 1,908 1,601' 1,707 1,505 1,505 1,406 1,305 1,205 1,424 1,367 1,310 1,253 1,196 1,139 1,082 1,025 968 797 854 654 740 683 PN80TQFP PM64TQFP 1,351 PH80 PQFP 1,490 1,430 1,371 D156SS0P 1,330 1,277 1,223 1,170 1,117 1,064 1,011 1,528 1,467 1,406 1,345 1,284 1,222 1,161 DW28S0IC DW24S0lC 4-30 1,297 1,243 1,189 1,135 1,081 1,311 1,251 1,027 973 1,710 1,509 1,509 1,408 1,308 1,207 919 1,192 1,132 1,073 1,013 904 1,100 1,039 1,420 1,364 1,307 1,250 1,193 1,136 1,060 1,023 966 957 811 811 757 703 894 894 834 775 649 715 798 798 745 691 917 852 917 656 795 638 733 852 795 739 682 Single-Bit FIFOs Page Sample Power-Dissipation Calculations for SN74ACT2229 ............................. Power Characteristics ....................... ,..................................... SN74ACT2226 and SN74ACT2228 Single FIFO Supply Current Versus Clock Frequency .. SN74ACT2227 and SN74ACT2229 Single FIFO Supply Current Versus Clock Frequency .. 4-33 4-34 4-34 4-34 . 4-31 Sample Power-Dissipation Calculations for SN74ACT2229 This example shows power-dissipation calculations for the SN74ACT2229 FIFO used in a system under the conditions below. On board the SN74ACT2229 are two separate and independent FIFOs. The calculation below is for only one FIFO: FIFOI is assumed active and FIF02 is assumed idle. Assumed Conditions • • • • • • Data input line (10) is driven by a TTL device. Control signals (1 WRTCLK, IRDCLK, IRESET, 1WRTEN, lOE, IRDEN) are driven by a CMOS device. The input (or output) is switching at a given time. The RDCLK clock rate is 33.3 MHz and the WRTCLK clock rate is 40 MHz. The SN74ACT2229 is used about 113 of the time by the system. The load capacitance on each output is about 30 pF, and the supply voltage is 5 V. The following parametric values are needed to calculate power dissipation: Vcc DCd fWRTCLK fRDCLK pF ICCI NTTL ICCp AlCC DCVIH CL fo - 5 V (assumed condition) 113 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.2 mAlMHz (worst-case approximation) (fWRTCLK + fRDCLK) x pF - 14.7 mA (calculated, see equation 2 in the introduction to Section 4) 1 (assumed condition) 20 mA @ 33.3 MHz (from active ICC versus frequency plot) 1 mA (see data sheet) 1 (assumed condition) 30 pF (assumed condition) (0.5) x (1) x (33.3 MHz) - 16.7 MHz (calculated, see equation 4 in the introduction to Section 4) Solution Quiescent Power PT - Vcc x [lccI + (NTTL x Alec x DCvm)] 5Vx [14.7 mA + (1 x 1 mAx 1)] 78.3mW (1) Dynamic Power Pq - Vcc x [ICCf+ (NTTLxaICC x DCvm)] + l;(CLXVCC2 xfo ) 5Vx [20 mA + (1 x 1 mAx 1)] +[30pFx(5V)2x 16.7 MHz] 117.5mW (2) Total Power PT - (Pd x DCd) + [Pq x (I-DCd)] (117.5mWx1l3)+(78.3mWx2/3) 91.4mW (3) Therefore, the total power is approximately 91 mW. 4-33 Power Characteristics The SN74ACI'2229 is available in the 24-pin SOIC and 28-pin SOIC packages. The maximum power-dissipation capabilities of these packages forvarying ambient temperatures and airflows can be found in Section 5 of this handbook. At an ambient temperature of 70°C with no air flow, the maximum power that each of these packages can dissipate to free air is 909 mW and 978 mW, respectively. The total power dissipation of the SN7 4ACT2229 with one input (or output) switching falls well within the thermal budget of either of these packages. 18 fl = 1/2 fclock TA=75°C CL=OpF 16 1I 14 'E 12 ~ :::I U 10 fa. 8 :::I UJ I ~ 8 4 2 0 5 0 10 15 20 25 'clock - Clock Frequency - MHz Figure 1. SN74ACT2226 and SN74ACT2228 Single FIFO Supply Current Versus Clock Frequency 40 35 1 30 ~ 26 I :::I u ~ 20 a. a. :::I UJ 15 I e: J~ 10 fl = 1/2 fclock --+----+--+ TA = 75°C 5 CL=OpF 0 0 10 20 30 40 50 60 fclock - Clock Frequency - MHz Figure 2. SN74ACT2227 and SN74ACT2229 Single FIFO Supply Current Versus Clock Frequency 4-34 36-Bit Clocked FIFOs Page Sample Power-Dissipation Calculations for SN74ABT3613 •............................ Power Characteristics ............................................................. SN74ABT3611 Supply Current Versus Clock Frequency ......•..................•..... SN74ABT3612 Supply Current Versus Clock Frequency ............................... SN74ABT3613 Supply Current Versus Clock Frequency ............................... SN74ABT3614 Supply Current Versus Clock Frequency ....•........•................. SN74ABT3614 Idle Current With ClKA Switching. Other Inputs at'O or Vee - 0.2 V and Outputs Disconnected ............•.•............•........................... SN74ABT3614 Idle Current With ClKB Switching. Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . .. Sample Power-Dissipation Calculations for SN74ACT3641 ..........................•.. Power Characteristics .....•.............••............•........................... SN74ACT3632 Supply Current Versus Clock Frequency ............•....•............. SN74ACT3638 Supply Current Versus Clock Frequency .........................•..... SN74ACT3631 and SN74ACT3641 Supply Current Versus Clock Frequency ............. SN74ACT3641 Idle Current With ClKA Switching ..................................... SN74ACT3641 Idle Current With CLKB SWitching ..................................... SN74ACT3641 Active Current With ClKA and ClKB Switching. Simultaneous ReacllWrite and ClKB as Data Output ....•......•........•........... 4-37 4-38 4-38 4-38 4-39 4-39 4-40 4-40 4-41 4-42 4-42 4-42 4-43 4-43 4-44 4-44 4-35 4-36 Sample Power-Dissipation Calculations for SN74ABT3613 This example shows the power-dissipation calculations for the SN74ABT3613 unidirectional FIFO used in a system under the following conditions. Assumed Conditions • • • • • • • Data input (AO-~S) ~2.con.,.!!:>l signals (CLKA, CLKB, CSA, CSB, ENA, ENB, FSO, FS1, MBA, RST, ODDIEVEN, W/RA, WIRB, BE, SIZO, SIZ1, SWO, SW1, PGA, PGB) are driven by a CMOS device. 36 inputs (or 36 outputs) are switching at a given time. The port-B clock rate is 33.3 MHz and the port-A clock rate is 40 MHz. The SN74ABT3613 is used about 60% of the time by the system. The load capacitance on each output is about SO pF and the supply voltage is 5 V. During the FIFO active period, the bus is enabled 75% of the time. When the bus is enabled, the output is in the high state 80% of the time. The following parametric values are needed to calculate power dissipation: - VCC VOH VOL DCd fCLKA fCLKB pF DCEN Iccr - ICCF IcCH IcCL Iccz NUNT NHINT CL fo - - - 5 V (assumed condition) VCC-1.3 V 0.3V 0.6 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.265 mAlMHz (assumption based on data for the SN74ABT3614) 0.75 (assumed condition) (fCLKA + fCLKB) x pF - 19.4 rnA (calculated, see equation 7 in the introduction to Section 4) 136 rnA@ 33.3 MHz (from active Icc versus frequency plot) 30 rnA (see data sheet) 130 rnA (see data sheet) 30 rnA (see data sheet) 0.2 (assumed condition) 0.8 (assumed condition) 50 pF (assumed condition) (0.5) x (1) x (33.3 MHz) - 16.7 MHz (calculated, see equation 9 in the introduction to Section 4) Solution Quiescent Power Pq - - VCC x [DCEN x (NH x IcCHINT+ NLx IcCLINT) + (l-DCEN) x Iccz + Iccd 5 V x [0.75 rnA x (0.8 x 30 rnA + 0.2 x 130 rnA) + (0.25) x 30 rnA + 19.4 rnA] 322mW (I) Dynamic Power Pq - - VCCxIcCf+l:[VCCxCLx(VOHxVOL)xfo1 5 Vx 136 rnA + (36 x 5 V xSOpFx (5 V -1.3 V -0.3 V) x 16.7 MHz) 1191 mW (2) Total Power For - (Pd x DCd) + [Pq x (l-DCd)] - (1191 mW) (0.6) + (322 mW) (0.4) 843.4mW (3) Therefore, the total power is approximately 843 mW. 4-37' Power Characteristics The SN74ABT3613 is available in the 120-pin TQFP and 132-pin PQFP packages. The maximum power-dissipation capabilities of these packages for varying ambient temperatures and airflows can be found in Section 5 of this handbook. At an ambient temperature of 70°C with no air flow, the maximum power that each of these packages can dissipate to free air is 1606 mW and 1610 mW, respectively. The total power dissipation of the SN74ABT3613, under the rather severe conditions of 36 inputs (or 36 outputs) switching simultaneously, falls well within the thermal budget of either of these packages. 400 'd~ = 1~2 'clo Ik I I VCC = 5.5 V / 350 ,TA=25°C CL=OpF 11 300 i,. 250 (.) >15. a. ~ 1 s: J~ 150 50 o o ./ / ". / 1/~ / ~ ~ V VCC =4.5V VCC=~Z 200 100 ~ / / ,~ 10 k% ~ V 20 30 40 50 60 70 80 'clock - Clock Frequency - MHz Figure 1. SN74ABT3611 Supply Current Versus Clock Frequency 400 1 'data = 112 'clock 350 r- TA = 25°C CL=OpF "' E 1 250 fa. 200 (.) :::0 1 s: J~ 150 100 50 o o ~ / V / / / )': V ~ ~ """""VCC = 4.5 V VCC=5Z ~ en / / 300 ,.~ _I, VCC=5.5V / ,~ 10 ~~ V 20 30 40 50 60 70 80 'clock - Clock Frequency - MHz Figure 2. SN74ABT3612 Supply Current Versus Clock Frequency 4-38 400 c( E I C 300 ~~ 250 ~ Do 200 0 II , _I. VCC=5.5V 150 _0 100 50 o o ~ l/ ~V . . ~ ~ VVCC = 4.5 V ~ I L L L I'" V L .L L VCC=~ II) s: 0 I 'data = 1/2 'clock 350 f-TA=25°C CL=OpF ,~ 10 ~~ V 20 30 40 50 60 70 60 'clock - Clock Frequency - MHz Figure 3. SN74ABT3613 Supply Current Versus Clock Frequency 1I ~ G b 8: ~ ~ 150 t--II--t-7F7""-P"-+--t--t--; J~ 100t--I1-~~~--r-+--t--t-~ 50t-~~-t--+--r-+--t--t-~ O~~--~--~--~~~~--~~ o 10 20 30 40 50 60 70 60 'clock - Clock Frequency - MHz Figure 4. SN74ABT3614 Supply Current Versus Clock Frequency 4-39 ~ I J J I B o~~~~--~--~--~--~--~ o 10 20 30 40 50 80 70 f - Frequency - MHz Figure 5. SN74ABT3614 Idle Current With CLKA Switching, Other Inputs at 0 or Vce - 0.2 V and Outputs Disconnected ~ I J J I B 5r-~r--;--~---+---+---r--~ O~~~~--~--~--~--~--~ OW" ~ 40 80 80 ~ f - Frequency - MHz Figure 8. SN74ABT3614 Idle Current With CLKB Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected 4-40 Sample Power-Dissipation Calculations for SN74ACT3641 This example shows power-dissipation calculations for the SN74ACT3641 unidirectional FIFO used in a system under the folIowing conditions. Assumed Conditions • • • • • • Data input (AO-A35) are driven~ TIL device. _ _ Control signals (CLKA, CLKB, CSA, CSB, ENA, ENB, FSO, FSI, MBA, MBB, RPM, RTM, RST, WIRA, WIRB) are driven by a CMOS device. 36 inputs (or 36 outputs) are switching at a given time. The port-B clock rate is 33.3 MHz and the port-A clock rate is 40 MHz. The SN74ACT3641 is used about 113 of the time by the system. The load capacitance on each output is about 30 pF, and the supply voltage is 5 V. The following parametric values are needed to calculate power dissipation: VCC DCd = fCLKA fCLKB pF(A) pF(B) DCEN ICCI NTIL ICCf Icc CL fo = - 5 V (assumed condition) 1/3 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.20 rnAlMHz (from idle Icc versus frequency plot, CLKA switching) 0.16 rnAlMHz (from idle ICC versus frequency plot, CLKB switching) 0.75 (assumed condition) [fCLKA x pF(A)} + [fCLKB x pF(B)] - 13.3 rnA (calculated, see equation 2 in the introduction to Section 4) 36 (assumed condition) 100 rnA@ 33.3 MHz (from active Icc versus frequency plot) 1 rnA (assumed condition) 30 pF (assumed condition) (0.5) x (I) x (33.3 MHz) - 16.7 MHz (calculated, see equation 4 in the introduction to Section 4) SolutIon Quiescent Power Pq - Vccx[IcCI+(Nrn...xAIccxDCVIH)] 5 V x [13.3 rnA + 36 x 1 rnA xl)] 246.5mW (1) Dynamic Power Pd - Vcc x [IcCf+ (NTTL x AIcc x DCVIH)] + l:[CL x (VCc2 x fo) 5 Vx [100 rnA + (36 x 1 rnA xl)] + [36 x 30 pFx (5 V)2 x 16.7 MHz) 1130.9mW - (Pd x DCd) + [Pq x (I-DCd)] (1130.9 mW x 113) + (246.5 mW x 213) 541.3mW (2) Total Power PT = - (3) Therefore, the total power is approximately 541 mW. 4-41 Power Characteristics The SN74ACT3641 is available in the 120-pin TQFP and 132-pin PQFP packages. The maximum power-dissipation capabilities of these packages for varying ambient temperatures and air flows can be found in Section 5 of this handbook. At an ambient temperature of 70°C with no air flow, the maximum power that each of these packages can dissipate to free air is 1606 mW and 1610 mW, respectively. The total power dissipation of the SN74ACT3641, under the rather severe conditions of 36 inputs (or 36 outputs) switching simultaneously, falls well within the thermal budget of either of these packages. 300 , . - - - , - - - , - - - , - - - , - - , - - , - - - - - , 1 ~ fdata = 1/2 'clock TA=75°C CL 0 pF +---t---+--__+--t---:~ = 250 I G I 150~-+_-+_-+-_.~~~~__+-_t I ~ 100 50 O~-~-~-~-~-~-~-~ o 10 20 30 40 50 60 70 fclock - Clock Frequency - MHz Figure 1. SN74ACT3632 Supply Current Versus Clock Frequency 250 1 fdata = 1/2 fclock TA = 25°C CL=OpF 200 I i :::I U 150 b a. a. ci I 100 e: u _u 50 O~--~--~--~--~--~--~--~ o 10 20 30 40 50 60 70 fclock - Clock Frequency - MHz Figure 2. SN74ACT3638 Supply Current Versus Clock Frequency 4-42 250 r----r----r----r----r----r----,..---, 'data =1/2 'clock 1 TA=25°C CL=OpF 200 I J 1501--t--t--t--hI'....,.h~+----f 1(/) J 100 1--t--t-~~'2I't--+--+----f I 501--~WI--t--t--t--+----f O~--~-~--~-~-~--~~ w o ~ ~ ~ 50 50 N 'clock - Clock Frequency - MHz Figure 3. SN74ACT3631 and SN74ACT3641 Supply Current Versus Clock Frequency 18 TA 18 !. 25° ~ 12 ./ 8 /. ~ 8 L. 4 o / / / ~~ / VCC=5V, 10 2 ~ VCC=5.5V 14 ./ o " 10 ~ V Vee = 4.5 V V ~ 20 ~ 1\ , ~ Slope = 0.20 40 50 50 70 ,- Frequency - MHz Figure 4. SN74ACT3641 Idle Current With CLKA Switching 4-43 14 12 1 10 I 8 i 6 I :::I () iI ~ 4 2 10 20 30 40 50 60 70 f - Frequency - MHz Figure 5. SN74ACT3641 Idle Current With CLKB Switching 10 20 30 40 50 60 70 f - Frequency - MHz Figure 6. SN74ACT3641 Active Current With CLKA and ClKB Switching, Simultaneous ReadlWrlte and CLKB as Data Output 1S-Bit Clocked/Strobed FIFOs Page Sample Power-Dissipation Calculations for SN74ABT7819 ............................. Power Characteristics ............................................................. SN74ABT7819 Supply Current Versus Clock Frequency ....•....•..................... SN74ABT7820 Supply Current Versus Clock Frequency ............................... Sample Power-Dissipation Calculations for SN74ACT7803 .........••.................. Power Characteristics .....•........................•....•...........•............. SN74ACT7804 Supply Current Versus Clock Frequency ....................•.......... SN74ACT7806 Supply Current Versus Clock Frequency ............................... SN74ACT7814 Supply Current Versus Clock Frequency ............................... SN74ACT7803 Supply Current Versus Clock Frequency ............................... SN74ACT7803 Idle Current With RDCLK or WRTCLK Switching ...•...•....••.......... SN74ACT7805 Supply Current Versus Clock Frequency ................•.•............ SN74ACT7813 Supply Current Versus Clock Frequency .................•............. 4-47 4-48 4-48 4-48 4-49 4-50 4-50 4-50 4-51 4-51 4-52 4-52 4-53 4-45 4-46 Sample Power-Dissipation Calculations for SN74ABT7819 This example shows power-dissipation calculations for the SN7 4AB17S19 unidirectional FIFO used in a system under the following conditions. Assumed Conditions • • • • • • • Data input lines (AO-AI7) and control signals (CLKA, CLKB, CSA, CSB, RENA, RENB, WENA, WENB, RSTA, RSTB, PENA, PENB, WiRA, W/RB) are driven by a CMOS device. IS inputs (or IS outputs) are switching at a given time. The port-B clock rate is 33.3 MHz and the port-A clock rate is 40 MHz. The SN74AB 17S19 is used about 60% of the time by the system. The load capacitance on each output is about 50 pF, and the supply voltage is 5 V. During the FIFO active period, the bus is enabled 75% of the time. When the bus is enabled, the output is in the high state SO% of the time. The following parametric values are needed to calculate power dissipation: VCC VOH VOL Ded fCLKA fCLKB pF DCEN ICCI - ICCf IcCH IcCL ICCZ NLINT NHINT CL fo - = 5 V (assumed condition) VCC-1.3 V 0.3V 0.6 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.2 mAlMHz (worst-case assumption) 0.75 (assumed condition) (fCLKA + fCLKB) x pF - 14.7 rnA (calculated, see equation 7 in the introduction to Section 4) 75 rnA @ 33.3 MHz (from active Icc versus frequency plot) 15 rnA (see data sheet) 95 rnA (see data sheet) 15 rnA (see data sheet) 0.2 (assumed condition) O.S (assumed condition) 50 pF (assumed condition) (0.5) x (1) x (33.3 MHz) - 16.7 MHz (calculated, see equation 9 in the introduction to Section 4) Solution Quiescent Power Pq - VCC x [DCEN x (NH x ICCHINT + NL x Iccr/NT) + (1- DCEN) x Iccz + IccIl 5 V x [0.75 x (O.S x 15 rnA + 0.2 x 95 rnA) + (0.25) x 15 rnA + 14.7 rnA] 20S.5mW (1) Dynamic Power Pd = VCC x ICCf+ l;[Vcc xCLx (VOH- VOL) x fo] 5Vx [100 rnA + (36 x 1 rnA x 1)] + [36x30pFx(5 V)2x 16.7 MHz] 630.5mW (2) Total Power PT - (PdxDCd)+[Pqx(I-DCd)] (630.5 mW) (0.6) + (20S.5 mW) (0.4) 461.7mW (3) Therefore, the total power is approximately 462 mW. 4-47 Power Characteristics The SN74ABTI819 is available in the 80-pin TQFP and 80!pin PQFP packages. The maximum power-dissipation capabilities of these packages forvarying ambient temperatures and airflows can be found in Section 5 of this handbook. At an ambient temperature of 70°C with no air flow, the maximum power that each of these packages can dissipate to free air is 911 mWand 954 mW, respectively. The total power dissipation of the SN74ABTI819, under the rather severe conditions of 18 inputs (or 18 outputs) switching simultaneously, falls well within the thermal budget of either of these packages. 160 I II TA=76°C CL=OpF Vcc =6.6 V. /V / V ,/V 140 1 120 i d 100 8: 60 I I ~ 60 40 "/. ~ ~ ,/ i-" V ,/ V ,/V , / ~ b ~ ,/ VCC=6V - / 10' ~ V ,/ VCC =4.6 V ~ ~V 20 10 16 20 26 30 36 40 46 60 55 60 66 70 'clock - Clock Frequency - MHz Figure 1. SN74ABT7819 Supply Current Versus Clock Frequency 160 140 1 120 ::I 100 i TA=76°C CL=OpF VCC=5V b Go Go ~ 60 I ~ 60 40 ~ ,/ , / '" , / ~~ t -'_ L V ~ ,/ " ,/ .~ V ,/i-" , / , / ~ V V ,/ V ,,- U I VCC=6.6V / ,/ ~VCC=4.5V 20 10 15 20 25 30 36 40 46 60 55 60 66 70 'clock - Clock Frequency - MHz Figure 2. SN74ABT7820 Supply Current Versus Clock Frequency Sample Power-Dissipation Calculations for SN74ACT7803 This example shows power-dissipation calculations for the SN74AC17803 unidirectional FIFO used in a system under the following conditions. Assumed Conditions • Data input lines (00-D17) are driven by a TIL device. • Control signals (WRTCLK, RDCLK, RESET, WRTEN1, =WR=TE=N:=2, OE1, OE2, RDEN, PEN) are driven by a CMOS device. • 18 inputs (or 18 outputs) are switching at a given time. • The RDCLK clock rate is 33.3 MHz and the WRTCLK clock rate is 40 MHz. • The SN74ACT7803 is used about 113 of the time by the system. • The load capacitance on each output is about 30 pF, and the supply voltage is 5 V. The following parametric values are needed to calculate power dissipation: VCC DCd fWRTCLK fRDCLK pF DCEN ICCI NTIL Iccf AlCC IcCL DCvm CL fo - 5 V (assumed condition) 113 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.12 mNMHz (from idle Icc versus frequency plot) 0.75 (assumed condition) (fwRTcLK + fRDCLK) XpF - 8.8 rnA (calculated, see equation 2 in the introduction to Section 4) 18 (assumed condition) 75 rnA @ 33.3 MHz (from active ICC versus frequency plot) 1 rnA (see data sheet) 95 rnA (see data sheet) 1 rnA (see data sheet) 30 pF (assumed condition) (0.5) x (1) x (33.3 MHz) - 16.7 MHz (calculated, see equation 4 in the introduction to Section 4) Solution Qulescsnt Power Pq - VCCX[IcCI+NTILXAIccXDCVIH)] 5 Vx [0.75 x (8.8 rnA + (18 xl rnA x 1)] 134mW (1) Dynamic Power Pd - Vcc x [IcCf+ (NTIL x AIcc x DCVIH)] + :E(CLx Vc,..z xfo) 5 V x [75 rnA + (18 x 1 rnA x 1)] + [18 x 30pFx (5 V)2 x 16.7 MHz] 690.5mW (2) Tota/Power PT - (Pd xDCd) + [Pqx(I-DCd)] (690.5 mW x 113) + (134 mW x 2/3) 319.5mW (3) Therefore, the total power is approximately 320 mW. 4-49 Power Characteristics The SN74ACT7803 is available in the 56-pin SSOP package. The maximum power-dissipation capabilities of these packages for varying ambient temperatures and air flows can be found in Section 5 of this handbook. At an ambient temperature of 70°C with no air flow, the maximum power that the 56-pin SSOP package can dissipate to free air is 851 mW. The total power dissipation of the SN74ACT7803, under the rather severe conditions of 18 inputs (or 18 outputs) switching simultaneously, falls well within the thermal budget of either of this package. 200 180 1I ~ V / Y 120 80 I ./ VCC=5V, 100 s: 8 I 140 a. a. ::I II) .- VCC=5.5V 160 G ~ I TA =75°C CL=OpF b ./ VCC=4.5V ~ ,~W 20 0 // / ~~ 40 /' / / / 80 I' o 20 10 30 40 50 60 70 'clock - Clock Frequency - MHz Figure 1. SN74ACT7804 Supply Current Versus Clock Frequency 200 180 1I ~ I TA=75°C CL=OpF VCC=5.5V 160 VCC=5V" 140 120 ::I 0 ~ a. ,jl I s: 8 / / Y / / L / 100 80 b 60 ~~ 40 l/ ./ VCC=4.5V ~ p 20 0 V/ ~ I' ", o 10 20 30 40 50 60 70 'clock - Clock Frequency - MHz Figure 2. SN74ACT7806 Supply Current Versus Clock Frequency 4-50 200 180 I TA = 75°C CL=OpF VCC=5.5V 180 VCC=5V" ""E 140 ~ 120 I G 100 ::s 80 I 60 ~ D. / V/ / 8 l/ / Y ./ / '/ VCC=4.5V ~~ 40 , P' 20 0 V V ~~ II) S ~ o 10 20 30 40 50 80 70 'clock - Clock Frequency - MHz Figure 3. SN74ACT7814 Supply Current Versus Clock Frequency 200 180 I TA = 75°C CL=OpF ""E ~::s I () itil I S 8 / / Y VCC=5V~ 140 120 V/ / 80 l/ ./ L "£ / 100 VCC=4.5V ~ Y'" ~~ 80 40 ,p 20 0 V Vcc = 5.5 V / 180 o 10 20 30 40 50 60 70 'clock - Clock Frequency - MHz Figure 4. SN74ACT7803 Supply Current Versus Clock Frequency 4-51 1 I J I I B 10 20 30 40 50 80 70 f - Frequency - MHz Figure 5. SN74ACT78031dle Current WlthRDCLK or WRTCLK Switching 200 180 I TA = 75·0 OL=OpF Vcc = 5.5 V ~ ~ 180 1 I I 3 i~ I ~ V ,/ / rY . / / / / /,/ /" Vcc = 4.5 V VOO=5V~ 140 120 100 80 h Y' 80 ~~ ~W 40 , 20 o o 10 20 30 40 50 60 70 fclock - Clock Frequency - MHz Figure 8. SN74ACT7806 Supply Current Versus Clock Frequency 200 180 I TA=75OC CL=OpF VCC=5.5V I 180 ~ I i B iIi: I ~ VCC=5V~ 140 100 h 80 ~~ ~~ 40 20 o V l/ / IY / / / / / / / VCC=4.5V 120 80 V ~ ~ r" o 10 20 30 40 50 50 70 fclock - Clock Frequency - MHz Figure 7. SN74ACT7813 Supply Current Versus Clock Frequency 4-53 4-54 9-8;t Clocked/Strobed FIFOs Page Sample Power-Dissipation Calculations for SN74ACT7807 ..........................