Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63 Wolf Research Whirlwind I Checkout Progress Report Jul63
Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63 Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63
Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63 Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63
User Manual: Wolf_Research_Whirlwind_I_Checkout_Progress_Report_Jul63
Open the PDF directly: View PDF 
.
Page Count: 45
| Download | |
| Open PDF In Browser | View PDF |
WHIRLWIND I MOVING, REASSEMBLY
ANI;? CHECKOUT
PROGRESS REPORT
JULY 1963
by
Albert V. Shortell, Jr.
WOLF RESEARCH AND DEVELOPMENT CORPORATIO\N,
Baker Avenue and Route 2
West Concord, Massachusetts
'w
0,1'
Research and
Development Corporation
P.O. Box 136; Baker Avenue, West Concord, Massachusetts
EMerson 9-211 1
August 16, 1963
Mr. Alfred J. Wis e
Special Representative
Office of Naval Research
Massachusetts Institute of Technology
Cambridge 39, Mass achusetts
Dear Mr.
Wi~e:
We are enclosing a report on the current st~tus' of the moving,
reassembly and check-out of the Whirlwind I Computer. This report
summarizes the activities and events compris'ing ·the moving and 1"eassemb,ly phase and the check-out phase. The activities and events of
these two phases are also portrayed by means of two PERT networks.
The PER T network descr-ibing the cheek-out phase presents' scheduled
dates for completing the remaining events prerequisite to complete
computer check-out. While these scheduled dates are based on our
present best estimates, unforeseen problems may cause a further slippage in this s ched ule .
Although this report provides considerable detail as to the present status of computer checkout, a visual demonstration of actual computer operation is more impress ive particularly now that we have a few,
of the Whirlwind display demonstration programs operating. We
would therefore like to extend an invitation to you, Mr. Healy, Professor
Wilson and other ONR personnel to visit us again soon for the purpose
of such a demonstration.,
'
Very truly
yours~
~${.4M.
William M.
President
AVS/ csn
Enc~osures
wOlf-'
TABLE OF CONTENTS
I
INTRODUCTION.
II
MOVING AND REASSEMBLY . .
III
. .
. .
1
A.
Rigging and Moving...
B.
C.
D.
Air Conditioning Installation
Power Equipment
·2
.
•
t.
2
•
3
. . . . . .
. . . .
3
Mechanical and Electrical Reassembly.
4
COMPUTER CHECKOUT . .
A.
Equipment Checked Out.
1.
13
13
.
Power Equipment
a.
13
Standby Alternator and Filament Power
Distribution System . . . . . . . .
b.
DC Power Supplies.
c.
DC Power Distribution System
v3.
4.
13
14
.
14
16
. . . .
.
17
Arithmetic Element
18
Storage Element.
19
Cont ro1 Element.
a.
/b.
c.
5.
/'
. . . . . .
Marginal Checking System
2.
,
.
Toggle Switch Storage and Parity Register .
Flip Flop Storage
. . ..
19
. . . ..
19
20
Core Memory . . .
.
21
a.
Control......
21
b.
Peripheral Equipment
21
Input -Output Element.
1)
Flexowriter Keyboard Input, Reader/Printer,
and Punch
. . . . . . .
. . . .
21
Photoelectric Tape Reader . . .
22
. 3)
Real Time Clock.
22
/4)
Test Control Display Scopes
22
Intervention: and Activate Registers
23
2)
5)
. . . . .
Table of Contents (Cont'd)
,Page
B.
1.
Power Equipment.
/ct.
b.
2.
b.
tiC.
V
.
. .
23
.
400 Amp Alternator and Filament Power
Distribution System . . . . . .
....
DC Powe r Dist ribution System.
Input -Output Equipment.
\/a.
IV·
23
Equipment to be Checked Out
Display Consoles.
c
>
23
24
.
25
.
25
.
Indicator Light Registers
25
Light Guns.
26
.
.
. .
.
d.
Magnetic Tape Units
26
(e.
IBM Card Machines
26
f.
Anelex Alphanumeric Printer
g.
Test Control Camera .
27
h.
Magnetic Drums . .
27
i.
Teletype Input and Output
.26
.
28
MAJOR PROBLEMS ENCOUNTERED TO DATE
33
A.
B.
Wire- Wound Resistors
33
400 Amp Standby Alternator Slip Rings .
35
C.
Air Conditioning Equipment Trouble
36
D.
Phenolic Breakdowns ..
E.
Power Failures
SUMMARY
.
. ..
'
.
38
40
42
LIST OF ILLUSTRATIONS
Figure 2-1
Figure
3~1
PERT Network WWI Com.puter Moving and Reassembly
12
PERT Network '?fWI Com.puter Checkout.
32
.
.
.
.
.
.
LIST OF TABLES
Table
2-1
Rigging and Moving (RM).
2-2
Air Conditioning Installation (AC) .
2-3
Power Equipment (PE)"
2-4
Mechanical and Electrical Reassembly (ME) .
3 -1
Check Out (CO) .
"0
.
•
. .
"
•
" . . . . . ,.
.
•
.
.
7
8
9
•
.
.
.
.
.
" " ..
" " lq;i~l
.
" " 29-31
1
I
INTRODUCTION
This memorandum summarizes the present status of Whirlwind I
I
computer checkout and de sc ribes the wo rk ac complished to date in moving,
reassembling and checking out the Whirlwind I computer in its new
In addition, items of
equipm~nt
locat~on.
yet to be checked out are desc ribed and
scheduled.
Section II describes the moving and reassembly phase which comm.enced on Decem.ber 6, 1961, and was com.pleted on Decem.ber 13, 1962.
A PERT network showing the interrelationships and completion dates of the
events comprising the m.oving and reas sem.bly phase is also presented.
Section III describes the computer checkout phase which com.m.enced
on Dec em.be r 19, 1962, and is scheduled for com.pletion on Novem.b e r 15,
1963.
This section describes both work accomplished to date and work to
be accom.plished by the scheduled date of Novem.ber 15, 1963.
Section IV describes the major problems encountered'to date in
checking out the com.puter.
2
II
MOVING AND REASSEMBLY
The moving and reassembly phase commenced on December 6, 1961,
and was completed on December 13, 1962.
Figure 2 -1 is a PERT network
which depicts the interrelationships of events comprising this phase.
Events
are described by the circular. disks and the activities leading to these events
are described by the printed paper tape whose length is proportional to the
time required for the activity.
Unprinted tape represents activities which
require no time while unlabeled disks are merely connectors and do not represent events.
Tables 2 -1 through 2 -4 inclusive identify each of the events
in this network.
The labeling of these events is similar to that used in our
memorandum "Schedule for Moving and Reassembling the Whirlwind I Computer"
dated October 6, 1960.
Modification and additions to the tasks listed therein
have been made where necessary.
A.
Rigging and Moving
The events associated with rigging and moving are listed in Table
2-1.
Moving of the air conditioning cooling towers (RMll) and ductwork
(RM13) was accomplished on December 8, 1961, and the air handling units
and associated controls (RM12A) were moved on January 16, 1962.
Due to
delays in building construction, the moving of the main computer racks
(RM3, RM4, RM5 and RM7), which was originally scheduled for December,
was delayed until February.
Following this, the end wall of the computer
section of the building was completed.
The moving of rotating machinery
and associated switch gear (RM14), DC power supplies (RM15), and air
conditioning compressors (RMlO) had to be delayed until the completion of
the mechanical room section of the building.
These latter units were moved
and set up in place by the latter part of March, 1962.
Due to internal con-
struction work such as electrical, plumbing and heating, and painting, the
equipment racks could not be spread out and positioned until early April.
