Baron Services DSSR-250C Pulsar Digital Solid-State Radar System User Manual

Baron Services Inc Pulsar Digital Solid-State Radar System

Plot Assisted Setups

Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–14. Plot-Assisted SetupsThe IFD receiver module replaces virtually all of the IF components in a traditional analogreceiver. The alignment procedures for those analog components are usually very tedious, andrequire continued maintenance even after they are first performed. Subtle drifts in componentspecifications often go unnoticed until they become so severe that the radar’s data arecompromised.The RVP8 makes a big improvement over this by providing an interactive graphical alignmentprocedure for burst pulse detection, Tx/Rx phase locking, and calibration of the AFC feedbackloop. You may view the actual samples of the burst pulse and receiver waveform, examine theirfrequency content, design an appropriate matched filter, and observe live operation of the AFC.It is a simple matter to check the spectral purity of the transmitter on a regular basis, and todiscover the presence of any unwanted noise or harmonics. Moreover, the RVP8 is able to trackand modify the initial settings so that proper operation is maintained even with changes intemperature and aging of the microwave components.The Plot-Assisted Setups are accessed using the various “P” commands within the normal TTYsetup interface. These commands are described later in this chapter. The RVP8 supportsopcodes that allow the host computer to monitor the data being plotted. The dspx utility candisplay these plots directly on the workstation screen, and thus, can carry out the graphicalcheckup and alignment procedures remotely via a network.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–24.1 P+ — Plot Test PatternThe RVP8 can produce a simple test pattern to verify that the display software is workingproperly. From the TTY monitor enter the “P+” command. This will print the message“Plotting Test Pattern...” on the TTY and then produce the plot shown in Figure 4–1.This display is actually an overlay of six different strokes: 1) bottom line, 2) middle line, 3) topline, 4) line sloping up, 5) line sloping down, and 6) the sine wave pattern. The later changesphase with each plot so that, with a little imagination, it appears to be radiating from the left sideof the display.Figure 4–1: The Test Pattern DisplayWhen you are satisfied that the plot is being drawn correctly, type “Q” or hit ESC to return tothe TTY monitor.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–34.2 General Conventions Within the Plot CommandsThe “Pb”, “Ps”, and “Pr” commands all have a similar structure to their TTY user interface.Each command begins by printing a list of subcommands that are valid in that context. Thesesubcommands are single keystrokes that are executed immediately by the RVP8 as they aretyped. The “ENTER” key is not required. The available subcommands are different for eachplot command; but, as much as possible, each key has a similar meaning across all commands.The working and measured parameters for each plot command are printed on the TTY as twolines of information following the subcommand list. The first line contains settings that onlychange when a subcommand is issued; but the second line is live and reflects the current statusof the burst input, the IF input, or the AFC output. The first line is printed just once, but thesecond line is continually overprinted on top of itself. This makes it appear as a live status linewhose values always remain up to date. The ”Pb”, ”Ps”, and ”Pr” commands will report ”NoTrigger” on the TTY status line whenever the external trigger is expected but missing.The TTY screen will scroll upward each time a new subcommand is executed, so that a historyof information lines and command activity can be seen on the screen. You may also use theCarriage-Return key to scroll the display up at any time. If the initial list of subcommandsdisappears off the top, you may type “?” to force a reprint. To exit the plot command entirelyand return to the TTY main menu type “Q” or ESC. These basic “help” and “exit” keystrokesapply everywhere within the RVP8 setup menus. To save space and minimize clutter on theTTY screen, they are not shown in the itemized list of subcommands.Most commands have a lowercase and an uppercase version. If a lowercase command doessomething, then its uppercase version does the same thing but even more so (or in reverse). Forexample, if the “w” subcommand widens something by a little bit, then “W” would widen it alot. This simple convention reduces the number of different subcommand keys that are needed,and makes the interface easier to memorize.The graphical display and TTY status lines are continually updated with fresh data several timesper second. Occasionally it is useful to freeze a plot so that it can be studied in more detail, orcompared with earlier versions. To accomplish this, every plotting command supports a “SingleStep” mode that is accessed by typing the “.” (period) key. This key causes the display and TTYstatus lines to freeze in their present state, and the message “Paused...” will be printed.Subsequently, typing another “.” will single step to the next data update, but the plot and printoutwill still remain frozen. Typing “Q” or ESC will exit the plot command entirely (as theynormally do). All other keys return the plot command to its normal live updating, but the key isotherwise discarded (i.e., subcommand keys are not executed while exiting from single stepmode).All of the plot commands support subcommands whose only purpose is to alter the appearanceof the display, e.g., zoom, stretch, etc. These subcommands make no changes to the actualworking RVP8 calibrations. However, the display settings are stored in nonvolatile RAM justlike all of the other setup parameters. This means that all previous display settings will berestored whenever you restart each plot command. This is very convenient when alternatingamong the various plots.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–4The “Pb”, “Ps”, and “Pr” commands are intended to be used together for the combined purposeof configuring the RVP8’s digital front end. You may, of course, run any of the commands atany time; but the following procedure may be used as a guideline for first time setups. The fullprocedure must be repeated for each individual pulsewidth that the radar supports.1. Use Mb to set the system’s intermediate frequency (See Section 3.2.6).2. Use Mt to choose the PRF and pulsewidth (See Section 3.2.4). Also, choose therange resolution now, as it may constrain the design of the matched filter later.3. Use Mt0,Mt1, etc., to set the relative timing of all RVP8 triggers that are used bythe the radar. Do not worry about the absolute values of the trigger start times.Just setup their polarity and width, and their start times relative to each other (SeeSection 3.2.5). Make an initial guess of FIR filter length as 1.5 times thepulsewidth.4. Use Pb to slew the start times of all triggers so that the burst pulse is properlysampled (See Section 4.3). Refine the impulse response length if necessary sothat all samples easily fit within the display window.5. Use Ps to design the matched FIR filter (See Section 4.4). Further refine theimpulse response length and passband width to achieve a filter that matches thespectral width of the burst, and that has strong attenuation at DC. If the FIRlength is changed, return to Pb to verify that the burst is still being sampledproperly.6. Continue using Ps and Mb to tune up the AFC feedback loop. The settings thatwork for one pulsewidth should also work for all others.7. Use Pr to verify that targets are being detected with good sensitivity (See Section4.5).Sometimes it is useful to run the Pb and Ps commands with samples from the IF-Input of theIFD, rather than from the Burst-Input. Likewise, it is sometimes useful to view the Pr plots onsamples of Burst data. The top-level “~” command (See Section 3.3.3) allows you to do thiseasily.