PDP-8/e Troubleshooting

PDP-8/E at the Computer History Museum (Tomwsulcer, CC0)

Fault-finding on the PDP-8/e, from the corrective maintenance chapter of DEC's PDP-8/E, PDP-8/F, PDP-8/M Maintenance Manual, Volume 1. The processor and memory tables apply equally to the PDP-8/F and PDP-8/m; their H740 supply is covered in PDP-8/m Troubleshooting. Check the H724 supply first, with PDP-8/e Power Supply Guide.

DEC's method

DEC lists seven phases: preliminary investigation, system troubleshooting, logic troubleshooting, circuit troubleshooting, repair, validation and a log entry. Start with the shortest MAINDEC that shows the fault, not with a large system program, and reduce the failing program to a short loop. Duplicating the loop in a different area of memory shows whether the fault depends on the memory location, which separates memory faults from processor faults. Disconnect any peripheral the test does not need: faults in equipment that sends or receives data, or a wrong connection, can look like a computer fault.

Diagnostics

Acceptance tests (DEC-8E-HMM1A, Table 2-4)

Program	MAINDEC	SA / SR	Accept time
Instruction test 1 and 2	8E-D0AA	200 / 7777	3 min
Instruction test 2	8E-D0BA	200 / 0000	3 min
Adder test	8E-D0CA	200 / 0000	35 min
Random AND, TAD, ISZ, DCA, JMP and JMP-JMS tests	8E-D0DA, D0EA, D0FA, D0GA, D0HA, D0JA	200 / 0000	3 min each
Basic JMP-JMS test	8E-D0IA	200 / 0000	3 min
Memory address test	8E-D1EA	200 / 0000	5 min
Memory checkerboard	8E-D1AA	200 / 0000	15 min
Teletype control test	8E-D2AA	200 / 0000	40 min

Processor

Processor troubleshooting (DEC-8E-HMM1A, Table 4-3)

Symptom	Likely cause	Module
An OMNIBUS signal is always low	Shorted diode from ground to the OMNIBUS on the bus loads	M8320
Cannot start (run light always off)	Missing MEM START L (a 300 to 500 ns pulse each time CONT, DEP or EXAM is pressed), or POWER OK delayed or stuck	M8330, KC8-E
Cannot change major states	CPMA LOAD L missing; MA, MS LOAD CONT L grounded	M8310
Cannot change any memory location	MB LOAD, memory direction stuck low, INHIBIT stuck low, WRITE L or SOURCE stuck low (M8330); FIELD L high (G104)	M8310, M8330, G104
Memory data does not reach the MB	Memory direction stuck high, no MB LOAD L, no TIME STROBE, WRITE L stuck high	M8330, M8310, G104
Loaded address differs from the switch register	LA ENABLE always high	KC8-E
MA changes to a wrong value on LOAD ADDR, DEP or EXAM	EN0, EN1, EN2 or LEFT L, RIGHT L, TWICE L at wrong levels	M8310
LOAD ADDR enters all ones in the MA	DATA T always high	M8310
MA decrements on EXAM or DEP	DATA F always low	M8310
MA bits 0 to 4 zeroed on EXAM, DEP, CONT or JMP	PAGE Z L high	M8310 (M8300 if only one bit)
CPMA, MB, PC or AC will not increment	CAR IN L always high to the adder	M8310
ISZ never skips	SKIP flip-flop will not clear; no carry out; no overflow	M8310
Teletype input loads into MA and PC	C2 L always low	M8320, M8310, M8330

Every OMNIBUS pin with the same letter is connected in every slot, and a signal can have more than one source. MD6, for example, is on pin BM1 of every slot and can be driven by either the M8300 or the G104. DEC's Appendix B lists the source and destinations of every signal in the basic computer.

Memory

Memory data errors (DEC-8E-HMM1A, Table 4-4)

Symptom	Cause	Module	Check
One bit stuck at 1 or 0	Inhibit driver	G104	Collectors of the 2007 transistors
One bit stuck at 1 or 0	Sense amplifier	G104	E31–E42 pin 8
Random 1s or 0s	Time strobe	G104	E9 pin 3
Random or all 0s	X/Y current low	G227	About 5.3 V between +5 V and pin JU2
Random or all 1s	X/Y current high	G227	As above
Random or all 1s	Slice voltage low	G104	About −5.3 V between ground and test point DA1
Random or all 0s	Slice voltage high	G104	As above
Random or all 1s	Inhibit current low	G227	−15 V supply
Random 0s	Inhibit current high	G227	−15 V supply

DEC's table prints about 5.3 V for both the current-control and slice checks; its Table 4-5 gives about 1.4 V at JU2 on the G227 and about −5.3 V at DA1 on the G104, and the current-control text gives about 3.5 V nominal between +5 V and JU2 at 25 °C. Check against a working board where possible.

Memory test point voltages (DEC-8E-HMM1A, Table 4-5)

Signal	Pin	Module	Approx.
Current control	HA1 / HV2	G104	1.2 V / 2.3 V
Current source	JU2	G227	1.4 V
Current source	FU2 / HU2	G104	0.25 V / 4 V
Current source	FA1 / FB1	G104	2.3 V / −4 V
Test points	DA1 / CB1 / DB1	G104	−5.3 V / 6 V / −6 V
Q17 emitter	–	G104	−4.8 V
Top of thermistor	–	Stack	2.5 V
Q18 collector	–	G104	−6 V
Q15, Q16 emitters	–	G104	+3 V
Q13 base	–	G104	+1.3 V

With power off and the stack removed, DEC gives these resistances: thermistor network about 150 Ω, a single thermistor out of circuit about 56 Ω, one bit's winding about 3 Ω (for example pins FA and FB for bit 6), and the FSA2501 diodes about 24 Ω forward and over 1 MΩ reverse. The memory troubleshooting flowchart leads, for faults common to many bits, to the G227 drivers or the diodes on the stack. Metal-can transistors on the memory boards have their case connected to the collector; keep probe grounds and neighbouring boards away from them.

Logic and circuit work

• Extend a suspect module with two double extender boards in its slot, and its H851 connector cables (BC08M) if it uses them. Use an IC pin extender for probing and keep the oscilloscope ground short.
• Within modules, unused gate inputs are held at +3 V.
• ICs are numbered from the contact edge of the board, starting at the upper right with the component side facing you and the handle in your left hand, increasing towards the handle and continuing row by row.
• DEC's ohmmeter guide for discrete parts: use the ×10 range, not the lowest or highest; a good diode reads about 20 Ω forward and over 1,000 Ω reverse; a good transistor reads open between collector and emitter both ways and 50 to 100 Ω forward on each junction.
• When soldering, heat-sink the leads, use a small iron with an isolation transformer, and remove solder with a desoldering tool or braid; never rap a board to clear solder.

Teletype

Most Teletype faults are in the 33 ASR keyboard or reader, the printer or punch, or the M8650 receiver/transmitter. Run the Teletype off line first: if it works locally, the fault is in the M8650 or the cable. If the printout and the punched tape agree and are both wrong, suspect the selector mechanism or the M8650; if the tape is right and the printout wrong, the fault is in the printer. The M8650 shifts bits from a crystal clock and needs no adjustment.

Related pages

• PDP-8/e
• PDP-8/e Power Supply Guide
• PDP-8/e General Maintenance
• PDP-8/e Documentation
• PDP-8/m Troubleshooting