•.. Power Characteristics ............................................................. SN74ACT7807 Supply Current Versus Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. SN74ACT7807 Idle Current With WRTCLK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected .......................... SN74ACT78071dle Current With RDCLK Switching, Other Inputs at 0 or Vee - 0.2 V and Outputs Disconnected .......................... SN74ACT7808 Supply Current Versus Frequency ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 4-57 4-58 4-58 4-58 4-59 4-59 4-56 Sample Power-DIssipation CalculatIons for SN74ACT7807 This example shows power-dissipation calculations for the SN74ACT7807 unidirectional FIFO used in a system under the following conditions. Assumed Conditions • • • • • • Data input lines (DO-D8) are driven by a TIL device. Control signals (WRTCLK, RDCLK, RESET, WRTEN1, WRTEN2, DE, RDEN1, RDEN2, PEN) are driven by a CMOS device. Nine inputs (or nine outputs) are switching at a given time. The RDCLK clock rate is 33.3 MHz and the WRTCLK clock rate is 40 MHz. The SN74ACT7807 is used about 113 of the time by the system. The load capacitance on each output is about 30 pF, and the supply voltage is 5 V. The following parametric values are needed to calculate power dissipation: vCC DCd fWRTCLK fRDCLK pF(WRTCLK) pF(RDCLK) DCEN IcC! - NTIL Icer - ~Icc IccL DCVIH CL fo = - 5 V (assumed condition) 113 (assumed condition) 40 MHz (assumed condition) 33.3 MHz (assumed condition) 0.07 rnA/MHz (from idle Icc versus frequency plot) 0.09 mAlMHz (from idle Icc versus frequency plot) 0.75 (assumed condition) (fWRTCLK x pF(WRTCLK» + (fRocLK x pF(RDCLK» - 5.8 rnA (calculated, see equation 2 in the introduction to Section 4) 9 (assumed condition) 65 rnA @ 33.3 MHz (from active Icc versus frequency plot) 1 rnA (assumed, from data sheet) 95 rnA (see data sheet) 1 rnA (assumed condition) 30 pF (assumed condition) (0.5) x (1) x (33.3 MHz) - 16.7 MHz (calculated, see equation 4 in the introduction to Section 4) Solution Quiescent Power Pq - Vcc x [Icc! + NTIL x ~Icc x DCVIH)) 5 V x [5.8 rnA + (9 x 1 rnA xl)] 74mW (1) Dynamic Power Pd - VCC X [Icer+ (NTILx ~Icc x DCVIH)] +l:(CL x VCc2 x fo) 5 V x [65 rnA + (9 x 1 rnA xl)] + [9 x 30 pF x (5 V)2 x 16.7 MHz] 482.7mW (2) Tota/Power PT - (Pd x DCd) + [Pq x (l-Ded)] (482.7mWx1l3)+(74mWx2l3) 21O.2mW . (3) Therefore, the total power is approximately 210 mW. 4-57 Power Characteristics The SN74ACT7807 is available in the 64-pin TQFP and 44-pin PLCC packages. The maximum power-dissipation capabilities of these packages for varying ambient temperatures and airflows can be found in Section 5 of this handbook. At an ambient temperature of70°C with no air flow, the maximum power that each of these packages can dissipate to free air is 1121 mWand 1244 mW, respectively. The total power dissipation of the SN74ACT7807, under the rather severe conditions of nine inputs (or nine outputs) switching simultaneously, falls well within the thermal budget of either of these packages. o~~--~--~--~~--~--~~ o 10 20 30 40 50 60 70 80 'clock - Clock Frequency - MHz Figure 1. SN74ACT7807 Supply Current Versus Frequency O~--~--~--~--~---L--~--~ o 10 20 30 40 50 60 70 , - FrequencY - MHz Figure 2. SN74ACT78071dle Current With WRTCLK Switching. Other Inputs at 0 or Vee - O.2V and Outputs Disconnected 4-58 ~ I J i I ~ 2r-~--~~~---+---+---+--~ o~~--~--~--~--~--~--~ ~ ~ ~ ~ ~ ~ o ro , - Frequency - MHz Figure 3. SN74ACT7807 Idle Current With RDCLK SWitching, Other Inputs at 0 or VCC - 0.2 V and Outputs Disconnected 1~ 1~ TA=75OC CL=OpF I , .'. / VCC=5V""\ ~ 120 I 100 I 3>0 ~ ~ ~ 1 / / ,1~ V ~ I S 8 ~ ~ o ~ 1/ VCC =5.5 V V VV ~/ / ' VCC=4.5V ~ """ o ro ~ ~ ~ ~ ~ ~ ~ 'clock - Clock Frequency - MHz Figure 4. SN74ACT7808 Supply Current Versus Frequency 4-59 4-60 5-1 3: CD n :::sO) _. :1 n 0) - 0) :1 Co -t :::s- ...CD3 0) 0' ...3 ...o_. :1 0) :1 5-2 Comparison Summary of Advanced Packaging Derating Curves (in Still Air) Introduction This section of the FIFO Designer's Handbook contains mechanical and thermal information and data relating to the packages used for Texas Instruments (TI) FIFOs. A series of application reports and papers that address such issues as package thermal resistance, package moisture sensitivity, fme-pitch packaging manufacturability, and measurement procedures of thermal resistance are included in this section. One of these papers, More Power in Less Space: A Thermal-Enhancement Solution for Thin Packages, was published in TI Technical Journal, Volume 11, Number 4 (July-August 1994). The paper discusses the design and development of the thermally enhanced thin quad flat package (TQFP TEP) that is capable of dissipating 2.4 watts of power in a 256-mm2 board area. The thermally enhanced, fme-pitch package is the result of the efforts of a cross-functional team at TI that included resources from package design, chip design, package assembly, reliability, and device testing. Following the application reports is the mechanical and thermal data for each FIFO package. The mechanical data consists of outline drawings of each package annotated with critical dimensions. Accompanying each package drawing is the associated thermal data. This data consists of measured thermal resistances and derating curves of maximum power dissipation versus ambient temperature for varying air flows. Figure I is a summary plot of the still-air derating curves for the advanced FIFO packaging options. The 120-pin TQFP TEP and the 132-pin PQFP packages are nearly equivalently capable of dissipating more power for a given ambient temperature than any of the other packages considered. 3000 --- 24-Pln SOIC (OW) -6- 28-Pln SOIC (OW) ....... 44-Pln PLCC (FN) -0- 56-Pln SSOP (OL) 64-Pln TQFP (PAG) ....... 64-Pln TQFP (PM) ___ 68-Pln PLCC (FN) + 2500 -b- SO-Pin PQFP (PH) -0- SO-Pin TQFP (PN) -+- 10o-Pln TQFP (PZ) ___ 12o-Pln TQFP (PCB) -D- 132-Pln PQFP (PQ) 2000 1500 1000 500 25 30 35 40 45 50 55 60 65 70 75 80 85 90 TA - Ambient Temperature _·C Figure 1. Comparison Summary of Advanced Packaging Derating Curves (In stili air) 5-5 Application Reports/Papers Page FIFO Surface-Mount Package Information .•........................................... 5-9 FIFO Memories: Fine-Pitch Surface-Mount Manufacturability ........................... 5-19 Package Thermal Considerations ................................................... 5-29 K-Factor Test-Board Design Impact on Thermal-Impedance Measurements .............. 5-41 More Power in Less Space: A Thermal-Enhancement Solution for Thin Packages ......... 5-55 FIFO Surface-Mount Packages for PCMCIA Applications. . . . . . . . . . . . . • . . . . . . . . . . . . . . . .. 5-69 5-7 FIFO Surface-Mount Package Information First-In, First-Out Technology Tom Jackson and Mary Helmick Advanced System Logic - Sem/conductor Group :lllExAs INSIRUMENTS SSPA001A 5-9 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by govemment requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications'~. TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 5-10 Contents Title Page Introduction ............................................................................. 5-13 Thermal Resistance ....................................................................... 5-13 Package Moisture Sensitivity ............................................................... 5-14 Shipping Metbods/Quantities/Dry Pack ..................................................... 5-15 Package Dimensions and Area Comparison ................................................... 5-15 Test Sockets .............................................................................. 5-17 5-11 5-12 Introduction Texas Instruments provides seven types of plastic surface-mount packages for CMOS FIFO memory devices. These packages and the data bus width that each package can provide are listed in Table 1. Table 1. Plastic Surface-Mount FIFO Packages PACKAGE 44-pinPLCC NO. OF DATA BITS 64-pinTQFP 9 9 56-pinSSOP • 18 68-pinPLCC 18 8().pin TQFP 18 8~inQFP 12O-pin TQFP SSOP. PLCC TQFP • QFP .. 18 32 or 36 shrink small-outhne package plastic leaded chip carrier thin quad flat package quad flat package This application report discusses several topics concerning the FIFO packages listed in Table 1: • The thermal resistance. RaJA, and the chip junction temperature of the device • The need for dry packing to maintain safe moisture levels inside the package • The three methods used by Texas Instruments for shipping FIFOs to customers • The package dimensions. including two-dimensional drawings that show areas, heights, and lead pitches • The area comparison of surface-mount packages used for commercial FIFO memories • The test sockets available for surface-mount FIFO packages Thermal Resistance Thermal resistance is dermed as the ability of a package to dissipate heat generated by an electronic device and is characterized by RaJA. RaJA is the thermal resistance from the integrated circuit chip junction to the free air (ambient). Units for this parameter are in degrees Celsius per watt. Table 2 lists RaJA for SSOP, PLCC, TQFP, and QFP packages under five different air-flow environments: 0, 100,200,250, and 500 linear feet/minute. The chip junction temperature (TJ) can be determined using equation 1. (1) Where: Tl - chip junction temperature (0C) RaJA - thermal resistance, junction to free-air (oC/watt) l>T - total power dissipation of the device (watts) TA - free-air (ambient) temperature in the particular environment in which the device is operating (0C) 5-13 Table 2. Thermal Resistance, RaJA, for FIFO Packages ReJA(OCIW) PACKAGE LEAD FRAME OLFPM 100LFPM 200LFPM 260LFPM 500LFPM 56-plnSSOP Copper 94.2 62.2 N/A 70 57.8 44-pin PLCC Copper 65 NlA NlA N/A NlA 68-pinPLCC Copper 47.2 43.4 NlA 32.7 27.8 64-pinTQFP Copper 92.5 87.8 N/A 72.9 57.8 8O-plnTQFP Copper 87.8 79.1 N/A 67.3 54.2 12Q-pln TQFPt Copper 49.6 44.3 N/A 38.3 28.6 8Q-pinQFP Alloy 42 80 67 61 NlA N/A t Heat slug molded inside the package N/A =not available The RaJA generally increases with decreasing package size; however, this is not true with the 120-pin SQFP package. A heat slug molded inside the package absorbs a large amount of heat dissipated by the device. As a result, this package provides a relatively low RaJA' Package Moisture Sensitivity When a plastic surface-mount package is exposed to temperatures typical of furnace reflow, infrared (IR) soldering, or wave soldering (215°C or higher), the moisture absorbed by the package turns to steam and expands rapidly. The stress caused by this expanding moisture results ininternal and external cracking of the package that leads to reliability failures. Possible damage includes the delamination of the plastic from the chip surface and lead frame, damaged bonds, cratering beneath the bonds, and external package cracks. To prevent potential damage, packages that are susceptible to the effects of moisture expansion undergo a process called dry pack. This dry pack process helps to reduce moisture levels inside the package. The process consists of a 24-hour bake at 125°C followed by sealing of the packages in moisture-barrier bags with desiccant to prevent reabsorption of moisture during the shipping and storage processes. These moisture-barrier bags allow a shelf storage of 12 months from the date of seal. Once the moisture-barrier bag is opened, the devices in it must be handled by one of the following four methods, listed in order of preference: The devices may be mounted within 48 hours in an atmospheric environment of less than 60% relative humidity and less than 30°C. The devices may be stored outside the moisture-barrier bag in a dry-atmospheric environment of less than 20% relative humidity until future use. The devices may be resealed in the moisture-barrier bag adding new fresh desiccant to the bag. When the bag is opened again, the devices should be used within the 48-hour time limit or resealed again with fresh desiccant. The devices may be resealed in the moisture-barrier bag using the original desiccant. This method does not allow the floor life of the devices to be extended. The cumulative {:xposure time before reflow must not exceed a total of 48 hours. All plastic surface-mount FIFO devices are tested for moisture sensitivity in accordance with Texas Instruments JESD A112 procedure. 5-14 Shipping Methods/Quantities/Dry Pack Three methods are used by Texas Instruments for shipping FIFOs to customers. These methods are tubes, tape/reel, and trays. The quantities for each of the shipping methods are listed in Table 3. The shipping quantity is dermed as the maximum number of packages that can be packed in a single shipping unit (e.g., the maximum number of 56-pin SSOP packages that can be packed in a tube is 20). Whether or not the packages require dry pack before shipping is noted in the dry-pack column. Table 3. Shipping Methods and Quantities PACKAGE SHIPPING METHOD TUBEt TAPE/REELt TRAYSt DRY PACK 56-pinSSOP 20 500 N/A No 44-pln PLCC 27 N/A 68-pin PLCC No Yes 64-pinTQFP 18119* NlA 500 250 80-plnTQFP NlA N/A 160 N/A NlA 119 Yes 120-pin TQFP N/A NlA 90 Yes 8O-pinTQFP N/A N/A 50 Yes Yes t Texas Instruments reserves the right to change any of the shipping quantities at any time without notice. * Eighteen packages can be packed In a single tube when pin is used as a tap or nineteen packages can be packed in a tube when plug is used as a tap. NlA =not applicable Package Dimensions and Area Comparison Figure I contains two-dimensional drawings of the seven available surface-mount FIFO packages. For detailed mechanical drawings of these packages, please refer to the mechanical drawing section of the 1994 High-Performance FIFO Memories Data Book, literature #SCADOO3B. 5-15 14-- 16.0 ---.j J+::: 14.0 -~ I 1 ISO-Pin 14.0 TQfP I 0 (PN) JE.. Area: 144.00 mm2 Area: 196.00 mm2 Height: 1.50 mm Height: 1.50 mm (PM) Height: 1.0 mm (PAQ) Lead Pitch: 0.5 mm 12Cl-pln TQFP 11(0 ;c Lead Pitch: 0.5 mm 0 (PCB) Area: 256.00 mm2 Height: 1.6 mm Lead Pitch: 0.4 mm "'14- - 26.2 - -.... ~ I+-- I --.j 17.6 44-pln 1 PLCC (FN) Area: 309.60 mm2 1 1 68-pln PLCC (FN) .: Area: 636.04 mm2 Height: 4.37 mm Height: 4.37 mm Lead Pitch: 1.27 mm Lead Pitch: 1.27 mm I+-- 23.6 - -.. I Area: 189.50 mm2 Height: 2.59 mm Lead Pitch: 0.635 mm 1 Area: 415.40 mm2 "ii8lghti2.96 mni Lead Pitch: 0.8 mm Figure 1. Package Dimensions ,. " 5-16 '- " Figure 2 shows the area comparison of surface-mount packages for FIFOs from Texas Instruments and other FIFO vendors. 144 mm2 84-Pln TQFP { 32-PlnPLcc • • • 11Jmm l 44-Pln PLCC • • • • • J.8mm2 . .PlnSSOP 18-BIt-Wlde { Data _1 1 .5 8O-Pln TQFP • • • 1J m~ mm~ 8O-Pln QFP • • • • • • 41 4 mm2 88-Pln PLCC • • • • • • • • • 32-or 38-Blt-Wlde { Data 120-PlnTQFP • • • • • • • • • 132-Pln QFp .. o 100 200 300 400 500Il00 700 800 IIiIo 1000 A.... (mm2) Figure 2. Surface-Mount Package Area Comparison Test SOckets For prototype development of a system, it is often an advantage to have sockets for surface·mo'\lllt products. Test sockets a....ailable for use with Texas Instruments FIFO packages are listed in Table 4. Only one manufacturer is listed for each socket type, although other vendors may offer comparable sockets. Table 4. Test Sockets for FIFO Packages PACKAGE 56-pinSSOP 44-pln PLOO 66-pln PLeO 64-pinTQFP 8O-pInTQFP 12D-pln TQFP 8O-pInQFP MANUFACTURER Yamalchl NEY NEY Yamalchl Yamalchl Yamalchi Yamalchi NUMBER 1051-0562-1387 6068 1051-0644-807 1051..()8()4-808 1051-1204-1596 DESCAIPTIOH Solder through hole Solder through hole Solder through hole Solder through hole Solder through hole Solder through hole 1051-0604-394 Solder through hole 6044 5-17 5-18 FIFO Memories: Fine-Pitch Surface-Mount Manufacturability First-In, First-Out Technology Tom Jackson Advanced System Logic - Semiconductor Group ~lExAs INSTRUMENTS SCZAOO3A 5-19 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to Its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products In such applications Is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC·sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or Infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, Is granted under any patent right, copyright, mask work right, or other Intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 5-20 Contents TItle Page Introduction .............................................................................. 5-23 Improved Function Density ................................................................ 5-23 Manufacturing ......................................................... . . . . . . . . . . . . . . . . .. 5-23 Palladium-Plated Lead Frames ............................................................. 5-25 Testabllity ............................................................................... 5-26 DesignlPreproduction Considerations ....................................................... 5-26 Conclusion .............................................................................. 5-27 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-27 List of Tables Table 1 2 3 4 5 ntle Page Fine-Pitch Packages ................................................................. 5-23 Defect Causes and Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-24 Results of Soldered Joint Strength ..................................................... 5-25 Lead-Frame Platings by Package Type .................................................. 5-26 Available Fine-Pitch Test Sockets and Mechanical Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-27 5-21 5-22 Introduction Recent advances in semiconductor processing and packaging have produced highly integrated, fine-pitch devices to satisfy the demand for smaller systems. With the trend towards higher chip complexity occupying less board space, device manufacturers must increase bit density while decreasing package size. To accommodate these requirements, manufacturers have two choices: increase bit density, keeping the number of pins constant while reducing pitch and area, or reduce the package lead pitch, keeping area constant while increasing pin count. Manufacturers of hand-held and laptop computers and data communications and telecommunications equipment require the use of [me-pitch packages to build and maintain a competitive advantage. Improved Function Density Texas Instruments (TI) provides five types of fine-pitch plastic surface-mount packages for its FIFO product line (see Table 1). Each of these surface-mount packages has lead-to-Iead spacing less than or equal to 0.635 mm (0.025 in.). All of these packages offer designers critical board-space savings that is required for advanced systems. Compared to the commonly used 68-pin plastic leaded chip carrier (PLCC) for 18-bitFIFOs, TI's Widebus™ package, in either the 56-pin shrink small-outline package (SSOP) or the 80-pin thin quad flat package (TQFP) , reduces board space by 70%. A 67% saving of board space is available with TI's 36-bit FIFO family in the 120-pin TQFP compared to the 132-pin plastic quad flat package (PQFP). Table 1. Fine-Pitch Packages THIN SHRINK SMALL-OUTLINE PACKAGE (SSOP) THIN QUAD FLAT PACKAGE (TQFP) Pin count Lead pitch (mm) Footprint (mm) Board area (mm2) Package suffix 64 80 0.5 12x 12 0.5 14x 14 120 0.4 16x 16 144 PM 196 256 PN PCB 132 0.635 28x28 784 PO 56 0.635 10.35 x 18.42 190,6 DL Manufacturing Manufacturers are currently employing high-volume board-assembly techniques using standard lead pitches of 0.5 mm (20 mils) and greater. However, as lead pitch continues to decrease, questions must be asked of both the'manufacturer and the supplier: Are [me-pitch packaging capabilities available? Does production equipment have sufficient accuracy to produce high-volume, high-quality parts? Do the manufacturing personnel have experience in high-volume, high-quality production using fine-pitch packaging? Have the testability issues of [me-pitch packaging been considered? Standard processing techniques such as those used with surface-mount rigid-lead packages become difficult with fine-pitch packaging. Manufacturing issues may arise from compromises in screen-printing techniques, solder board/lead coplanarity, placement-accuracy requirements of components, and solder deposition methods (e.g., mass reflowing). All of these factors can result in shorts or opens due to poor placement, too much solder, or not enough solder. These issues influence the overall yield and reliability of the product. Widebus is a trademark of Texas Instruments Incorporated. 5-23 Equipment for the placement of fine-pitch packaging must feature a highly accurate positioning system. Placement accuracy for fme-pitch packages must increase as lead pitch decreases. Misaligned packages and boards greatly reduce production yields as well as throughput. Systems that feature state-of-the-art machine vision, align and inspect leads, and calculate registration with an extremely high degree of accuracy and repeatability, ensure high production yields. There must also be careful control over the Z-axis pressure when placing these fme-pitch packages to' protect the lead coplanarity. Currently, there are systems available with accurate placement as fme as O.I-mm pitch. One of the most critical issues facing the manufacturer is the reliability of the footprint design. Constraints include the length and width of the footprint and the amount of solder paste used to produce a good joint. If too much solder is used, the footprint can bridge, causing a short (see Table 2). The minute dimensions associated with fme-pitch packages require that the footprint be drawn to the highest level of accuracy in order to ensure consistent reliability. Board assemblers must be able to match the footprint with the same level of accuracy and repeatability. Table 2. Defect Causes and Effects DEFECT CONTROL Solder bridging Control the solder-paste quantity Open circuits Control solder-paste thickness and maintain lead coplanarity Shorts and opens Control equipment accuracy in the placement of parts As previously discussed, the key to ensuring high yield is an accurate footprint pattern. Many manufacturers request footprint pattems and dimensions to assist in their board assembly. There are several factors to consider when designing a footprint pattern to ensure reliability: • Device design - JEDEC or EIAJ Standard • PWB - foil thickness, number of layers, supplier's capabilities • Solder paste - type, solder mesh • Printer - manufacturer, standoff control, squeegee pressure • Print mask - type (stencil/mesh), tension, bias • Reflow process - preheat, temperature, dwell, etc. The key dimensions for designing an accurate footprint layout is shown in Figure 1. A = B1 B2 = L Distance Package Edge to End of Pad = Pad Extension Beyond Heel of Foot Pad Extension Beyond Toe of Foot = Lead Foot Length P = Lead Pitch S = Distance From Center of Pin to Center of Pin Figure 1. Footprint Diagram 5-24 Palladium-Plated Lead Frames Another area for manufacturers to investigate is metallization, or bonding of the leads to the circuit board with solder. There are several widely used localized reflow techniques including hand soldering, hot bar, focused infrared (IR), and laser. With each technique, heat is applied to the leads until the solder melts. When the heat source is removed, the solder cools forming the joint. Each manufacturer must make the choice between precision point-to-point systems (one chip at a time) and the speed of gang bonding (multiple chip bonding). Another area ofmeta1lization to consider is preplating of the leads by the device manufacturer. TI has begun to implement palladium (Pd) lead plating on many fme-pitch packages. These efforts began with joint testing of palladium-plated leads with several large computer and telecom customers in 1987. Since then, TIhas begun high-volume manufacturing with over five billion palladium-plated devices in the field. Palladium preplating is essentially a nickel- (Ni) plated lead frame that has a minimum of3 micro inches (0.076 micron) of Pd. The Pd fmish protects the Ni from oxidation and eliminates the need for silver spotting. Silver (Ag) spots are used to attach the fine wires from the die to the lead frames. However, the silver can migrate over time to form extraneous electrical contacts that greatly impact reliability. Many problems associated with fine-pitch manufacturing can be eliminated with palladium preplating: • • • • • • • • Reduces excess solder Excellent Pd wetting characteristics Reduced handling Improved package integrity Reduced mechanical damage Tarnish resistant Compatible with existing assembly processes Excellent adhesion to mold compounds Table 3 shows the results of a solder-joint strength test comparing Pd solder joints to traditional solder joints. The results demonstrate an equal performance between the two techniques. Palladium preplating also exhibits adhesion to most mold compounds, which reduces moisture ingress and plastic-to-Iead-frame delimitation. Table 3. Results of Soldered Joint Strength SAMPLE HOURS OF HEAT AGING OHR aHR 3 microinches Pd 5.171bf 5.951bf 5.85lbf 4.711bf 16HR 24HR Solder dip 5.071bf 4.511bf 5.551bf 5.50 Ibf In many cases, the cause for shorts and opens can be attributed to lead coplanarity, or the extent to which all leads lie in a single plane. This holds especially true for fine-pitch packaging due to the smaller geometries and delicate leads. Traditional solder-dipped leads tend to have more pin-to-pin alignment problems than the Pd-plated leads. The Pd-preplated leads have a more conformal and uniform coating than those that are solder dipped since the plating is performed prior to the packaging process (see Figure 4). An increase in coplanarity improves overall circuit reliability. The excellent wetting characteristics of Pd improve the wicking effects of solder and form a better solder joint/fillet. The thin Pd coating and minimal handling reduce the chance of coplanarity problems (i.e., shorts and opens) and also produce uniform solder joints with a minimum amount of solder. Table 4 lists TI's fme-pitch packages that implement Pdplating. 