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–54.3 Pb — Plot Burst Pulse TimingFor magnetron radars the RVP8 relies on samples of the transmit pulse to lock the phase of itssynthesized “I” and “Q” data, and to run the AFC feedback loop. The “Pb” command is used toadjust the trigger timing and A/D sampling window so that the burst pulse is correctly measured.4.3.1 Interpreting the Burst Timing PlotThe display plot will ultimately resemble Figure 4–2, which shows a successful capture of thetransmitter’s burst pulse. The horizontal axis of the display represents time, and the overall timespan from the left edge to the right edge is listed as “PlotSpan” on the TTY.Figure 4–2: Successful Capture of the Transmit BurstThe upper portion of the plot shows the sampling window wherein the burst pulse is measured.The duration of this window is determined by the impulse response length of the matched FIRfilter. This is because the same FIR coefficients that compute “I” and “Q” are also used tocompute the reference phase vectors for the burst pulses. The A/D samples of the RVP8/IFD’sburst input are plotted (somewhat brighter) within the sample window.The RVP8 computes the power-weighted center-of-mass (COM) of the burst pulse envelope.This allows the processor to determine the location of the “middle” of the transmitted pulsewithin the burst analysis window. The Pb plot displays small tick marks on the top and bottomof the burst sample window to indicate the location of the COM. These markers are onlydisplayed when valid burst power is detected. A second “error bar” is drawn surrounding thetick mark to indicate the uncertainty of the mark itself. This error interval is used by the burstpulse tracking algorithm to decide when a timing change can be made with confidence.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–6It is possible to independently choose a subinterval of burst pulse samples that are used by theAFC frequency estimator. Thus, the AFC feedback loop is not constrained to use the same set ofsamples that are chosen for the FIR filter window. The FIR window typically is longer than theactual transmitted pulse, and thus, the samples contributing to the frequency estimate willinclude the leading and trailing edges of the pulse. These edges tend to have severe chirps andsidebands, compared to the more pure center portion of the pulse. The AFC frequency estimate(which is power weighted) could be mislead by these edges and might not tune to the optimumcenter frequency if they were included.The lower portion of the plot shows the six triggers that are output by the RVP8. Trigger #0 is atthe top, and Trigger #5 is on the bottom. They are drawn in their correct polarity and timingrelative to each other, and relative to the burst sample window. Note that the sample window isalways drawn in the center of the overall time span. Thus, depending on the PlotSpan andlocation of the six trigger’s edges, triggers that do not vary within the plotted time span willappear simply as flat lines.The RVP8 defines “Range Zero” to occur at the center of the burst sample window. This alsodefines the zero reference point for the starting times of the six programmable triggers. Forexample, a trigger whose starting time is zero will be plotted with its leading edge in the exacthorizontal center of the display. Knowing this convention makes the absolute value of thetrigger start times more meaningful.4.3.2 Available Subcommands Within “Pb”The list of subcommands is printed on the TTY:Available Subcommands within ’Pb’: I/i Impulse response length Up/Dn A/a & S/s Aperture & Start of AFC window L/l & R/r Shift triggers & RVP8/Tx waveform left/right, T/t Plot time span Up/Dn Z/z Amplitude zoom B/b BP Tracking On/Off (temporary) + Hunt for missing burst . Single StepThese subcommands change the matched filter’s impulse response length, shift the radartriggers, and alter the format of the display. I/i The “I” command increments or decrements the length of the matchedfilter’s impulse response. Each keystroke raises or lowers the FIRlength by one tap.A/a & S/s These commands raise/lower the aperture/start of the subwindow ofburst pulse samples for AFC. If you never use these commands, thenthe full FIR window will be used; however, shortening the AFCinterval will result in two sample windows being drawn on the plot.The smaller AFC window should be positioned into the center portionof the transmitted pulse, where the burst amplitude and frequency arefairly stable.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–7L/l & R/r These two commands shift the entire group of six RVP8 triggers leftor right (earlier or later in time). The lowercase commands shift in0.025 msec steps, and the uppercase commands shift in 1.000 msecsteps (approximately). The reason for shifting all six triggers at onceis that the relative timing among the triggers remains preserved.However, the absolute timing (relative to range zero) will vary, andthis will cause the burst pulse A/D samples to move within the samplewindow. T/t The “T” command increments or decrements the overall time span ofthe plot. The available spans are 2, 5, 10, 20, 50, 100, 200, 500, 1000,2000 and 5000 microseconds. The value is reported on the TTY as“PlotSpan”. Z/z The “Z” command zooms the amplitude of the burst pulse samples sothat they can be seen more easily. The value is reported on the TTY as“Zoom”.B/b These keys temporarily disable or re-enable the Burst Pulse Tracker.The tracker must be disabled in order for the L/R keys to be used toshift the nominal trigger timing. The “b” key disables tracking andsets the trigger slew to zero; the “B” key re-enables tracking startingfrom that zero value.+The “+” subcommand initiates a hunt for the burst pulse. Progressmessages are printed as successive AFC values are tried, and thetrigger slew and AFC level are set according to where the pulse wasfound. If no burst pulse can be found, then the previous trigger slewand AFC are not changed.4.3.3 TTY Information Lines Within “Pb”The TTY information lines will resemble:Zoom:x2, PlotSpan:5 usec, FIR:1.36 usec (49 Taps)Freq:27.817 MHz, Pwr:–53.9 dBm, DC:0.14%, Trig#1:–5.00, BPT:0.00 Zoom Indicates the magnification (in amplitude) of the plotted samples. Azoom level of “x1” means that a full scale A/D waveform exactly fillsthe height of the sample window. Generally, the signal strength of theburst pulse will not be quite this high. Thus, use larger zoom levels tosee the weaker samples more clearly. You may zoom in powers of twoup to x128.PlotSpan Indicates the overall time span in microseconds of the complete scopedisplay, from left edge to right edge. FIR Indicates the length of the impulse response of the matched FIR filter,and hence, the duration of the burst pulse sample window. The lengthis reported both as a number of taps, and as a time duration inmicroseconds.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–8 Freq Indicates the mean frequency of the burst, derived from a 4th ordercorrelation model. The control parameters for this model are set usingthe “Mb” command (Section 3.2.6). Pwr Indicates the mean power within the full window of burst samples.DC offsets in the A/D converter do not affect the computation of thepower, i.e., the value shown truly represents the waveform’s(Signal+Noise) energy. DC Indicates the nominal DC offset of the burst pulse A/D converter. Thisis of interest only as a check on the integrity of the front end analogcomponents. The value should be in the range 2.0%. Trig#1 Indicates the starting time of the first (of six) RVP8 trigger outputs.This number will vary as the “L” and “R” subcommands cause thetriggers to slew left and right. Note that if the radar transmitter isdirectly fired by an external pretrigger, then the pretrigger delay (inthe form “PreDly:6.87”) will be printed instead.BPT This shows the present value of timing slew (measured inmicroseconds) being applied to track the burst. The slew will be zeroinitially when the RVP8 is first powered up, meaning that the normaltrigger start times are all being used.4.3.4 Recommended Adjustment ProceduresThe burst pulse timing must be calibrated separately for each individual pulsewidth. Repeat thefollowing procedure for each pulsewidth that you plan to use. Each iteration is independent.It is first necessary to setup the proper relative timing for all RVP8 triggers that are being used.The six trigger output lines are completely interchangeable, and each may be assigned to anyfunction within the radar system. For example, Trigger #0 might be the transmitter pretrigger,Triggers #2 and #3 might synchronize external displays, and Triggers #1, #4, and #5 might beunused.Choose an initial impulse response length of 1.5 times the transmit pulsewidth. This length willbe refined later when the matched filter is designed (See Section 4.4). Adjust the plot time spanto view a small region around the sample window, and set the initial amplitude zoom to x16.This assures that the plotted waveform will still be noticeable even if the burst signal strength isvery weak.Verify that the transmitter is radiating, and observe the burst pulse samples on the display. Usethe “L” and “R” commands to shift the timing of all six triggers relative to range zero. This hasthe effect of moving the burst sampling window relative to the transmitted pulse. Depending onwhether the triggers are set properly, you may at first see nothing more than a flat line ofmisplaced A/D samples. However, after a few moments of hunting, the burst pulse shouldappear on the display screen. Fine tune the triggers so that the burst envelope is centered in thewindow, and adjust the amplitude zoom for a comfortable size display.The clean center portion of the burst pulse should then be isolated to a narrower subwindow ofthe overall FIR interval. Use the”A” and “S” commands to change the aperture and start of thenarrowed region from which the AFC frequency estimator’s data will be derived.