This was done during the period April 5, 1962 through April 13, 1962,
inclusive, and is described by events RM6, RM8 and RM17.
listed under rigging
an~
Other events
moving in Table 2 -1 were accomplished at our con-
venience as the equipment was needed but were scheduled so as' to vacate
Navy Building 61 in South Boston before termination of our lease extension
on June 30, 1962.
3
During the moving of additional equipment, it soon became
obvious that additional storage space more convenient to Concord was a
necessity.
Consequently, 8400 square feet of warehouse space was leased
in Maynard, Massachusetts.
While not shown explicitly on the PERT network,
the space at Building 61 in South Boston was vacated on or about May 29 J 1962.
Events RMl and RM16 represent, respectively, the moving from Maynard to
Concord of panels removed from the computer and the completion of the moving
of additional equipment stored in our Boston office and in quarters adjac ent to
our new building in Conc 0 rd.
B.
Air Conditioning Installation
Table 2-2 describes the events associated with air conditioning
equipment installation.
The installation of the air conditioning equipment com-
menced on May 23, 1962, with the installation of ductwork (AC3A) and electrical switchgear and control wiring (AC4A).
Installation of refrigerant piping
(ACl) was started early in June and was completed on July 31.
Cornpletion of
the rnajority of the ductwork (AC3C) and electrical wiring (AC4B) as well as
pneumatic control installation (AC2B) was accomplished by August 2.
Charging
of the system with Freon and check out of the basic system (AC5B) was
accomplished on August 4, 1962.
Completion of remaining air conditioning
work had to be delayed until receipt of a
which provides room air cooling.
n~w
cooling coil for the air handler
This coil was received during the latter
part of September and completion of all air conditioning work (AC2C,
AC3D, AC4C, AC4D, and AC5C) was accomplished on October 18, 1962.
C.
Power Equipment
The tasks involved in installation of power equipment are listed
in Table 2 -3.
Two of the twelve events listed in this table were purpos ely
ornitted from the reassembly phase of the project in order to perrnit an
earlier start of the computer checkout phase.
These two events, namely..
the wiring and checkout of the 400 amp alternator (PEl and PE5B) affect only
the special input-output equipment, particularly the magnetic drums and thus
were not essential to the checkout of the main computer.
These events are
yet to be accomplished but are scheduled on the PERT network presented in
4
Section III.
A third event listed in Table 2 -3 (PEB) has been completed but
is shown on the PERT network of Section III since it is more properly considered as a checkout event rather than a reassembly event.
The remaining events listed in Table 2 -3 are portrayed in the
PERT network of Figure 2-1., These commenced during the latter part of
March, 1962 and were completed on December 13, 1962.
These tasks in-
volved the setup and wiring of the 400 amp standby alternator, associated
power control racks and contactors (PE2, PE3A, PE3B, PE4, and PE5A),
installation and wiring of computer DC power supplies (PE6A, PE6B, PE7,
and PE9) and wiring, checkout, and- distribution of laboratory DC power
(PEIO and PE 11).
D.
Mechanical and Electrical Reassembly
The events associated with mechanical and electrical reassembly
of the computer are listed in Table 2 -4.
All but the last four events in this
table were completed during the moving and reassembly phase and are portrayed in the PERT network of Figure 2-1.
on the PERT network of Figure 3 -1.
The last four events are scheduled
One of these events (ME23) has been
completed and the remaining three (ME24, ME25 and ME26) are scheduled
for completion within the next ten days.
These four events were purposely
eliminated-from the moving and reassembly phase since they could be
con~
veniently postponed without delaying the start of the checkout phase of the,
project.
The remaining events associated with mechanical and electrical
reassembly were started early in April, 1962 and were completed on November 15, 1962.
It should be pointed out that one of these events (ME14) was
intentionally deleted from the PERT chart, since after a few weeks of checking
it was concluded that the error rate of Northeastern students in reconnecting
cables was sufficiently low to make exhaustive checking of cables unnecessary.
It was decided that if power connections were checked out in a fail-safe manner
that signal cabling errors could be tolerated without damage to the computer.
This decision to bypass checking of wire and cable reconnections in great
detail saved considerable time and has been proven to be a correct decision
5
during the checkout phase which is described in Section III.
While exact
figures on the error rate in reconnecting cables have not been kept, our
best estimate is that this error rate was not more than 50/0 and was probably_
as low as 2 to 30/0.
Considering the semi-skilled personnel perform.ing' this
task, the fact that two different groups disconnected and reconnected the
cables, and inevitable human error in recording and interpreting cable
termination data, we feel that this record is extremely good.
While the mechanical and electrical reassembly phase consisted
of about 25 different tasks, the major portion of the job involved reconnection of wires and cables (ME13).
on April 30, was
This task, which was started
cOlTIpleted on October 29.
It shoul? be pointed out here,
that thfs reconnection included all the ca_ble s which were disconnected at
Cambridge, not only in the main computer but in the input-output equipment
area associated with drums, radar and teletype inputs, and digital outputs.
Thus, this"task involved about twice as much work as that defined as MEl3
in our memorandum nSchedule for Moving and Reassembling the Whirlwind I
Computer" dated October 6, 1960, wherein we had planned to reconnect only
those cables associated with the central computer.
Other factors tending to
lengthen this task included the fabrication of new cables to replace ones
which were badly frayed and tangled at Cambridge, extending or replacing
cables which proved, in a few cases, to be too short due to our relocation of
test control relative to the main computer, and time spent in pulse checking
each video cable prior to reconnection.
Pulse checking of video cables has
proven valuable during the checkout phase since the only video cable failures
found to date during checkout have been in cables which were disconnected
only at one end and which could not be pulse checked since their destinations
were not known without considerable checking of MIT records.
All other events associated with mechanical and electrical
reassembly required relatively short times for their completion.
For example,
events MEl through ME7 inclusive were completed during the month of
April, 1962.
Other events which appear later during the schedule shown in
the PERT network of Figure 2 -1 involved effort of a few days to two weeks
and were accomplished as required and as time permitted.
The activity
paths,to most of these events represent time delays primarily.
For example,
6
about 22 weeks elapsed between the installation of computer wireways and
ground busses (ME5) and the completion of reconnecting computer wires
which we had cut in Cambridge (ME 11).
This latte r event (ME 11) required
about a week of effort for its accomplishment. but due to manpower limitations was not started until the ,middle of September, ,1962.
This manpower
limitation was dictated primarily by efficiency considerations, since the
size of the
work~ng
crew was about optimum, an increase in its size ..would
have resulted in decreased effectiveness of the group.
7
TABLE 2-1
Rigging and Moving (RM)
Event No.
Desc ription
Date Completed
or Scheduled
RM1
Removed Panels Moved
July 21, 1962
RM2
Transformer Panels Moved
April 15, 1962
RM3
Test Control, Core Banks and
February 13, 1962
Drums Moved
RM4
Computer Racks 0-7 Moved
February 9, 1962
RM5
Computer Racks 8-15 Moved
February 9, 1962
RM6
Computer Racks Positioned
April 10, 1962
RM7
I/O 'Racks Moved
February 9, 1962
RM8
I/O Racks Positioned
April 13, 1962
RM9
Wire Boxes Moved
April 15, 1962
RMIO
Compressors Moved and Set Up
March 26, 1962
RMll
Evaporators Moved and Set :UP
December 8,1961
RM12A
Air Handling Units and Control
January 16, 1962
Boards Moved
RM12B
Air Handling Units Set Up
February 7, 1962
RM13
Ductwork Moved
December 8, 1961
RM14
Rotating Mac;:hinery and Switch-
March 26, 1962
gear Moved and Set Up
RM15
DC Supplies Moved
March 2'3, 1962
RM16
Remaining Equipment Moved
June 12, 1962
RM17
TC Rack$ Positioned
April 5, 1962
8
TABLE 2-2
Air Conditioning Installation (AC)
Event No.