5-25 COPLANARITY vs %OFVALUE 40 35 I 30 25 o PdUnlts 20 • Dipped Units 15 10 5 o 0.2 0.4 0.6 0.8 n ,.., 1.2 .. 1.4 1.8 Coplanarlty (mm) Figure 2. Coplanarlty Results Table 4. Lead-Frame Platings, by Package Type PACKAGE SUFFIX LEAD FRAME 132-pin POFP PO Palladium 12o-pin TOFP PCB Palladium 8D-pinTOFP PN Solder 64-plnTOFP PM Solder 56-pinSSOP DL Palladium Testability Another issue introduced by the onset offme-pitch surface-mount packages involves testing circuit boards. With denser printed-circuit boards heavily populated with fme-pitch surface-mount packages, the issues involved with functional testing should be addressed. One of the most cost-effective solutions is the implementation of boundary-scan methodology defmed by the joint test action group (JTAG) and adopted by the mEE 1149.1 committee. JTAG devices incorporate on-chip test points called boundary-scan cells and utilize a serial-scan protocol tQrough the device. Devices with JTAG can be designed into the datapath and provide the controllability and observability needed to troubleshoot manufacturing defects. Design/Preproduction Considerations For designers who wish to implement fme-pitch packaging, TI provides an easy alternative for the development of prototypes and breadboarding. TI has worked with several test-socket manufacturers who provide accurate and easy-to-use through-hole test sockets for all of their surface-mount packaging. In addition to test sockets, TI also offers mechanical packages. These are packages that include lead frames without the silicon and meet all mechanical specifications. Mechanical packages provide an inexpensive means for manufacturing capability studies, machine setup, personnel training, and process-development work (see Table 5). 5-26 Table 5. Available Fine-Pitch Test Sockets and Mechanical Packages SOCKET TYPE MANUFACTURER PART NUMBER DESCRIPTION 64-pinTQFP Yamaichi 1051-0644-807 Through hole 56-pinSSOP Yamaichi IC51-0562-1514 Through hole 8o-pinTQFP Yamaichi IC51-0804-808 Through hole 12o-pin TQFP Yamaichi IC51-1204-1596 Through hole 132-pin PQFP Yamaichi. I051-828-KS12338 Through hole PACKAGE TI PART NUMBER 64-pinTQFP SN700870PM 56-pinSSOP SN250011 DLR 8o-pinTQFP SN700871PN 12o-pin TQFP SN700782PCB Conclusion Designs that incorporate fine-pitch packages have the advantage of critical board-space reduction. As designers continue to implement higher levels of integration, board space remains at a premium. With the implementation of concurrent engineering practices from design to test to manufacturing, many packaging difficulties can be overcome. Fine-pitch packaging is the designers' easiest option to reduce critical board space without the loss of higher chip integration. References Abbott, D.C., Brook, R.M., McLelland, N., Wiley, J.S., "Palladium as a Lead Finish for Surface Mount Integrated Circuit Packages," IEEE Transaction on Components, Hybrid Manufacturing Tech., Vol. 14, No.3, Sept. 1991. Romm, D., McLellan, N., "Evaluation of Water Soluble and No-Clean Solder Pastes with Palladium Plated and Solder Plated SMT Devices." 5-27 5-28 Package Thermal Considerations Darla WeI/heuser Advanced System Logic - Semiconductor Group ~1ExAs INSTRUMENTS SCZA002A 5-29 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to Its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to th'e specifications applicableatthe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products In such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 5-30 Contents TItle Page Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-33 Introduction ............................................................................. 5-33 Reliability ............................................................................... 5-34 Power Consumption ...................................................................... 5-34 Power Calculations ....................................................................... 5-36 CMOS ................................................................................ 5-37 BiCMOSlBipolar .....•.....•........................................................... 5-37 Thermal-Resistance Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5...:J7 Summary ................................................................................ 5-40 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-40 Figure List of IDustrations TItle Page 1 Advanced Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-33 2 3 Icc Versus Frequency (One Switching, Unused Outputs Low) ............................... Icc Versus Frequency (All Outputs Switching) ..•................................•....... Icc Versus Frequency (All Switching, 50% Duty Cycle Enabled) . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Icc Versus Duty Cycle Enabled (25 MHz) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . .. 4 5 6 7 8 5-35 5-35 5-35 5-36 48-Pin SSOP 9JA Versus Trace Length ...........................................•..... 5-38 48-Pin SSOP 9JA Versus Air Flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-38 48/56-Pin SSOP K-Factor Board Modeling ..........................•................... 5-39 5-31 5-32 Abstract To meet current and future system requirements of increasing speed and decreasing size, integrated circuit manufacturers are pushing the edge on existing packaging technology. A component's perfonnartce is detennined by process technology and the thennallimitations of its package. As a leader in package technology, Texas Instruments (TI) has introduced a number offme-pitch packages and is acutely aware of the thermal considerations that must be examined by systems designers. This paper is intended to create awareness and understanding of thennal issues and to .explore factors that influence thermal perfonnance. Introduction Thennal awareness became an industry concern when surface-mount (SMT) packages began replacing through-hole (DIP) packages in PCB designs. Circuits operating at the same power enclosed in a sma1lerpackage meant higher power. To add to the issue, systems required increased throughput, which resulted in higher frequencies, increasing the power density even further. Not only are these same concerns haunting designers today, they are progressively getting more severe. Figure 1 shows part of the reason for increased attention to thennal issues. As a baseline for comparison, the 24-pin small-outline integrated circuit (SOIC) is shown along with several fme-pitch packages supplied by n, including the 24-pin SSOP (shrink small outline), 48-pin SSOP, and the lOO-pin TQFP (thin quad flat pack). The 24-pin SSOP (8, 9, and 10 bits) allows for the same circuit functionality of the 24-pin SOIC to be packaged in less thanha1fthe area, while the 48-pin SSOP (16, 18, and 20 bits) occupies just slightly more area but has twice the functionality of the 24-pin SOIC. This same phenomena is expanded even further with the lOO-pin TQFP (32 and 36 bits), which is the functional equivalent of four 24-pin or two 48-pin devices, with additional board savings over that of the SSOP packages. As the trend in packaging technology continues toward smaller packages, attention must be focused on the thermal issues that are created. D 24-PlnSOIC Area = 165 mm2 D 48-PlnSSOP Area = 171 mm2 24-PlnSOIC ) , , l Height = 2.65 mm Volume = 437 mm3 Lead Pitch = 1.27 mm 0 24-PlnSSOP Area=70mm2 ~ Helght=2mm Volume = 140 mm3 Lead Pitch = 0.65 mm 10o-Pln TQFP 48-PlnSSOP Height = 2.74 mm Volume = 469 mm3 Lead Pitch = 0.635 mm 24-PlnSSOP 1CJO.Pln TQFP and 100-Pln Cavity TQFP Area = 262 mm2 Height = 1.6 mm Volume = 419 mm3 Lead Pitch = 0.5 mm Figure 1. Advanced Packages 5-33 Reliability The overriding effect of increased power densities in integrated circuits is a decrease in overall system reliability. A direct relationship exists between junction temperature and reliability. Table 1 provides an example of a device with an initial junction temperature of 150°C and the calculated failure-rate decrease as the in-use junction temperature is lowered. The data in Table 1 indicates that lower junction temperature results in increased system reliability. Table 1 TEMPERATURE ·C % FRt 150 96 140 80 130 46 120 11 110 1 100 0.02 t Failure rate at 100,000 hours A better understanding of the factors that contribute to junction temperature (Tl) provides a system designer with more flexibility when attempting to solve thermal issues. Device junction temperature is determined by equation 1: (1) Where: TJ TA - junction (die) temperature (0C) - ambient temperature (0C) alA - thermal resistance of the package from the junction to the ambient (OC/W) P-r - total power of the device (W) Junction temperature can be altered by lower chip power consumption, longer trace length, heat sinks, forced air flow, package mold compound, lead-frame size and material, surface area, and die size. Some of these are mechanically inherent to a particular package while others are controlled by the designer and are application specific. Understanding which variables can be influenced by practicing good thermal-design techniques requires a more detailed investigation of power considerations as well as thermal-resistance measurements. Power Consumption One way to lower the junction temperature (TJ) of a device, thus improving reliability, is to lower the power consumption. A vaiiety of options are available to help achieve this, such as low-power process technologies, reduced output swing, and reduced power-supply voltage. A closer look at the power performance and advantages of several popular logic families can assist the designer when choosing what best fits his/her needs. The choices available from TI for high-speed bus interfaces range from standard bipolar (F) to advanced CMOS (ACL/ACT) to state-of-the-art BiCMOS (BCT) and advanced BiCMOS (ABT). Figures 2 through 4 show comparisons of current (ICC) consumption of'244 functions for these technologies across frequency. As expected, the bipolar device consumes more current than the CMOS device at lower frequencies, but as frequency increases, this relationship no longer holds true. In fact, there is a region in the frequency range where the CMOS device consumes more current than the bipolar device. The point where they are equal is referred to as the crossover frequency. 80 70 ~ -- - -- - -- - - ~ - - ECr !It"" 50 40 I - -- -- F 60 ABT 30 20 ~ 10 o -1....--' o 10 - ....--'r 20 40 30 50 ~ 60 t-- il'CT 70 80 100 90 Frequency - MHz Figure 2. Icc Versus Frequency (One Switching, Unused Outputs Low) 180 CT V 160 / 140 c( E 100 I 80 8 60 40 ~CT V 120 k:==e'! tr ~ k-:1 7 - .,6 V 20 o~ o 10 20 ~ _r- ~ :.---- ABT 40 30 ----' ~~ F 50 60 70 80 90 100 Frequency - MHz Figure 3. Icc Versus Frequency (All Outputs Switching) 100 F 80 ~ 60 I ~ 40 !-----"" ~ ..-;if.--l ~ CT ~ 20 ~-- o ~ / ---? V~ o 10 20 V" V V -- --- f..-"" v--' ~ 30 40 V r-- 50 60 70 ~ ~ - --- - -BCT~ ABT 80 90 100 Frequency - MHz Figure 4. Icc Versus Frequency (All SWitching, 50% Duty Cycle Enabled) 5-35 Typical applications for bus-interface devices require them to be disabled or in the standby mode during certain periods of time, for instance, while other devices access the bus. This can result in a large decrease in current consumption for ABT, BCT, and ACT devices, which have low-standby current. These values are given in the data sheets as ICC for ACT and Iccz for ABT (250 IJA) and BCT (~ 10 mA). Current-consumption data versus percent duty cycle enabled is shown in Figure 5. The frequency of the data is held constant at 25 MHz and all outputs are switching. 80 - 70 F 80 50 ~ t-- .......CT 40 30 20 10 .~ .......V V o --"" o 10 :,..::-- . ..- f.-- f-" 20 30 40 ,-- . .. . i-- I-- .....- f-- 60 60 70 . . .. . -- ACT ... P --ra; 80 90 100 Duty Cycle Enabled - % Figure 5. IcC Versus Duty Cycle Enabled (25 MHz) The power-consumption data provided is limited to a small range of variations. However, using this data, along with standard formulas, power consumption can be calculated for specific applications. Power Calculations When calculating the total power consumption of a circuit, both the static and the dynamic currents must be taken into account. Both bipolar and BiCMOS devices have varying static-current levels, depending on the state of the output (IcCLo IcCH, or ICCz), while a CMOS device has a single value for Icc. These values can be found in the individual data sheets. ACT and ABT inputs, when driven at TTL levels, also consume additional current because they may not be driven all the way to VCC or GND; therefore, the input transistors are not completely turned off. This value is known as Mcc and is provided in the data sheet. Dynamic power consumption results from charging and discharging of both internal parasitic capacitances and external load capacitance. The parameter for ACT and AC devices that accounts for the parasitic capacitances is known as Cpd' It is obtained using equation 2 and is found in the data sheet. (l) Where: fi - input frequency (Hz) Vcc - supply voltage (V) CL - load capacitance (F) Icc - measured value of current into the device Although a Cpd value is not provided for ABT, BCT, or F devices, the Icc versus' frequency curves display essentially the same information. The slope of the curve provides a value in the form of mA/(MHz x bit), which when multiplied by the nUmber of outputs switching and the desired frequency, provides the dynamic power dissipated by the device without the load current. Equations 3 through 7 can be used to calculate total power for CMOS, bipolar, and BiCMOS devices: P,. - PS(static) + PD(dynamic) 5-36 (3) CMOS AC (CMOS-level inputs) Ps -Vee x lee PD - [(Cpd + cr) x vee2 x ftl Nsw (4) ACf (TIL-level inputs) Ps - Vee lice + (N'ITLxAlee x DCd>] PD - [(Cpd + Cr) x Vee2 x fdNsw (5) BICMOs/Blpolar Ps - Vee [DCen(NH x leeHINT + NL x leaJNT) + (I-DCen)Iccz] + (N'ITL x Alee x DCd> (6) Note: alee - 0 for bipolar devices Po - [DCen x Nsw x Vee x fl x (VOH- VOr) x CLI + [DCen x Nsw x Vee f2 x (mAlMHz x bit)] x 10-3 (7) Where: Vee lee IeCL IeCH lecz alee DCen DCd NH NL Nsw NT fl f2 VOH VOL CL .= mA/(MHz x bit) - Supply voltage (V) Power-supply current (A) (from the data sheet) Power-supply current (A) when outputs are in low state (from the data sheet) Power-supply current (A) when outputs are in high state (from the data sheet) Power-supply current (A) when outputs are in high-impedance state (from the data sheet) Power-supply current (A) when inputs are at a TIL level (from the data sheet) % duty cycle enabled (50% - 0.5) % duty cycle of the data (50% - 0.5) Number of outputs in high state Number of outputs in low state Total number of outputs switching Total number of outputs Operating frequency (Hz) Operating frequency (MHz) Output voltage (V) in high state Output voltage (V) in low state Extemalload capacitance (F) Slope of the lee versus frequency curve Thermal-Resistance Values Design trends requiring board size reduction have made way for circuit manufacturers to produce fme-pitch packages that appear to threaten the reliability of systems due to further thermal constraints. As a leader in packaging technology, TI has done considerable research into the validity of traditional thermal measurements and data provided by circuit manufacturers. Unlike data-sheet parameters, where the industry has adopted a standard load for measurement (50 pf, 500 0), the measurement of alA has no standard to which all manufacturers comply. The problem facing the designer wishing to make comparisons of thermal data from several manufacturers is that this could be an apples-to-oranges type comparison. As a result, a software package has been developed at TI to allow designers to obtain thennal data based on their specific application. The validity and usefulness of the traditional approach to presenting alA values became a pressing issue when TI and another manufacturer measured an identical package and obtained results that varied by 40%. Extensive research led 5-37 to the conclusion that the methodology used to measure ElJA did not cause the discrepancy but the physical aspects such as trace length, trace width, number of devices per board, and proximity of the other devices did. To demonstrate the extreme impact of trace length alone, Figure 6 shows the EllA values forTI's 48-pin SSOP at 0 LFMP and 250 LFMP with varying trace lengths. The 48-pin SSOP is shown in Figure 1 for a side-by-side comparison with the standard 24-pin SOle, the 24-pin SSOP, and the I OO-pin TQFP. The data in Figure 6 clearly shows the need for more complete thermal data, not simply a single data point. ,, 130 110 90 A_ -.... OLFPM .......... -A _ ~ --- --- --........... -- - 250 rPM 70 50 ~ o 0.2 0.4 0.8 0.8 1.0 1.2 1.4 1.8 1.8 2.0 Trace Length -In Figure 6. 48-Pln SSOP ElJA Versus Trace Length There are other methods to lower the EljA of a device. Using heat sinks or blowing air across a device certainly improves the ability to remove heat from its surface. Figure 7 provides EllA data for the 48-pin SSOP with trace lengths of200 mils and 1 inch while varying the amount of air flow. Although many applications tend to limit the amount of air flow, excellent benefits are possible with increased air flow. 120 . ... 100 80 80 --- .. . . .. . ... r--, ~ r--. r-- 200 Mils • i'-" ~ 1000 Mils .. -r---,r-- .:.:..:; 40 20 o 50 100 150 200 250 300 350 400 450 500 Air Flow - LFPM Figure 7. 48-Pln SSOP ElJA Versus Air Flow 5-38 Several variables that have a direct effect on @JA values were compared and results are shown in Figure 8. Surprisingly, the major contributing factor is trace length, not air flow. Once again, this validates the need for improvement not necessarily in the test methodology used to calculate @JA values, but certainly in the way those values are provided. D VARIAIU.s BAmlE ~ QQ!::!IBIBUIIQN Trace Length 75 mils - 2000 mils 41.4 AirFlow o LFM - 500 LFM 28.8 Board Extension ~ After Trace o mils -400 mils 6.5 ~ Board Extension After Package End o mils - 755 mils 2.6 ~ Trace Thickness 1oz-2oz 2.2 Trace Width 3 mlls-15 mils 2.1 Power 0.5W-1.5W 0.1 Total Interactions Betwaen Factors 16.3 a-HI EI -- Figure 8. 48-/56-Pln SSOP K-Factor Board Modeling TI provides @JA values for a variety of packages (including the SOle, SSOP, and QSOP) in a user-friendly software . package. The program allows designers to specify their conditions, such as trace length, air flow, proximity of other devices, and trace width in order to obtain realistic thermal solutions. 5-39 Summary How a system can avoid being a reliability nightmare in today's world where: • • • Eight-bit devices are being replaced by 16 and 32 bits in a single package, increasing the power. Higher operating frequencies add to the increase in power. Fine-pitch packages are reducing the amount of available surface area to remove heat from a device. Semiconductor manufacturers must take the first step and provide realistic and useful thermal information that will provide designers key variables to focus on for thermal management. References Thermal Software Contact the factory at (903) 868-7682. Power Dissipation Advanced CMOS Logic Designer's Handbook, Texas Instruments Incorporated, 1988, literature number SCAAOOIB SSOP Designer's Handbook, Texas Instruments Incorporated, 1991, literature number SCYAOOI 6-40 K-Factor Test-Board Design Impact on Thermal-Impedance Measurements LarryW. Nye Advanced System Logic - Semiconductor Group :lllExAs INSTRUMENTS SCAA022A 5-41 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable at the time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI prodl,.lcts in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize Inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 5-42 Contents ntle Page Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-45 Introduction ............................................................................. 5-45 Modeling Approach . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. S-46 Design of the Experiment .................................................................. 5-47 Orthogonal·Polynomial Expansion .....•..•...............................•...•............. 5-49 Verificadon: Equation Versus Actual Measurements ........................................... 5-52 Conclusion .............................................................................. 5-53 Acknowledgements .................................•........................... ,.......... 