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–9Check that the burst pulse signal strength is reasonably matched to the input span of theRVP8/IFD’s A/D converter. The maximum analog signal level is +4dBm. Exceeding this levelproduces distorted samples that would seriously degrade the algorithms for phase locking andAFC. However, if the signal is too weak, then the upper bits of the A/D converter are wastedand noise is unnecessarily introduced. We recommend a peak signal level between –6dBm and+1dBm, i.e., a signal that might be viewed at x2 or x4 zoom. Take note of the burst energy levelreported on the TTY; it will help decide the minimum energy threshold for a valid burst pulse,which is needed in Section 3.2.6.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–104.4 Ps — Plot Burst Spectra and AFCOnce the transmit burst pulse has been captured the next step is to analyze its frequency contentand to design a bandpass filter that is matched to the pulse. In a traditional analog receiver thematched filters use discrete components that can not be adjusted, and the transmit spectrum cannot be viewed unless a spectrum analyzer is on hand. The RVP8 eliminates all of theserestrictions via its “Ps” command, which plots the burst spectrum, designs the bandpass filter,plots its frequency response, and also helps with alignment of the AFC.4.4.1 Interpreting the Burst Spectra PlotsAn example of a plot from the Ps command is shown in Figure 4–3. The display screen isdivided into two independent areas. The major portion (the lower seven eighths) is devoted topower spectrum plots of the burst pulse and/or the matched filter response. The top portion(single line) serves as a visual indicator of the present AFC level.Figure 4–3: Example of a Filter With Excellent DC RejectionThe horizontal axis of the spectrum plot represents frequency. The overall span from the leftedge to the right edge is 36MHz for the RVP8 CAT-5E IFD, and 18MHz when the legacy RVP7FibreOptic IFD is used. The remainder of this chapter will refer only to the newer RVP8 unit.The exact endpoints of the plot depend on which alias band the radar’s intermediate frequencyfalls in. For example, a 30MHz IF would imply a horizontal axis range of DC to 36MHz,whereas a 60MHz IF would make the range 36MHz to 72MHz. The frequency span is printedon the TTY when the command is first entered. Since the left edge of the spectral plot always
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–11represents an integer multiple of 36MHz, either the left side or the right side will always be amultiple of 36MHz. This is important to remember when designing the matched filter, sincefixed DC offsets in the A/D converters will appear aliased at these 72MHz multiples.The vertical axis of the spectrum plot is logarithmic and is marked with faint horizontal lines in10-dB increments. An overall dynamic range of 70 dB can be viewed at once. The horizontallines also contain major and minor tick marks to help calibrate the frequency axis. Major marksare small downward triangles that represent integer multiples of 5MHz; minor marks are inbetween and represent 1-MHz steps. The power spectrum example in Figure 4–3 is from asystem with an intermediate frequency of 30MHz. Thus, the left edge of the plot begins at DC,and the graph is centered on the sixth major tick, i.e., 30MHz.Two types of spectra can be plotted on the screen: 1) the frequency response of the FIR filter,and 2) the frequency content of the burst pulse itself. The burst spectrum is computed by firstapplying a Hamming window to the raw samples. You may choose to view either plotindividually, or both at the same time.Figure 4–3 is an example of a single filter response plot, whereas Figure 4–4 shows a combineddisplay of both spectra. The combined display makes it easy to compare the filter beingdesigned with the live waveform that it is intended to selectively pass. Note that the filter’sfrequency response is always drawn with its passband peak touching the top of the plot. Thevertical height of the burst spectrum, however, will vary with signal strength but can be adjustedusing the “Z” subcommand.The horizontal line at the top of the plotting area is also marked with an upward pointing majorand minor tick. These indicate the present value of the burst pulse frequency estimator. Themajor tick is a triangle whose position along the horizontal axis corresponds directly to theestimated frequency. It should always be positioned directly over the main lobe of spectralpower. The minor tick gives finer scale resolution by indicating the fractional part of each1-MHz multiple.It is helpful to read the minor tick relative to the ten horizontal division lines that are present onmost scopes. Motion of the minor tick is apparent even with very small changes in burst pulsefrequency; a change of just 5 KHz can easily be seen. This means that you can observe thefrequency drift of the magnetron in great detail, and also watch the AFC’s behavior in real time.The horizontal line at the very top of the display (above the spectra plot) serves to indicate thepresent value of the AFC control voltage. The line contains an upward pointing major andminor tick, similar to the ones used to represent the burst frequency estimate on the line below.However, the horizontal axis now represents voltage rather than frequency, and the overall spanis the complete range of the AFC’s digital-to-analog converter. The major tick will move fromthe left edge to the right edge as the AFC varies from its minimum to maximum value. Theminor tick will traverse the screen at ten times this rate.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–124.4.2 Available Subcommands Within “Ps”The list of subcommands is printed on the TTY:Frequency span of the plot is 36.0 MHz to 72.0 MHz.Available Subcommands within ’Ps’: I/i Impulse response length Up/Dn N/n & W/w Filter bandwidth Narrower/Wider U/u & D/d MFC Up/Down (On/Off ’=’ , Test ’|’) A/a & S/s Aperture & Start of AFC window # Print filter coefficients $ Search for an optimal filter V/v Number of spectra averaged Z/z Amplitude zoom <space> Alternate Plots % Toggle between dual receivers . Single StepThese subcommands change the design of the matched FIR filter, assist with calibration of theAFC loop, and alter the format of the display. I/i The “I” command increments or decrements the length of the matchedfilter’s impulse response. Each keystroke raises or lowers the FIRlength by one tap. Often the matched filter’s characteristics can bevery much improved merely by changing the FIR length by one or twotaps. Be sure to experiment with this as you design your filter.N/n & W/w The “N” and “W” commands change the passband width of thematched filter, making it narrower or wider. The lower casecommands make changes in 1KHz steps, and the upper casecommands use 100KHz steps. The value is reported on the TTY as“BW”. Often a small change in passband width will shift the exactlocations of the filter’s zeros, and possibly improve the DC rejection.U/u & D/d The “U” and “D” commands implement the Manual FrequencyControl (MFC) override, and allow the RVP8/IFD’s AFC outputvoltage to be manually set to any fixed level. The lower casecommands make changes in 0.05 D-Unit steps, and the upper casecommands use 1.0 D-Unit steps. The value is reported on the TTY as“AFC”.=MFC mode is toggled on and off using the “=” key. A warning will beprinted if the Ps command is exited while MFC is enabled, and youwill be given a second chance to reenable AFC.|The AFC test submode is entered by typing the “|” key. The followinglist of keybindings will be shown, and will remain in effect until thetest mode is exited by typing “Q”.AFC Test Mode Subcommands W Use Walking–Ones pattern P Toggle Pin/Bit numbering 0–9,A–O Toggle AFC Bits 0–24 (Pins 1–25)