ACI/
AC2A
~/
AC2B
Description
Date Completed
or Scheduled
Refrigerant Piping Com.pleted
July 31, 1962
Sta rtConnecting Pneum.atic
Controls
August 1, 1962
~\
Com.plete all Pneumatic Controls
August 2, 1962
Except Room and Mechanical Room.
AC2C
Complete all Pneumatic Controls
September 27, 1962
AC3A//,
Start Ductwork Installation
May_ 28, 1962
AC3B
Complete Fresh Air And First
JUly 6, 1962
J
Section Supply Ducts
AC3C
Complete all Ductwork Except
August 2, 1962
Room and Mechanical Room
AC3D /'
AC4A
,ft_/'
Complete all Ductwork
September 27, 1962
Start Electric Switchgear and
May 28, 1962
Control Wiring
AC4B
Complete all Wiring Except Room
July 29, 1962
and Mechanical Room Controls
AC4C
Start Remaining Wiring
October 12, 1962
AC4D
Complete all Wiring
October 15, 1962
1/'-
Start System Checkout
August 3,' 1962
1-
Complete Basic System. Checkout
August 4, 1962
Complete System Checkout
October 18, 1962
.AC5A
AC5B
AC5C
9
TABLE 2-3
Power Equipment (PE)
Desc ription
Event No.
* PEll
Date Completed
or Scheduled
400A Alternator Wired
August 14, 1963
400A Standby Alternator Wired
June 22, 1962
PE3A I-
Power Control Racks Set Up
April 6, 1962
c/
Power Control Racks Wired
October 15, 1962
Filament Contacto r s Wired to
October 26, 1962
PE2
;PE3B
~
PE4
Busses
1"/
PE5A
* PE5B
PE6A
j--'
PE6B
/'
PE7
//'
** PEB
PE9
PEI0
y"
v'
Standby Alternator Checked Out
December 13, 1962
400A Alternator Checked Out
August 20, 1963
DC Supplies Installed in Racks
June 9, 1962
DC Supplies Wired
October 12, 1962
AC Inputs to DC Supplies Wired
October 12, 1962
DC Supplies Checked Out
January 17, 1963
DC Supplies Wired to Busses
August 17, 1962
Lab. DC Supplies Wired and
June 29, 1962
Checked Out
PE11
Lab. DC Distribution Completed
October 16, 1962
*Incomplete as of July 31, 1963. Desc ribed and scheduled in Section III.
** Des(f:ribed and scheduled in Section III.
10
TABLE 2-4
Mechanical and Electrical Reas sembly (ME)
Event No.
Dese ription
Date Completed
or Scheduled
MEl
v
Floor Marked for Computer Racks
April 6, 1962
MEl
r
Floor Marked for TC Racks
April 2, 1962
Floor Marked for I/O Racks
April 10, 1962
Junction Box Mount Completed
April 18, 1962
Computer Wireway and Ground
April 26, 1962
ME3
)//
ME4
r
ME5
j/'
Bus Installed
ME6A
Start 'Test Control Wireways
April 16, 1962
ME6B
Test Control Wireways Installed
April 27, 1962
ME7
I/O Wireways and Ground Bus
April 27, 1962
Installed
ME8
'
.
Removed Panels Replaced
JUly 26, 1962
ME9A
,!,.,.,.".
Transformer Panels Replaced
April 20, 1962
ME9B
l:r
Transformer Panels Rewired
August 10, 1962
Filter Panels Replaced
November 15, 1962
Bus Ba r s Installed
July 18, 1962
Computer Cut Wires Reconnected
September 26, 1962
I/O Cut Wires Reconnected
September 27, 1962
Start Reconnection of Wires and
April 30, 1962
ME9C':,MElO
MEll
v
","'"
V
MEl2
MEl3A
t
Cables
MEl3B
ME14 .
~.
f-.-/
Reconnection of Cables Completed
October 29, 1962
Reconnectionof Cables Checked
Deleted
Out
ME15
Interco;rn Reconnected and Checked
Out
September 14, 1962
11
TABLE 2-4
Mechanical and Electrical Reassembly (ME) (Cont'd)
Event No.
ME16
Date Completed
or Scheduled
Description
j,-/
Transformers in Banks A and B
October 25, 1962
Replaced
ME17
ME18
p/
j/
Core Stacks A, Band C Installed
October 29, 1962
Drum Bearings Checked and
September 24, 1962
Replaced if Required
ME19
["./
ME20
Drums Installed
September 28, 1962
AC and DC Power for TC
October 3, 1962
Installed
ME2l
v
ME22
J unction Boxed Repo s itioned
October 5, 1962
Test Bench for Drum Chassis
September 24, 1962
Installed
** ME23
1-/
I
Construct Anelex Transition Panel
April 11, 1963
* ME24
y/ Install Anelex Cabling "
August 9, 1963
* ME25
y
August 9, 1963
Repair Brush Box,es on 400A
Alternator
*
ME26
/ ' Recable 3 Display Consoles
*Incomplete
;
as of July 31,
1963.
August 2, 1963
De,scribed and scheduled in Section III.
>:C*Described and scheduled'in Section III.,
I
I.,
13
III
COMPUTER CHECKOUT
The checkout phase commenced on December 19, 1962 with the
completion of checking the standby alternator cycling system from the
alternator output to the computer busses and thence to the computer racks.
That is, filament power was applied to all of the main computer racks.
Checkout ,is scheduled to be completed on Novembe r 15, 1963 with the com -:.
pletion of checkout of the teletype input subsystem.
'.
Presented in Figure 3-i
is a PERT network which describes the activities and events comprising
this phase of the project.
listed in Table 3 -1.
All but seven of the events in this network are
As mentioned in Section II, the remaining seven event's,
(PEl, PE5B, PEB, ME23, ME24, ME25, and ME26) have been carried over
from the moving and reassembly phase as their accomplishment was not
essential to the start of the checkout phase and it was more convenient to
defer action on these tasks.
2 -4 of Section II.
These tasks are described in Tables 2-3 and
Two of these seven events (PEB and ME23) have already
been accomplished; the remaining five are scheduled to be completed within
the next three weeks.
It should be noted that our October,
~960
moving
schedule memorandum has no counterpart of Table 3 -1, since the 1960
memorandum scheduled only the moving and
reas~embly
!
I
but not the checkout
of the computer.
Section A below describes equipment already checked out during the
checkout phase, while Section B describes tasks involving equipment yet to
be checked' out.
A.
Equipment Checked Out
1.
Power Equipment
a.
Standby Alternator and Filament Power Distribution System)
As mentioned above, the checkout phase began on December
19, 1962 with the
success~ul
completion of checking of cycling of the standby
alternator to the computer busses (COl) and thence to the computer racks
(C02).
These two events were completed with very little difficulty.
Power
was cycled on to groups of eight racks at a time with several personnel
14
st~nding
by to warn of trouble so that power could be reITloved immediately,
thereby minimizing damage to the computer panels.
As power was applied
succes sfully to each group of eight racks, the filament voltage was then
cycled down, eight more racks were turned on and observed for signs of
trouble as the filaments were cycled back on.
This procedure of adding
modules of eight racks to previously checked racks with alternate cycling on
and off of the filament power was repeated until filament power was successfully applied to all the racks in the main computer including the core memories.
The success of this operation is indicated by the fact that the entire process
was completed in a period of about four hours.
Other than a difficulty due to
voltage breakdown in the standby alternator slip rings, which occurred on
April 12 a.nd which is described in Section IV -B, asic input-output devices.