5-S3 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5-53 5-44 Abstract The rapid advancement in semiconductor-device technology is placing unprecedented demands on device-packaging technology. In an effort to meet system requirements for increased speed in smaller footprints, integrated circuit manufacturers are pushing existing packaging technology to new limits. Product performance is a function of both device and packaging technologies. In many instances, the thermallirnitations of the packaging system can severely restrict the performance of the device, thus limiting systems applications. System designers and integrated circuit manufacturers are becoming increasingly more concerned about accurate thermal characterization. There are several indices typically reported to reflect the thermal performance of a package. Thermal impedances, ElJA (junction to ambient) and ElJC (junction to case) are the most frequently used throughout the industry. Although there are several specifications on the administration of these tests and measurements, there is no universally accepted industry-wide standard. This lack of standardization promotes an apples-to-oranges comparison of published data, as well as inaccurate estimation of application performance. This paper focuses on the impact of the wind-tunnel k-factor test-board design parameters on reported ElJA results. By employing statistical experimental design techniques and finite element analysis (PEA), equations are derived that can be used to quickly normalize reported ElJA values under various test-board conditions. These mathematical equations are shown to correlate well with empirical wind-tunnel results. A computer program, THETACAVM, has been developed by Texas Instruments (TI) to assist system designers in understanding and comparing the thermal capabilities of packages sourced from various integrated-circuit manufacturers. Introduction The use of statistical design of experiment (DOE) techniques combined with PEA provides the engineering community with valuable tools for forecasting the behavior of a system or process. A natural marriage, the DOE and PEA combination allows the engineer to study a range of boundary conditions for numerous design factors and to analyze the impact and associated response for each factor and interaction within the system. With the use of orthogonal polynomial expansion techniques, experimental results can be effectively transformed into mathematical equations based on the strength of the various factors and associated interactions. These equations are useful for performing what-if analyses on a system or process. The thermal impedance (k-factor) of a package is defined as the increase injunction temperature above the ambient due to the power dissipated by the device and is measured in degrees Celsius per watt. There are two indices commonly used to describe the thermal characteristics of an integrated-circuit package: ElJA and ElJC' ElJC is the thermal impedance from the integrated-circuit die junction to the package external case and is typically measured in a circulating bath of an inert liquid simulating an infinite heat sink. ElJA is the most widely used and least understood measurement utilized in package selection for application design criteria. There are actually two 9lA measurements commonly reported: Socket mounted and measured in 1 cubic foot of still air Board mounted and measured in a wind tunnel at various air flows In case #1, the type of socket and socket manufacturer should be noted when comparing reported values, as they can significantly impact the reported ElJA values. The focus of this report and the development of the THETACAL software tools are targeted to address the problems associated with board-mounted ElJA values reported under wind-tunnel conditions (case #2). As indicated earlier, there are several specifications on the administration of wind-tunnel tests and measurements, but there is no universally accepted standard. Wmd-tunnel dimensions and k-factOl;' board construction techniques vary widely from manufacturer to manufacturer and can dramatically impact the reported ElJA values. Figure 1 shows the differences noted on an identical package measured on two different integrated-circuit manufacturer's k-factor boards in the same wind tunnel. THETACAL is a trademark of Texas Instruments Incorporated. 5-45 In Figure I, there is roughly a 45°C/W difference (still air) between manufacturers induced by the k-factor board alone. Considering this fact, it is possible for a wide range of reported k-factorvalues to exist for any given package, depending on the test-board design employed by the manufacturer. Systems designers unaware of these differences can be artificially restricted in packaging selections based only on reported thermal-impedance values. 120 11 " ... ~ 100 90 80 -- -- ___ TI Short-Trace Board (actual) COmpetItor's Long-Trace Board (actual) -+I--- ~ 70 6 o~~ 50 ---- t--- --~ ~ ----- 40 30 o 50 100 150 200 250 300 350 400 450 500 AlrFlow-lfm Figure 1. Competitor Versus TI K-Factor Board (52-pin· MQFP) In an effort to provide system designers with an accurate tool for estimating the impact ofk-factor test-board designs on reported thermal impedance values (aJA), TI has developed the THETACAL software package. By employing statistical experimental design techniques combined with finite element analysis tools, equations can be derived to perform what-lfanalyses varying single or multiple input parameters simultaneously to accurately estimate their impact on ~e desired response (aJA)' This tool allows designers to compare the thermal performance of a given package sourced from various manufacturers on an apples-to-apples basis. In addition, the software can be utilized to better understand the influence of the various board-related parameters and their impact on aJA. Modeling Approach The following four-phase methodology was utilized to develop the THETACAL software package: 1. Design of the experiment using orthogonal arrays 2. Modeling the package using PEA (ABAQUSTM) tools 3. Expansion of the matrix results into an orthogonal-polynomial equations 4. Mathematical simulation of the thermal response (aJA) for performing what-if analyses OnCe the orthogonal-polynomial equations are completed and verified, they are incorporated into the THETACAL software environment. Each package type and pin count are evaluated separately to ensure accuracy in the equations. ABAQUS is a trademark of Texas Instruments Incorporated. 5-46 Design of the Experiment Typical k-factor boards are constructed using an FR4 or polyimide-composite substrate. Copper traces of varying dimensions are fabricated upon this substrate for package mounting and to complete the electrical connections required for k-factor testing. The parameters in Table 1 are most typically varied in k-factor board construction (see Figure 2) and are the focus of this study and the THETACAL software development. Table 1. Evaluation Parameters TRACE LENGTH (mils) AIRFLOW (11m) POWER (watts) 50 750 0 500 0.5 1.5 Low High TRACE WIDTH (mils) BOARDZ EXTENSION (mils) BOARDY EXTENSION (mils) TRACE THICKNESS (mils) 3 0 550 0 550 2.8 15 1.4 Board Extension After Trace (or V) T Trace Length -r~~~~~~~~~ OUT o Board Extension After Package (or Z) Trace Width NOTE: Y and Z are used for PLCC and QFP packages. Figure 2. Test Board An L16 orthogonal arrayl was selected as the design vehicle used to evaluate the impact of all identified main factors and their expected interactions within the k-factor test-board system. Since the focus of this study is to derive mathematical equations to be utilized for estimation purposes, it is extremely important to properly define the layout of the experiment to capture all sources of variability; i.e., the accuracy of the equation is best when the unresolved variability is minimized. Once the experiment has been properly defined, appropriate models are prepared per the matrix and processed through the fmite element analysis thermal solver (ABAQUS). A typical data set, as returned by the FEA software, is shown in the far right column of Table 2. Statistical analysis is done and orthogonal-polynomial equations can be derived from the completed data set. 5--47 Table 2. Matrix Definition and Results RUN 1 2 3 4 5 6 7 8 9 10 11 12' TRACE LENGTH (mils) AIRFLOW (lfm) POWER (watts) 50 50 0 0 0.5 50 50 50 0 0 500 50 50 50 750 750 750 500 500 500 0 0 13 750 750 14 15 750 750 16 750 0 0 500 500 0.5 1.5 1.5 0.5 0.5 1.5 1.5 0.5 0.5 1.5 1.5 0.5 500 0.5 1.5 500 1.5 TRACE WIDTH (mils) BOARDZ EXTENSION (mils) BOARDY EXTENSION (mils) TRACE THICKNESS (mils) MODEL 3 15 0 550 0 550 3 15 0 550 550 0 151.3 100.6 119.2 121.2 3 15 3 15 3 15 550 0 550 0 0 0 550 550 0 0 550 1.4 2.8 2.8 1.4 2.8 1.4 1.4 2.8 2.8 1.4 3 15 550 0 550 3 15 550 0 0 550 3 550 550 2.8 2.8 15 0 0 1.4 550 0 1.4 2.8 1.4 ("~A W) 66.5 68.6 61.6 75.5 77.7 71.6 84.0 67.5 56.1 50.0 53.5 52.5 By using statistical tools such as the effects table (see Table 3) and analysis of variance (ANOYA) table (see Table 4), one can analyze the impact of each individual factor and associated interactions within the system. In Table 4, the air-flow parameter accounts for 47.3% of the total variability measured within the ranges probed, followed closely by the trace length parameter with a 31.4% contribution. In addition, the trace length by air-flow interaction accounts for 8.6% of the variability measured in the system. In total, the airflow, trace length, and the interaction between these two factors accounts for 87.3% of the total variability of the system. The sum of the contribution of the remaining factors is a mere 12.7%. When analyzing the effects table (see Table 3), one can see that employing a longer trace length (see high-level average) on a k-factor board achieves basically the same impact as using a high velocity of moving air across the package. Longer trace lengths on a k-factor board essentially act as built-in heat spreaders on the board and are one of the primary reasons for the dramatic differences noted between manufacturers in reported aJA values on identical packages. In light of the fact, it is essential that designers understand the measurement conditions employed when determining the fitness for use of a package for a given application. Table 3. Effects Table SUM OF SQUARES LOW-LEVEL AVERAGE °ClW HIGH-LEVEL AVERAGE °ClW EFFECToClW Airflow 5959.84 3956.41 1079.12 402.00 99.14 95.56 60.54 64.11 -38.60 Trace length 88.05 84.85 71.63 74.83 84.20 83.74 83.54 75.48 75.94 76.14 76.31 76.76 FACTOR Trace length by air flow BoardZ ext. Trace length by board Y ext. Trace width Board Yext. Trace thickness Trace length by board Z ext. Trace length by trace thickness Trace width by trace thickness Trace length by power Air flow by power 5-48 304.50 243.36 219.04 198.81 151.29 83.36 82.91 ,40.32 81.43 15.60 8.41 7.84 80.83 80.56 80.54 78.25 78.85 79.11 79.14 -31.45 -16.42 -10.02 -8.72 -7.80 -7.40 -7.05 -6.15 -3.18 -1.98 -1.45 -1.40 Table 4. Analysis of Variance (ANOVA) (50-mil thru 7So-mll trace lengths) FACTOR DOF SUM OF SQUARES MEAN SQUARES F·TEST % CONTRIBUTION STATISTICAL SIGNIFICANCE Airflow 1 5959.84 5959.84 2995 47.3 99% Trace length 1 3956.41 3956.41 1988 31.4 99% Trace length by airflow 1 1079.12 1079:12 542 8.6 99% Board Z ext. 1 402.00 402.00 202 3.2 99% Trace length by board Y ext. 1 304.50 304.50 153 2.4 99% Trace width 1 243.36 243.36 122 1.9 99% BoardYext. 1 219.04 219.04 110 1.7 99% Trace thickness 1 198.81 198.81 100 1.6 99% Trace length by board Z ext. Trace length by trace thickness 1 1 151.29 151.29 76 95% 40.32 40.32 20 1.2 0.3 Trace width by trace thickness 1 15.60 15.60 8 0.1 Trace length by power 1 8.41 8.41 4 0.1 Air flow by power 1 7.84 7.84 4 Power 1 3.42 Trace length by trace width 1 95% Error pool: Residual 0.56 0.005 Total residual 2 3.98 Total 15 12590.54 1.99 0.2 100 Orthogonal-Polynomial Expansion The results of an orthogonal array can be easily expanded into a powerful orthogonal-polynomial equation. Employing statistical principles, it is possible to construct an accurate mathematical model to quickly estimate the thennal response of a package by analyzing the impact of critical design parameters and key interactions. There has long been a need for an effective thennal calculator that can accurately and reliably estimate the thennal condition of a package under various design parameters, i.e., to analyze the impact of various material, dimensional, and air-flow conditions. Orthogonal-polynomial equations provide such a tool. Utilizing this approach, the solving power of the PEA software can be effectively transfonned into a mathematical equation or a system of equations for the desired response for performing what-if analyses within the ranges probed. The general equation fonnat for a two-level orthogonal array with interactions follows. Higher-order equations for nonlinear responses, not described in this work, can also be employed using this technique2. General Equation Format a JA = 8M: +bl(a)(a-a) [linear tenns for factor a] + bl(b)(b - b) + bll(a _ x _ b)ea - a)(b - b) .. · [linear tenns for factor b] [interaction tenns for factor a - x -b] 5-49 Where: aJA - Predicted matrix response 8M: - Average matrix response b I(i) - Coefficient of the linear response for i lvl(i) - Number of factor levels at indicated setting for factor i h(i) - Il setting between factor levels b II (i) - Coefficient of interaction response for i = Factor input variable for what-if analysis a a - Average of factor settings Factor CoeffIcIents Linear coefficient: b 1(a) _ -AI + Ai -::-;,f--_-:--'" r * ""'(a)h(a) - Where: Ai - Sum of factor response at level indicated r - Number of runs per factor level AS - (1 for a 2-level factor) (2 for a ;3-level factor) heal = Factor Il setting Interaction coefficient: b 11 (A-x-B) _ [(AI - + B I) - (AI + B 2)] - [(A 2 + B I) - (A 2 r*AsAhA * ASBhB + B 2)] Once the equation is derived, the initial test conditions can be plugged into the equation to check for accuracy against the original modeled parameters as shown in Table 5. If the experiment has been properly designed to capture all significant sources of variability, the equation results should closely match the modeled results. If the error term is minimal (see Table 4), as in this case, the equation matches the modeled results exactly. As the error increases, the accuracy of the equation decreases. At this point, the power of the FEA is transformed into a simple mathematical model for this response within the ranges probed for all parameters. It is now possible to vary individual or multiple input parameters (within the ranges studied) for performing what-if analyses. As with any simulation, equation results should be tested against empirical results to ensure proper accuracy. If the desired accuracy is not achieved, the input models should be reevaluated and adjusted as required. HO Table 5. Equation Versus Model RUN 1 2 3 4 5 6 7 TRACE LENGTH (mils) AIR FLOW (lfm) POWER (watts) 50 50 50 50 50 50 0 0 0 0 0.5 0.5 1.5 1.5 500 500 500 0.5 0.5 1.5 1.5 0.5 8 9 10 11 50 50 50 750 ' 750 12 13 750 750 14 15 750 750 750 16 500 0 0 0 0 500 500 500 500 TRACE WIDTH (mils) BOARDZ EXTENSION (mils) BOARDY EXTENSION (mils) TRACE THICKNESS (mils) MODEL 8JA ("C/W) 0 0 550 1.4 2.8 2.8 1.4 2.8 151.3 100.6 119.2 121.2 66.5 1.4 1.4 2.8 2.8 1.4 1.4 68.6 61.6 75.5 77.7 71.6 84.0 2.8 67.5 84.0 67.5 1.4 2.8 2.8 56.1 50.0 53.5 56.1 50.0 53.5 1.4 52.5 52.5 3 15 550 0 550 3 15 3 15 550 0 3 15 550 0 550 0 0 550 550 0 0.5 1.5 3 15 3 0 550 0 0 550 550 1.5 0.5 15 3 550 550 0 0 0.5 1.5 1.5 15 3 15 0 550 550 550 0 0 EQUATION 8JA{"C/W) 151.3 100.6 119.2 121.2 66.5 68.6 61.6 75.5 77.7 71.6 120 11 .I v~~ 100 ~ ~ 90 80 ------ 7n. ...... TI Short-Trace Boerd (actual) ___ TI Short-Trace Boerd (equation) ___ Competitor's Long-Trace Board (actual) ~ Competitor's Long-Trace Boerd (equatIon) ~ ~~ _jt-r-- 60 50 40 30 o 50 100 -- ----.....:: ........... r-- I--......... ""'":l~ r-- 150 ~ 200 250 300 350 400 450 .......... -:ll' 500 AlrFlow-lfm Figure 3. 52-Pin MQFP (actual versus equation) 5-51 Verification: Equation Versus Actual Measurements The air-flow and trace-length parameters were the most dominant influences noted with respect to 8JA' Wind-tunnel measurements taken on k-factor boards with both short- and long-trace conditions were compared against the orthogonal-polynomial equation. Figure 3 shows that the equation accurately estimates the 8JA under radically different k-factor board conditions. Individual FEA-model runs in the orthogonal array used to derive the equation can take several hours of computer processing time. depending on the type of workstation used. In an effort to minimize the modeling time required for each package. 2-1evel orthogonal arrays were used to minimize the number of runs required to approximate the system. Due to the extremely broad range of trace lengths that can be employed in k-factor test board design (50 mils to 2000 mils). and considering the dramatic influence of the trace-length parameter on 8JA. two separate matrices are evaluated for each package. The impact on the 8JA induced by the trace-length parameter is quite dramatic on most packages between 50 mils and 750 mils. From 750 mils to 2000 mils. the impactis less dramatic. In an effort to achieve acceptable equation resolution using linear approximations. two matrices were evaluated for each package. The first matrix focused on the shorter trace lengths and the second matrix focused on the longer traces. In the mETACAL software. this is transparent to the user upon input; however. minor discontinuities may be noted where the equations converge. The equation(s) can now be utilized to perform what-if analyses on the various input parameters. Figure 4 indicates the impact that the trace length has on 8JA at various wind-speed conditions. The knee of the curve at approximately 0.75 inches (750 mils) is the point where the equations from the two matrices converge. The mETACAL software can be used to evaluate any of the parameters considered in the study in a similar fashion. The equations used in the mETACAL software package provide the end users with the power of FEA capabilities (for the 8JA response) instantaneously and requires no workstation or special skills to use. 140 ..... StliIAlr ____ 250lfm . . . 500lfm - 120 ~~ 100 ~ -- ---.~ ~ 60 ... - 60 40 20 o 0.2 0.4 ~ ..... ..... --- - --- 0.6 0.8 1.2 ..... ..... ..... "Y "Y "Y - - - 1.4 Trace Length -Inches Figure 4. 52·Pin MQFP (trace length effect) - 1.6 1.8 2 Conclusion Thennal management of semiconductor packages is becoming increasingly more critical with the move to smaller package geometries and higher power requirements. In order to meet increasingly challenging design goals, systems designers and end users of integrated-circuit packages must be able to make infonned decisions on the fitness for use of a package based on thennal considerations. K-factor test-board construction can dramatically impact reported ElJA results and promote restrictions in package selection and system-perfonnance specifications. The THETACAL software package provides users with an effective tool to nonnalize reported ElJA values and assist in making infonned decisions on package selection'to reach design goals. Acknowledgements The author wishes to thank the following persons for their contributions to this work: Paul Noel for his exten~jve efforts in developing the THETACAL software enviromnent, Dr. Jerald Parker, Professor Jim Cutbirth, and the team of engineering students at Oklahoma Christian University of Science and Arts for their efforts in perfonning the thennal simulations, and Jim Fielding of Texas Instruments for his extensive modeling efforts. In addition, the author would like to thank Darvin Edwards, Steve Groothius, and Lon McKinstry of Texas Instruments for their support in the areas of finite element analysis and k-factor measurement. References 1. Genichi Taguchi, "Introduction to Quality Engineering," page 183, UNlPUB/Kraus International Publications, 1986. 2. Genichi Taguchi, "Systems of Experimental Design," VolUme 2, pages 599-622, UNlPUB/Kraus International Publications, 1988. 5-53 5-54 More Power in Less Space: A Thermal Enhancement Solution for Thin Packages Mary Helmick, Larry Nye, and Edgar Zuniga Advanced System Logic - Semiconductor Group Reprinted from Texas Instruments Technical Journal; Vol. 11, No.4, July-August 1994 ~lExAs INSTRUMENTS SCAU001A 5-55 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of its semiconductor products and related software to the specifications applicable atthe time of sale in accordance with Tl's standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use of TI products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications asSistance, customer product design, software performance, or infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. Copyright © 1996, Texas Instruments Incorporated 5-56 ENGINEERING TECHNOLOGY More Power in Less Space: A Thermal Enhancement Solution for Thin Packages Integrated circuit packaging technology can no longer be treated as a secondary consideration to circuit design. The packaging system is an integral part of the device function, composed of sophisticated materials and complex assembly processes that balances many diverse factors, dramatically impacting the device performance. Often, package design must now be done concurrent with circuit design, understanding the limitations of each on the other. Recent advances in wafer fabrication technology have forced IC package designers to provide packaging solutions for higher power in smaller spaces. Traditionally, shrinking the size of plastic packages restricts the device performance due to thermal constraints. Using the equivalent 8- and 16-bit functions as the board space benchmark, a package was needed to accom- Abstract: T RENDS toward higher device functionality in smaller space have driven the development of space-efficient packages. Integrated circuit (lC) devices have evolved from lowpin-count, coarse-pitch, throughhole packages to high-pin count, fine-pitch, surface-mount packages. This evolution has placed new thermal management demands on IC packaging technology. As plastic packages shrink in area and thickness, thermal impedances increase, limiting the power and frequency at which devices can operate. To take advantage of the increasing capability of IC devices, solutions must be found for smaller packages that can dissipate high power. A traditional option for thermal management is to add external heat sinks to conventional plastic packages, modate 32- and 36-bit logic devices. The resulting package size presented thermal management problems that had to be addressed with some revolutionary approaches. Additional design goals were the equivalent package reliability to existing packages and the ability to produce the package at an acceptable cost for the target market. A cross-functional team was structured to include resources from package design, chip design, package assembly, reliability and device testing. Design for Manufacturability concepts were used to meet six sigma process capability on all aspects of the packaging system. The resulting design was a thermally enhanced thin quad flat package (TQFP TEP) that can dissipate 2.4 watts of power in a 256-mm2 board area, assuming 25°C ambient temperature and 150°C maximum junction temperature. but emerging applications (e.g., laptop and notebook PCs) have placed additional clearance constraints, preventing this approach. Customer requirements for this package design were: meet existing package reliability levels, 2.4 W power dissipation, footprint smaller than the equivalent function in multiple packages and manufacturable at acceptable cost. A design team was structured to include chip designers, assembly process engineers, package design engineers and key component suppliers to facilitate the development process. Design Criteria A new package was required for a family of 32- and 36-bit bus interface logic components. Several constraints were placed on the design of the package: the package must have 100- and 120-pin configurations; the board area must be no larger than that required for the equivalent function in 8- or 16bit versions; the package must dissipate 2.4 watts without an external heat sink; and the manufacturing cost must be competitive in the marketplace. Design Approach Considering the design constraints, two form factors were analyzed; a dual in-line design and a quad design. Device architecture preferred the in-line design, but considering available die bonding and leadframe manufacturing capabilities, an in-line design that met the board space constraint would require a lead pitch of 0.3 mm. Research throughout the marketplace indicated that by 1993, O.4-mm pitch Mary Helmick, Larry Nye and Edgar Zuniga JULV-AUGUST 1994 5-57 ENGINEERING TECHNOLOGY was acceptable, but O.3-mm pitch was not compatible with current board mount capabilities. The 0.4mm outer lead pitch quad design was chosen. Two approaches to thermal enhancement were considered (Figure 1) a cavity-type lidded plastic package and an exposed heat slug plastic package. Two heat slug options were evaluated; a heat slug attached to a conventional die pad and direct die attach to a heat slug. All designs were evaluated in chip-up and chip-down configurations. Thermal models and analytical data showed that both design approaches can dissipate the required power. For each design, the chip-down configuration maximizes the heat transfer from the package, especially with forced convection. In this configuration, the customer has the option to use an external heat sink if the application demands (Table D. The cavity design had addi- Table I. Modeled Power Dissipation of Various Package Configurations. ..........