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–13 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 – – – – – – – – – – – – – – – – – – – – – – – – – O N M L K J I H G F E D C B A 9 8 7 6 5 4 3 2 1 0The Ps command continues to run normally during the AFC testmode. The customary AFC information will be replaced with ahexadecimal readout of the present 25-bit value. Your live displaymay look something like:Navg:3, FIR:1.33 usec (48 Taps), BW:1.000 MHz, DC–Gain:ZEROFreq:26.610 MHz, Pwr:–64.6 dBm, AFC–Test:0000207F (Bits)Initially, a walking-ones bit pattern will be output in lieu of the normalformatted AFC value. This test pattern shifts a single “1” downwardthrough the AFC word, making a transition approximately every 4ms.It is intended to help ring out and test the wiring for digital AFCinstallations. The walking-ones test is handy as an oscilloscopediagnostic, and you may return to it at any time by typing “W”.Typing any of the characters “0” through “9” or “A” through “O” willenter a new mode in which a static 25-bit digital AFC pattern iscontrolled directly. Each key toggles its corresponding bit, assummarized in the keybindings printout. Any 25-bit pattern can bemade by toggling the appropriate bits (initially all zero) to one.Within any particular pattern, it is also easy to toggle a particular bitOn/Off in order to verify its function.The “P” command lets you decide whether the 25-bit word representsa numeric AFC span that is mapped into pins via the pin-map table inthe Mb menu; or whether it represents those pins directly. The printedhex test value will be followed either by “(Bits)” or “(Pins)”accordingly. When in “Pins” mode, the “0” key toggles Pin-1, the “1”key toggles Pin-2, etc. When in “Bits” mode, the “0” key toggleswhatever pin or pins have been designated to be driven from Bit-0 ofthe numeric AFC. The “Pins” mode is useful when you are doing theinitial electrical tests of the wiring of each pin. After the pin wiringhas been verified and the Mb mapping table has been created, then the“Bits” mode allows you to test the complete digital AFC interface.#The “#” command results in a printout of the coefficients of thecurrent FIR filter. The values are scaled by the coefficient with thelargest absolute value, so that they all fall within the –1 to +1 range.This detailed information may be used to model the behavior of thefilter for point targets that fall in between discrete range bins, e.g., aswill happen when performing a radar sphere calibration. See Section5.1.1 for the exact definition of these coefficients.$The “$” command performs an automatic search for optimal (DC gainof zero) filters in the vicinity of the current one. As an example,suppose that we wanted an optimal filter that was approximately 2.2msec long and 650 KHz wide. We would first use the “I/i” and
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–14“W/wN/n” subcommands to manually move to that starting point.Typing “$” would then print a dialog line in which the search spanlength and width are chosen. You may keep the indicated values ortype in new ones, just as for all RVP8 setup questions. The searchbegins when the spans are accepted.The search procedure may require a few seconds to a few minutes,depending on the length and width spans that are being scanned.During this time, a progress message is printed showing the length andwidth currently under examination. You may type “Q” to abort thesearch and retain the original filter settings. When the searchcompletes normally, it will print “Done” and replace the old filtersettings with the best ones that could be found.In dual-receiver mode, the ”$” command will search for a filter thatminimizes the maximum width and DC offset at both receiver’sintermediate frequencies. The final filter will be the one that has thebest simultaneous performance at both IFs. V/v The “V” command increments or decrements the number of burstpulse spectra that are averaged together to create the plot. The countranges from one (no averaging) to 25, and is reported on the TTY as“Navg”. Z/z The “Z” command zooms (i.e. shifts on a logarithmic scale) in 1.0-dBsteps the amplitude of the burst pulse spectra. This is useful when theoverall 70dB plot span is not sufficient to hold the full range. Zoomcan also be used to line up the burst spectrum with the filter responseso that the two can be compared. The zoom level is not printed on theTTY because there is nothing useful that could be done with it. <space> The space bar alternates among three choices for the type of spectrathat are plotted: 1) FIR frequency response, 2) Burst pulse spectrum,and 3) Both.%In dual-receiver mode, the “%” command toggles between eachreceiver. The printed status line is prefixed with ”Rx:Pri” or ”Rx:Sec”according to which receiver is selected. Specifically, typing “%” willtoggle the plot of the FIR filter’s frequency response, and the printoutof its DC–Gain. However, the plotted spectrum and printed powerlevels are always based on the sum of all input signals, and thus do notchange with “%“.4.4.3 TTY Information Lines Within “Ps”The TTY information lines will resemble:Navg:3, FIR:1.33usec (48 Taps), BW:1.000, MHz, DC–Gain:ZEROFreq:30.027 MHz, Pwr:–64.2 dBm, Loss:1.2dB, AFC:23.05% (Manual) Navg Indicates the number of burst spectra that are averaged together priorto plotting. Larger amounts of averaging increase the ability to seesubtle spectral components, but the display will update more slowly.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–15 FIR Indicates the length of the impulse response of the matched FIR filter.See description on Page 4–7. BW Indicates the actual 3dB bandwidth of the matched filter. This is thecomplete width of the passband from the lower frequency edge toupper frequency edge. Note that the filter’s center frequency is fixedat the radar’s intermediate frequency, as chosen in the “Mb” setupcommand.DC-Gain Indicates the filter’s response to DC (zero frequency) input. The valueis a negative number in decibels, or the word “ZERO” if the filter hasa true zero at DC. The filter’s DC gain should be kept at a minimumso that fixed offsets in the A/D converters will not propagate into thesynthesized “I” and “Q” values. Freq Indicates the mean frequency of the burst. See description on Page4–8. Pwr Indicates the average power in the full burst sample window. Seedescription on Page 4–8.Loss The filter loss is a positive number in deciBels, and is only displayedif the overall burst power exceeds the minimum valid burst thresholdset in the Mb command (clearly, it would not be possible to computethe filter loss when the burst waveform is missing). The filter loss isdiscussed further in Section 4.4.4. AFC Indicates the level and status of the AFC voltage at the RVP8/IFDmodule. The number is the present output level in D-Units rangingfrom –100 to +100. The shorter “%” symbol is used since percentageunits correspond in a natural way to the D-Units.An additional number in square brackets will be printed to the right ofthe AFC level to show the encoded bit pattern which corresponds tothat level. This will only appear when the RVP8 deduces that a specialdigital format is being used, i.e., when the backpanel phase-out lineshave been configured for AFC, or when any of the following are nottrue: a) the low and high numeric AFC span is –32768 to +32767, b)the uplink is enabled, c) the uplink format is binary, and d) pinmapprotocol is OFF. Binary format is printed in base-10, BCD format isprinted in Hex, and 8B4D format is printed with the low 16-bits (fourBCD digits) in Hex and the upper bits in base-10.The AFC mode is shown to the right of the numerical value(s), andcan take on the following states.(Disabled) Indicates that neither AFC nor MFC are enabled. Theoutput voltage remains fixed at 0% (center of its range).(Manual) Manual Frequency Control (MFC) is overriding AFC.The “U” and “D” commands can be used to slew thevoltage up and down.Whenever any of the following four states appears, it implies that AFCis enabled and that MFC is disabled.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–16(NoBurst) The energy in the burst is below the minimum energythreshold for a valid pulse (See Page 3–25). The AFCloop remains idle.(Wait) The burst pulse has become valid just recently, but theAFC loop is idle until the transmitter stabilizes (SeePage 3–26)(Track) The burst pulse is valid, and the AFC loop is trackingin order to bring the burst frequency within the innerhysteresis limits.(Locked) The burst pulse is valid and the AFC loop is locked.The burst frequency is now within the outer hysteresislimits and has previously been within the inner limitswhile tracking. This is the stable operational mode inwhich data acquisition should take place.4.4.4 Computation of Filter LossThe Ps printout displays the power loss (calibration error) that results when the given filter isapplied to the given transmit burst waveform. This allows you to correct for the differencebetween what a broad-band power meter measures as the overall transmit power, and what theRVP8 narrow-band receiver will detect within its passband. The filter loss is a subtle quantitythat depends on the combined characteristics of both the transmit waveform and the receivermatched filter.The filter loss is zero if the burst waveform consists of a pure sinusoid at the designatedintermediate frequency. It is also very near zero as long as most of the burst energy is confinedwithin the passband of the RVP8’s filter. The filter loss will increase as the bandwidth of theburst waveform increases and begins to spill out of that passband. Typical losses for awell-matched filter are in the 0.5–1.8dB range, depending on the FIR length and other designcriteria.As an example, consider how the RVP8 filters would respond to a simple rectangular pulse ofenergy lasting To seconds. For this discussion we can ignore the sinusoidal IF carrier that mustalso be present within the pulse, and just focus on the rectangular envelope. This is validbecause the signal bandwidth, and hence the filter loss, is determined entirely by the shape of themodulation envelope. For a pulse of length To to have unit-energy it must have an amplitude of1ńToǸ . By centering this pulse at time zero the power spectrum is easily computed using areal-valued integral:S(f)+ȧȧȡȢŕToń2–Toń21ToǸcos( 2pft)dt ȧȧȣȤ2+sin2(pfTo)p2f2Towhere f is the frequency in Hertz. This is the familiar “synch” function, whose main frequencylobe extends from –1ńTo to 1ńTo Hertz, and whose total power integrated over all frequenciesis 1.0.