IBM PS/2 Model 80 Capacitor Replacement Guide

This guide covers capacitor diagnosis and replacement on the IBM PS/2 Model 80 (machine type 8580, all submodels). It covers the planar, the tower power supply, the diskette drive and the fixed-disk subsystem. The Model 80 shares its chassis and its power-supply form factor with the Model 60 and the Model 65 SX, so most of what follows applies to those too.

The Model 80 floor-standing tower. The power supply sits at the top of the chassis, above the drive support structure, and its fan blows across the planar.

Read this first. IBM published no capacitor values for any part of the 8580: the technical reference, the hardware maintenance manual and the parts catalogues all work at FRU level, and the planar schematics are net-level without a valued parts list. Every value below comes either from a reverse-engineered schematic of a physical supply or from a documented community teardown, and is labelled as such.

Safety Warning

The tower supply is mains-rectified. Its two primary reservoir capacitors are connected in series across the rectified mains so the unit can accept 220 V input, with a chain of balancing resistors across them; between them they store a great deal of energy and they do not self-discharge quickly. Before any work inside the supply:

1. Power off and unplug the mains lead.
2. Discharge each reservoir capacitor individually through a 1 kΩ / 5 W resistor across its own terminals — discharging across the pair is not enough if one has drifted.
3. Confirm with a meter that both read near zero before you touch the board.
4. Do not probe a powered primary side without an isolation transformer.

There is no CRT in the system unit; the matching IBM displays are serviced separately. See CRT Discharge Procedure if you are working on one.

The Planar Does Not Need Recapping

The 8580 planar does not carry surface-mount aluminium electrolytic capacitors. Photographs of the Type 1 and Type 2 planars show through-hole dipped tantalum beads — the orange and yellow teardrops scattered between the logic — together with surface-mount ceramic chips for high-frequency decoupling. There are no electrolytic cans on the board to leak.

This matters, because the standing advice for PS/2 machines is often "recap the planar on sight". On this board there is nothing to recap, and lifting dozens of healthy tantalums off a thirty-five-year-old multilayer board is a reliable way to turn a working machine into a donor.

Two claims in particular should not be applied here:

• There is no published designator-to-value map for this planar, and the values are not on the silkscreen. The "10 µF / 16 V marked 106 16V" tantalum quoted for IBM PC, XT and AT motherboards is documented for those through-hole boards only and has never been shown to apply to a PS/2 planar.
• IBM factory rework wire is present on many 8580 planars. The yellow wires visible on photographs of the 20 MHz board are how it left the factory, not damage.

What to do instead

The commonest planar-level fault on this machine is not electrical at all. The supply's fan draws household dust directly onto a sensitive area of the board, and that dust is often conductive; current leaking between IC pins produces lost configuration, a clock showing nonsense after a long power-off, and disk read/write errors. The affected area holds the 8042 keyboard controller, the MC146818A real-time clock, the diskette controller, the NS16550 UART, the RTC crystal and sometimes an 8259A. Strip the machine, blow the supply out somewhere you do not mind making filthy, then brush and vacuum the board and check between the IC pins.

If you suspect a shorted tantalum

Tantalums fail short far more often than open, and a shorted decoupling part drags its rail down and can make the supply latch off.

1. Unplug and discharge the supply.
2. Set a multimeter to its lowest resistance range or to diode test and probe from ground to the rail side of each part. A good one settles high after a brief charging pulse; a shorted one sits near 0 Ω.
3. Everything on a rail is in parallel, so a low reading indicts the rail, not the part. Lift one end and measure again before replacing anything.
4. Replace with an equivalent tantalum or a low-ESR ceramic of the same value and equal or higher working voltage, read from the part you removed.

Power Supply

Three ratings were fitted across the tower family, and IBM's own technical reference lists heat outputs for all three. Only one of them has been documented at component level.

Supplies fitted to the 8580

Rating	Identification	Component-level data
225 W	IBM P/N 72X6665, REV A58491, PEC 3978; also seen as FRU 15F6548, and as P/N 90X9109 / FRU 15F6551 on Schrack-built units. Red switch	Reverse-engineered schematic available for the 72X6665 unit
242 W	Later 8580s, including the 25 MHz models	None published
250 W	FRU 57F1600 — the part the October 1994 maintenance manual lists as the replacement across 60/65/80	None published

More than one manufacturer built these. Al Savage's teardown of a failed 225 W unit found a Schrack EP071063-A with screwed case halves, while another supply carrying the same IBM part numbers was an Italian Plessey with riveted halves. The internals differ. Read values off your own board.

How the 225 W supply works

Knowing the topology tells you which capacitors matter. Eric Schlaepfer's reverse-engineering of the 72X6665 establishes it as a resonant converter with a separate auxiliary bias supply — effectively two supplies in one box:

• Inrush into the two series reservoir capacitors is limited by thermistors RT1 and RT2 with resistors R117 and R118, and the pair is balanced by the resistor chain R100–R114.
• The bias supply uses a triac preregulator that chops part of each mains cycle, so the controller's supply rail sits at roughly the same voltage on 120 V or 220 V input.
• All three outputs are monitored for over- and under-voltage. An over-voltage latches the supply off; an under-voltage simply de-asserts Power Good.
• The front-panel power LED is driven by Power Good. The hard-disk LED is driven by the planar, not by the supply.