More
detailed checking of the multiply, divide and shift operations was again undertaken on July 15 and was completed on July 30, 1963.
D':lring the initial check
19
of Arith:metic Element operation (CO 17 A) most of the troubles were corrected
.by replacing all the 2000 ohm ten watt wire -wound resistors with 2000 ohm
twelve watt wire-woundresistors and replacing several of the matched pairs
of flip-flop tubes.
The few remaining
troubl~s
were corrected by replacing a
small quantity of gate tubes and rectifying a few video cabling errors.
During
the recent final check of the Arithmetic Element (C017B) troubles with the
multiply, divide, and shift instructions were found to be caused by pulse
repetition frequency sensitivity when shifting alternate l's and a's through
the flip-flops of the B-Register and Accumulator.
Further investigation
revealed that this sensitivity could be eliminated by replacing the 1600 ohl;'n
I
ten watt resistors in the flip-flop circuits with 1600 ohm twelve watt resistors.
Since multiplication is basically an add and shift operation, and since divid,e
is basically a subtract and shift ope ration, elimination of the shifting probl'erns.
also cured the multiply and divide problems.
4.
Storage Element
a.
Toggle Switch Storage and Parity Register
Checkout of Toggle Switch Storage and the Parity
Register (GalS) as well as the Storage Switch was performed concurrently
with the initial check of the Arithmetic Element and the checkout of FlipFlop Storage.
This event was accomplished on April 22, 1963.
Troubles
with the Toggle Switch Storage Registers and the Storage Switch were corrected by replacing ten watt resistors with twelve watt resistors, replacing
several defective 6Y6 tubes, and by replacing several wires which had been
cut when the computer was moved and which had been poorly butt connected.
Troubles with the Parity Register were cleared up mainly by replacing flipflop tubes and/or flip-flop plug-in units.,
b.
Flip-Flop Storage
Checkout of Flip-Flop Storage (CO 16) also proceeded
concurrently with the initial check of the Arithmetic ·Element.
As with the
Arithmetic Element, most of the troubles with Flip-Flop Storage were corrected by replacing 2000 ohm ten watt resistors with twelve watt resistors
and replacing several matched pairs of 6145 flip-flop tubes.
A few other
troubles were corrected by replacing defective crystal diodes and correcting
two wiring errors in the Flip-Flop Storage reset switches.
20
c.
Core Memory
AC and DC power were initially applied to core memory
Banks A and B on May 15 and to core memory Bank C two days later.
After
correcting two video cabling errors and replacing a defective video cable,
proper operation of the Memory Address
R~gisters
was accomplished.
At
this point it was then possible to attempt writing into and reading from registers
in each core bank.
Parity errors were encountered in Banks A and B but a
check of a few registers in each of the four core fields in Bank C seemed to
indicate that this bank was working perfectly.
The initial success of operation
of Bank C and initial failure of Banks A and B was due, in pa rt, to adjustment
of read and write currents.
Read and write current controls in Banks A and
B had been misadjusted during the move) whereas the controls in Bank Care
locked and had not been disturbed by the moving process.
The current controls
in Banks A and B had been roughly reset in accordance with some photographs
we had taken in Cambridge prior to computer disassembly.
More precise
adjustment of the read and write current controls in Banks A and B with an
oscilloscope made some improvement in operation of these cores but did not
entirely correct the problems.
It was then discovered that most of the
remaining troubles in Banks A and B were due to faulty operation of digit plane
driver plug-in units which were failing to provide proper inhibit currents and
\
were causing parity alarms.
Troubles with these plug -in unit digit plane
drivers were found to be due to open ten watt non-inductively wound resistors.
Since we had good operation from the 4000 core registers in Bank C at that
point, it was decided to reduce our efforts on checking core memory so that
checkout of the input-output system, the Flexowriter and photoelectric tape
reader could be expedited.
Further checking of Bank B
suddenly revealed that the
parity error problems were not entirely due to faulty digit plane driver
operation when it was discovered that the cables from the digit plane drivers
to the core. planes had been completely reversed.
Correcting this cabling
error and repairing the digit plane drivers seemed to clear up all core
memory troubles for a period of ahout two weeks.
However, with the onset
of very hot weather toward the latter part of June, troubles developed in all
21
core banks and we were suddenly forced to concentrate all our efforts on
checking core memory again.
Some of the troubles were evidenced by
blown fuses, probably due to overheating, which resulted from a lack of
Freon in the air conditioning system.
The system, as described in Section IVe,
had been found to have lost all Freon during the latter part of May.
Over-
heating problems during the hot weather at the end of June subsequently
revealed that we had underestimated the Freon capacity of the system and had
replaced only about half of the required amount of Freon.
During the first
week of July, a more exhaustive check of all core banks was made.
Troubles
with Bank C were cleared up by replacing two broken tubes, two blown fuses,
correcting an open circuit in a common ground line on a selection plane driver
due to a poor solder joint, and replacement of a burned out 5998 tube in a
selection plane driver panel.
5.
Input -Output EleITlent
a.
Control
Checkout of the In -Out Element (CO 19) was accomplished
with very little difficulty by June 5, 1963.
The only difficulty encountered with
the In-Out Register was the failure of one flip-flop which was corrected by replacing the flip-flop tubes with a new matched pair of 6145 IS.
Troubles with the
In-Out Delay Counter were corrected by replacing a missing gate tube and two
defective video cables which were responsible for failure to terITlinate block
transfer. instructions.
Difficulties with the In-Out Switch were corrected by
replacing a missing 6Y6 tube which drives the clear line to the flip-flops of
the In-Out Switch and by replacing an open 22, 000 ohm twelve watt wirewound resistor in a cathode
follower.
This cathode follower was failing to
shut off a gate tube in the clear line to the In-Out Switch flip-flops and was
thus permitting spurious clear
In-Out Switch.
puls~s
to interfere with proper operation of the
Operation of the In-Out Element to date has been excellent
and no failures have been encountered since the above troubles were corrected.
b.
Peripheral Equipment
1)
Flexowriter Keyboard Input, Reader/Printer, and
Punch
Checkout of the Flexowriter direct keyboard input,
direct printer and 'punch output and mechanical paper tape reader input (C022)
was accomplished on June 5, 1963.
The only difficulties encounteredhere
were caused by two wires which were incorrectly connected.
22
2)
Photoelectric Tape Reader
Checkout of the Ferranti Photoelectric Tape Reader
(C023) was also accomplished with very little difficulty.
Three troubles
were encountered here, two due to missing DC voltages, the third due to
failure of one of the digits to read to the In-Out Register.
One of the missing
voltages was due to a wiring error, the other was due to a broken wire.
The
failure to read one of the digits to the In-Out Register was corrected by
tightening a tube which was loose in its socket in an In-Out Register Input
Mixer panel.
3)
Real Time Clock
Checkout of the Real Time Clock (C032) was com-
pleted on June 13, 1963.
Trouble with the clock was corrected by rectifying
one video cabling error and replacing an indicator light bracket.
4)
Test Control Display Scopes
Successful completion of event C027 A was accom-
plished on June 19, 1963.
This task included checkout of all the control
elements o.f the display subsystem as well as the two display scopes in Test
Control.
Troubles encountered in this area included failure of the vertical
deflection signal which was caused by a bad 5651 voltage regulator tube in the
Vertical Decoder, failure of vertical deflection on one scope due to a loose
wire on a binding post, and a cabling error which caused the horizontal and
vertical deflection signals on one scope to be interchanged.
Troubles with the
.individual scopes were encountered in the Dumont high voltage power supplies
and in the deflection amplifiers.