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–17We can now examine what the filter loss (dBloss ) would be if this pulse were applied to abandpass filter. The filter loss is simply the ratio of the power that is passed by the filter, dividedby the total input power (1.0 in this case). Assume for the moment that the filter is an idealbandpass filter centered at zero Hertz (corresponding to how S(f) was defined) and having abandwidth Bw, then:dBloss +–10 log10ȧȧȡȢŕBwń2–Bwń2S(f)df ȧȧȣȤThis integral can be computed for a few “interesting” filter bandwidths, yielding filter losses of0.44dB, 1.11dB, and 3.31dB when Bw is 2ńTo, 1ńTo, and 1ń2To respectively. These threeexample bandwidths correspond to filters that pass the entire main frequency lobe, half of thatlobe, and one quarter of it.You can experimentally verify these results using the RVP8 as follows:SUsing the Mt0 command, setup a To+0.5 msec trigger pulse from the RVP8 in thevicinity of range zero, and use that trigger to gate a signal generator whose output isapplied to the RVP8/IFD Burst Input. Also setup 125-meter range resolution, and arather long 6.0 msec impulse response length. The long length will make the transitionedges of the matched filter as steep as possible, so that it becomes a reasonably goodapproximation to the ideal bandpass filter used in the above analysis.SUse the Pb command to verify that the burst pulse is present, and position the triggersleft and right until the pulse is centered exactly at zero.SUse the Ps command to examine the frequency spectrum of the pulse. You should see amain energy lobe that is 4MHz wide and centered at the radar’s IF. There should also beweaker lobes spaced 2MHz apart on both sides of the main lobe. If the lobe spacing doesnot look quite right, it may be because the signal generator has slightly shortened orlengthened the trigger gate.SContinue using Ps to examine filters that are 4MHz, 2MHz, and 1MHz wide at their 3dBpoints. You should see filter losses reported that are very close to the theoretical valuesfor the ideal bandpass filter.In the above analysis we have assumed that S(f) is the idealized power spectrum of a continuoustime signal. Of course, the RVP8 filter loss algorithm can only work from an estimate of S(f)that is obtained from a finite number of samples. The filter loss calculation thus becomes morecomplicated than the above example in which we integrated an idealized filter response over anidealized power spectrum.Let B^(f) denote the estimated power spectrum of the continuous-time Tx burst waveform, forwhich we have only a finite number of discrete samples {bn}. For purposes of this discussionwe can assume that the frequency variable f is continuous. Furthermore, let C^(f) denote apower spectrum estimate that is derived in an identical manner using the same number of
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–18samples, but of a pure sine wave at the radar’s IF. The RVP8 determines B^(f) according to itssampled measurement of the transmitted waveform; however it can calculate C^(f) internallybased on an idealized sinusoid. The reported filter loss is then:dBloss +–10 log10ȧȧȧȧȡȢŕ|H(f)|2B^(f)dfŕB^(f)dfBŕ|H(f)|2C^(f)dfŕC^(f)dfȧȧȧȧȣȤWhere |H(f)|2 is the spectral response of the RVP8 IF filter, and the integrals are performed overthe Nyquist frequency band that is implied by the RVP8/IFD sampling rate. Note that the twointegrals involving C^(f) will have constant value and need only be computed once. They serveto normalize the B^(f) integrals in such a way that the filter loss evaluates to 0dB whenever thetransmit burst is a pure tone at IF.This normalization is necessary for the filter loss values to be meaningful. Regardless of thebandwidth and center frequency of H(f), the filter loss should be reported as 0dB whenever theTx waveform appears to have zero spectral width, i.e., is indistinguishable from a pure IFsinusoid. Of course, the real Tx waveform has only finite duration, and thus should never looklike a pure tone as long as the RVP8 is able to “see” the entire Tx envelope. For this reason, it isimportant that the filter’s impulse response length be set long enough (using the Pb plot) toinsure that all of the details of the Tx waveform are being captured. If the entire Tx envelopedoes not fit within the FIR filter, then the filter loss will be underestimated because the Txspectrum will appear to be narrower than it really is.The RVP8’s calculation of digital filter loss is very similar to how the loss of an analog filterwould be measured on a test bench. Suppose we are given an analog bandpass filter and areasked to determine its spectral loss when a given waveform is presented. We could use a powermeter to measure the waveform power before and after the filter is inserted, and compute theratio of these two numbers. This corresponds to the first integral ratio in the above equation.However, this is not by itself an accurate measure of filter loss because it does not take intoaccount the bandwidth-independent insertion loss. Put another way, a flat 3dB pad would seemto produce a 3dB filter loss in the above measurement, but that is certainly not the result that wedesire. The remedy is to make a second pair of power measurements of the filter’s response to aCW tone at the passband center. This serves to calibrate the gain of the filter, and allows us tocompute a filter loss that captures the effects of spectral shape independent of overall gain. Thisnormalization step corresponds to the second integral ratio in the above equation.If your radar calibration was performed using CW waveforms, then the reported filter lossshould either be added to the receiver calibration losses, or subtracted from the effective transmitpower; the net result being that dBZ0 will increase slightly.In dual-receiver systems the filter loss is computed for the primary and secondary channels usingonly the portion of bandwidth that is allocated to that channel. For example, if the two IFs are24MHz and 30MHz, then the filter losses for each channel would use the frequency intervals21–27MHz and 27–33MHz respectively. This is necessary to avoid picking up energy from the
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–19other receiver and interpreting it as out-of-band input power. A consequence, however, is thatthe real out-of-band power is underestimated, i.e., the filter loss itself is underestimated. Werecommend temporarily switching dual-receiver systems back to single-receiver mode when thefilter loss is being measured. This is easily done by changing the Mc setup question back to“single”, and disconnecting the secondary burst input to the RVP8/IFD.4.4.5 Recommended Adjustment ProceduresThe Ps command should be used only after the burst pulse has been successfully captured byway of the Pb command. Use the <space> key to display the burst spectrum plot by itself, anduse the “Z” key to shift the entire graph into view. You are now looking at the actual frequencycontent of the transmitted pulse. The plot should show a clean main power lobe centered at thereceiver’s intermediate frequency. Check the spectrum for spurious harmonics, excessive width,and other out-of-band noise. Make any adjustments in the transmitter that might give a sharpermain lobe or reduced spurious noise.Once we know the power spectrum of the transmitted pulse we can begin designing the matchedFIR filter. Use the <space> key to display both the filter response and the burst spectrum on thesame plot. Use the “Z” key to shift the burst’s main lobe up to the top horizontal line of thegraph. This makes it level with the filter’s peak lobe, which is always drawn tangent to the sametop line.Figure 4–4: Example of a Poorly Matched Filter
