PDP-8 Power Supply Restoration

DEC drawing RS-C-708, the Type 708 power supply of the original PDP-8 (F-87, p. 10-8)

The original PDP-8 runs from a single DEC Type 708 power supply in the base of the machine. It provides unregulated +10 V and −15 V for the logic from a ferroresonant transformer, a 40 V floating supply feeding two series-regulated supplies for the core memory's inhibit and read/write currents, and a 0 to 20 V marginal-check supply with its own meter and polarity switch. DEC documented it in the PDP-8 Maintenance Manual (F-87), with the schematic as drawing RS-C-708. Two detailed restorations of 708 supplies have been published: David Gesswein's on his own Straight-8, and the University of Iowa's on serial number 85.

Warning: the 708 holds 210,000 µF of filter capacitance on the −15 V supply and 105,000 µF on the +10 V supply, enough to keep the computer running through a 50 ms mains interruption. The ferroresonant transformer's resonating capacitor sits on a high-voltage AC winding, and DEC marks the supply's input capacitors C18 as 600 V DC parts. Unplug the machine and discharge the capacitor banks before working in the base.

Outputs

Type 708 outputs (DEC F-87, Table 9-3, p. 9-7)

Terminals	Output	Range	Max. current	Max. ripple (p-p)
Blue TAB	−15 V	−14.5 to −16.5 V	printed as 15 mA	700 mV
Red TAB	+10 V	+9.5 to +11.5 V	2 A	700 mV
Orange (+) to green (−)	Marginal check	0 to 20 V	2 A	700 mV
Orange (+) to green (−)	Inhibit (+35 V)	27 to 37 V	2 A	under 50 mV
Red (+) to blue (−)	Read/write (+35 V)	27 to 37 V	1.5 A	under 50 mV

DEC notes that the inhibit and read/write supplies are outside their regulating range if the ripple exceeds 50 mV. The 15 mA figure for the −15 V output is not a usable rating: Gesswein measured 9.7 to 9.9 A on that output with the computer connected, and found no corrected value in any manual. The +10 V and −15 V supplies have no adjustment, so DEC treats an output or ripple figure outside the table as a defective supply.

How it works

• A tapped winding on the ferroresonant transformer T1 feeds full-wave rectifiers for the unregulated +10 V and −15 V supplies.
• An independent winding with taps at 40, 42 and 45 V feeds diodes D1 and D2 and the 34,400 µF bank C1 to C4, giving 40 V floating for the memory supplies.
• Series pass transistors Q1 (inhibit) and Q2 (read/write) regulate that 40 V under the control of two G808 power supply control modules. Each G808 compares its output against a Zener reference, reduces the voltage as the memory array warms (via a positive-coefficient thermistor of 350 Ω at 25 °C in the array) and has adjustment potentiometer R16. The output range is 27 to 37 V.
• A G809 relay driver watches the −15 V line and energises relay K1 once it reaches −14 V. K1's contacts connect the inhibit and read/write supplies to the memory. If −15 V falls below −14 V the relay drops out and the OK line goes to ground, halting the processor.
• The marginal-check supply is a variable transformer T2, half-wave rectifier D3 and filter C5, switchable to either polarity and metered on the front of the supply.

Capacitor list

From DEC drawing RS-C-708. Values are in µF; DEC prints a voltage rating for only some positions.

Type 708 capacitors (RS-C-708, F-87 p. 10-8)

Reference	Value	Rating as printed	Circuit
C1, C2, C3, C4	8,600 µF	50 V	40 V floating memory supply (34,400 µF in total)
C5	35,000 µF	25 V	Marginal-check supply
C6 to C11	35,000 µF each, 25 % tolerance	not printed	−15 V supply (210,000 µF in total)
C12, C13, C14	35,000 µF each, 25 % tolerance	not printed	+10 V supply (105,000 µF in total)
C15, C16	950 µF	75 V	Inhibit (C15) and read/write (C16) regulated outputs
C17	Resonating capacitor, supplied with the transformer	not printed	Ferroresonant transformer T1
C18	Two 1 µF "bathtub" capacitors	600 V DC	Mains input
C19, C20	0.01 µF	not printed	Q2 and Q1 regulator circuits

The groupings follow DEC's text (210,000 µF on −15 V, 105,000 µF on +10 V, 34,400 µF on the 40 V supply) and the Iowa log, which reformed C6 to C11 as the −15 V bank. Check the parts in the supply in front of you: Gesswein found that some capacitors in his were soldered in with bleed resistors across their terminals.

Taking the supply out

1. In a cabinet machine, support the logic assembly before drawing it out on its slides. DEC warns that a PDP-8 in a radio rack can topple when pulled forward, and the Iowa team made a wooden prop to carry the extended computer while the rack sides were off.
2. Photograph and label every wire before removing the harness. Iowa removed most of the wiring harness to get the supply onto a bench.
3. Expect pop rivets on the back panel. Iowa drilled them out and fitted swage nuts so the panel can be removed again.
4. Remove any capacitors that are not soldered in, and clean their terminals. Gesswein used a wire brush in a Dremel on corroded terminals.