Troubles with the Dumont high voltage supplies
were corrected by replacing a broken high voltage coil and two open wire -wound
potentiometers.
were
corr~cted
Malfunctions in one of the horizontal deflection amplifiers
by replacing defective tubes. After correcting these dif-
ficulties, prope r operation of point displays, characte r displays, and vector
displays was accomplished by aligning the deflection decoders, the decoder
output amplifier, and the scope deflection amplifiers in accordance with
procedures established at MIT.
23
5)
Intervention and Activate Registers
The Intervention and Activate Registers (C02S)
were also checked out on June 19.
The only difficulties encounte'red with
these subsystems were caused by missing voltages on the switches in, Test
Control, two video cabling errors, and a fuse blowing problem associated
with the cathode followers which enable the read-out gates of the individual
Intervention Registers.
The fuse involved kept blowing repeatedly and the
trouble was finally cleared up by replacing it with a larger size fuse after
determining that the fuse rating was too low for the amount of current being
drawn by these cathode followers.
B.
Equipment to be Checked Out
Several more units of input -output equipment remaiI). to be checked
before the computer can be considered to be completely checked out.
These
include principally the magnetic tape units, the IBM card machines, the 'Anelex
printer, three additional display consoles, and the two magnetic drums.
In
addition, the 400 amp alternator which provides filament power to the magnetic
drums and associated input -output equipment requires some repair and must be
rewired.
Also, the power distribution system, both AC and DC, for the
magnetic drums and associated equipment must be checked out.
Some checkout
work has been done in each of these areas with the exception of ,the drums.
Many of these jobs are scheduled to be completed within the next few weeks
and present indications are that all work should be completed on or about
November 15, 1963.
1.
Powe'r Equipment
a.
400 Amp Alternator and Filament Power Distribution
System
Work on repairing the brush boxes on the 400 amp
Alternator (ME25) is scheduled to start the latter part of this week.
Prelim-
inary investigation into this problem indicates that replacing of the slip ring~
may be a more desirabl,e long-term solution.
We are awaiting information
from the manufacturer, which will determine just how much wear on the
,
rings has occurred.
sl~p
\
After talking with the manufacturer's local representativ.~,
24
there isa slight possibility that the solution to the problem may be merely
the use of a different type of brush than that used by MIT.
Data from the
factory giving original dimensions of the slip rings will be the deciding factor
on whether or not these will be replaced.
If it is necessary to replace the
slip rings, there is a good pos sibility that a new pair may be available from
stock.
~
If the slip rings are not available from stock, then a two week delivery
period is quoted.
In this latter event, it would be necessary to postpone further
work on the slip rings for two weeks.
However, other tasks as sociated with
the alternator could be performed sooner so that the two week delivery delay
.will not impose any slippage on the existing schedule.
Reconnecting the cables from this 400 amp alternator
(PE I) has been planned and presently is scheduled to be accomplished after
the slip ring problem is corrected..
However, if it is necessary to wait the
two weeks for slip ring delivery, the wiring of the alternator can be done
first.
Checkout of the 400 amp alternator (PE5B) is presently
scheduled to be completed by August 20.
No difficulty is ant,icipated in
accomplishing this task as scheduled, even if it is necessary to wait two
weeks for delivery of slip rings.
After checking out the operation of the 400 amp alternator, the next tasks will be to check the 400 amp alternator cycling system
and connections to the in-out filament busses (C034) and the filament connections to the individual in-out equipment racks and to the drum bays (C035).
With the exception of the feeder cables from the 400 amp alternator, all the
necessary filament power cycling and distribution wiring was completed
during the moving and reassembly phase last year.
Thus, little difficulty is
foreseen in completing these two tasks by the end of August.
b.
DC Power Distribution System
After checking filament power connections to the in-out
equipment racks and drums, it will be necessary to check the DC connections
to these equipments.
Two tasks are involved-the first, checking of the DC
supplies to the in-out equipme~t busses (C036) and the second, checking the
25
DC power distribution to the individual racks and drum bays (C037).
The
only difficulties anticipated in connection with these two tasks are in checking
DC power connections to the drum bays and the connection 9f rack DC switches
which control the outputs of the power distribution panels to the associated
racks and drum bays.
Connection of the outputs of the power distribution
panels to the individual racks should give no trouble, since unlike the racks
in the main computer, most of this DC wiring did not have tobe disconnected.
The scheduled date of September 20 for completion of DC checking should be
accomplished without difficulty.
2.
Input -Output Equipment
With the exception of the Anelex printer, magnetic drums,
teletype input, and teletype output on which no checking has yet been done,
checkout of remaining input-output units has been done to some degree.
a.
Display Consoles
Three display consoles, in addition to the two in Test
Control already checked out, are planned.
and checkout is nearly completed.
just been started.
Two of these have been recabled
The third display console recabling has
When cabling of this scope is completed, event ME26 will
have been accomplished.
Completion of this event is scheduled for August
2, but may slip one or two days.
Complete checkout of these three display
consoles will accomplish event C03l 'on or about August 7.
The only diffi-
culties anticipated in checking these three displays are troubles with the
Dumont high voltage power supplies, alignment problems in the deflection
amplifiers, and a few minor wiring errors.
b.
. Indicator Light Registers
Completio~
of checkout of all Indicator Light Registers
required at present (C030) is expected to occur concurrently with the completion of the checkout of three display consoles mentioned in a. above.
Most
of these registers have been checked and it is felt that remaining troubles are
due to defective thyratrons in these registers or to one or two· wiring errors.
26
c.
Light Guns
Three light guns are being checked out for use with the
three display consoles mentioned in a. above.
Two light guns have been
reconnected and one of these has been checked out completed while the third
is being connected with the third display console.
Completion of light gun
checkout (C029) is also scheduled to coincide with the completion of checkout
of the three display consoles.
No difficulties other than tube failure and minor
wiring errors are anticipated.
d.
Magnetic Tape Units
Checkout of the znagnetic tape units (C026) is scheduled
to be completed on August 12.
are all the manual controls.
Mechanically these units are operational as
Thus, all the remaining troubles with these units
are electronic and it is not anticipated that any serious difficulties will be
encountered here.
e.
IBM Card Machines
All the wiring for the IBM card machines was checked
very thoroughly during the moving and reassembly phase last fall.
The IBM
523 gang summary punches used with this system are in good mechanical
shape.
These two machines, which were specially modified by MIT for use
with Whirlwind, were being leased from IBM by MIT.
The Wolf Research and
Development Corporation took over the lease from MIT and has since purchased
the machines outright from IBM.
Since the wiring associated with these card
machines has been thoroughly checked, checkout of the Whirlwind card machine·
system (C024) should be a fairly straightforward task and should be completed
around August 15.
f.
Anelex Alphanumeric Printer
The Anelex Alphanumeric Synchroprinter used with
Whirlwind was not acquired from Lincoln Laboratory until September, 1962.
As the task of reconnecting computer wires and cables was nearly completed
at that point, it was decided to defer recabling of the Anelex printer until the
checkout phase.
All units of the Anelex printer system were acquired from
Linc9ln with the exception of the transition panel which could not be located in
their warehouse.
Thus, it was necessary to build a new transition panel and
27
this task (ME23) was com.pleted on April 11 of this year.
Recabling of the
Anelex equipment has been partially completed over the past couple of months
as time permits.
Presently, this task (ME24) is about half completed and is
scheduled for final completion on August 9. Operational checkout of the Anelex
system will be accomplished during the last two weeks of August with a scheduled
date of August 30.
Here again, little difficulty is anticipated particularly
since one of our engineers has had considerable experience with the operation
of this printer at MIT.
g.
Test Control Camera
Some difficulties
wi~h
the checkout of the Test Control
)
camera subsystem were encountered during June when the Test Control
Display Scopes were checked out.