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–20Begin with the FIR length that was chosen previously in the Pb command, and use the “N” and“W” keys to set an initial bandwidth equal to the reciprocal of the pulsewidth. The main lobesof the two plots should more-or-less overlap. Experiment with changing the FIR length andbandwidth to achieve a filter with the following properties.SThe filter width should be no greater than the burst spectral width. A wider passbandwill reduce the SNR of the received signal because out-of-band noise would be allowedto pass.SThe DC gain should be as small as possible, preferably less than –64dB (See discussionbelow).SIf there are conspicuous interference spikes at particular frequencies, try to adjust thelocation of the filter’s zeros so that the interference is maximally attenuated.The filter should not pass any frequencies that do not actually contain useful information fromthe original transmitted pulse. If anything, choose a filter whose width is slightly narrower thanthe burst’s spectral width. Figure 4–4 shows an example of a filter that is poorly matched to thepulse. Although the filter has fairly good DC rejection, it passes frequencies that are outside ofthe transmitter’s broadcast range. These frequencies contribute nothing but noise to thesynthesized “I” and “Q” data stream.There are two procedures for optimizing the performance of the FIR filter:SManual Method –– The process of arriving at a nearly optimal filter will require a fewminutes of hunting with the “I”, “W”, and “N” keys. Every time you press any of thesekeys the RVP8 designs a new FIR filter from scratch, and displays the results.Fortunately, the DSP chips are fast enough that this can be done quickly andinteractively. Even though the user must still control two degrees of freedom (length andbandwidth), the RVP8’s internal design calculations are actually performing severalhundred iterative steps each time, which preferentially select for the best filter. Becausethe FIR coefficients are quantized in the filter chips themselves, the process of finding anoptimal filter becomes quite nonlinear.SAutomatic Method –– Simply type the “$” command and let the RVP8 do all of thework (See description in Section 4.4.2).The offset error of the RVP8/IFD’s A/D converter is at most 10mV, i.e., –27dBm into its 50Winput. If we wish to achieve 85-dB of dynamic range below the converter’s +4dBm saturationlevel, then we expect usable “I” and “Q” values to be obtainable from a (sub-LSB) input signalat –81dBm. This is 54dB below the interference that would result from the worst-case A/Doffset. But a weak input signal at –81dBm would still be damaged by even an equal level of DCinterference. Therefore, adding another 10dB safety margin, we get –64dB as the recommendedmaximum DC gain of the matched filter. This DC gain should be reduced even further if it isknown that coherent leakage is present in the receive signal at a level greater than the –27dBmworse-case A/D offset.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–21Figure 4–5 shows a 60MHz filter with particularly poor (–42dB) DC rejection. The frequencyrange of the plot is 36–72MHz; hence, DC appears aliased at the right edge and we can see apeak in the filter’s stopband at DC. Contrast this with the filter shown in Figure 4–3 that has atrue zero at DC. In general, a poor filter can be converted into a “nearby” good filter by makingonly incremental changes to the impulse response length and/or desired bandwidth.Figure 4–5: Example of a Filter With Poor DC Rejection
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–224.5 Pr — Plot Receiver WaveformsThe “Pb” and “Ps” commands described in the previous sections have been used to analyze thesignal that is applied to the “Burst-In” connector of the RVP8/IFD receiver module. The taskthat remains is to checkout the actual received signal that is connected to “IF-In”. The goal is toverify that the received signal is clean and appropriately scaled, and that nearby targets can beseen clearly. The “Pr” command is used to make these measurements.4.5.1 Interpreting the Receiver Waveform PlotsAn example of a plot from the Pr command is shown in Figure 4–6. The horizontal axisrepresents time (range) starting from a selectable offset and spanning a selectable interval. Thedata are acquired from a single transmitted pulse, are are plotted both as raw IF samples and asthe LOG of the detected power using the FIR filter for the current pulsewidth.Figure 4–6: Example of Combined IF Sample and LOG PlotThe IF samples are plotted on a linear scale as signed quantities, with zero appearing at thecenter line of the scope. Any DC offset that may be present in the A/D converter is notremoved, and will be seen as a shift in the baseline at higher zoom levels. For example, theconverter’s worst case DC offset of 10mv would appear as a 91-count offset in the 12-bit rangespanning –2048 to +2047. At the x32 or higher zoom scales, this offset would peg the sampleplot off scale. Typically the DC offset will be much less than this worst case value; but theRVP8 preserves the DC term in the Pr sample plot so that its presence is not forgotten.The “AC” amplitude of the IF samples will increase wherever targets are present. On top ofthese samples is drawn the detected power on a logarithmic scale. Each horizontal linerepresents a 10dB change in power. The graph is scaled so that the LOG power reaches the topdisplay line when the samples occupy the full amplitude span. Using Figure 4–6 as an example,
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–23the two equal-power targets just to the left of center are approximately 18dB down from the top.The amplitude of the samples is thus 10(*18ń20) +0.13, i.e., 13% of full scale. Thiscorrespondence between the LOG scale and the amplitude scale applies regardless of the plot’szoom level. As the IF samples are zoomed up and down by factors of two, the LOG plot willshift up and down in 6dB steps.The LOG plot is obtained by convoluting the FIR filter coefficients with the raw IF datasamples, and then plotting log(I2)Q2) at each possible offset along the sampling interval.This convolution produces only (1 + N – I) output points, where N is the number of samplepoints and I is the length of the FIR filter. For this reason the LOG plot begins approximatelyI/2 samples from left side and ends approximately I/2 samples from the right.The LOG points are computed at each possible offset within the raw IF samples. At the nominal72MHz sampling rate the spacing between LOG samples will be a mere 4.17 meters. Thus, theLOG plot gives a very detailed view of received power versus range. Of course, successiveLOG points will be highly correlated because successive input data intervals differ by only onesample point. This is why the LOG plots appear smooth compared to the instantaneous variationof the raw IF samples.As the starting offset of the Pr plot is decreased to range zero you will begin to see part of theburst pulse (the second half of it) appear at the left edge of the plot. This is because the burstdata samples are multiplexed onto the same fiber cable that carries the IF data samples. Zerorange is defined to occur at the center of the burst window; hence, the later half of the burstpulse will be visible when the plot begins at range zero.A second type of Pr display is shown in Figure 4–7. This plot shows a frequency spectrum ofthe received data samples in a format that is nearly identical to the Ps display. The horizontalaxis represents the same band of frequencies (half the sampling rate), and the vertical axisrepresents power in 10dB steps. The entire vertical axis is used so that an overall span of 80dBis visible. This particular plot was made with the time span set to 50 msec, and with a 1-meterantenna attached to the IF input so that a broad range of signals (radio stations, electrical noise,etc.) would be detected.The purpose of the Pr power spectrum is to check for spurious interference in the IF signal fromthe radar receiver. The spectrum should be viewed with the transmitter turned off, and with thestarting range moved out so that the burst samples are not mixed in with the receiver data. Thepower spectrum is computed using the complete interval of raw IF samples which, depending onthe chosen time span, may contain many hundreds of points. The frequency resolution of the Prspectrum can therefore be quite fine; making it possible to discern any interfering frequencieswith some detail.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–24Figure 4–7: Example of a Noisy High Resolution “Pr” SpectrumThe Pr spectrum plot will properly show a 0-Hz peak from any DC offset in the A/D converter,and is thus consistent with how the DC offset is presented in the Pr sample plot. Both of theseplots preserve the DC component of the IF samples so that it can be monitored as part of theroutine maintenance of the receiver system. This is one of the few places in the RVP8 menusand processing algorithms where the DC term deliberately remains intact.4.5.2 Available Subcommands Within “Pr”The list of subcommands is printed on the TTY:Available Subcommands within ’Pr’: L/l & R/r Start range Left/Right T/t Plot time span Up/Dn V/v Number of spectra averaged Z/z Amplitude zoom <space> Alternate Plots % Toggle between dual receivers . Single StepThese subcommands change the start time and span of the IF sampling window, and alter theformat of the display.L/l & R/r The “L” and “R” commands shift left and right the starting point ofthe window of IF samples. The lower case commands shift in 0.25msec steps, and the upper case commands use 10 msec steps. The
