Jump to content

PDP-11/34 Power Supply Restoration

From RetroTechCollection
An H744 +5 V regulator after cleaning, showing the two screw-terminal output capacitors and the large input capacitor (Jonathan Chapman, Glitch Works)

A PDP-11/34 in the 10½-inch BA11-K box is powered by the H765 power system: a chassis across the rear of the box carrying a large transformer, an AC power control, a power line monitor board and up to four plug-in regulators.[1][2] The smaller BA11-L box uses the H777 supply instead, which is not covered here.[3] The H765 and its regulators are shared with the PDP-11/04 and with many expansion and memory boxes, so restoration reports on those machines apply here too.[2][4]

Warning: the H765 runs from mains, the AC power control and the cap/MOV board are at mains potential, and the large capacitors in each regulator hold their charge after switch-off. Unplug the box and discharge the capacitors before working on it.[4][5]

What is in an H765

[edit | edit source]

DEC's power system description lists the parts of the basic H765 chassis:[1]

H765 basic chassis (EK-1134A-TM-002, note to Table 1-2)
Assembly DEC number
AC input box 70-09811-1 (120 V) or -2 (240 V)
AC power control board 54-10993
Power distribution board 54-10864-0, or -YA in the 11/34A
AC power cord 70-12500
+15 V DC power line monitor 54-11086
Transformer assembly with fans 70-11486

[1] The transformer steps the mains down to about 28 V AC for the regulators; the regulators rectify it themselves.[4] The 5411086 makes +15 V, the line-time-clock signal and the AC LO and DC LO signals.[4][6] DEC's print set also lists a cap/MOV board (54-11689) on the transformer; one restorer found it to carry CDE 0.1 µF 220 V AC capacitors across the line in parallel with metal-oxide varistors.[7][8]

Regulators

[edit | edit source]

In the BA11-K, regulator slot 1 holds an H745 (−15 V), slots 2 and 3 hold +5 V regulators, and slot 4 may hold an H754 or, with battery backup, an H785.[2][1] Regulator 2 feeds the processor backplane and regulator 3 all the other backplanes.[1]

Regulators used in the H765
Regulator Output Rating Used in
H744 +5 V 25 A; DEC tolerance ±5%, 200 mV p-p ripple PDP-11/34 slots 2 and 3
H7441 +5 V 32 A PDP-11/34A slots 2 and 3; needs the 54-10864-YA distribution board
H745 −15 V 10 A; ±5%, 450 mV p-p ripple Slot 1
H754 +20 V and −5 V 8 A and 1 A Slot 4 when fitted
H785 +5 V, +15 V and −15 V "B" supplies for MOS memory refresh – Slot 4 in an 11/34A with H775 battery backup

[1][9][10] DEC's PDP1134A regulator configurations, each with one H745 and two H7441s, are 70-13323-00 (120 V), -01 (120 V, with H785), -02 (120 V, with H754), and -03, -04 and -05 for 240 V in the same pattern.[1]

The H744 and H7441 +5 V regulators

[edit | edit source]

Each regulator takes about 20–30 V AC from the transformer, rectifies it with its own bridge, filters it, and drops it to +5 V with a buck converter: a large switching transistor and choke controlled by an LM723 voltage regulator wired with positive feedback, so the switch turns on below one threshold and off above another.[11][4][12] The output therefore carries a sawtooth ripple, which is normal; DEC allows 200 mV peak to peak.[12][10] AC enters on connector pins 6 and 7, +5 V leaves on pins 2 and 5 (and pin 1 on the H7441), and ground on pins 3 and 4.[11]

H744 electrolytic and tantalum capacitors (DEC parts list, April 1974 drawings)
Ref Value as printed DEC part no.
C1 31K µF 50 V (31,000 µF), input filter 10-10858
C8, C9 6K µF 10 V (6,000 µF), output 10-10704
C2 15 µF 20 V 10% 10-04812
C6 2.2 µF –

[5] The parts list also gives F1 as a 15 A fuse.[5] Rob Jarratt describes the output capacitors as screw-terminal 6,000 µF 10 V parts, agreeing with DEC's list. Exact replacements are scarce; on the cctalk list Matt Burke reported using Mouser part 80-PEH169HA460AQU2, and others suggested a higher-voltage screw-terminal part or snap-in capacitors with wires soldered on, noting that ripple current rating matters as much as voltage and that the original parts had wide tolerances.[13]