This is not a very' complicated system, but
MITis drawings and records associated with this system are in very poor shape
and apparently a few wiring errors were made in reconnecting the camera.
At present, the camera can be indexed manually but not by program command.
One of the known troubles is that about six plug-in units are missing from this
subsystem and were missing, according to our records, when we acquired
responsibility for the computer.
Also, because of the poor records involved,
about two dozen wires will have to be traced manually before the cam.era can
be checked out.
Since this is a relatively low priority item, it has been
deferred until the first week of September and this task (C027B) is scheduled
for completion on September 4.
While this job will not take more than two
or three days, the amount of effort is out of proportion relative to the complexity of the system due to the poor records available.
h.
Magnetic Drums
Checkout of the Auxiliary Drum (C038) is scheduled for
October 11 while checkout of the Buffer Drum (C039) is scheduled for
November 8.
These two tasks cannot be started until AC and DC checkout
I
(C034, C035, C036, and C037) of the drum bays and associated input-output
equipment is completed.
It is somewhat more difficult to anticipate what
troubles may be encountered in checking out the drums, but possible trouble
sources include adjustment and alignment of read and write heads, cabling
28 '.
errors between the drums and the main computer, troubles with the electronic
chassis in the drum bays, and possible difficulties with drum bearings.
A preliminary check of the condition of the drum. bearings
was made during the moving and reassembly phase.
This check indicated
that the drum bearings are in good shape and that the oil leakage problem. was
due to over filling of the oil reservoirs
~t
MIT.
Although we have not been able
to obtain mechanical assembly drawings for the drums from Sperry Rand, we
do know that the bearings are commercially available.
We have also been
advised by Sperry Rand that the run down time from full speed to rest when
power is removed should not be less than 2-1/2 to 3 minutes after the drum
has been running for one hour.
bearing troubles.
A shorter run down time would indicate
As for bearing life, Sperry Rand indicated that the bearings
should be good for at least another ten years on the basis of usage to date by
MIT.
We feel that our present schedule makes ample provision for any diffi-
culties which may be encountered in checking out the drums.
i.
Teletype Input and Output
Thorough checking of teletype input and output subsystems
requires a teletype transmitter distributor, a typing reperforator, and a Type
28RO teletypewriter.
Since we have no immediate plans for use of these sub-
systems, we will be able to check out only those portions of these subsystems
which can be checked without the teletype equipment mentioned above.
A
reasonably thorough check of the electronics can be made by means of the
Teletype Test Message Generator on a closed loop basis with a computer
program. and this type of checking is all that is planned for the accomplishment
of events C033 and C040.
Checking of the teletype input subsystem which uses
slots on the buffer drum cannot be started until checkout of the buffer drum. is
completed, however.
29
TABLE 3-1
Check Out (CO)
Event No.
COl
Desc ription
Standby Alternator Cycling to
Date Completed
or Scheduled
December 19, 1962
Computer Bus
C02
AC to Computer Racks
December 19, 1962
C03
DC Supplies to Computer Busses
January 21, 1963
C04
DC to Compute r Racks
February 28, 1963
COS
Ma rginal Checking System
April 30, 1963
C06
Pulse Generator, Frequency
February 28, 1963
Divider, and Restorer Pulse
Generator
C07
Clock Pulse Control
February 28, 1963
C08
Time Pulse Distributor
February 28, 1963
C09
Test Control Synchronizers
February 28, 1963
COlO
Alarm System
May 10, 1963
COlI
Control Switch
March 15, 1963
COl2
Ope ration Matrix.
March 15, 1963
COl3
Control Pulse Output Units
March 15, 1963
COl4
Program Counter
April 23, 1963
COIS
Toggle Switch Storage and Parity
April 22, 1963
. Register
C016
Flip Flop Storage
May 16, 1963
C017A
Initial Check Arithmetic Element
May 16, 1963
C017B
Final Check Arithmetic Element
July 30, 1963
C018A
Initial Che'ck Core Memory
June 4, 1963
C018B
Final Check Core Memory
July 3, 1963
30
TABLE 3-1
Check Out (CO) (Cont'd)
Event No.
Description
Date Completed
or Scheduled
C019
In-Out Element
June 5, 1963
C020
Check Register
May 10, 1963
C021
Start I/O Equipment Check
June 5, 1963
C022
Flexowriter Keyboard Input,
June 5, 1963
Punch, Reader and Printer
>:<
>!<
>!<
>!<
C023
Photoelectric Tape Reader
June 5, 1963
C024
IBM Card Machines
August 15, 1963
C025
Anelex Printer
August 30, 1963
C026
Magnetic Tape Units
August 12, 1963
C027A
Test Control Displays
June 19, 1963
C027B
Test Control Camera
September 4, 1963
C028
Intervention and Activate
June 19, 1963
Registers
* C029
>!<
>:(
Light Guns
August 7, 1963
C030
Indicato'r Light Registers
August 7, 1963
C031
Display Consoles (3)
August 7, 1963
C032
Real Time Clock
June 13, 1963
C033
Teletype Output
September 13) 1963
C034
400A Alternator C'ycling to Bus
August 23, 1963
C035
AC to I/O Racks
August 30, 1963
C036
DC Supplies to I/O Busses
September 6, 1963
~:c
>!<
>!(
>!<
31
TABLE 3-1
Check Out (CO) (Cont'd)
Event No.
* C03?
De sc ription
Date Completed
or Scheduled
DC Supplies to 1/0 Racks
September 20, 1963
C038
Auxiliary Drum
October 11" 1963
C039
Buffer Drum
November 8, 1963
~O40
Teletype Input
November 15" 1963
~:c
):c
):c
*Incomplete as
of July 31" 1963
33
IV
MAJOR PROBLEMS ENCOUNTERED TO DATE
During the checkout phase, we have encountered several major
problems.
Two of these, namely troubles with defective wire-wound
resistors and phenolic breakdowns had been anticipated.
Two other
difficulties, one with slip rings on the 400 amp standby alternator,
and the other, troubles with the air conditioning equipment, occurred
unexpectedly.
We felt that we had solved the wire-wound resistor
problem last Summer when we replaced all the existing 8 watt wirewound resistors in the computer, but unexpected difficulties were
encountered with these new res istors.
Phenolic breakdowns did not
occur initially and for several months.
We had finally reached a point
where we felt the problem did not exi st when a rash of such trouble s
occurred.
Except for wire-wound resistor problems, failure rates
of other electronic components have been much lower than we had
anticipated.
In particular, vacuum tube failure has been extremely
low as has failure of crystal diodes.
Relay operation has also been
excellent and we have encountered trouble with only one relay out of
more than 2000.
A.
Wire-Wound Resistors
Tests of wire -wound resi stors at Navy Building 61
during the Summer of 1961 indicated a need for replacing all such
resistors in the computer panels.
The sample tested indicated about
75% of these resistors had opened up since the computer was shut
down in May, 1959.
During the Summer of 1962, new military RW3lG
ten watt wire-wound resistors were purchased to replace the 8 watt
resistors in the computer panels.
These resistors were carefully
bridged on a Wheatstone Bridge and matched to within 10/0 tolerance of
nominal value and of each other as specified in Whirlwind records.
Two students from Wentworth Institute were hired for the Summer and
were assigned the task of replacing these resistors.
Although MIT
had replaced some 6f the 8 watt resistors with 12 watt resistors of a
type similar to that which we purchased, our decision to purchase 10
watt rather than 12 watt resistors was based on two factors.
34
First, the 10 watt resistors which we purchased were the same
physical size as the 8 watt resistors to be replaced, whereas the
12 watt resistors are longer and present some mechanical problems
in :mounting.