·,.~,_-~~~:1.1.•~ ·muIm~J~""~·.. CbipOQWll t-ieatSI\4Sty1e ChIp.U,p . Cavity Package - Chip Down Heat Slug Package (Die Pad Attached) - Chip Up Heat Slug Package (Direct Attach) - Chip Up Heat Slug Package (Die Pad Attached) - Chip Down Heat Slug Package (Direct Attach) - Chip Down Figure 1. Package Configurations Considered. TITECHNICAL JOURNAL 5-58 ENGINEERING TECHNOLOGY tional process complexity and possible reliability problems due to moisture ingress, thermal mismatch and bond wire movement. The heat slug design used conventional processes and equipment. The heat slug design in the chipdown configuration was chosen based on its ability to meet the design criteria, compatibility with existing processes, potential for improved reliability performance and more cost-effective manufacturing flow. The package was designed to meet the emerging JEDEC standard for 1.4-mm thick thin quad flat packages (MO-136) (Figure 2). Material Selection Thermal constraints drove the use of a copper leadframe and a copper heat slug (Figure 3). Based on positive past results with finepitch packages, a palladium preplated leadframe was selected. The heat slug was directly attached to the leadframe due to the 1.4-mm thickness of the package. This design, as opposed to a slug attached to the die pad or a drop in heat spreader, gives a more direct heat path to the outside of the package and simplifies the assembly process. The heat slug is attached to the lead frame using a two-sided adhesive film. A polyimide film was chosen due to its stability at elevated processing temperatures and its low moisture absorption. The adhesive film also provides electrical isolation between the heat slug and the lead fingers. The heat slug required a surface treatment to inhibit corrosion since it was exposed to air. A proprietary surface treatment was selected based on its superior 1-._ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _--1 adhesion performance with molding compounds. Shear tests were Figure 3. Leadframe with Attached Heat Slug. JUL V-AUGUST Package Cross Section Figure 2. Package Outline Drawing -100 TQFP TEP. o o Bottom View Top View 1994 5-59 ENGINEERING TECHNOLOGY run to compare the adhesion of "Button" of various mold compounds to cop¥Mold Compound per treated with this proprietary process and standard leadframe materials by shearing "buttons" of mold compound off of material samples (Figure 4). The "buttons" of compound are molded to the heat slugs or conventional dieattach pads and subjected to the t:;CCCCCCCCCCCYv,~Treated Heat Slug normal post-mold cure process or Die Attach Pad before being sheared from the slugs. The shear force required to remove the "button" from the Figure 4. "Button" Shear Test Setup. sample provides a relative adhesion value for different compounds and leadframe or heat slug materials (Table ID. Finite element analysis was Table II. Average Shear Force Normalized for Button Area (psi). used to determine the von Mises stress levels in the silicon chip, at the mold compound/chip surface 31 interface and at the die attach/ " . heat slug interface using different B size and shape heat slugs (Figure 5). Heat slugs with troughs were considered in an attempt to deflect the point of maximum stress away from the ball bond area. For exam- . extra processing step was not The leadframe design employs pIe, in Cases 1 and 3 (Figure 5), required to obtain the desired a preplated palladium finish. The while the stress at the ball bond package reliability. slug is attached to the leadframe was reduced from 7 K psi to 5.9 K using a two-sided adhesive film psi with the addition of a trough, Design Features applied at high temperature. The stresses at the die attach/heat slug The heat slug was designed to film is cut in a "window-frame" interface increased from 62 K psi maximize the exposed surface configuration, providing support to 70 K psi. The die attach/heat area for heat flux and to accom- for all lead fingers. The package slug interface is the weakest "link" modate the largest possible chip. assembly operation starts with an in the package, therefore the The heat slug was designed with a assembled lead frame, which is trough design was not chosen. flange on the top surface. This processed using the same equipHeat slug thickness and shape flange provides both a locking ment and flows as conventional was optimized by balancing ther- mechanism with the mold com- plastic packages. mal and stress responses to meet pound and a longer surface interThe ~hip-down configuration of the design guidelines. face between the compound and this package maximizes the therMold compound candidates the slug. If moisture penetrates mal performance. The leadframe were evaluated for wire sweep along the slug-mold compound design allows the package to be and autoclave performance and interface, this moisture would processed conventionally (chipgold wire was chosen to minimize have a longer path to reach the die up) through the molding operawire sweep in this package. Sili- surface. The slug shape also tion (Figure 7). The package oriencone die coating was evaluated for allows for easy orientation in tation is changed before lead form, enhanced die corrosion resistance, automated assembly processes but this change is transparent to but test results showed that this (FigUre 6). the manufacturing process. ~. TI 5-60 TECHNICAL JOURNAL ENGINEERING TECHNOLOGY Figure 7. Assembly Process Flow. JULy-AUGUST 1994 5-61 ENGINEERING TECHNOLOGY Thermal Performance on the system conditions. (See Fig- Both model and analytical data shows that the heat slug package can meet the design constraint of dissipating 2.4 W in either the chip-up or chip-down configurations. The chip-down style was chosen due to its thermal efficiency when airflow is present in the system. When considering if the required heat could be dissipated using a thin package, different thicknesses of heat slugs were modeled. The results showed that a heat slug as thin as 0.015" could be used without sacrificing the desired performance (Table lID. Actual power diSSipation of all packages is dependent on the thermal impedance of the package, the system ambient temperature, the maximum allowable junction temperature and the available airflow. The TQFP TEP package dissipates between 1 and 5 watts, depending ure 8 and Figure 9.) The slug provides a spreading effect for the heat generated by the chip. Since the chip is attached directly to the heat slug, the die attach material is the only therrri.al resistance interface between the chip and the slug. This interface is relatively short, typically 0.001" and the heat is easily drawn away from the chip. Conventional plastic packages. receive some heat spreading effects from the die attach pad, but the heat still has to pass through a layer of plastic molding compound before reaching the ambient air. Compared to the conventional plastic package of the same outline, the heat slug package thermal impedance is 30% lower (Figure 10). ~sseDlblyProcess DevelopDlent Design for manufacturability concepts were used to meet six sigma process capability in every assem- Table III. Maximum Power Dissipation of 1.4mm Heat Slug Package (Watts) - Thermal Model Data. (Assumes 25°C Ambient Temperature and 150°C Maximum Junction Temperature.) 1.4mm Heat Slug Power Analysis (100/120 Pin) 50 5 49.8 4.5 45 I'" ..... 25°CAMB A 35°CAMB o 45°CAMB .55°CAMB 6. 65°CAMB Figure 8. Power Dissipation at Scheduled Ambient Temperatures. TI TECHNICAL JOURNAL 5-62 ENGINEERING TECHNOLOGY 1.4mm Heat Slug Power Analysis (100/120 Pin) 50 5.5 49.8 5 45 ~ ~ 4.5 4 40 3.5 35 3 .~ (!> 2.5 30 2.5 2.1 1.7 n--_ _ _ _ ==~~~~~~-fr9-2":~ 25 2 __ ~~1J 1.3 . . 0.9 fr "iii ~ ~ D.. 1.5 24.2 ---ht"'c.r 20 0.5 oLFM 100 LFM 500 LFM 250 LFM Airflow (LFM) • Sja <> 150°C Max Tj .. 130°C Max Tj o • 90°C MaxTj 6. 70°C Max Tj 110°C Max Tj Figure 9. Power Dissipation at Selected Maximum Junction Temperatures. Actual K-Factor 8or-----------------------------------------, 72.7 70 =- 60 ~ E 50 40 30 2o~----~--------------~------------~~--~ o 250 Airflow (LFM) • Conventional 1.4mm o 1.4mm Heat Slug Figure 10. Thermal Impedance of Conventional vs. Heat Slug Packages. JULy-AUGUST bly process operation. Stress reduction and good thermal dissipation required a low stress die attach material with minimum voiding, thick bond line (>0.001") and excellent adhesion characteristics. Several die attach materials, dispensing needles and parameters were evaluated to find the best combination. The evaluation efforts combined with the improved adhesion characteristics of the heat slug surface treatment yielded a substantial improvement in die shear strength compared with conventional Pd plated lead frames, as Table IV shows. Optimum epoxy application parameters and customized dispenser needles designs allowed us to meet 100% epoxy coverage, less than 10% epoxy voiding and a minimum bond line thickness of 0.001" (Figure 11). 1994 5-63 ENGINEERING TECHNOLOGY Table IV. Average Die Shear Force in Kg. DIe Pad Die . . . Coating Fcirce standard ....... Pd Plated Die Pad 13.0 2.7 41.2 8.6 Heal Slug wIIh Surfaoe treatment DIe . . . precise dam bar removal tooling was required due to the O.4-mm outer lead pitch. A "cam form" process was used to form leads, preventing damage to the preplated 5-mil-thick leads. Optical lead inspection was employed to confirm outgoing coplanarity and lead true position. Package Reliability One of the design constraints for developing a thin thermallyenhanced package was that its reliability meet the current reliability level of conventional plastic packages. The heat slug package was tested using the same qualification requirements for all plastic packages. All electrical tests were preconditioned using 168 hours exposure at 85°C/60%RH followed by two reflow operations. Table VI Figure 11. Epoxy Dispensing Pattern. The heat slug leadframe subassembly required substantial wire bond process development. Heater block design was critical due to the presence of adhesive film in the bonding areas. The fine pitch leadframe fingers and the chip bonding pads required a new capillary design and utilization of advanced wire looping techniques (Figure 12). Mold compounds were evaluated for wire sweep, voiding and moldability (Table V). The highadhesion properties of the mold compound candidates required a unique mold die design, to assist in mold release. Both bond process and mold process optimization were used to minimize wire deflection during the encapsulation process (Figure 13). Devices are symbolized using a proprietary laser process. Ultra Figure 12. Wire Looping -120 TQFP. ---------------------------------------------TI TECHNICAL JOURNAL 5-64 ENGINEERING TECHNOLOGY Table V. Mold Compound Evaluations 100 Pin TQFP Packages.) .... -CompoI.m4I Cempound A '~ 1.57 ~.(m!n) ~~) 7.02 :. 81U'.~J' 1.88 Wire Sweep. ("Wire Sweep % Deflection for Various Mold Compounds on ,'I1"\d .. ~ c.mpDIIiId 1.4 1.9 D .. ~ r .. 5.11r . ; .... , ..... un I. 8 1.16 COmpOunct ~ COIIIpDUIId F 1.4 1.11 1.9 • S. 1m ., '~!16 1.07 • "" 6.9'4 .-'.' 1.'. ping and storage using a "dry" packaging scheme. This package provides an excellent opportunity to eliminate the need for desiccated packaging and special handling in a customer's factory. Conclusions Figure 13. Molded Wires. describes the results of this reliability qualification testing. The moisture sensitivity tests were performed per the proposed IPC-SM-786A Levell and Level 2 conditions, 168 hours of 85 °C/85 %RH and 168 hours of 85 °C/60 %RH, respectively. The heat slug package passed both Levelland Level 2 moisture sensitivity tests, showing no degradation in delamination nor internal cracking after stressing. This is superior performance to a conventional plastic JULy-AUGUST TQFP package. The photos in Figure 14 show a typical crack in a TQFP package after Levell testing and the heat slug package after identical stress conditions (Figure 14). Level 1 conditions equate to an unlimited exposure time at factory floor conditions of 30°C/ 60%RH and no special "dry" packaging. The Level 2 conditions equate to an exposure time of one year at 30°C/60%RH, providing the devices are protected during ship- A thermally-enhanced plastic IC package has been developed by a cross-functional team from engineering and manufacturing to meet the constraints of space and power required by the customer. The package was developed with a primary goal of efficiency of manufacturing and compatibility with current plastic package assembly equipment and processes. The TQFP TEP uses a leadframe subassembly, with a heat slug attached by two-sided adhesive film. This leadframe is processed through conventional assembly processes and equipment, with a change of orientation after the mold process. The proprietary surface treatment of the copper heat slug provides a tenfold improvement in mold compound adhesion and a threefold improvement in die attach adhesion compared to conventional leadframe surface finishes. The slug also provides stability to the thin package, preventing warpage. This package represents a substantial step forward in the quest for plastic IC 1994 5-65 ENGINEERING TECHNOLOGY packages that can meet the everincreasing power and board space requirements of the electronics industry. Acknowledgments Typical Crack in TQFP Package Resulting from "Popcorn Effect" The authors would like to express their sincere appreciation to the following people, without whose contributions the development. of this package would not have been possible: Bobby O'Donley, Mario Magana, John Tellkamp, Pedro Cabezas, Mike Pomeroy, John Wiley, RIch Brook, Ray Purdom, TQFP TEP Package with No Cracks Figure 14. Cross Sections of TQFP Packages After 168 Hours 85°C/85%RH. Table VI. 120 Pin TQFP TEP Environmental Test Data. , (.," . ." "'-' ... ~~ ... ' .. ' J~., . . ,. .,~ ·· . Ws·· . ' ..... ,. /~' ': .,"" .... " . ':' '.". .... " ,.."".:,;.. ,'. . . ,,' .. ,.'< .; . ,0 . .: "':':"''':~:'' , .. ,' ' . MO ONO ONO TI TECHNICAL JOURNAL 5-66 ENGINEERING TECHNOLOGY Terrill Sallee, Jim Fielding, Tomas Luna, Brenda Gogue, Steve Groothius, Dick Shaw, Jim Sisco, Jay Alexander, Archie Sutton, Herb Wyman and others who may have been omitted. Edgar R. Zuniga Edgar Zuniga has been involved with the development of new packages since 1989, first as part of the group that developed the 48- and 56-pin SSOP packages. For the last two years he has served as the team leader in the development of the 100- and 120-pin TQFP TEP packages. Before this, Edgar spent two years as engineering section head manager for the assembly operations in the Flexible Assembly Module in Sherman (FAM) and one year as assembly process engineer. He joined TI in 1975 at the TI plant in EI Salvador as a process engineer in the assembly operation. He transferred to Sherman in 1986. Edgar received a B. S. in electrical engineering in 1977 from the Catholic University, San Salvador, EI Salvador, and an M. S. in engineering science from the University of Texas at Dallas in 1993. JULy-AUGUST LarryNye Larry Nye is currently a member of the package development team within the Advanced System Logic Organization. Since 1983, Larry has been heavily involved in die attach and bonding process development activities including programs such as palladium bonding process development, copper wire bonding, low temperature bonding, hermetic chip, bonds over active circuits (BOAC), rapid cure process (RCP), and others. Besides his development responsibilities, Larry serves as the ASL representative on the Worldwide Mount Commodity Team, the Worldwide Gold Wire Commodity Team and the Worldwide Bond Process Standardization Team. Additional responsibilities include thermal and stress modeling activities along with material characterization and qualification. He joined TI in 1983. Larry has presented papers twice at the international electronic packaging conference on thermal analysis of semiconductor packages. He has written other papers on palladium bonding and thermal modeling of semiconductor packages. In addition, Larry has served as a co-author on papers dealing with thermal modeling, experimental design, semiconductor package design, and bonds over active circuits (BOAC). He currently has two patents pending in the U.S. patent office. Mary Helmick Mary Helmick is part of the package development group for the Advanced System Logic Department in Sherman. Mary joined TI in 1990 as part of the 48- and 56-pin SSOP package development team in the ASL Test/Finish area (formerly FAM). Since 1991 she has been part of ASL packaging engineering, working on the 100/120 pin Heat Slug Package Development Team. She received her bachelor of science in mechanical engineering from Purdue University in 1986 and her master's degree in computer integrated design and manufacturing from George Washington University 0 in 1989. 1994 5-67 5-68 FIFO Surface-Mount Packages for PCMCIA Applications Tom Jackson Advanced System Logic - Semiconductor Group ~TEXAS INSTRUMENTS SDMA001A 5-69 IMPORTANT NOTICE Texas Instruments (TI) reserves the right to make changes to its products or to discontinue any semiconductor product or service without notice, and advises its customers to obtain the latest version of relevant information to verify, before placing orders, that the information being relied on is current. TI warrants performance of Its semiconductor products and related software to the specifications applicableatthetime of sale in accordance with Tl's Standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except.those mandated by government requirements. Certain applications using semiconductor products may involve potential risks of death, personal injury, or severe property or environmental damage ("Critical Applications"). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, INTENDED, AUTHORIZED, OR WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT APPLICATIONS, DEVICES OR SYSTEMS OR OTHER CRITICAL APPLICATIONS. Inclusion of TI products in such applications is understood to be fully at the risk of the customer. Use ofTl products in such applications requires the written approval of an appropriate TI officer. Questions concerning potential risk applications should be directed to TI through a local SC sales office. In order to minimize risks associated with the customer's applications, adequate design and operating safeguards should be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance, customer product design, software performance, or Infringement of patents or services described herein. Nor does TI warrant or represent that any license, either express or implied, Is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process In which such semiconductor products or services might be or are used. Copyright @ 1996, Texas Instruments Incorporated 5-70 Contents Title Page Introduction ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.•• 5-73 PCMCIA •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.••.••••••••••• 5-73 Packaging •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.••••••• 5-74 Power ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.••••••••••.••••• 5-78 Thermal Resistance ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••. 5-79 Conclusion •••••••••••••••••••••••••••••••••••'........................................... 5-80 List of Dlustrations Figure 1 2 3 4 5 6 7 8 9 10 Title PCMCIACardDimensions ........................................................... Portable Computers Supporting PCMCIA Cards ....................................... : .. FIFO Package Dimensions ........................................................... Area Versus Package Option .......................................................... TQFP Package Height ............................................................... Type II PCMCIA Package Height Versus Dual Package Solutions ............................ Typical Package Height Versus Height Variance .......................................... Active ICC Versus Frequency ......................................................... Multiple 9-Bit Solution Versus Single 36-Bit FIFO ........................................ RaJA for FIFO Packages in Still Air ................................................... Page 5-73 5-74 5-75 5-76 5-76 5 -77 5 -77 5 -78 5-79 5 -80 5-71 5-72 Introduction As today's applications become more complex and integrated, there is a continuing need to reduce board space without sacrificing functionality. This need has never been greater than for Personal Computer Memory Card International Association (PCMCIA) cards. The appeal for PC card adapters in large area network (LAN) is due to the small form factor and performance. PCMCIA card designers face the challenge of reducing current half-size boards (typically 4.5 in x 8 in) such as those used in present deslaop systems, to the size of a credit card. With the emergence of notebook computers, personal data assistants (PDAs), and wireless communications, designers are turning to PCMCIA cards to meet the growing demand for more flexibility. This trend has driven the chip-set manufacturers to reduce the number of add-on features normally designed into desktop computers to fit the·board confines of portable systems. PCMCIA cards provide an alternative: standard add-on features in a miniature-portable format. Until the introduction and standardization ofPCMCIA cards, portable systems were left virtually unconnected to other systems and peripherals. Many of the PCMCIA designs provide the needed interconnectedness between systems by performing inputloutput (1/0) data functions. Typical 110 adapters found today are EtherNet, faxes, SCSI, and modems, to name a few. As bus widths and data speeds increase, so does the chance of data bottlenecks and latency. Specialty memories such as FIFOs are required for either rate matching or clock partitioning from data buses and processors. Until now, many designers either had to compromise performance or increase device count due to the lack ofPCMCIA-compliant FIFO packaging. Texas Instruments (TI) has met the challenge of reduced board area and increased integration with their advanced FIFO memories. TI offers 9-, 18-, and 36-bit high-performance FIFO memories in the PCMCIA-compatible thin quad flat package (TQFP). PCMCIA PCMCIA was founded in 1989 to·define and set PC-card standards. In today's market, there are three widely accepted standards: Type I, ratified in June 1990; Type II, ratified in September 1991; and Type m, which is pending. PCMCIA cards all share a common length and width, differing only in their height (see Figure 1). IOII~I----- 85.6mm ----~~I T 54.0mm PCMCIA I!:::::=:====::!..L...J _1_ H&19ht:J. ~--------------------"'L::! - T Figure 1. PCMCIA Card Dimensions Due to the increased popularity ofPCs (Le., laptops, notebooks, andPDAs), a demand for the same functionality as found in larger PC counterparts has arisen. To keep pace with the growing market demand, portable-computer manufacturers have begun to support PCMCIA card ports on nearly all new designs (see Figure 2). 5-73 80,000 70,000 6' co ~ .!S. CD 60,000 E 40,000 i 30,000 ::I II 50,000 PCMCIA Supported PCMCIA Unsupported 20,000 10,000 0 1992 1993 1994 1995 1996 1997 t Gartner Group Figure 2. Portable Computers Supporting PCMCIA Cardst In the past, PCMCIA cards were seen as a way to easily upgrade system memory without adding storage overhead to , the already compact chip sets. Today, PCs are replacing hardwired-desktop networked systems as the main computing unit. The need to provide for system interconnection and data communications has targeted new PCMCIA designs for LAN, fax, and modem-adapter cards. These applications typically are found in Type II PCMClA cards. Other designs such as subminiature disk drives and wireless radio frequency communication adapters typically found in Type ill PCMClA cards (see Table 1). Table 1. PCMCIA Card Applications TYPE HEIGHT IPC 3.3mm Memory devices: Flash, DRAM, OTP, and high·speed add-ons APPLICATION II PC Smm 110 devices: Fax, modems, and LAN adapters III PC 10.Smm Wireless devices: RF-communications devices and submini disk drives Type I cards are focused mainly for plug-in memory. Type II and Type ill cards are gaining ground in many new applications. Type II cards are used mainly for II 0 applications such as those listed in Table 1. The growing acceptance of PCMClA cards for II 0 interface has caused manufacturers ofDSPs, CODECs, bus-interface devices, and ASlCs to begin producing PCMCIA-compliant devices. For example, TI's Rio Grande chip set for PCl features ports to support two PCMClA cards. There is a demand for these features in a portable package the size of a PCMCIA card and an even greater demand for devices in PCMClA-compatible packages. These devices must provide the needed features, consume less power, and require less critical board space. Packaging The primary obstacle facing many designers is obtaining packages small enough to incorporate into their PCMCIA designs. To shrink a current adapter-card design and have it fit into the small form factor of a PCMCIA card requires all components, not just the printed circuit board, to be reduced in size. To ensure functionality is not lost, many designers implement multilayered boards to help increase integration. Some boards have ten layers and measure only 0.03 in thick. Multilayered boards are only part of the solution; both active and passive components must be dual-side mounted for maximum chip count and overall integration. To ensure the entire board fits into a PCMClA form requires specialized packaging from the device manufacturers. Since FIFOs playa key role in the functionality of many of these designs, TI has utilized board-space-saving TQFP packaging across 9-, 18-, and 36-bit FIFO product lines (see Figure 3). 5-74 12o-Pln TQFP (PCB) 36 Bit 10o-Pln TQFP (PZ) 18 Bit '~ 0,40 t 14,0 Area of Package 224mm2 Area of Package 224mm2 0 0 ~16'O Package Height Typical 14,0 ~16'O 1,5 + J GBUUUIIIUUUUHIUUUUUU'uUdt. f Package Height Typical 1,5 t .rGAAIlPIlPWIIIIIIlIlUUUWWh. f SO-Pin TQFP (PN) 18 Bit 0,50 Area of Package 168 mm2 1,5 Package Height Typical J~ Package Height Typical n Area of Package 120mm2 10,0 ~,.'~ ~ f 64-Pln TQFP (PAG) 9 Bit 64-Pln TQFP (PM) 9 Bit Area of Package 120mm2 t 1,5 t f Package Height Typical n 10,0 ~,.'~ ~ 1,1 t f NOTE: Dimensions are in millimeters. Figure 3. FIFO Package Dimensions 5-75 TQFP packaging not only reduces critical board area and height; it also offers increased performance and reliability due to TI's advanced CMOS and BiCMOS processing. There are a number of players in the FIFO market today that employ either plastic-leaded chip camer (FLCC) or leadless chip camer (LCC) packages as the smallest option for any organization. Due to the larger size of these older packages, many designers that otherwise would have chosen a FIFO for a design have been forced to design without FIFOs in their PCMCIA designs, incurring higher integration cost and increased board space. By comparing total package area by FIFO organization, it is obvious that TI offers the smallest package option for each of the popular FIFO organizations (see Figure 4). 1,000 898.2 .. ~, ! 800 r 600 r 784.56 638.67 400 , , o n, 9Blt • 18Blt 38Blf mm 407.75 200 144 o 'KK 256 256 198 ~t .... ~ % " '" , , 84 TQFP 32 PLCC 52 PLCC 88 PGA 120 TQFP 32 LCC 80 TQFP 88 PLCC 100 TQFP 132 PQFP Figure 4. Area Versus Package Option The board-space savings are even more dramatic when considering cascaded multiple 9-bit FIFOs in 32-pin PLCC packages to construct an 18- or 36-bit FIFO solution. Tl's 9-, 18-, and 36-bit FIFOs offered in the TQFP packages not only eliminate the need to cascade devices, but reduce board space. An example of this board-space savings is the conventionall8-bit package, the 68-pin PLCC. Tl's 18-bit FIFOs in the 80-pin TQFP reduce board space by 80%. The 68-pin PLCC has a nominal package height of 4.38 mm versus 1.5 mm for all of Tl's TQFP packages. This is better shown by Figure 5 , which illustrates two TQFP packages with a nominal height of 3 mm. An additional 0.5 mm for PC board thickness makes the total height 3.5 mm. This is not only 20% thinner than PLCC, but also meets PCMCIA TYPeIlspecDication. 6-mm Type II PCMCIA Figure 5. TQFP Package Height Figure 6 shows a comparison of FIFO surface-mount packages versus the PCMCIA TYPe II specification. Packages below the reference line meet the TYPe II specDications. Those packages above the reference line exceed the maximum height requirements. All values are calculated based on double-side mounting (two packages) and do not include the PC board thickness (nominally 0.5 mm). Only the TQFP and 9-bit 32 LCC packages pass the TYPe II PCMCIA specifications. 5-76 0 - D 8 • 18Blt - • 36 Bit 6 Type II ----------- ----- -- -- r------- 9Blt PCMCIA . Maximum Height Specification - 4 2 I 0 I I I I 64 TQFP (PAG) 32 LCC 80 TQFP 68 PLCC 100 TQFP 132 PQFP 64TQFP (PM) 32 PLCC 52 PLCC 68 PGA 120 TQFP Figure 6. Type II PCMCIA Package Height Versus Dual Package Solutlonst t Board thickness is not included. Although the LCC package may physically fit the requirements for PCMCIA cards, it does not permit optimum system design, since so much of the board area is occupied by one device. TI FIFOs available in TQFP packages allow a designer to choose the architecture and features that best meet the design criteria and dramatically reduce critical board space. Another critical point to consider in PCMCIA design is package dimensioning and tolerances. As previously stated, the nominal values specified for a package may appear to meet PCMCIA card specification. However, due to different techniques involved in the mold processes of different packages, designers must carefully review all specifications given in the mechanical drawings for each package. In the case of the 32-pin LCC, the height tolerance varies greatly from a minimum of 1.27 mm to a maximum of 2.2 mm, a variance of 57% (see Figure 7). 6 5 E E 4 +++ + I :c f iii .5 E 3 I 2 T~~~~~~ 0 z o I I I I I I I I I I I I 68 PLCC 132 PQFP 32 PLCC 64 TQFP (18 Bit) 80 TQFP 120 TQFP 100 TQFP 64 TQFP (PAG) 52 PLCC 68 PGA 32 LCC 64 TQFP (PM) Figure 7. Typical Package Height Versus Height Variance 5-77 Figure 7 shows thatTI's TQFP packages hold the tightest mechanical tolerances for all dimensions, ensuring the package clearance desired never varies by more than 0.05 mm from the nominal value. Tight control on package height is critical because clearance of a full circuit board is very tight to begin with, without factoring in variations in circuit board, solder thickness, and the PCMCIA card. TI's control of the TQFP package dimensions allows mechanical conformity without package inspection and sorting.. Power The PCMCIA card specifications have forced all aspects of devices to be reduced in size. Equal to the need for smaller packaging is the need for reduced device power consumption. Since PCMCIA cards are closed systems, cooling fans are size restricted. Designers must carefully review device power consumption because PCMCIA cards are used primarily with portable systems that are battery powered. Reducing power consumption is especially critical to increased system battery life. FIFO power consumption depends on several factors. Most of the power consumed by a FIFO is used in charging the CMOS circuit while performing reads and writes, sometimes referred to as duty cycle. The speed at which a FIFO operates affects the amount of power consumed. As speed increases, so does the frequency of reads and writes. To assist designers in calculating power, TI provides an ICC versus frequency plot for each FIFO in the Sept. 1994 High-Performance FIFO Memories data book (literature number SCAD003B). Because the duty cycle and clock frequency at which a FIFO is operated depend on the design, TI has implemented a unique circuit feature on its advanced FIFOs, i.e., dynamic-sense amplifiers. Dynamic-sense amplifiers draw power only during a read or write operation; otherwise, they are idle, drawing less than 400 ~. Conventional FlFOs implement static-sense amplifiers that draw power even when the device is idle (approximately 50 rnA). TI's dynamic-sense amplifiers are designed to optimize maximum performance without any degradation of propagation delay times. Figure 8 is a comparison of TI's 9-, 18-, and 36-bit clocked FlFOs with several conventional synchronous 9-bit FIFOs. The synchronous 9-bit FlFOs all draw more than 50 ~ when idle. As system speed increases, so does the amount of power consumed. As speeds approach 60 MHz, the 9-bit synchronous FIFO draws more power than TI's clocked 36-bit FlFOs. ~Or-----------------------------------------------~ ____ Synchronous 1/2/4K x 9 ......... ACT7807 2K x 9 ""*- ACT7803512x18 ...... ABT7819512x18x2 ACT3632512x36x2 ",200 E -+- I i.. a 150 ~ a. a. ~ 100 I ~ 50 O~------~------~------~------~------~------J o 10 20 30 40 f - Frequency - MHz Figure 8. Active ICC Versus Frequency 5-78 50 60 Using the following equation, an accurate power calculation can be made for any FIFO device. PT - Vee x [lee + (N x Alee x dc)] + :E(Cpd x Vee2 x fi) + :E(CL x Vee2 x fo) Where: lee - power-down supply current maximum N - number of inputs driven by a TIL device Alee - increase in supply current de Cpd CL fi fo duty cycle of inputs at a TIL high level of 3.4 V power dissipation capacitance output capacitance load data input frequency data output frequency - The power consumption of a single 9-bit device is an important consideration, since many designs require 18- or 36-bit FIFO solutions. Since power consumption is primarily a factor of the number of outputs switching, reducing power consumption with wider-word-width FIFOs is critical [for example, the total power consumption of a lK x 36 FIFO when constructed by cascading four lK x 9 FIFOs (see Figure 9)]. TI's single-chip solution, the SN74ACT3641, not only saves 65% board space, but reduces power consumption by 78% when operating at 60 MHz. 1,000 Synchronous ACT3641 ___ 1/214Kx9 31: ........ ACT7807 2Kx9 800 E I c 0 i 800 J 400 c I lII. 200 0 0 10 20 30 40 50 60 f - Frequency - MHz Figure 9. Multiple 9·Blt Solution Versus Single 36-Blt FIFO Thermal Resistance As with any small package, thermal considerations must be taken into account Any heat generated by a device must be dissipated to ensure proper operation. The heat dissipation of a package is measured in terms of theI1Il!ll resistance (RS). RSJA is defmed as the thermal resistance from the die junction to ambient air. Figure 10 shows a listing of all TI's surface-mount packages and their associated RaJA values. The listed values are measured in still air, which is a better representation of the true operating conditions of a FIFO in a PCMCIA card. The 120-pin TQFP is an example of TI's new thermally enhanced packaging (TEP) technology. The 120-pin TQFP has a heat spreader mounted to the top of the package and the die is mounted underneath in a dead-bug fashion. The 120-pin TQFP has the same RaJA characteristics as the larger 132-pin PQFP and is 67% smaller. The heat dissipation similarities between the two packages are due to the addition of a heat spreader built into the 120-pin TQFP. 5-79 3000 - - - 24-Pln SOIC (OW) --.Ar- 28-Pln SOIC (OW) ....... 44-Pln PLCC (FN) 58-Pin SSOP (OL) 64-Pln TQFP (PAG) -+- 64-Pln TQFP (PM) -III- 68-Pln PLCC (FN) -b- SO-Pin PQFP (PH) -0- SO-Pin TQFP (PN) --t- 100-Pln TQFP (PZ) - . - 120-Pln TQFP (PCB) -a- 132-Pin PQFP (PQ) -<>-+- 2500 ~I c 0 i j 2000 c I E ::I E 'R 1\1 :Ii 1500 I ~ 1000 500 25 35 30 40 45 50 55 60 65 70 75 80 85 90 TA - Ambient Temperature - ·C Figure 10. RSJA for FIFO Packages in Still Air The RSJA characteristics in Figure 10 are measured in still air (no laminar flow), which best represents the conditions of a closed PCMCIA card. The RaJA values are calculated using the following equation: Ra1A = T1 - TA power Effective heat dissipation is needed as power increases to reduce junction temperature of the die. The increased temperature can cause drift and even device failure, which dramatically decreases mean time between failure (MTBF). TI has improved device reliability by combining decreased power and effective packaging. Conclusion As the demand forPCMCIA card continues, so does the demand for lower power and more space-saving packages. TI has met both demands ·with TQFP packages for their advanced CMOS and BiCMOS FIFOs. The TQFP package dramatically reduces board space over conventional packaging and eliminates the need to cascade multiple FIFOs to create 18- and 36-bit FIFO solutions. The implementation of dynamic-sense amplifiers on all advanced FIFOs reduces power consumption and; therefore, improves system reliability and provides longer battery life for portable systems. 5-80 Package Outlines and Thermal Data Page Introduction ...................................................................... Plastic Dual-In-Line Package (PDIP) N/R-PDIP-T16 (16-pin DIP) ................................................. N/R-PDIP-T20 (2o-pin DIP) ................................................. N/R-PDIP-T28 (28-pin DIP-600 mil) .......................................... N/R-PDIP-T40 (40-pin DIP) ................................................. NP/R-PDIP-T28 (28-pin DIP-300 mil) .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. NT/R-PDIP-T24 (24-pin DIP) ................................................ Plastic J-Leaded Chip Carrier (PQCC) FN/S-PQCC-J20 (2o-pin PLCC) FN/S-PQCC-J28 (28-pin PLCC) FN/8-PQCC-J44 (44-pin PLCC) FN/S-PQCC-J68 (68-pin PLCC) RJ/R-PQCC-J32 (32-pin PLCC) 5-83 5-84 5-86 5-88 5-90 5-92 5-94 ............................................. 5-96 ............................................. 5-98 ............................................ 5-100 ............................................ 5-102 ............................................ 5-104 Plastic Wide-Body Small-Outline Package (PDSO) DW/R-PDSO-G16 (16-pin SOIC) ............................................ DW/R-PDSO-G20 (20-pin SOIC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. DW/R-PDSO-G24 (24-pin SOIC) ...................................'. . . . . . . .. DW/R-PDSO-G28 (28-pin SOIC) ............................................ 5-106 5-108 5-110 5-112 Plastic Small-Outline Package (PDSO) DV/R-PDSO-G28 (28-pin SOIC) ............................................ 5-114 Plastic Shrink Small-Outline Package (PDSO) DUR-PDSO-G56 (56-pin SSOP) ........................................... 5-116 Plastic Quad Flatpack (PQFP) PAG/S-PQFP-G64 (64-pin TQFP) . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. PCB/S-PQFP-G120 (120-pin TQFP) ........................................ PH/R-PQFP-G80 (80-pin PQFP) ............................................ PM/S-PQFP-G64 (64-pin TQFP) ............................................ PN/S-PQFP-G80 (80-pin TQFP) ............................................ PQ/S-PQFP-G132 (132-pin PQFP) ......................................... PZ/S-PQFP-G100 (100-pin TQFP) .......................................... 5-118 5-120 5-122 5-124 5-126 5-128 5-130 5-81 5-82 MECHANICAL AND THERMAL DATA Introduction The remainder of this section contains mechanical and thermal data for each package offered for TI's FIFOs. The mechanical data consists of drawings of each package annotated with critical dimensions. These drawings typically include the following dimensions: lead pitch (tip to tip); body width and length; shoulder-to-shoulder insertion width; lead width, thickness, and angles; and package maximum height and stand-off clearances from seating plane to bottom of the package. For packages designed in English units, inch dimensions are shown first followed by millimeter dimensions in parentheses. A period is used as the English units decimal point and a comma as the metric units decimal point. The official JEDEC descriptor is used to identify each package type. The thermal data consists of the thermal resistances from junction to ambient (ElJA) measured in either a one-cubic-foot box or in a wind tunnel under varying air velocities. Values of thermal resistance from junction case (ElJC) also are included for some package types. Derating curves of maximum power dissipation versus ambient temperature for varying air flows are provided for each package. ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75266 MECHANICAL AND THERMAL DATA N/R·PDIP·1** PLASTIC DUAL·IN·LINE PACKAGE 16 PIN SHOWN C'~ T 0.260 (6,60) 0.240 (6,10) ~~~~~~~~ '8 0.070 (1,778) MAX 0.035 (0,89) MAX JL -./ I." 0.014 (0,356) 0.010 (0,254) 0.100 (2,54)TVP (see Note C) J 0.125 (3,18) MIN II - ~6)' 0.010 (0,254) \.- 0°-15° 4040049/8-10/94 NOTES: A. All linear dimensions 'are in inches (millimeters). B. This drawing is subject to change without notice. C. Each lead centerline is located within 0.010 (0.254) of its true longitudinal position. ~1ExAs 5-84 INSTRUMENTS ' POST OFACE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA i6-Pln DIP (N) Package Power Dissipation Derating In Stili Air 2000 1800 ~I c: i ~.. I E ;, E "5C III 1600 1400 1200 1000 800 800 :& 400 200 0 25 30 35 45 40 60 80 55 65 70 75 80 65 Ambient Temperature _·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE 0 0 100 250 500 1 ft 3 box 67 Wind tunnel - NlA NlA NlA NlA NlA NlA 26 NlA Wind tunnel Wind tunnel Wind tunnel N/A t Mounted on pnnted-clrcUit board ~1ExAs INSTRUMENTS POST OFFICE BOX 665303 • DAu.AS. TEXAS 75265 ("C/W) 90 MECHANICAL AND THERMAL DATA N/R·PDlp·T** PLASTIC DUAL·IN·L1NE PACKAGE 16PINSHOWN CA~ T 0.260 (6,60) 0.240 (6,10) ~~~~~~~~ 8 0.070 (1,778) MAX ~ 0.310(7,87) 0.035 (0,89) MAX JL - .\ I.- Ir--ll ,.... 0.100 (2,54)TVP (see Note C) 0.125 (3,18) MIN 0.014 (0,356) 0.010 (0,254) j ll _ \.~6) V.." 0°_15° 0.010 (0,254) 4040049/8-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Each lead centerline is located wHhin 0.010 (0.254) of its true longitudinal position. ~TEXAS . 5-86 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 2o-Pln DIP (N) Package Power Dissipation Derating at Varying Air Flows 3000 Oft/min 100ft/min 250ft/min 500ft/min 2800 2600 2400 2200 ~ E 2000 I c 0 1600 'OJ .!!! 1600 iQ. .. c I 1400 E ::I E 1200 :IE 1000 '; 800 600 400 200 0 25 30 35 45 40 60 60 55 65 70 75 60 85 90 Ambient Temperature _·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) 0 0 100 250 500 N/A ENVIRONMENT 1 1t3 box 9 JUNCl'lON TO AMBIENT ("C/W) 9 JUNCTION TO CASE ("C/W) N/A Wind tunnel 67 90 75 58 48 N/A N/A 33 Wind tunnel Wind tunnel Wind tunnel NlA NlA NlA N/A t Mounted on printed-CircUit board ~TEXAS ' INSTRUMENTS POST OFFICE BOX 655303 • DALLAS, TEXAS 75265 5-87 MECHANICAL AND THERMAL DATA N/R·PDlp·T** PLASTIC DUAL·IN·LINE PACKAGE 24 PIN SHOWN· r A ------1-3tJ11 24 0.560 (14,22) 0.520 (13,21) ~'i"i"""'i'T"'i""'i"'i'''i'''''i''i"''i''''''i''i''''~~ 12 0.060 (1,52) TYP 0.020 (0,51) MIN - 0.200 (5,08) MAX UUU~ UUUU~ U~ U~I~~.~~~~~--~-~ JL-.I I.-- Semlng Plane 0.100 (2,54) TYP 0.125 (3,17) MIN 0.021 (0,533) 0.015 (0,381) JL 0°-15° 0.014 (0,355) 0.008 (0,203) 404OO53IB-10/94 NOTES: A. All linear dimensions are In inches (millimeters). B. This drawing is subject to change without notice. :lllExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75266 MECHANICAL AND THERMAL DATA 28-Pln DIP (N) Package Power Dissipation Derating In Stili Air 3000 2800 2800 2400 2200 ~I c 0 :;:: t. 2000 1800 1600 is ; 1400 E :. E 1200 ~ I 1000 800 600 400 200 0 25 30 45 40 35 50 60 55 65 70 75 60 65 Ambient Temperature _·C Thermal Resistance (e) Measurementst AIRFLOW (ft/mln) 0 0 100 250 500 NlA ENVIRONMENT e JUNCTION TO AMBIENT e JUNCTION TO CASE 1 ft 3 box 53 NlA Wind tunnel NlA Wind tunnel - Wind tunnel - NlA Wind tunnel - NlA NlA N/A 20 (·C/W) t Mounted on pnnted-cIICUIt board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALlAS, TEXAS 75265 (·C/W) NlA 90 MECHANICAL AND THERMAL DATA N/R·PDlp·T** PLASTIC DUAL·IN-LINE PACKAGE 24 PIN SHOWN A ------1-3+1.1 0.560 (14,22) 0.520 (13,21) ~'i'T""i'T""~~i"""i"i""'TT~~ 12 0.060 (1,52) TYP 0.020 (0,51) MIN - 0.200 (5,08) MAX ~IDJ_~;--r-. UUUUUUUUUUUU7 ' ~~ Jl' Seating Plane JL -.J I.