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–25starting point is displayed both in microseconds and kilometers on theTTY, and is not allowed to be set earlier than range zero. T/t The “T” command increments or decrements the time duration of thewindow of IF samples. The window is not allowed to become shorterthan the impulse response length of the FIR filter, since that wouldpreclude calculating even a single LOG power point. The value isreported in microseconds on the TTY, and the largest permitted span is50 msec. V/v The “V” command increments or decrements the number of powerspectra that are averaged together to create the plot. The count rangesfrom one (no averaging) to 25, and is reported on the TTY as “Navg”. Z/z The “Z” command zooms the amplitude of the IF samples by factorsof two from one to 128. The LOG plots are shifted in corresponding6dB increments as the amplitude is zoomed up and down. The zoomlevel is reported on the TTY so that absolute power levels and A/Dusage can be assessed. <space> The space bar alternates among three choices for the type of data thatare plotted: 1) Received Samples, 2) Received Samples and LOGPower, and 3) Received Power Spectrum.%In dual-receiver mode, the “%” command toggles between eachreceiver. The printed status line is prefixed with ”Rx:Pri” or ”Rx:Sec”according to which receiver is selected. Specifically, typing “%” willtoggle the LOG plot of the received power, and the printout of the“Total”, “Filtered”, and “Midpoint” powers. However, the plots ofpower spectra and raw IF data samples are always based on the sum ofall input signals, and thus do not change with “%”.4.5.3 TTY Information Lines Within “Pr”The TTY information lines will resemble:Zoom:x1, Navg:4, Start:0.00 usec (0.00 km), Span:5 usecTotal:–63.3 dBm, Filtered:–77.6 dBm, MidSamp:–77.4 dBm Zoom Indicates the magnification (in amplitude) of the plotted samples. Azoom level of “x1” means that a full scale A/D waveform exactly fillsthe vertical height of the plot. Generally, the IF signal strength willnot be quite this high. Thus, use larger zoom levels to see the weakersamples more clearly. You may zoom in powers of two up to x128. Navg Indicates the number of spectra and/or LOG powers that are averagedtogether prior to plotting. Larger amounts of averaging increase theability to observe subtleties of the signals, but the display will updatemore slowly.Start Indicates the starting time of the IF sample window relative to rangezero. The time is shown both in microseconds and in kilometers.Span Indicates the time span of the IF sample window in microseconds.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–26Total Indicates the total RMS power that is being detected by the IF-InputA/D converters. This total is computed using all of the raw IF samplesin the chosen interval, and is the sum of power at all frequencies otherthan 0 Hz (and its aliases).Filtered Indicates the RMS power that falls only within the passband of theFIR filter for the current pulsewidth. This is computed using all of theraw IF samples in the chosen interval.MidSamp Also indicates the RMS power within the passband of the FIR filter,but using only the raw IF samples in the exact center of the choseninterval.The computation of “Total Power” is performed using the same subset of central IF samples thatare used to compute “Filtered Power”. This smaller subset of IF samples comes about becausefiltering the data requires a convolution with the current FIR filter, and this computation doesnot produce results all the way to the edges of the input data. This is the same reason that theLOG plots do not extend across the full screen.Because of this definition, it is valid to intercompare the “Total Power” and “Filtered Power”.The two numbers will match exactly as long as all of the incoming power falls within thepassband of the FIR filter. The difference between the two powers can be used as a measure ofthe “filter loss” for a given pulse shape, i.e., the portion of signal that is lost outside of thefilter’s passband.Note: The “Total”, “Filtered”, and “MidSamp” values represent true RMSpower (i.e., variance), and not merely a sum-of-squares. Thus, any DC offsetpresent in the A/D converter will not affect these power levels.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–274.6 Pa — Plot Tx Waveform AmbiguityWith the introduction of the RVP8/Tx Digital Transmitter PCI Card it is now possible for theRVP8 to make radar observations using compressed pulse waveforms. This opens up many newopportunities within the weather radar community for using low-power solid-state transmittersthat employ very long pulse lengths (20–80μsec) . Transmitters of this kind are less expensive,both to build and to maintain, compared with traditional Magnetron or Klystron systems.However, the signal processing and waveform design that are required to make good use of theselong transmit pulses is also much more complex. To help with this, the RVP8 provides the Pa(Plot Ambiguity) command in which compressed transmit waveforms can be designed, studied,and optimized. Within the Pa plots you can experiment with different waveform designs, try outvarious bandwidth and pulsewidth options, and examine and optimize the range/time sidelobesof your waveform.4.6.1 Interpreting the Ambiguity PlotsFigure 4–8 shows one form of Pa plot in which the magnitude of the Tx/Rx range sidelobes aredrawn on a log scale having 10dB vertical ticks. The horizontal span of the plot is equal to thelength of the pulse, and consequently, only half of the complete ambiguity diagram is shown.This was done to make the plots more viewable; and no information lost since the zero-Dopplerresponse (white plot) can safely be assumed to be symmetric. In this example the pulsewidth is30μsec, bandwidth is 3MHz, PSL is –61.2dB and ISL is –50.8dB, Doppler shift +/– 50 KHz.Figure 4–8: Ambiguity Diagram of a Compressed Tx Pulse
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–28Also shown in yellow and green are the Tx/Rx responses when the overall waveform is modifiedby a 50KHz target Doppler shift. Real weather targets would never have such a large Dopplercomponent, but the Pa menu allows you to study its effect anyway.An alternate form of Pa plot of the same Tx waveform is shown in Figure 4–9. The horizontalaxis again represents time, but now spans the entire duration of the pulse. Three different plotsare drawn, hence the vertical axis is interpreted differently in each case:SThe instantaneous frequency across the full length of the pulse is shown in white. Thevertical scale is normalized to hold the overall frequency span, which is also shownnumerically in the Pa TTY output.SThe waveform baseband phase is shown in green, and is normalized so that the verticalaxis holds the full span of values. Note that the phase, which generally spans a fewthousand degrees, is “unwound” in this plot so that you can see its behavior nicely.SThe amplitude of the Tx waveform envelope is shown in yellow. It is drawn using alinear vertical scale which occupies only the middle half of the plot. This is to avoidcreating too much “plotting clutter” in the corners.Figure 4–9: Frequency, Phase and Amplitude of a Compressed Tx PulseFrom Figure 4–9 we can see that the waveform consists of a linear FM chirp that occupies about87% of the central pulse duration. The frequency remains nearly constant in the leading andtrailing edges, hence the overall label “Non-Linear FM”. The central chirp is contained within asomewhat larger amplitude modulation envelope that applies full scale power within the middle82% of the pulse, and also provides bandlimited shaping of the leading and trailing edges.