Reforming the electrolytics

Both restorations reformed the original capacitors before applying mains. Gesswein's method:

1. Use a bench supply with an adjustable current limit, set to the capacitor's rated voltage, with the limit around 20 mA.
2. Feed each capacitor through its own resistor (he used 4 kΩ) so the leakage of each can be read and the current shared between them. Keep within the resistor's power rating.
3. Watch the leakage fall. His best capacitors reached almost nothing; the worst stayed at 1.4 to 3.4 mA after hours at rated voltage.
4. Replace any capacitor that will not come down. Gesswein's leakiest parts began to bulge at the vents after two days at rated voltage, and he replaced every capacitor whose leakage stayed above 1 mA.
5. Reform capacitors that have bleed resistors across them in place, with the G808 and G809 control cards removed and the current limit 5 to 10 mA above the bleed current.

For comparison, Gesswein notes that current capacitors of similar value are specified at 4 to 6 mA maximum leakage after five minutes. The Iowa team reformed the banks over about five weeks; C15, one of the 950 µF 75 V output capacitors, would not reform and was replaced.

Connections, relay and transistors

• Clean and tighten every 0.25-inch quick-disconnect (Faston) terminal. On his PDP-8/I Gesswein found corroded Fastons dropping enough voltage under load to cause trouble. On the 708 he looked for unusual voltage drops along the load path with a meter, and for ripple on capacitor terminals with an oscilloscope, which found a slightly loose riveted terminal and one intermittent Faston.
• Inspect relay K1 (Potter & Brumfield KRP14DG, 12 V DC coil) and its socket. Gesswein's had corroded pins and a damaged contact; he cleaned the socket and fitted a new old-stock relay.
• Q1 and Q2 are socketed on the chassis. Gesswein reseated them with new heat sink compound, then found one bad during testing and replaced it with a DEC-labelled transistor of the same type.
• Replace the AC resonating capacitor. Gesswein bought a new one because old ones tend to fail, can burst and may contain PCBs; Iowa also replaced C17.

Faults found on restored supplies

Faults reported on Type 708 supplies

Symptom	Cause found	Source
Ripple on the 40 V memory supply not symmetrical between half-cycles	The two rectifier diodes were wired to different transformer taps (40 V and 45 V); moving both to the 40 V tap made the ripple symmetrical	Gesswein
Read/write output does not change when the thermistor input is varied	Shorted pass transistor and an open resistor on a G808 control card	Gesswein
Relay chatters at power-on and power-off	Design fault in the G809 revision B; the schematic in F-87 shows a later corrected version	Gesswein
Marginal-check meter reads high	Ageing of the meter's internal series resistor; Gesswein added a series resistor (meter about 15 % high), Iowa fitted a trimmer (about 10 % high)	Gesswein, Iowa
Neon indicator on the back flickers or does not light	Aged neon; Gesswein names the VCC 1050A1 as a close match for the original Industrial Devices 1020C55	Gesswein
Power OK never asserted	G808 and G809 cards in the wrong slots	Iowa

Testing

Test the supply on resistive dummy loads before reconnecting the computer. Gesswein used potentiometers in place of the memory-stack thermistors and loaded the supply to about half its rating; Iowa built a dummy load and measured every output under it.

Gesswein's 708 measurements on resistive load (119 to 122 V input)

Output	Voltage	Load	Ripple (p-p)	Ripple frequency
+10 V	10.67 V	1.4 A	192 mV	Twice line
−15 V	−14.95 V	7.5 A	190 mV	Twice line
Marginal check	−9.97 V	1 A	436 mV	Line
Read/write	30.5 V	0.9 A	8 mV	Twice line
Inhibit	30.9 V	0.9 A	8 mV	Twice line
40 V memory capacitors	41.2 V	1.8 A	303 mV	Twice line

Then check the supply against DEC's limits:

1. Measure each output with a multimeter without disconnecting the load, and the ripple with an oscilloscope, against Table 9-3.
2. Check the marginal-check supply: with all marginal-check switches at normal, connect the meter between +10 MC and ground, swing the control through its range in each polarity, and confirm that the panel meter and the multimeter agree within 1 V. Return the control fully anticlockwise and the toggle switch to its centre off position.
3. Set the inhibit and read/write supplies only through DEC's memory current check, against the currents on the memory array label, with the machine warmed up for about an hour (see PDP-8 General Maintenance).
4. After replacing any filter capacitor, repeat the ripple check for that section, as DEC instructs.

Related pages

• PDP-8
• PDP-8 General Maintenance
• PDP-8 Troubleshooting
• PDP-8 Documentation
• Capacitor Failure Symptoms