-- 0.100(2,54)TYP 0.125 (3,17) MIN 0.021 (0,533) 0.015 (0,361) 0°-15° 0.014 (0,355) 0.008 (0,203) 4040053/8-10194 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. ~TEXAS 5-90 INSTRUMENTS POST OFFICE BOX 665303 • DAUAS. TEXAS 752e5 MECHANICAL AND THERMAL DATA 4o-Pln DIP (N) Package Power Dissipation Derating In Stili Air 3000 2800 2600 2400 2200 ~ E 2000 I c 0 i ~ 1800 1600 J 1400 = 1200 E E I 1000 800 600 400 200 0 25 30 40 35 45 50 60 55 65 70 75 80 65 90 Ambient Temperature _·C Thermal Resistance (e) Measurementst AIRFLOW (ft/mln) ENVIRONMENT e JUNCTION TO AMBIENT rC/W) e JUNCTION TO CASE (·C/W) 1 ft 3 box 43 Wind tunnel 0 0 100 250 500 Wind tunnel - Wind tunnel - Wind tunnel - NJA NJA NJA NJA NJA NJA N/A N/A 12.5 t Mounted on pnnted-clllluit board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-91 MECHANICAL AND THERMAL DATA NP/R·PDlp·T28 PLASTIC DUAL·IN·LINE PACKAGE 1.375 (34,93) 1.345 (34,16) 0.295 (7,49) 0.270 (6,68) ~~~~~~~~ 14 0.030 (0,76) 0.015 (0,38) 0.180 (4,57) 0.145 (3,68) Seating Plane I Jl 0.110 (2,79) ~ 0.090 (2,29) 0·_15· 0.015 (0,38) 0.008 (0,20) 404007518-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Body dimemsions do not include mold flash or protrusion. ~1ExAs 5-92 I INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 28-Pln DIP (NP) Package Power Dissipation Derating at Varying Air Flows 5000r-------------------------------------~-------------------, Oftlmln 4500 ~I _. - 4000 3500 __ c o J J E :::I E I 200 ft/mln 400 ftlmln I - . __ r- 3000 2500 2000~ ~ -- -- .--.--. _ --.--.--- --.- -- ----.-_ _ • __ • -- 1500 - - . __ . ~ -......;; - 1000 500 __~~___l~~I~~L___~~__~I~I__~I___~I__~I~I__~ O~~-L ~ ~ ~ ~ 40 ~ ~ ~ ~ ro ~ ~ ~ ~ Ambient Temperature _·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT rC/W) 9 JUNCTION TO CASE (·C/W) 0 1 1t3 box - 0 Wind tunnel 63.5 200 Wind tunnel 43 400 Wind tunnel 36.5 NlA NlA NlA NlA NlA - N/A NlA t Mounted on pnnted-clrcult board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • OAUAS. TEXAS 75265 5-93 MECHANICAL AND THERMAL DATA NT/R-PDIP-T** PLASTIC DUAL-IN-LiNE PACKAGE 24 PIN SHOWN A-~~I 13 T 0.280 (7,11) 0.250 (6,35) ~~~~~~~~~~~~ 12 0.200 (5,06) MAX 0.020 (0,51) MIN Seating Piline H 0.100 (2,54) 1 JL . 0.125 (3,18) MAX ~:~~! ~~::: r-I."""I0-.0-10-('-O,25)=--;;:@:-11 Jl 0°_15° ~ 0.010 (0,254) NOM 4040050/B-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. ~1ExAs 5-94 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 2S-Pln DIP (NT) Package Power Dissipation Derating at Varying Air Flows 3000 Oft/mln 2800 2600 - 2200 c 0 'ia. ~ I :. E ::I E 2000 1800 1600 1400 500ft/mln .............. .... "".." ..".. ..".. as .............. ................ ............ .................... ............................ ............................ ~ ................ ~ .............. ..................... 1200 '5C ~ 100 ft/mln 250ft/mln .............................. 2400 ~I -- 1000 "".... .............. .............. .............. 800 600 400 200 OL---~--~--~--~--~--~--~--~--~--~--~--~~ 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Ambient Temperature - ·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) 0 0 100 250 500 N/A 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE ("C/W) Windlunnel 67 81 72 55 46 NlA NlA NlA NlA NlA N/A N/A 25 ENVIRONMENT 1 ft 3 box Wind tunnel Windlunnel Windlunnel t Mounted on pnnted-clrcult board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-95 MECHANICAL AND THERMAL DATA , PLASTIC J·LEADED CHIP CARRIER FN/8-PQc6.J** 20 PIN SHOWN D-----~ 01 0.020 (0,51) MIN 0.008 (0,20) NOM ~ 4 E D2/E2 ----t -.1 I E1 lJ 0.032 (0,81) 0.028 (0,66) D2/E2 ~~ 14 9 13 0.013 (0,33) I'EfH 0.007 (0,18) ® I DIE NO. OF PINS·· MIN D2/E2 D1/E1 MAX MIN MAX MIN MAX 4040005/8-10194 NOTES: A. All linear dimensioris are in inches (millimeters); B. This drawing is subject to change without notice. C. Falls within JEDEC MS-018 ~1ExAs 5-96 INSTRUMENTS POST OFFICE BOX 656303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 20-Pln PLCC (FN) Package Power Dissipation Derating at Varying Air Flows 3000 oft/mln 2800 _. - 100ft/mln 250 ft/mln 2600 ---- 500 ft/mln 2400 2200 ~I 2000 i I 1600 c 1400 ~ 1200 I """ ... ... """ ...... .. ~... 1600 I E ...... ......... """ ... ... """ ... "'""" """ --- --- .........~.. ~... --- --- ------ ~... ~ """ ...... ------ ------ ~ .. --- --- --- --- ---. ...... ~ ........ ~ 1000 ... 800 800 400 200 0 25 Ambient Temperatura _·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 0 9 JUNCTION TO AMBIENT 9 JUNCTION TO CASE (·C/W) (·C/W) 1 ft3 box 99 0 Wind tunnel 98.6 86.1 100 Wind tunnel 250 Wind tunnel 70.7 500 Wind tunnel 58.1 NlA NlA NlA NlA NlA N/A 23 NlA NlA t Mounted on printed-cIICult board :lflExAs INSTRUMENTS POST OFFICE BOX 655303 • DAUAS. TEXAS 75265 5-97 MECHANICAL AND THERMAL DATA PLASTIC J·LEADED CHIP CARRIER FN/S·PQCC·J** 20 PIN SHOWN D-----~ 0.008 (0,20) NOM 4 E 0.020 (0,51) MIN J I E1 U D2/E2 0.032 (0,81) (0,66) -*- ~.026 9 D2/E2 t:~~ 14 13 0.013 (0,33) 1+1 0•007 (0,18) @I NO. OF PINS" DIE D1/E1 D2/E2 404000518-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Falls wHhin JEDEC MS-018 ~1ExAs 5-98 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 28-Pln PLCC (FN) Package Power Dissipation Derating at Varying Air Flows 3000 oft/mln 2800 100 ft/mln 2600 500ft/mln 250ft/mln .... 2400 2200 ~I 2000 0 1800 i 1600 c iCI. ... ~... .............. Q I .. '" ................. ............. ~ ....... ....... .............. 1400 IL :::I E E ;c 1200 ::Ii! 1000 . .. ........... ...~ ~ ....... ....... .. ............. ...~ ~ .............. .. .. ............. .. ...~ ~ ........... ................. ~ .............. .... ... ....... .............. 800 600 400 200 0 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Ambient Temperature - ·C Thermal Resistance (9) Measurementst t AIR FLOW (ft/mln) ENVIRONMENT 0 9 JUNCTION TO AMBIENT e JUNCTION TO CASE (·C/W) (·C/W) 1 ft 3 box 95.1 NlA 0 Wind tunnel 87.7 N/A 100 Wind tunnel 69.6 N/A 250 Wind tunnel 61 N/A 500 Wind tunnel 50 N/A NlA N/A N/A 26.7 Mounted on pnnted-clrClult board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-99 MECHANICAL AND THERMAL DATA FN/8-PQCC-J** PLASTIC J-LEADED CHIP CARRIER 20 PIN SHOWN O-----~ 01 0.008 (0,20) NOM 3 J I E 1J °2 /E2 ----t -.1 0.032 (0,81) 0.026 (0,66) 02 /E2 ~ 14 13 9 0.013 (0,33) 1-.1 0•007 (0,18) @I NO. OF PINS" OlE 01 /E1 02 /E2 4040005/8-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without noti~. C. Falls within JEDEC MS-018 ~1ExAs 5-100 INSTRUMENTS POST OFRCE BOX 865303 • DALlAS. TEXAS 75285 MECHANICAL AND THERMAL DATA 44-Pln PLCC (FN) Package Power Dissipation Derating at Varying Air Flows 5000 ------- 4500 4000 ~I ---- ---------- o 3000 __ 2500 ----__ ............ 2000 --. ,. E E I .. ...................... Q ~ 250ft/min 500ft/min 350Q _____ _ I: t :; I Oft/min 100 ft/mln - ----- __ _ _ 1500 - -- - -------. ------ 1000 ----- .......... - _ _ _ _ _ - __ _ - - • -..;: - 500 0~~1-~1-_~1-1~~1-~1-_~1--1~~1-_~1-~1~1-~ ~ ~ ~ ~ ~ 50 ~ 50 M ro n 50 ~ 00 Ambient Temperatura - ·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE ("C/W) 0 1 ft 3 box 68.1 N/A 0 Wind tunnel 64.3 N/A 100 Wind tunnel 54.9 N/A 250 Wind tunnel 44.2 N/A 500 Wind tunnel 35.8 N/A N/A 22 NlA N/A t Mounted on pnnteck:lrcuit board -!I1TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DAu.AS. TEXAS 75265 5-101 MECHANICAL AND THERMAL DATA FN/8-PQCC-J** PLASTIC J-LEADED CHIP CARRIER 20 PIN SHOWN D-----~ o.ooa (0,20) NOM ~ 4 E I El U 0.032 (0,81) 0.026 (0,66) 14 13 9 1-$-10.007 (0,18) ® 1 NO. OF .. DIE PINS 4040005/B-10/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Falls within JEDEC MS-018 ~1ExAs 5-102 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 68-Pln PLCC (FN) Package Power Dissipation Derating at Varying Air Flows 5000 Oft/min 4500 200ft/min 400ft/min 400ft/min 4000 ~I c 3500 0 11a. 'iI is J E ::0 E '5C ::Ii '" 3000 -- -- -- --- 2500 2000 1500 1000 500 0 25 30 35 40 45 50 60 55 65 70 75 80 85 90 Ambient Temperatura - °C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT (OC/W) 9 JUNCTION TO CASE (OC/W) 0 1 1t3 box 51.3 N/A 0 Wind tunnel 51.3 NlA 100 Wind tunnel 43.4 N/A 250 Wind tunnel 32.7 NlA 500 Wind tunnel 27.8 NlA NlA N/A N/A 14.5 t Mounted on pnnted-clrcult board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS, TEXAS 76266 5-103 MECHANICAL AND THERMAL DATA RJ IR·PQCC·J32 PLASTIC J·LEADED CHIP CARRIER 1 4 - - - - - 0.495 (12,57) _ _ _~ 0.485 (12,32) 1 4 - - - - 0.453 (11,51) _ _-.! 0.449 (11,40) 0.045 (1,14) x 45° 5 29 0.595 (15,11) 0.565 (14,86) "--_""--1_' 0.400 (10,16) 0.553 (14,05) 0.549 (13,94) TYP L.,---J'"T""" -.l 0.530 (13,46) 0.490 (12,45) 21 14 20 0.050 (1,27) TYP 0.032 (0,81) 0.026 (0,66) t 0.021 (0,53) 0.013 (0,33) 0.012 (0,30) 0.008 (0,20) Seating Plane-- I J+- 0.300 (7,62) TYP 1 4 - - - - 0.430 (10,92) I ---+I _ _--+I 0.390 (9,91) 4040077/A-10/93 NOTES: A. B. C. D. All linear dimensions are in inches (millimeters). This drawing is subject to change without notice. Body dimensions do not include mold flash or protrusion. Formed leads shall be planar with respect to one another within 0.004 (0,10) at the seating plane. ~TEXAS INSTRUMENTS 5-104 POST OFFICE BOX 655303 • DALLAS, TEXAS 75265 MECHANICAL AND THERMAL DATA 32·Pin PQCC (RJ) Package Power Dissipation Derating at Varying Air Flows 5000 oft/mln 4500 _. - 200 ft/mln 400 ft/mln 4000 ~I c - ---.-...-.-.... -....-... --- -.-.... -....-... --.....:.-.... - ...::::-..-... .-... 3500 0 :; .. "0; is ... J E :::I E .. >< 3000 2500 2000 1500 ::I 1000 500 0 25 30 35 40 45 50 60 55 65 70 75 60 85 90 Ambient Temperature - 'C Thermal Resistance (9) Measurementst AIR FLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ('C/W) 9 JUNCTION TO CASE ("C/W) 0 1 ft3box - N/A 0 Wind tunnel 50.5 NJA 200 Wind tunnel 40.7 NJA 400 Wind tunnel 36.8 N/A N/A N/A N/A - t Mounted on printed-circu~ board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-105 MECHANICAL AND THERMAL DATA DWIR-PDSO-G** PLASTIC WIDE-BODY SMALL-OUTLINE PACKAGE 20 PIN SHOWN -x------t 20 10 1-$-1 ---=-..:...' 0.010 (0 25) @ 1 -'--"""'-'. L .." ' - ' - . 0.050 (1,27) 0.016 (0,40) 404000018-10194 NOTES: A. All linear dimensions are In inches (millimeters). B. This drawing is subject to change without notice. C. Body dimensions do not include mold flash or protrusion not to exceed 0.006 (0,15). ~1EXAS 5-106 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 16-Pln sOle (OW) Package Power DIssipation Derating at Varying Air Flows 2000 oft/mln 1800 _. - 1600 ...... --- --- --.-----. --. -- ----- --- ~ E I 1400 c 0 I ~ E :::I E 'lC ftI :IE ... ................ 1200 J --. ... ...... - .... ... 100 ft/mln 250 ft/mln 500ft/mln ... ~ ... ---~ ................ ................... --- -----.:.------. 1000 800 600 --. .... ..... ...... ..:::;-- - 400 200 0 25 30 35 45 40 50 60 55 65 70 75 80 85 90 Ambient Temperature _·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) 0 0 100 250 500 NlA t 9 JUNCTION TO AMBIENT (·C/W) 9 JUNCTION TO CASE (·C/W) N/A N/A Wind tunnel 130 123 102 91 78 N/A N/A 42 ENVIRONMENT 1 ft 3 box Wind tunnel Wind tunnel Wind tunnel NlA N/A NlA Mounted on printed-circuit board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALlAS, TEXAS 75265 5-107 MECHANICAL AND THERMAL DATA PLASTIC WIDE·BODY SMALL·OUTLINE PACKAGE DW IR·PDSO·G** 20 PIN SHOWN r 20 ,,------t 0.419 (10,65) 0.400 (10,15) 0.299 (7,69) 0.293 (7,45) 0 L~-----I 10 r 0.104(2,65) IL 0.093 (2,35) ! t--!fiUHUUUUUill-p- L..L-Ji :::~::: l JL·. . . . ,' 1=1 ........ 0.014 (0,35) 1 ~:::~::::: ' ., I Jl fl IOIIIT~:::= 1-$-1 0.010 (0 25) @ 1 ~.---=--'-,.....;....... _"-'. L... 0.050 (1,27) 0.016 (0,40) 4040000lB-l0/94 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Body dimensions do not include mold flash or protrusion not to exceed 0.006 (0,15). ~1EXAS 5-108 INSTRUMENTS POST OFACE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 2o-Pln sOle (OW) Package Power Dissipation Derating at Varying Air Flows 2000 .... ................ 1800 -"""".................. -""""................ ........ ... """"-.. ... ................. ....... ....... """"-- """".............. ........ ----.. 1600 ~I c Oft/min 100ft/min .. .. --....... ....... 1400 0 250ft/min 500ft/min iDo 1200 .. 1000 .............. ... ... """".............. ....... .............. """"- -""""-- 800 -- ....... 70 75 j C I E ::I E iC III - .... -- . - -~ """"-" -""""- ....... 600 ::E 400 200 0 25 30 35 45 40 50 55 60 65 80 85 90 Ambient Temperature - °C Therm!ll Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE (OC/W) 0 1 1t3 box 110 N/A 0 Wind tunnel 110 N/A 100 Wind tunnel 85 N/A 250 Wind tunnel 74 N/A 500 Wind tunnel 66 N/A NJA NJA N/A 39 t Mounted on pnnted-clrcuit board ~1EXAS INSTRUMENTS POST OFFICE BOX 655303 • DALlAS. TEXAS 75265 5-109 MECHANICAL AND THERMAL DATA DWIR-PDSO-G** PLASTIC WIDE-BODY SMALL-OUTLINE PACKAGE 20 PIN SHOWN I-$-I 0.010 (0 25) ®1 ....:...._::.L.l. .L....:!;....L.----.;...:...' 0.050 (1,27) 0.016 (0,40) 404000018-10/94 NOTES: A. All linear dimensions are In inches (millimeters). B. This drawing is subject to change without notice. C. Body dimensions do not include mold .flash or protrusion not to exceed 0.006 (0,15). 5-110 -!llExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 24·Pln sOle (OW) Package Power Dissipation Derating at Varying Air Flows 3000 Oft/min 2800 2600 -_. - 100ft/min 250 ft/mln ---- 500 ft/mln 2400 2200 ~I 2000 0 1800 f 1600 J 1400 c i c E ::I E .;c 1\1 ::i 1200 1000 800 600 400 200 0 25 30 35 45 40 50 55 60 65 75 70 80 85 90 Ambient Temperature - ·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE (·C/W) 0 1 ft 3 box 92 NlA 0 Wind tunnel 88 N/A 100 Wind tunnel 69 250 Wind tunnel 57 NlA NlA 500 Wind tunnel 48 N/A N/A N/A N/A 25 t Mounted on printed-circuit board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-111 MECHANICAL AND THERMAL DATA DW/R-PDS()'G** PLASTIC WIDE-BODY SMALL-OUTLINE PACKAGE 20 PIN SHOWN 0.419 (10,85) 0.400 (10,15) r· 20 0.299 (7,59) 0.293 (7,45) 0 L~-----t 10 r- 0.104(2,85) ~::~:::~ 1_1'" 0.093 (2,35) Jl 1t-1ijijihililiUil_AM. L.L-A . ,,- l JL 0.004 (0,10) 0.020 (0,51) 0.014 (0,36) 1+1 0.010 (0 25) fL 1~10.004(O'10)IITO.01~(O'30) 0°-8" @) 1 _.J:..L.~;.;..:.;;:.;';';':"'_::o.LI. L.. 1 ., I 0.009 (0,23) 0.050 (1,27) 0.016 (0,40) 404000018-10194 NOTES: A. All linear dimensions are in inches (millimeters). B. This drswing is subject to change without notice. C. Body dimensions do not include mold flash or protruSion not to exceed 0.006 (0,15). ~1ExAs 5-112 INSTRUMENTS POST OFACE BOX 856303 • DALlAS. TEXAS 76266 MECHANICAL AND THERMAL DATA 28-Pln sOle (OW) Package Power Dissipation Derating In Still Air 2000 1800 ~ E 1600 I 1400 ia. 1200 c 'iii .!! Q J E :::I E ill III :::E 1000 800 600 400 200 0 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Ambient Temperature - °C Thermal Resistance (a) Measurementst AIRFLOW (ft/mln) ENVIRONMENT a JUNCTION TO AMBIENT COC/W) a JUNCTION TO CASE (OC/W) 0 1 ft 3 box 81.8 N/A 0 Wind tunnel 100 Wind tunnel 250 Wind tunnel 500 Wind tunnel - NlA NlA N/A N/A N/A N/A N/A 15.4 t Mounted on printed-circuit board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-113 MECHANICAL AND THERMAL DATA DVIR·PDS()"G28 PLASTIC SMALL·OUTLINE PACKAGE 1 ~ 0.050 (1,27) I ~O~ __________________~ 1 0.350 (8,89) 0.340(8,64) ~ 0.478(12,14) 0.462 (11,73) ~~~~~~~~~~ 0.010 (0,25) 0.006 (0,15) 1 4 - - - - - - 0.728 (18,49) _____~ 0.718 (18,24) 0.120 (3,05) 0.110 (2,79) IT~---r---r-Sea--=tlng'--Pla ....---~'----h.. \8 t __ ne 1.010.004(0,10) 1 . . . 0.014 (0,36) 0.005 (0,13) 4040076/8-10194 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. ~1ExAs 5-114 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS, TEXAS 75265 MECHANICAL AND THERMAL DATA 28-Pln sOle (DV) Package Power Dissipation Derating at Varying Air Flows 3000 Oftlmln 2800 - - 200 ftlmln - -- 400 ftlmln 2600 2400 ... 2200 ~ E ........ I c 0 'ia.ii :~Q I 1800 -= 1000 ::E ... ................. ... ................ ........ 1400 1200 ... ................ ........ 1600 E ::I E " " " "'" " ........................ 2000 ... ................ ~ ... ........................ 800 600 400 200 0 25 30 35 45 40 60. 55 60 65 70 75 80 85 90 Ambient Temperature - °C Thermal Resistance (e) Measurementst AIRFLOW (ft/mln) 0 0 200 400 NlA ENVIRONMENT e JUNCTION TO AMBIENT e JUNCTION TO CASE 1 ft 3 boX - N/A Wind tunnel 63 Wind tunnel 56 N/A NlA Wind tunnel 51 NlA NlA N/A - (OC/W) (OC/W) t Mounted on printed-circu~ board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-115 MECHANICAL AND THERMAL DATA DL/R·PDSO·G** PLASTIC SHRINK SMALL·OUTLINE PACKAGE 48 PIN SHOWN ~ 0.299 (7,59) 0.291 (7,39) 0.420 (10,67)L 0.395 (10,03) 0 ~~~~~""""~~~~~~ 0.110 (2,79) 'T_'~ ~,..--::..:,=.:~:-=.:..:....--r-~o==f--+ ~se _'02_0_N~_t:_1-1=) " ___ 0.016 (0,406) 0.008 (0,203) NOTES: A. B. C. D. 40400481B-l0/94 All linear dimensions are In inches (millimeters). This drawing is subject to change without notice. Body dimensions do not Include mold flash or protrusion. not to exceed 0.006 (0,15). Foot length is measured from lead top to point 0.01 0 (0.254) above seating plane. ~1EXAS 5-116 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 56-Pin SSOP (DL) Package Power Dissipation Derating at Varying Air Flows 3000 oft/mln 2800 - 2600 -- 100 ft/mln -- 250 ft/mln ---- 500 ft/mln 2400 2200 it E ................... 2000 ............... I c 0 1800 "ii 1800 ~ 1400 E ::I E 1200 ~ Q. is b .... -............. -... -... -... Do. .... ............. .. -............. -...-... . ---...-... ........ . "5C III :::IE 1000 ............. .. . ............... .. .. -............... ............. .. ............... .... -...-... -............. -... -... -...-... .. .. -...-... ~ --- ~ -...-... 800 800 400 200 0 25 30 35 45 40 50 55 60 65 70 75 80 85 90 Ambient Tempereture - °C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT @ JUNCTION TO AMBIENT (OC/W) @ JUNCTION TO CASE (OC/W) 0 1 ft 3 box 94 NJA 0 Wind tunnel 94 N/A 100 Wind tunnel 82 N/A 250 Wind tunnel 70 NJA 500 Wind tunnel 58 NJA N/A N/A N/A 22 t Mounted on printed-circuit board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DAUAS. TEXAS 75266 5-117 MECHANICAL AND THERMAL DATA PAG/S·PQFp·G64 PLASTIC QUAD FLATPACK 11. 0,271.1 o,08@1 0,17 33 32 49 o 64 0,137 0,117 17 '$TVP lii~. J 10,10 sa 9,90 1 4 - - - - - 12,10 sa ----~ 0,05 MIN 11,90 t~ - Seating Plane 1--------rI-~-tI-"-o,=08..,1 - 1,20 MAX 4040282/8-10194 NOTES: A. All linear dimensions are in millimeters. B. This drawing is subject to change withQut notice. ~TEXAS 5-118 INSTRUMENTS POST OFFice BOX 655303 • DALLAS. TexAS 75265 MECHANICAL AND THERMAL DATA 64-Pln TQFP (PAG) Package Power Dissipation Derating at Varying Air Flows 3000 Oft/min 2800 100ft/min 2800 500ft/min 250ft/min 2400 2200 ~ E 2000 I I: 0 1. I J E 1800 1800 1400 l:I 1200 I 1000 I 800 800 400 200 0 25 30 35 45 40 50 60 55 66 70 75 80 66 90 Ambient Temperature _·C Thermal Resistance (9) Measurementst e JUNCTION TO CASE AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 0 1 ft3 box 75.7 0 Wind tunnel 71.3 N/A 100 Wind tunnel 61.9 NlA N/A (·C/W) NlA 250 Wind tunnel 52.6 500 Wind tunnel 44.6 N/A N/A N/A N/A 15.1 t Mounted on pnnted-clrcult board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 ---------- -.--_._- - --_.__._------_._...._.. 5-119 MECHANICAL AND THERMAL DATA PCB/8-PQFp·G120 PLASTIC QUAD Fl,ATPACK 61 90 91 60 0,25 0,17 120 o I-$-I o,07@1 31 mmHmf I 14'10SQ~ ""TYP 13,90 16,10 SQ _ _ _ _ _ _~ 15,90 Seating Plane 1.0.10,07 I 4040202/B-1O/94 NOTES: A. B. C. D. All linear dimensions are in millimeters. This drawing is subject to change without notice. Body dimensions do not include mold flash or protrusion. Allowable protrusion is O,25mm maximum per side. Thennally enhanced molded plastic package (HSP). ~1ExAs 5-120 INSTRUMENTS POST OFFICE BOX 865303 • DALLAS, TEXAS 75265 MECHANICAL AND THERMAL DATA 12o-Pln TQFP (PCB)t Package Power Dissipation Derating at Varying Air Flows 5000 Oft/min 4500 100 ft/mln 4000 500ft/min 250ft/min ~I c --- --- --- --- 3500 0 t c I Do E 3000 2500 2000 :::I --- --- --- --- E iC :I 1500 --------- ---- 1000 500 0 25 30 35 45 40 50 60 55 65 70 75 80 85 90 Ambient Temperature _·C Thermal Resistance (e) Measurements* AIRFLOW Cft/mln) ENVIRONMENT e JUNCTION TO AMBIENT e JUNCTION TO CASE ("C/W) ("C/W) a a 1 ft 3 box 52.4 N/A Wind tunnel 49.8 NlA N/A 100 Wind tunnel 44.2 250 Wind tunnel 34.3 N/A 500 Wind tunnel 24.7 NlA NlA N/A N/A 3.3 t Thermally enhanced package (TEP) :j: Mounted on printed-circuit board ~TEXAS ' INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-121 MECHANICAL AND THERMAL DATA PH/R·PQFP·G80 1r- ~:: 1;1 PLASTIC QUAD FLATPACK 0,18 @I ------------------~ 65 ~l 12,00TYP 14,20 18,00 13,60 17,20 nJ 60 14------ 18,40TYP - - - - - - - - . ! 20,20 _ _ _ _ _ _--.! 19,60 24,00 23,20 --------~ .l~ r 0,10 MIN 2,70TYP Seating Plane 404oo11/B -10/94 NOTES: A. All linear dimensions are in millimeters. B. This drawing is subject to change without notice. ~1ExAs 5-122 INSTRUMENTS POST OFACE eox 665303 • DALLAS. TEXAS 76265 MECHANICAL AND THERMAL DATA SO-Pln PQFP (PH) Package Power Dissipation Derating In Stili Air 2000 1800 ~I c 1800 1400 0 I :IQ J I E 1200 1000 800 ::I 600 400 200 0 25 30 35 45 40 50 80 55 65 70 75 80 65 90 Ambient Temperature _·C Thermal Resistance (9) Measurementst AIRFLOW Cft/mln) 0 0 100 250 500 NlA ENVIRONMENT 9 JUNCTION TO AMBIENT ,"C/W) 9 JUNCTION TO CASE C·C/W) NlA N/A 1 1t3 box 83.9 Wind tunnel - Wind tunnel NlA Wind tunnel - N/A N/A 15.1 Wind tunnel N/A NlA t Mounted on printed-ciICuit board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-123 MECHANICAL AND THERMAL DATA ' PLASTIC QUAD FLATPACK PMlS·PQFP·G64 1r-°. 271 • 1 0.08®1 0.17 33 49 32 64 17 ~ __________ 12'~SQ __________~ 11.80 Seating Plane 4040152/B-10/94 NOTES: A. All linear dimensions are in millimeters. B. This drawing is subject to change without notice. C. Falls within JEDEC MO-136 ~1ExAs 5-124 INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 64-Pln TQFP (PM)t Package Power Dissipation Derating at Varying Air Flows 3000 ------ 2800 2600 Oft/min 100ft/min 250ft/min 500ft/min 2400 2200 ~I 2000 I 1800 c 1600 c J ~ 1200 ! ... ""'-.. ...... ""'-.. ... ~ ... ~ ""'- 1400 E ""'-.. ... __ __ ... ""'-.. ...... ""'-.. ... ~ ... ---- -- 1000 ...... ""'-.. ... ~ ... ""'-.. ...... ""'-.. "'""",--- ............................ ~ ---- ---- ---- ---... -~ ... ...... ~ ... 800 ...... 600 400 200 o~--~--~--~--~----~--~--~--~--~--~~--~--~--~ 25 30 40 70 75 65 90 35 45 50 55 60 65 80 Ambient Temperature - °C . Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT (OC/W) 9 JUNCTION TO CASE (DC/W) 1 ft3 box 0 0 Wind tunnel 95.5 92.5 NJA N/A N/A 100 Wind tunnel 87.8 250 Wind tunnel 72.9 NJA 500 Wind tunnel 57.8 N/A N/A NJA N/A 10.4 t Mounted on pnnted-clI'Cult board ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DAUAS. TEXAS 75265 5-125 MECHANICAL AND THERMAL DATA PN/8-PQFP-G80 PLASTIC QUAD FLATPACK 61 40 80 21 Seating Plane 404013518-10194 NOTES: A. All linear dimensions are in millimeters. B. This drawing is subject to change without notice. C. Falls within JEDEC Mo-136 ~ThxAs 5-126 INSTRUMENTS POST OFFICE BOX 855303 • DALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA SO-Pin TQFP (PN) Package Power Dissipation Derating at Varying Air Flows 3000 oft/mln 2800 100ft/mln 2600 500ft/mln 250ft/mln 2400 2200 ~I 2000 0 1800 c: i 1800 Q J 1400 E :::a E 1200 -= 1000 ::E 600 800 400 200 0 25 30 35 45 40 50 60 55 65 70 75 80 85 90 Ambient Temperature - ·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT ("C/W) 9 JUNCTION TO CASE ("C/W) 100 Wind tunnel 79.1 250 Wind tunnel 67.3 500 Wind tunnel 54.2 N/A N1A N1A N1A N1A N1A N/A N/A 10.6 0 1 ft3 box 89.2 0 Wind tunnel 87.8 t Mounted on prnrted-clrcult board ~TEXAS INSTRUMENTS POST OFFICE BOX 655303 • DAUAS. TEXAS 75265 5-127 MECHANICAL AND THERMAL DATA PQl8-PQFp·G*** PLASTIC QUAD FLATPACK 100 LEAD SHOWN 13 f 0.012 (0,30) 0.008 (0,20) "03"TYP 0.025 (0,635) t 39 1 + - - - - - "01" SQ M------ "0" SQ 0.150 (3,18) 0.130 (3,30) 63 • ----~ ---------lM 1 + - - - - - - "02" SQ - - - - - - + 1 f Seating Plane r---rlo.---rl0""'.0::-:0~4""(0""',10~)1 0.180 (4,57) MAX 40400451B-10194 NOTES: A. All linear dimensions are in inches (millimeters). B. This drawing is subject to change without notice. C. Falls within JEDEC MO-069 ~1ExAs 5-128 INSTRUMENTS POST OFFICE BOX 655303 • OALLAS. TEXAS 75265 MECHANICAL AND THERMAL DATA 132·Pln PQFP (PQ) Package Power Dissipation Derating at Varying Air Flows 5000 --- --- ------- --- ------ --- --- --- ----- - -- ----- --- --. --.---- --- --- . --- -- --.--. --Oft/min 4500 100 ft/mln 250ft/min 500 ft/mln 4000 ~ E I 3500 i 3000 .. 2500 c is -= ~ E :::I E I 2000 1500 1000 500 0 25 30 35 40 45 50 60 55 65 70 75 80 65 90 Ambient Temperature - ·C Thermal Resistance (G» Measurementst AIRFLOW (ft/mln) ENVIRONMENT G> JUNCTION TO AMBIENT G> JUNCTION TO CASE (·C/W) (·C/W) 0 1 ft 3 box 49.6 0 Wind tunnel 49.7 100 Wind tunnel 42.6 NJA N/A NJA 250 Wind tunnel 34.1 N/A 500 Wind tunnel 27.2 NJA 9.S N/A N/A t Mounted on pnnted-clrcUit board N/A ~1ExAs INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-129 MECHANICAL AND THERMAL DATA PLASTIC QUAD FLATPACK PZ/8-PQFP·G100 76 60 100 26 16,20 SQ _ _~_--.! 16,80 1,46 1,36 Seating Plane 1,60 MAX 4040149/B-10/94 NOTES: A. All linear dimensions are In millimeters. B. This drawing is subject to change without notice. C. Falls within JEDEC MO-136 ~1ExAs 5-130 INSTRUMENTS POST OFFICE BOX 665303 • DAU.AS. TEXAS 76266 MECHANICAL AND THERMAL DATA 10o-Pln TQFP (PZ) Package Power Dissipation Derating at Varying Air Flows 3000 Oft/min 2800 100 ft/mln 2800 500ft/min 250ft/min 2400 2200 ~ E 2000 I c 0 ia. ·21 is I 1800 1600 1400 D. E ::I E "5C 1\1 :::I! 1200 1000 800 600 400 200 0 25 30 40 35 45 50 55 60 65 75 70 80 85 90 Ambient Temperature - ·C Thermal Resistance (9) Measurementst AIRFLOW (ft/mln) ENVIRONMENT 9 JUNCTION TO AMBIENT (·C/W) 9 JUNCTION TO CASE (·C/W) 0 1 tt 3 boX 79 N/A 0 Wind tunnel 72.7 N/A 100 Wind tunnel 65.2 N/A 250 Wind tunnel 54.5 NlA 500 Wind tunnel 43.6 N/A N/A N/A N/A 11.6 t Mounted on pnnted-clrcult board -!111EXAS INSTRUMENTS POST OFFICE BOX 655303 • DALLAS. TEXAS 75265 5-131 5-132 6-1 c ~ _. (') CD i: o Co CD - en 6-2 VHDL Models VHDL models for the FIFOs listed below are available through your local Texas Instruments sales office: SN74ACT2226 (Dual 64 x 1, Dual 256 x 1 Clocked FIFO) SN74ACT3632-15/20/30 (512 x 36 x 2 Clocked FIFO) SN74AcT3651-15/20/30 (2048 x 36 Clocked FIFO) SN74ACT7881-15/20/30 (1024 x 18 Clocked FIFO) SN74ACT7804-20/25/40 (512 x 18 Strobed FIFO) SN74ACT7813-15/20/25/40 (64 x 18 Clocked FIFO) SN74ACT7814-20/25/40 (64 x 18 Strobed FIFO) Logic-Modeling Behavioral Models 6-6 LOGIC-MODELING BEHAVIORAL MODELS The Logic Modeling Group continually updates model libraries. If you do not see a TI FIFO model that you need, contact a local Logic Modeling Group representative or call 1-800-34MODEL (1-800-346-6335) for current availability. For international inquiries, call 503-690-6900. Behavioral models of the following TI FIFOs are available from the Logic Modeling Group, 19500 N.W. Gibbs Drive, P.O. Box 310, Beaverton, OR 97075, 503-690-6900, fax: 503-690-6906, Technical Product Support: 1-800-445-1888. DEVICE ARCHITECTURE ORGANIZATION SN74ACT2226 Clocked 64 x 1 SN74ACT2228 Clocked 256 x 1 SN74ACT2227 Clocked 64 x 1 SN74ACT2229 Clocked 256 x 1 SPEED SORTS SN74ABT3611 Clocked 64 x 36 -15, -20, -30 SN74ABT3612 Clocked 64x36x2 -15, -20, -30 SN74ABT3613 Clocked 64 x 36 -15, -20, -30 SN74ABT3614 Clocked 64x36x2 -15, -20, -30 SN74ACT3632 Clocked 512 x 36 x 2 -15, -20, -30 SN74ACT3641 Clocked 1K x 36 -15, -20, -30 SN74ACT7813 Clocked 64 x 18 -15, -20, -25, -40 SN74ACT7803 Clocked 512 x 18 -15, -20, -25, -40 SN74ABT7819 Clocked 512 x 18 x 2 -12, -15, -20, -30 SN74ACT7811 Clocked 1K x 18 -15, -18, -20, -25 SN74ACT7807 Clocked 2K x 9 -15, -20, -25, -40 SN74ACT7814 Strobed 64 x 18 -20, -25, -40 SN74ACT7806 Strobed 256 x 18 -20, -25, -40 SN74ACT7804 Strobed 512 x 18 -20, -25, -40 SN74ABT7820 Strobed 512 x 18 x 2 -15, -20, -25, -40 SN74ACT2235 Strobed 1K x 9 x 2 -20, -30, -40, -60 SN74ACT2236 Strobed 1K x 9 x 2 -20, -30, -40, -60 SN74ACT7808 Strobed 2K x 9 -20, -25, -30, -40 SN74ACT7201 Asynchronous 512 x 9 -15,-25,-35,-50 SN74ACT7202 Asynchronous 1K x 9 -15, -25, -35, -50 SN74ALS2233 Asynchronous 64 x 9 SN74ALS2238 Asynchronous 32x9x2 SN74ALS2232 Asynchronous 64 x 8 SN74ALS236 Asynchronous 64 x 4 SN74ALS232 Asynchronous 64 x 4 ~1ExAs ' INSTRUMENTS POST OFFICE BOX 655303 • OALlAS. TEXAS 75265 6-7 Quality and Reliability Assurance 7-1 D r:::: D) 7-2 Qualification of Products and Processes 7-3 7-4 Introduction A significant change has recently occurred in the way our customers qualify products. At one point, virtually all customers were spending millions of dollars annually, duplicating supplier qualification tests. However, as years of improving qUality and reliability have raised the level of customer confidence and satisfaction, much of this in-house customer testing was eliminated and customers began to rely on Texas Instruments (TI) test results. This approach also became applicable to major process changes. If a major process change is made to a product, TI qualifies the new process through extensive testing and transmits the data to customers for approval. Both of these qualification approaches form the basis of the quality and reliability assurance practices for TI's FIFO products. 7-6 7-<3 Quality and Reliability Assurance in Integrated-Circuit Design Page General Quality and Reliability Rules .................................................. 7-9 Process-Specific Design Rules ...............................................•...... 7-9 Methodology-Oriented Design Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 7-9 Change Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 7-9 7-7 7-8 Quality and Reliability Assurance in Integrated-Circuit Design Quality and reliability of the products are important elements in achieving customer satisfaction. These elements must be designed in during the design phase of the products and built in during fabrication. To ensure that this occurs, all integrated-circuit designs at Texas Instruments (TI) follow a three-tier design-rule structure as follows: • • • General company-wide quality and reliability rules Process-specific design rules Methodology-oriented design rules General Quality and Reliability Rules The process-independent company-wide quality and reliability rules are defined and, as necessary, adapted to new requirements by a committee that is organized worldwide. These rules include directives for: • • • • • ESD protection Latch-up protection Electromigration constraints on current density Maximum junction temperature Hot-electron injection Strict application of these rules results in high process yields and high quality and reliability. These rules are company standards at TI' the result of many years of experience in semiconductor production. Process-Specific Design Rules These rules are created by a design council that is composed of members from different areas including process development, product development, wafer fabrication, test development, and quality and reliability engineering. The design council meets on a regular basis to generate and periodically update the design notebook, which contains all the process-specific design rules. The design council acts as a forum for worldwide sharing of experience, knowledge, and problems related to a specific process. The design notebook is maintained on a central computer, which can be accessed by any authorized user at any worldwide TI facility; therefore, the latest version is instantly available for worldwide access. Major chapters of the notebook are: • • • • • • • • • • General guidelines for circuit design Process parameters Simulation models for processes Simulation instructions and guidelines Design methodology Design documentation Process-layout rules Reliability rules Electrical models for circuit packages Requirements for photomask generation in wafer fabrication Methodology·Orlented Design Rules The guidelines for design methodology are basically the same for all types of semiconductors at TI and are rigidly adhered to. This gives the designer a high confidence level that the target specifications are maintained, the functions of a device are correctly implemented, and a high standard of reliability is achieved. Designing integrated circuits is now a classic application for computer-aided design. Each designer has a network at their disposal for circuit entry, circuit simulation, mask layout, and circuit verification. Change Control Continuous improvement at TI is recognized as a key method of offering greater value to our.1ong-te~ customers. A major responsibility is to ensure that improvements in the materials and processes do not adversely affect our customers. One of TI's missions is to assure our customers of a source of qualified supply and to provide a stable market to our suppliers. Through joint development efforts with strategic suppliers, TI has provided our customers with a steady stream of improvements to the materials that are used in volume production. In addition, to remain competitive, new equipment 7-9 -- - - - - - - ------- and process methods are constantly being evaluated. It is TI's goal to keep our customers infonned of all changes identified as requiring notification. The change-control system is contained in TI's quality system and covers all operations worldwide. The list of changes requiring notification (see Table 1) is based upon a dynamic composite of documented customer requirements and experience. The required qualification testing and process-capability studies also are dermed in the quality system. The change notification contains a description of the change, reason for change, anticipated impact on customers, if any, supporting reliability and applicable electrical-characterization data, and effective timing. Table 1. Examples of Major Changes That can Require User Notification PROCESS 7-10 MAJOR DESIGN CHANGE WaferFab Wafer-fab site Prooessflow Gate-oxide materials Dielectric material Metallization material Passivation material Die-coating material Wafer diameter Diffusion dopant Gate-oxide thickness Polysillcon-dopant type Metallization thickness Passivation thickness Die-coating thickness Assembly Assembly site Plating material Wire-bond material Sealing material Marking method Leadframe-base material Pleting method Mold-compound material Die-attech material Marking appearance Test Elimination of test steps Electrical Specification Relaxation of AC specification Relaxation of DC specification Mechanical Specification Case outline Loosening tolerance(s) Packing! Shipping! Labeling! Environment Carrier (real, tray) dimensions Maximum storage temperature Drypack requirements Quality and Reliability Monitoring Page Environmental Laboratory.; ...............•........................................ Methods of Measuring Component Reliability ......................................... Failure-Rate Calculations for FIFO Products . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Qualification Data for FIFO Products. . . . • . . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. Life-Test and ESD-Characterization Data. . . . . . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 7-13 7-13 7-13 7-13 7-14 7-11 7-12 Quality and Reliability Monitoring Environmental Laboratory The environmental-test laboratories provide environmental-test services, both climatic and mechanical, for device qualification, engineering evaluation, and acceptance purposes to operating entities within Texas Instruments (TI). TI acquires and maintains suitable calibrated equipment with which the test and inspections required by external customers and internal specifications can be performed. A properly trained and certified staff is maintained to perform the required tests and inspections in a timely, cost-effective manner. The TI laboratories worldwide are maintained in a certified/ approved status for those customers and agencies requiring this condition. TI conforms to national and international standards and is certified as an approved self-qualification lab for several major customers. Lab personnel interface closely with equipment manufacturers and standards bodies for maintenance of test capabilities. Environmental-test methods and specifications are developed and controlled for TI worldwide. For continuous improvement in maintaining world-class status, the environmental lab uses self-measurement indices by tracking cycle time and customer satisfaction, both internal and external. Methods of Measuring Component Reliability Product-reliability tests are performed at high-stress conditions so that performance levels can be established during a relatively short test duration. Specific stress conditions are chosen because they represent accelerated versions ofvarious device-application environments and allow meaningful extrapolations to lower stress levels. These reliability-test conditions are held constant so that product improvements or deficiencies can be readily discerned by comparing current test data with the historical database on identical tests. However, extreme care is exercised to avoid any overstress condition that could cause a device failure not related to the final device application. The reliability stresses used most widely at TI are: • • • • • Life stress Biased temperature and humidity Biased highly accelerated stress test (HAST) Nonbiased autoclave Temperature cycling (air-to-air) Failure-Rate Calculations for FIFO Products The failure-rate performance for FIFO products has been calculated to be 15 FITs (failures per 106 device hours). This calculation is based upon the following assumptions: • • • • • Applied supply voltage: 5 V Junction temperature; 125°e or 1500 e Activation energy; 0.7 eV Derating temperature; 55°e ehi2 upper confidence level: 60% Qualification Data for FIFO Products Qualification data has been gathered for each of the key design sets in TI's FIFO product line. This data represents the results from life test and ESD characterization (see Table 1). 7-13 Table 1. Life-Test and ESD-Characterlzatlon Data LIFE TEST DESIGN SERIES QUALIFICATION VEHICLE PACKAGE TYPE SN74ABT3612 SN74ABT3614 PCB SN74ABT3613 SN74ABT3614 PCB/PO 125°C, 68HRS DYNAMIC 126"C, 1000HRS DYNAMIC 01116 SN74ABT3614 SN74ABT3614 PCB SN74ABT7819 SN74ABT7819 PH/PN SN74ABT7820 SN74ABT7819 PH/PN SN74ACT2235 SN74ACT2235 FN SN74ACT2236 SN74AqT2235 FN SN74ACT7808 SN74ACT7808 PAG 01116 SN74ACT2226 SN74ACT2229 OW 01116 SN74ACT2227 SN74ACT2229 OW SN74ACT2228 SN74ACT2229 OW SN74ACT2229 SN74ACT2229 OW SN74ACT3632 SN74ACT3632 PCB/PO SN74ACT3641 SN74ACT3632 PCB/PO SN74ACT7803 SN74ACT7803 OL SN74ACT7804 SN74ACT7803 OL SN74ACT7805 SN74ACT7803 OL SN74ACT7806 SN74ACT7803 OL SN74ACT7807 SN74ACT7807 FN 01116 SN74ACT7811 SN74ACT7801 FN/PN 01116 SN74ACT7813 SN74ACT7803 OL SN74ACT7814 SN74ACT7803 OL 01116 01116 01116 01116 01116 ' HUMAN·BODY MODEL (V) 7-14 ESDLEVEL 160°C, 300HRS STATIC MACHINE MODEL (V) Levell 0-1999 0-199 Level II 2000-3999 200-399 Level III 4000 or greater 400 or greater HUMAN· BODY MODEL MACHINE MODEL NOTES NOTES NOTES NOTES NOTES NOTES NOTES NOTES NOTES TI Worldwide Sales and Representative Offices AUSTRALIA I NEW ZEALAND: Texaalnatruments Austrella Lid.: Melbourne [61]3-696-1211, Fax 3-696-4446; Sydney 2-910-3100, Fax 2805-1186. BELGIUM: Texas Instruments Belgium S.AJN.V.: Brussels [32] (02) 728 75 80, Fax (02) 726 72 76. BRAZIL: Texas Instrumentos Electronlcos do Breall LIds.: Sao Paulo [55]11-535-5133. CANADA: Texas Instruments Canads LId.: Montreal (514) 421-2750; Ottawa (613) 726-320·1; Fax 726-6363; Toronto (905) 864-9181; Fax 8840062. DENMARK: Texas Instruments AlS: Ballerup [45] (44) 68 7400. FRANCEIMIDDLE EAST/AFRICA: Texas Instruments Frence: VellzyVillacoubley [33](1) 30 701001, Fax (1) 30 70 10 54. GERMANY: Texas Instruments Deutachland GmbH.: Frelslng [49] (08161) 800, Fax (08161)804516; Hannover (0511)904960, Fax (0511)6490 331; Ostflldern (0711) 340 30, Fax (0711) 340 32 57. HONG KONG: Texas Instruments Hong Kong LId.: Kowloon [852] 29567286, Fax 2956-2200. HUNGARY: Texas Instruments Representation: Budapest [36] (1) 269 63 10, Fax (1) 267 1357. IRELAND: Texas Instruments Ireland LId.: Dublin [353] (01) 475 52 33, Fax (01) 478 14 63. ITALY: Texas Instruments Italla S.p.A.: Agrete Brlanza [39](039) 684 21, Fax (039) 684 29 12; Rome (06) 6572651. JAPAN: Texas Instrumenta Japan Ltd.: Kanazawa [81] 0782-23-5471, Fax 0762-23-1583; Klta Kanto 0485-22-2440, Fax 0485-23-6787; Kyoto 075341-7713, Fax 075-341-7724; Kyushu 0977-73-1557, Fax 0977-73-1563; Matsumoto 0263-33-1060, Fax 0263-35-1025; Nagoya 052-232-5601, .Fax 052-232-7886; Osaka 06-204-1861, Fax 06-204-1895; Tachlkawa 0425-276760, Fax 0425-27-6426; Tokyo 03-3769-8700, Fax 03-3457-6n7; Yokohama 045-338-1220, Fax 045-338-1255. KOREA: Texas Instruments Korea Lid.: Seoul [82]2-551-2804, Fax 2551-2828. MAINLAND CHINA: Texas Instruments Chins Inc.: Beijing [86]10-50G2255, Ext. 3750, Fax 10-500-2705. MALAYSIA: Texas Instruments Malaysia Sdn Bhd: Kuala Lumpur [60]3208-6001, Fax 3-23G-6605. MEXICO: Texas Instruments de Mexico S.A. de C.V.: Colonia dsl Valle [52] 5-639-9740. NORWAY: Texas Instruments Norge AlS: Oalo [47] (02) 264 75 70. PHILIPPINES: Texas Instruments Asia Ltd.: Metro Manila [63]2-836-0980, Fax 2-831-7702. SINGAPORE (& INDIA, INDONESIA, THAILAND): Texas Instruments Singapore (PTE) Ltd.: Singapore [65] 39G-7100, Fax 39G-7062. SPAIN/PORTUGAL: Texas Instruments Espatla S.A.: Madrid [34] (1) 372 8051, Fax (1) 307 68 64. SUOMUFINLAND: Texas InstrumentaJOY: Espoo [358] (0) 43 54 20 33, Fax (0) 46 73 23. SWEDEN: Texas Instruments International Trede Corporetlon (Sverigefilialen): Klsta [46] (06) 7525800, Fax (08) 751 97 15. SWITZERLAND: Texas Instruments Swltzarland AG: Dletlkon [41]886-23771450. TAIWAN: Texas Instruments Taiwan Limited: Taipei [886] 2-378-8800, Fax 2-377-2718. THE NETHERLANDS: Texas Instruments Holland, B.V. Amsterdam [31] (020) 5469800, Fax (020) 846 31 36. UNITED KINGDOM: Texas Instruments Ltd.: Northampton [44] (01604) 66 3000, Fax (01604) 66 30 01. UNITED STATES: Texas Instruments Incorporated: ALABAMA: Huntsville (205) 430'()114; ARIZONA: Phoenix (602) 224-7800; CALIFORNIA: Irvine (714) 660-1200; Los Angeles (818) 704-8100; San Diego (619) 278-9600; San Jose (408) 894-9000; COLORADO: Denver (303) 488-9300; CONNECTICUT: Wallingford (203) 269'()074; FLORIDA: Fort Lauderdale (214) 644-5580 orlando (407) 867-5308; Tampa (813) 573-0331; GEORGIA: Atlanta (770) 682-7967; ILLINOIS: Chicago (708) 5174500; INDIANA: Indianapolis (317) 573-6400; KANSAS: Kansas City (913) 451-4511; MARYLAND: Baltimore (410) 312-7900; MASSACHUSETTS: Boston (617) 895-9100; MICHIGAN: DetrOit (810) 305-5700; MINNESOTA: Minneapolis (612) 828-9300; NEW JERSEY: Edison (908) 90S'()033; NEW MEXICO: Albuquerque (505) 345-2555; NEW YORK: Long Island (516) 454-6600; Poughkeepale (914) 897-2900; Rochester (716) 385-6170; NORTH CAROLINA: ChartoHe (704) 522-5487; Raleigh (919) 876-2725; OHIO: Cleveland (216) 328-2149; Dayton (513) 427-6200; OREGON: Portland (503) 643-6758; PENNSYLVANIA: Philadelphia (610) 825-9500; PUERTO RICO: Hato Ray (809) 753-6700; TEXAS: Austin (512) 250-5769; Dallas (214) 917-1264; Houston (713) 778-6592; Midland (915)561-6521; WISCONSIN: Milwaukee (414) 798-5021. North Amsrlcan AuthOrized Distributors COMMERCIAL 80().426-1410 180().452-9185 Oregon only Almac I Alrow Anthem Electronics 800-826-8436 800-777-2776 Arrow I Sdlweber Future Electronics (canada) 8OG-368-8731 80G-332-8638 . HamiHon Hallmark 8OO-522'()084 or www.marshall.com Marshall Industries 8Q().414-4144 Wyle OBSOLETE PRODUCTS Rochestar Electronics 508-482-9332 MILITARY Alliance Electronics Inc 800-508-9494 Future Electronics (Canada) 8OG-388-8731 HamiHon Hallmark 80G-332-8638 800-524-4735 Zeus, An Arrow Company TI DIE PROCESSORS Chip Supply (407) 298-7100 (818) 768-7400 Elmo Semiconductor Minco Technology Labs (512) 634-2022 CATALOG Allied Electronics 800-433-5700 Arrow Advantage 800-m-2776 Newark Electonics 800-367-3573 For DlSI11butors outside North Amertca, contact)'OlJr local Sales OfIIos. Important NoIIco: T.,... Instrumonta (TI) reaerves tho righllo make changes 10 or 10 diacontinue any product or service idenllfied in III~ pubicatiorl wiIhout nollce. TI advise. is ,uslome.. 10 obIaln!he I_I version of the relevant information to verify, before pfac.ing orders, that the information being relied upon Is current Ploue be advised lIIal TI warranls lis ....mconduclO( producls and related soIIwarelo lIIe specification. applicable allllo time 01 saloln accordance willi TI"lIIandard warranly. Tlassumas no llabiily lor applications assl.lanCe. soIIwa.. performance. or 1II1rd-par\y product InIormallOn. or lor Inlllngemoni 01 palenls or aervlces deoc:rlled in IIIIs publcation. Tlassumea no rasponslbllily lor ,uslome..•appIlcallons or l~roducldeolgnL A022396 @ 1996 Texes Instruments Incorporated Printed In the USA • TEXAS INSTRUMENTS ~lExAs INSlRUMENTS Printed in U.S.A. 0496-CP SCAA012A
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.3 Linearized : No XMP Toolkit : Adobe XMP Core 4.2.1-c041 52.342996, 2008/05/07-21:37:19 Create Date : 2017:08:03 08:45:42-08:00 Modify Date : 2017:08:03 09:49:13-07:00 Metadata Date : 2017:08:03 09:49:13-07:00 Producer : Adobe Acrobat 9.0 Paper Capture Plug-in Format : application/pdf Document ID : uuid:27f83c66-f921-8c41-b470-b1fa11e4baa1 Instance ID : uuid:38eb862c-bc20-6f41-a987-96438a462161 Page Layout : SinglePage Page Mode : UseNone Page Count : 482EXIF Metadata provided by EXIF.tools