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–29This waveform was designed using the “$” automatic search-and-optimize command in the Pamenu. For a given pulse length and bandwidth of the Tx waveform, this command allows you totry thousands of combinations of FM shape and amplitude shape, searching for the combinationthat minimizes the sum of PSL and ISL (in dB). This gives the best overall waveform forweather radar observations in which both the PSL and ISL are important.4.6.2 Available Subcommands Within “Pa”The list of subcommands is printed on the TTY:Available Subcommands within ’Pa’: S/s & L/l Pulse Length Shorter/Longer N/n & W/w Bandwidth Narrower/Wider 1/2/3 Select Tuning Parameter to Change D/d & U/u Selected Tuning parameter Down/Up V/v Doppler Frequency Shift Up/Down Z/z Amplitude Zoom $ Search for Optimal WaveformThese subcommands change the bandwidth, pulsewidth, and shaping parameters of the transmitwaveform, and alter the format of the display.S/s & L/l The “Shorter” and “Longer” commands decrease or increase the timeduration (i.e., pulsewidth) of the transmitted pulse. The lower casecommands shift in 0.05 msec steps, while the upper case commandsuse 1 msec steps.N/n & W/w The “Narrower” and “Wider” commands decrease or increase thebandwidth of the transmitted pulse. The lower case commands shift in10KHz steps, while the upper case commands use 200KHz steps.1 & 2 & 3 Typing one of these numbers chooses which of the three waveformtuning parameters will be altered by the “D” and “U” keys.D/d & U/u The “Down” and “Up” commands decrease or increase the waveformtuning parameter that has been chosen by the most recent “1”, “2” or“3” key. The lower case commands shift in 0.001 (dimensionless)steps, while the upper case commands use 0.05 steps. Present valuesof all three parameters will be printed each time a Down/Up key ispressed, e.g.,: Tuning parameters: 0.9000 1.0000 0.1770Please see Section 3.2.5.1 for a description of how each of the tuningparameters are used.V/v The “Velocity” command allows you to investigate how the overallcompressed pulse Tx/Rx system will respond to the the effects oftarget Doppler shift. Frequency shifts as large as 100KHz can beintroduced in order to see their effect on the Range/Time sidelobes.The Pa plot will show the effects of +V and –V shifts as green andyellow plots in addition to the standard white (zero Doppler) plot.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–30Z/z The dynamic range of the Pa sidelobe plot is 80dB. Usually this willgive plenty of room to examine the properties of the waveform. Butfor very wide dynamic range pulses, you can shift the plot up/down in10dB steps using these “Zoom” keys.$Designs an optimal compressed waveform. For a given pulsewidthand bandwidth of the Tx waveform, this command allows you to trymany thousands of combinations of FM shape and amplitude shape,searching for the one that minimizes the sum of PSL and ISL (in dB).This gives the best overall waveform for weather radar observations inwhich both the PSL and ISL are important.The following dialog appears in response to the “$” command: TuningParam #1 (0.9000) –– 1 Grid Point from 0.9000 to 0.9000 TuningParam #2 (1.0000) –– 1 Grid Point from 1.0000 to 1.0000 TuningParam #3 (0.1774) –– 1 Grid Point from 0.1774 to 0.1774As the line is printed for each parameter, you may either accept thedefault single grid point (no search), or enter a desired number of gridsearch points followed by a span of values within which to search.For example, typing:200 .9 .95will request that the parameter be searched using 200 evenly spacedgrid points lying between 0.9000 and 0.9500 inclusive. After all threeparameter spans have been entered, the RVP8 will begin searching forthe optimum waveform. Progress messages are printed on the TTY,and the plot will update every time a better waveform is discovered.In this way it is easy to tell whether the search is converging.The process normally runs to completion on its own; but if the searchis taking too long, or you’ve changed your mind about which intervalsto search, typing “Q” will exit right away. In either case, the prompt:Keep this waveform ? [Y]will appear. Typing “Y” (or Enter) will keep the optimized tuningparameters that were just discovered, overwriting whatever startingvalues were originally there. Typing “N” will discard the searchresults and return to the original settings, as if “$” had never beentyped.4.6.3 TTY Information Lines Within “Pa”The TTY information lines will resemble:BW:3.40MHz PW:29.99usec PSL:–61.2dB ISL:–51.3dB TxLoss:0.5dB RxLoss:2.4dBBW Bandwidth of the Tx waveform in MegaHertz.PW Pulsewidth (pulse length) of the Tx waveform in microseconds.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–31PSL Peak Sidelobe Level of the ambiguity diagram, expressed in deciBelsrelative to the main lobe level. This is the peak height of the strongestrange/time sidelobe, and measures the ability of the compressed pulseto distinguish a given target from a small number of individual pointtargets that also lie within the pulse volume. The waveform’s abilityto “see” between clutter targets is largely determined by the PSL level.ISL Integrated Sidelobe Level of the ambiguity diagram, expressed in dBrelative to the main lobe. This is the total power in all range/timesidelobes divided by the total power in the main lobe. ISL measuresthe ability of the compressed pulse to distinguish a given target fromother distributed targets (such as rain) that also lie within the pulsevolume.TxLoss The TxLoss is calculated as the total power in the transmit waveformdivided by the power that would be contained in an equal length idealrectangular pulse. It is a measure of how much power does not gettransmitted due to the amplitude shaping of the synthesized waveform.TxLoss should be included in the computation of the radar constant,since the latter is based on a nominal pulsewidth equal to the overalllength of the entire Tx waveform (including the amplitude tapering).RxLoss The RxLoss is a measure of how much information is “thrown away”by the receiving filter in order to achieve the desired level of sidelobesuppression. These two quantities often trade off against each other inreceiver systems, so that optimum range/time sidelobes can only beachieved at the expense of a few deciBels of loss of sensitivity. Thereceiver filter loss is calculated as:dBloss +–10 log10ȧȧȧȧȡȢŤŕT(t)R(t)dt Ť2ŕ|T(t)|2dt ŕ|R(t)|2dtȧȧȧȧȣȤWhere T(t) is the complex-valued transmit waveform, and R(t) is thecomplex-valued filter being used to receive it. When R(t) is designedto be the complex conjugate of T(t), we have the ideal matched filtercase whose receive loss is 0dB. However, this matched filter hasrather poor sidelobe behavior that makes it unsuitable for use directlyin the receiver. Instead, a windowed version (Hamming, Blackman,etc.) of the ideal matched filter is used to achieve the desired sidelobelevels. Of course, that windowing operation also has the effect ofdiscarding some valid information in the leading and trailing portionsof the pulse. Hence, there is a loss in receive sensitivity whenever awindow is applied.Note: Given a compressed transmit waveform, the RVP8 designs theappropriate “mismatch” Rx filter automatically, using an optimizedBlackman window in all cases. Code developers can also access the
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–32internal APIs directly to design any desired transmit waveform alongwith the associated FIR filter to receive it.4.6.4 Bench Testing of Compressed WaveformsWorking with compressed pulse waveforms can be tricky, so it is reassuring to run some simplebench tests to verify that things are working properly. Once the Tx waveform has been designedit can be injected into the IFD for testing with the Pr command. This not only verifies that theanalog waveform is generated properly, but also that the matched filtering on the RVP8/Rx cardis able to deconvolve the compressed information.To setup the test, simply connect the Channel #1 or Channel #2 output of the RVP8/Tx card tothe IF-Input of the IFD. Use whichever RVP8/Tx channel has been configured for waveformsynthesis in the Mz menu, and set the Zero Offset of the Transmitter Pulse in the Mt<n> menuto, perhaps, 50μsec. The latter step will shift the waveform out in range so that the Pr plot isable to see it.Figure 4–10: IFD Sampling of Optimized Compressed Tx WaveformFigure 4–10 shows an actual Pr plot of a 40μsec, 5MHz optimized waveform generated by theRVP8/Tx card and fed into the IFD. In this example, the ideal Tx waveform has a Peak SidelobeLevel (PSL) of –76.7dB and an Integrated Sidelobe Level (ISL) of –62.3dB. The measuredtestbench performance is several dB short of this, probably because of the uncompensated analogbandpass filters on the RVP8/Tx and IFD. These filters have several tenths of a dB of amplituderipple as well as minor deviations from linear phase within the 5MHz signal bandwidth. Theeffect is that the sampled analog waveform is not quite identical to the ideal waveform.
Plot–Assisted SetupsRVP8 User’s ManualOctober 20044–33The Ps plotting command can also be used to examine the ideal transmit spectrum and actualreceived spectrum of compressed pulses. An example is shown in Figure 4–11 below for a60MHz, 40μsec linear FM pulse having a bandwidth of 2MHz. The energy in the pulse is bothsharply contained within and uniformly distributed over the 2MHz frequency interval centeredon the IF carrier. This demonstrates the ability of synthesized transmit waveforms to remaincleanly within their allocated bounds.Figure 4–11: Ideal and Actual Linear-FM Spectrum Displayed in Ps Plot

Navigation menu