Faults reported by restorers

[edit | edit source]
  • Bridge rectifier. Mattis Lind, who repaired at least ten H744-type regulators for the Update Computer Club, found the bridge rectifier the commonest fault; one had blown its metal housing apart and burned an AC input contact.[12]
  • Whining regulator. Rob Jarratt traced a whining H744 to output capacitors whose ESR was high and unstable. Matt Burke's scope traces of a noisy regulator showed the switch transistor turning partly on, and Jon Elson read the switching frequency as about 5 kHz, low enough to be audible.[13][14]
  • Indicator lamp. The indicator lamps in the regulators burn out. Replacements suggested on cctalk were the CM7381 lamp or an LED with a series resistor.[12][4]
  • Crowbar. DEC's field guide describes a regulator that has crowbarred as reading about 0 V. Power the machine down (the crowbar will not reset otherwise), turn the regulator's adjustment pot half a turn anticlockwise, power up and readjust while watching the backplane voltage on a scope. If it crowbars again, something is pushing the output high: DEC names poor connections in the power harness connectors, a loose wire shorting backplane pins, and bent backplane pins.[15]

Bench testing a regulator

[edit | edit source]

Restorers test the regulators out of the box with a separate transformer and a dummy load:

  • Jonathan Chapman used a 24 V AC, 8 A transformer, a small harness ending in a Mate-N-Lok connector and dummy load resistors.[4]
  • Mattis Lind used 30 V AC from a variac and a 0.4 Ω load, and checked that nothing but the load resistor got hot.[12]
  • Rob Jarratt reformed the three large capacitors first, then ran an H744 from a 20 V DC bench supply into a 1 Ω load (5 A), drawing just under 2 A from the bench supply.[16]

The 5411086 power line monitor

[edit | edit source]

The 5411086 takes its AC through two contacts of its edge connector. Its fuse is after the bridge rectifier and bulk filter capacitor of the +15 V section, so a shorted bridge is unprotected and can burn the AC traces at the connector.[4][6]

5411086 with burned AC input traces after its bridge rectifier shorted (Jonathan Chapman, Glitch Works)

Chapman's repair of this failure in a BA11-K:[4]

  1. Bench-test the regulators to clear them, then pull the sub-chassis with the 5411086 and the AC control.
  2. Replace the shorted 35 A bridge. The original is no longer made; Vishay GBPC3506W-E4/51 (35 A, 600 V) fits with three #6 flat washers under it to bring it to the height of the shared heat sink.
  3. Replace the 500 µF output capacitor (he used a low-ESR Nichicon part, since the output is switched).
  4. Cut back the damaged edge-connector traces and solder a short AC pigtail to the board, with an in-line 15 A fast-blow fuse.
  5. Test with a 24 V AC transformer on a variac and a 15 Ω load on +15 V, and check the line clock and that AC LO and DC LO change state at the right line voltages.

The Computer History Wiki recommends the same in-line fuse for every H765, noting that the H7420 has a fuse in its harness for this board and the H765 does not.[6] On reassembly, check how the board is held: one restorer found his chassis had only a slot for it, where another's had screw tabs at all four corners.[17]

AC power control

[edit | edit source]
AC power control board with its replaced 50 µF capacitor (Jonathan Chapman, Glitch Works)

The 50 µF electrolytic in the AC control box dries out with age. As it does, ripple on the supply to the AC contactor rises until the contactor buzzes instead of closing at power-up. It is easy to replace with the H765 apart and awkward once the machine is reassembled.[4]

Distribution board and backplane jumpers

[edit | edit source]

The 54-10864-0 distribution board was designed for 25 A H744s; the -YA version carries 32 A from the H7441s and adds a +5B bus bar for MOS memory with battery backup.[1] In a 10½-inch box without battery backup, the MOS refresh buses +5B, +15B and −15B on the backplanes must be jumpered to the normal rails with #20 insulated wire, or they float.[1] With a -YA board, put the +5 V to +5B jumper only on the CPU backplane: DEC warns that a second +5 V to +5B jumper anywhere else parallels the two +5 V regulators. This is most likely after an FP11-AU upgrade, when every backplane already has the jumpers fitted.[1]