Second, it was known that 12 watt wire -wound resistors
were a standard stock item at Lincoln Laboratory and were therefore
easier to obtain than the 10 watt resistors.
Although it was obvious
that 10 watt resistors would certainly handle the power dissipation
since the original resistors were only rated at 8 watts, we did compute
the power dissipation in these resistors and found that the. results,
namely dis sipations ranging from 2 to 4 watts, confirmed our conclusions.
At this point we felt that we had solved the problem and
due to other urgent tasks we assigned the resistor replacement job
to the Wentworth students.
During the latter part of the Summer of
1962·, some pulse checking was done on the Control Element and
the resistors which had been replaced here seemed to give very
satisfactory performance.
However, in the Spring of 1963 while checking the operation
of the Program Counter, difficulty was encountered with self-sustained
os cillations in many of the flip-flops in the Program Counter.
After
a few days of intensive bench testing of one of the Program Counter
panels, and review of early MIT records and memos for indications
of such troubles in the past, it was determined that the cause of the
trouble was primarily the 2000 ohm 10 watt resistors.
Although each
flip-flop contains two, 2000 ohm resistors, two 5000 ohm resistors,
and two 1600 ohm resistors, in the vast majority of cases proper
operation of flip-flops with voltage margins of plus or minus 30 to 40
volts has been accomplished merely by replacing the 2000 ohm ten watt
resistors with 2000 ohm twelve watt resistors.
Recently, however,
it has been neces sary to change the 1600 ohm resistors in the Accumulator Parital Sum flip-flops and the B -Register flip-flops to obtain
proper shifting operations.
35
A series of tests on ten and twelve watt wire wound resistors run
by the manufacturer at our request revealed the probable cause of the problem.
These tests measured the resistance and inductance of both ten and twelve
watt resistors at l, 2 and 5 megacycles.
The results
sho~ed
that the effective
AC resistance as a function of frequency increased much faster in the case of
the ten watt resistors than in the case of the twelve watt resistors.
This was
particularly true in the case of the 2000 ohm resistors and the trouble is assumed
to be due to the smaller wire size used to get the same resistance in the
smaller ten watt resistors.
While most computer operations are performed at
a maximum speed of 1 megacycle, shifting in the Accumulator and B -Register
is done at 2 'megacycles and thus the 1600 ohm resistors also had to be replaced
. ,to cure this latter problem.
While most of the open wire-wound resistors found to date have
been of the eight watt Navy-type, there have been other instances of open
wire-wound resistors.
Several commercial ten watt, non-inductive wire-
wound resistors in the digit plane drivers of Core Memory Ban-ks A and B
were found to be open, and one 22, 000 ohm, twelve watt resistor' used in a
cathode follower in In-Out Control was also found to be open.
Just this week,
trouble developed in an A-Register panel and this was traced to an open 2000
ohm twelve watt resistor which had been installed at MIT in place of an eight
watt resistor.
Environmental conditions to which the computer was subject
during the storage period are felt to be the cause of open wire -wound resistors.
Since the computer was stored in unheated space, it is felt that the low
temperatures encountered during the winter months caused the wire to become
brittle and that contraction of the wire at these low temperatures caused it to
break.
B.
400 Amp Standby Alternator Slip 'Rings
On April 12, after nearly two months of trouble-free operation,
difficulty was encountered with the 400 amp standby alternator.
This trouble
was evidenced by a higher than normal field current which could not be adjusted
downward to give proper operation.
Field resistance from pole to pole was
checked and found to be additive in a non-linear fashion, despite the fact that
the individual pole resistances were measured to be two ohms, which is the
corr¢ct value.
The inescapable colusion was that somehow additional parallel
resistance was being interposed between the two field terminals.
Further
checking revealed that the screws supporting the slip rings were too long ,and
36
nearly shorted the slip rings together.
A carbon buildup on the ends of these
screws was producing a twelve ohm resistance between the slip rings with
the field disconnected.
The screws were removed and replaced with shorter
ones and this corrected the trouble.
While the long screws were obviously
a manufacturing defect, the elapsed time meter indicated that the machine had
run for some 17, 000 hours prior to occurrence of this trouble.
While this
trouble would have occurred eventually, its onset m.ay have been hastened by
an accident which occurred a few days earlier .. At that time, it was noticed
that power was being applied to a relay as sociated with the magnetic tape
racks.
When the filament, switch for the tape racks was switched to what was
thought to be the off position to turn off this relay, two No. 10 filament wires
suddenly burned up. . When this switch was thrown, it effectivley placed a
short across the output of the 400 amp standbyalternator\ due'to some
temporary jumpers which had been installed last summer and which had not been
removed. 'This sudden overload on the alternator burned up these wires,
rather than blowing the 600 amp fuse in the output of the alternator.
Thu~,
this accidental overload probably hastened the breakdown between slip rings.
Although the slip ring breakdown was in the ;motor exciter, investigation of
the alternator exciter revealed that oversized screws had been used there as
well.
To prevent future troubles, the screws in the alternator exciter were
also replaced with shorter ones at that time.
c.
Air Conditioning Equipment Trouble
During the week of May 20 to May 24 we encountered our first
heat wave of the summer season.
Twice during the week the computer shut
itself down as a result of excessive temperature in the vicinity of the overheat
thermostat.
We concluded, at first, that the trouble was due to dirty filters
in the liquid line which were causing inefficient ope ration of the air conditioning
system.
On May 24, however, the air conditioning contractor, who had
installed the system, came to replace the filter elem.ents.
We were told
that the difficulty was due to a nearly complete los s of Freon in the system..
Consequently, the system had to be completely leak checked and then refilled
with Freon .. While leak checking, a few small leaks were found and corrected
37
and it was concluded that the Freon loss :must have occurred over a long
period of ti:me due to these s:mall leaks.
To prevent future losses of such
proportion l a schedule of periodic leak checking was instituted.
Prior to the Freon loss 1 the syste:m was shut down each evening l
:merely by turning off the :main· circuit breaker to the air conditioning syste:m.
After replenishing the Freon supply 1
this procedure resulted in severe
slugging of the co:mpressors with liquid Freon during the night.
To alleviate
this slugging proble:m 1 it was decided to let the syste:m run continuouslYI
even though unattended
safe devices.
l
since it is :more than adequately protected with fail-
On MondaYI June 171 exactly one week after instituting twenty-
four hour operation, we arrived to find the syste:m shut down and the floor
covered with oil.
So:meti:me during the week-end the nu:mber 4 co:mpressor
had blown a flare fitting in an oil line which resulted in co:mplete loss of oil
in the entire syste:m and very possibly si:multaneous loss of a. few cylinders of
Freon.
The oil fitting was replaced and fifty gallons of Capella oil were
added to replenish the oil supply.
Following the oil loss, twenty-four hour
operation of the air conditioning syste:m was abandoned and instead the syste:m
is shut down each evening after pu:mping the Freon back into the receiver to
prevent slugging of the co:mpressors.
This procedure has worked reasonably
well, but this was not the end of our air conditioning troubles.
During a heat wave at the end of June, it beca:me obvious that the
air conditioning syste:m was still not operating properly and the co:mputer again
shut itself off due to overheating.
After checking several possible causes of
the trouble, it was finally concluded that additional Freon was required.
More
careful checking revealed that we had replenished only half the original supply
of Freon at the ti:me of the loss in May.
Additional Freon was added to bring
the system up to capacity and, in addition, the cooling coils were washed down
and several adjust:ments :made to the syste:m.
These :measures have resulted
in excellent operation of the air conditioning syste:m during the :month of July
and particularly during the excessively hot days from. Wednesday, July 24
through Friday, July 26.
38
During July we have had several occassions on which we thought
additional Freon loss might have occurred.
and subsequent
~fficient
However, thorough leak
checkin~
operation indicate that these were false alarms.
The
most recent of these occassions was on'Monday, July 29, the sixth and final
day of a period of unseasonably hot weather.
Due to the hot weather over the
week-end with the system shut down, the system was taxed to its utmost to
overcome the accumulated heat load of the week-end, without the additional
heat load of the compute r.
Two small leaks we re found on Monday and
corrected, but low Freon indication at that time was follo.wed by extremely
good operation on Tuesday, which was a much cooler day.
Low Freon indica-
tion on Monday is now felt to be due to some los s of Freon at the time of the
oil leak in June, which means that two or three cylinders of Freon should
probably be added to bring the system up to full capacity.
D.
Phenolic Breakdowns
Our first encounter with what recent experience indicates was
probably a phenolic breakdown occurred on May 20, nearly three months
after power was first applied to the computer panels.
No further trouble was
encountered until July 11, when we encountered the first of six phenolic breakdowns in a two week period, four of which occurred in one panel.
The phenolic
breakdown problem had been a matter of considerable concern to MIT and,
as mentioned in previous progres s reports, was felt by us to be a major
difficulty to be encountered during computer checkout.
However, since nearly
five months of continuous power application had elapsed since turning on the
computer with only one possible phenolic breakdown, we had thought that this
would not be a serious problem.
However, recent experience has reversed
our thinking on this problem and we now feel that phenolic breakdown may be
a serious difficulty which may increase in frequency as time goes on.
Investigation of this problem. during July has resulted in the conclusion, however, that the phenolic breakdown problem is nota disastrous
one because there is a relatively inexpensive solution.
The conclusion that
this problem is tending to increase in frequency is based on a fairly thorough
search of MIT r s records which failed to reveal any two week period in which as
many, as six breakdowns occurred.
Our investigation generally concurs with
39
earlier investigation by MIT which indicated that phenolic breakdown is due to
silver migration and that this is predominantly a surface effect.
The first
four breakdowns we encountered in July on one panel occurred on the component
side of the phenolic and tended to confirrrl an MIT opinion that cool air heavily
saturated with moisture at the air duct outlet was a major contributing factor.
MIT records on this particular panel reveal about a half dozen previous breakdowns, whereas records on an identical panel at the top of the same rack show
that no breakdowns have ever occurred.
Consequently, the flow of cold air
to the component side of this panel was blocked and no further breakdowns have
. occurred.
We feel, on the basis of the two more recent breakdowns, however,
that cool air laden with moisture may not be as important a factor as we had
originally thought.
The two latter cases occurred on the back side of the phenolic
board in panels which were remotely located from the cool air ducts.
The
latter of these two panels, B -Register digit 15, was removed from the computer
and examined closely.
After removing the phenolic board from the aluminum
panel it was evident that several potential trouble spots existed which werel.n.ot.
in evidence on the component side of the phenolic.
Both sides of the phenolic
panel are covered with dirt which is probably soot from Cambridge and a film
whose composition is unknown.
Cleaning with a good solvent removes both the
soot and the film as well as deposits of silver, and in areas where breakdown
has occurred, cleaning of the surface appears to restore the insulation resistance of the phenolic.
Cleaning is a ve ry tedious task when done manually due
to the necessity for opening up the panel and the complex layout of components
and wires.
Thought was given, therefore, to cleaning in an ultrasonic bath,
but pricing of this procedure reveals it to be extremely expensive and cleaning
of all computer panels would be costly in terms of time as well.
Furthermore,
MIT's chemical analysis revealed that some. silver migration occurs within the
phenolic material itself, although to a considerably lesser degree.
Since
cleaning is an expensive process, and since body migration would not be cured
by cleaning, consideration was given to c;t more effective way of eliminating
the problem entirely.
40
MITts solution of correcting each breakdown as it occurs by
replacing one of the affected lugs with an insulated standoff lug is acceptable
provided the frequency of occurrence does not increase.
Phenolic breakdown
can be located and corrected quickly by this :means without excessive COIllputer
down ti:me.
required.
However) if the frequency increases a Illore effective solution is
ExaIllination of the panel taken from B -Register digit 15 revealed
that the probleIll exists only between terIllinals having a relatively high
potential difference in the order of 150 to 250 volts.
Thus) a solution which
effectively inhibits Illigration between such terIllinal pairs will eliIllinate the
probleIll.
After further consideration) it was concluded that the IllOSt effective
and siIllplest :method for eliIllinating Illigration between these terIllinal pairs
is to cut a slot in the phenolic board between these terIllinals thus iIllposing
an air gap between them.
While this in itself would be a fairly tiIlle consuIlling
task) it is felt to be the fastest, :most effective) and least expensive solution.
While there are sOIllewhere in the order of 170 panels affected by this probleIll,
there are only about 35 different types of panels.
Thus, it would be a fairly
reasonable task to check the asseIllbly drawings for each of these panel
type s, locate all potential trouble spots, and then proceed to cut air gaps
between all potentially troublesome terminal pairs in all panels.
This task)
which Illight require in the order of one or two weeks of cOIllputer shutdown,
will be instituted if it becomes necessary due to a sharp increase in the
frequency of occurrence of phenolic breakdowns.
In the meantime) however)
we plan to follow MITts procedure of correcting each trouble as it occurs.
E.
Power Failures
Two outside power failures have occurred during the checkout
period.
The first of these occurred on May 20 when some of Boston Edison's
underground power lines were flooded in Maynard.
The second failure occurred
on July 18 when lightning struck a Boston Edison substation in Burlington)
Massachusetts.
On both occassions the computer under-voltage sensing
mechanism worked correctly and cycled the computer down in the proper fashion.
Both of these failures were external to the building and were due to circumstances beyond ourco,ntrol.
Such failures cause a certain amount of incon-
venience and delay as they did at Cambridge.
This problem could be eliminated
41
by the installation of diesel generators, but this is an expensive solution to
be resorted to only if the priority of future work demands it.
No action on
this matter is contemplated at the present time and this subject is mentioned
here only for the purpose of documenting a major trouble encountered during
checkout.
to date no
Another reason for including it here is to emphasize the fact that
~nternal
power failures have occurred.
42
V
SUMMARY
This progress report describes, in some detail, the moving, reassem-
bling, and checkout of the Whirlwind I computer from the commencement of
its move from Navy Building 61 in South Boston on December 6, 1961 until the
present.
Also presented are our plans for completing the checkout of the
computer by November 15, 1963.
PERT networks showing the events of the
moving and reassembly phase and the checkout phase, their interrelationships
and completion dates are presented and described.
To date this project has encountered several major problems, some of
which were expected while others were unexpected.
described as have their solutions.
These problems have been
Solutions to all problems have been effected
with the exception of that for the phenolic breakdown problem and this will be
effected when and as required.
No further major problems are anticipated
but if new ones should occur, they are not expected to be insurmountable.
Thus, we feel that successful completion of checkout will be accomplished on
November 15, 1963.
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-c043 52.372728, 2009/01/18-15:56:37 Create Date : 2010:06:30 19:29:41-08:00 Modify Date : 2010:06:30 19:32:42-07:00 Metadata Date : 2010:06:30 19:32:42-07:00 Producer : Adobe Acrobat 9.32 Paper Capture Plug-in Format : application/pdf Document ID : uuid:acc95290-2418-4c66-8422-72951bb74f42 Instance ID : uuid:8aa2308f-7b96-4f15-81b4-0e5730c55f53 Page Layout : SinglePage Page Mode : UseNone Page Count : 45EXIF Metadata provided by EXIF.tools