Restoring a stored H765

[edit | edit source]
  1. Unplug the box and discharge the regulator capacitors.
  2. Remove the regulators. Clean them; their clear covers collect dust.[4]
  3. Check each regulator's fuse and bridge rectifier, then reform or test its large capacitors and bench-test it under load as above.[12][16]
  4. Inspect the 5411086 edge connector and bridge rectifier; add the in-line AC fuse.[4][6]
  5. Replace the 50 µF capacitor on the AC power control board.[4]
  6. Check the backplane +5B, +15B and −15B jumpers against the distribution board version.[1]
  7. Refit the regulators in their slots, power up with only the processor, terminator and memory installed, and measure +5 V and −15 V at the backplane with a scope, ground lead on the backplane, against ±5%.[10]
[edit | edit source]

References

[edit | edit source]
  1. ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10 Digital Equipment Corporation, PDP11/34A Power System, EK-1134A-TM-002, second edition July 1977, sections 1.4 to 1.6 and 2.3 to 2.7, Tables 1-2, 1-3 and 2-1, pp. 1-5 to 2-12 (File:PDP-11-34A Power System Description (EK-1134A-TM-002).pdf).
  2. ↑ 2.0 2.1 2.2 "H765 Power System", Computer History Wiki, gunkies.org/wiki/H765_Power_System.
  3. ↑ Digital Equipment Corporation, PDP-11/34 System User's Manual, EK-11034-UG-001, section 1.2.7, p. 1-9 (File:PDP-11-34 System User's Manual (EK-11034-UG-001).pdf).
  4. ↑ 4.00 4.01 4.02 4.03 4.04 4.05 4.06 4.07 4.08 4.09 4.10 4.11 4.12 Jonathan Chapman, "Repairing the PDP-11/10 Power Supply", Glitch Works, 5 June 2019, users.glitchwrks.com. PDP-11/10S in a BA11-K with an H765.
  5. ↑ 5.0 5.1 5.2 Digital Equipment Corporation, H744 regulator engineering drawings, April 1974, assembly parts list (File:H744 Regulator Engineering Drawings (1974).pdf).
  6. ↑ 6.0 6.1 6.2 6.3 "DEC Power Line Monitor", Computer History Wiki, gunkies.org/wiki/DEC_Power_Line_Monitor.
  7. ↑ Digital Equipment Corporation, H765 power supply print set, revision R, September 1977, drawing directory (File:H765 Power Supply Print Set (Rev R, 1977).pdf).
  8. ↑ Fritz Mueller, "DEC H765 line interference caps", cctalk mailing list, 22 December 2020, mail-archive.com.
  9. ↑ DEC, EK-11034-UG-001, appendix C, p. C-5. H744, H745 and H754 ratings.
  10. ↑ 10.0 10.1 10.2 Digital Equipment Corporation, PDP-11 Mainframe Troubleshooting Guide, December 1976, DECAID PWRB-1, regulator specifications and backplane voltage checks (File:PDP-11 Mainframe Troubleshooting Guide (December 1976).pdf). Written for the PDP-11/05 to 11/45; the regulators are the same types.
  11. ↑ 11.0 11.1 "H744 +5V Regulator", Computer History Wiki, gunkies.org/wiki/H744_+5V_Regulator.
  12. ↑ 12.0 12.1 12.2 12.3 12.4 12.5 Mattis Lind, "Re: Testing a H744 Regulator on The Bench", cctalk mailing list, 30 December 2021, mail-archive.com.
  13. ↑ 13.0 13.1 "Source for replacement caps in H744 regulators", cctalk mailing list thread, 6 January 2022: Rob Jarratt, Matt Burke, Dave Wade and Jon Elson, mail-archive.com and replies msg57871, msg57873, msg57874, msg57881.
  14. ↑ Jon Elson, "Re: Source for replacement caps in H744 regulators", cctalk, 6 January 2022, quoting Matt Burke's scope traces, mail-archive.com.
  15. ↑ DEC, PDP-11 Mainframe Troubleshooting Guide, DECAID PWRB-3, crowbar problems.
  16. ↑ 16.0 16.1 Rob Jarratt, "Testing a H744 Regulator on The Bench", cctalk mailing list, 30 December 2021, mail-archive.com.
  17. ↑ Fritz Mueller and Paul Birkel, "securing 15V pwr monitor board in H765 supply", cctalk mailing list, 18–19 January 2021, mail-archive.com and msg53122.
Retrieved from "https://wiki.retrotechcollection.com/index.php?title=PDP-11/34_Power_Supply_Restoration&oldid=21546"