The practical reading: a tower that clicks and dies is usually something shorting a rail, not a dead supply; a machine whose power light never comes on has a supply that is running but not meeting its own under-voltage thresholds; and a supply that latches off and stays off until the switch is cycled is doing what it was designed to do.

225 W supply — electrolytic capacitors

These values are read from the reverse-engineered schematic of the 72X6665 / PEC 3978 unit, hosted here: File:IBM PS2 225W Power Supply 72X6665 Reverse-Engineered Schematic (Tube Time 2024).pdf. They are not an IBM specification, and they apply to that unit only — not to the 242 W or 250 W supplies, and not necessarily to a Schrack- or Plessey-built 225 W. Check each part against your own board before ordering.

Electrolytics in the IBM 225 W supply, P/N 72X6665

Designator	Value	Position
C5, C6	1000 µF / 200 V	Primary reservoir pair, in series across the rectified mains, balanced by the 220 kΩ chain. These are the lethal ones.
C16, C40	4700 µF / 10 V	+5 V output reservoir, ahead of the output chokes
C41	470 µF / 35 V	+5 V output, after the chokes
C18	4700 µF / 16 V	+12 V output reservoir
C17	470 µF / 35 V	−12 V, ahead of the LM7912 regulator
C19	47 µF / 25 V	−12 V, after the regulator
C20	2200 µF / 16 V	Preregulator / auxiliary bias supply
C13, C14	22 µF / 25 V	Bias and control
C34, C56, C57	10 µF / 50 V	Control and reference

Note what this table does not say. The +5 V reservoirs are 4700 µF rated at 10 V, not the 2200 µF / 16 V often quoted, and the primary reservoir is a series pair of 1000 µF / 200 V, not a single can. Anyone who tells you the primary bulk in this supply is a single 220 µF / 200 V part has not looked at one.

The mains-side non-electrolytics on the same sheet include 1 µF / 250 V parts and C26 at 0.1 µF / 500 V, with 560 pF line-to-earth parts at C3, C4, C52 and C53. These are safety-critical positions — replace them only with correctly rated X- or Y-class safety capacitors, never with ordinary film parts. Any RIFA-branded part with crazing in its epoxy should be replaced on sight. Several designators on the sheet are marked "?" because the author could not read the original part; those are yours to read off the board.

A documented failure

Al Savage's 8580 died with a chirping supply. Disconnecting the outputs did not change the symptom. The cause was a shorted 3300 µF / 16 V output filter capacitor — one of an identical pair, the upper one — on a Schrack EP071063-A board with no silkscreened component locations. He also found a cold solder joint on the solder side.

That value does not appear on the Tube Time schematic of the 72X6665, which is exactly the point: these are different physical designs sharing IBM part numbers. A chirping or ticking supply with a symptom that does not change when the outputs are disconnected is an internal fault, and the output filter capacitors are the first place to look.

Recapping the supply

1. Remove the rear drive carrier from the drive support structure, unplug any 4-pin drive power leads and the planar power plug, take out the three screws holding the supply, and draw it out of the top of the chassis.
2. Discharge each reservoir capacitor individually and confirm with a meter.
3. The case screws are security Torx (a pin in the centre of the drive). On a Schrack unit, disconnect the short ground lead routed through the top cover before lifting it; the PCB is notched so the cover comes straight up after sliding it about an inch.
4. Photograph the board from both sides before touching anything, and write down every value, working voltage, diameter and polarity.
5. Replace like for like: same capacitance, equal or higher working voltage, 105 °C rating, low-ESR where the original was low-ESR.
6. Replace mains-side X and Y capacitors only with correctly rated safety parts.
7. Bench-test the outputs under a modest resistive load before refitting, against IBM's service limits on the maintenance page.

Before you open it at all: about half of failed Model 80 supplies failed on a broken power switch, not on the electronics — the red switches more often than the white. The same switch was used in IBM 327x and 525x terminals and was a stocked spare.

A crackling noise from a machine that is switched off is not automatically a capacitor. It has been traced to the switch itself (the front switch cover acts as an echo chamber and throws the sound to the rear), to an insulation fault in the AC line filter, and to a loose mains lead.

Diskette Drive Logic Board

This is the one assembly on the machine with a documented capacitor fault and documented values.

The 8580 used two interfaces. The 25 MHz -Axx machines use the 34-pin style drive; every earlier model uses the card-edge style, P/N 90X6766 with FRU 72X8523. Both were built by Mitsubishi, ALPS and others to a common IBM specification, and all take power through the interface with no separate power connector. 2.88 MB drives were released for later PS/2 systems and should not be used here.

Al Savage found that all three of his 8580 diskette drives would not read, and that the cause was surface-mount electrolytics that were open-circuit or very high-ESR. On the Mitsubishi MF355W-99M3 (P/N 90X6766, FRU 72X8523, EC A58560, dated August 1989) he identified and marked five parts:

Failed surface-mount electrolytics, Mitsubishi MF355W-99M3

Value	Quantity	Position
22 µF / 16 V	2	Beside the card-edge connector, at the left of the logic board
1 µF / 50 V	1	Centre of the logic board
0.22 µF / 50 V	2	Centre of the logic board, to the right of the 1 µF

Revisions differ. Another account of the same family replaced two parts with 10 µF; the 2.88 MB drives are different again. Read the values off your own board.

Alan Douglas, who did this work on 8580 drives, used surface-mount electrolytics of the same type for the 22 µF and 1 µF positions; 0.22 µF electrolytics are hard to obtain, so ceramics went in there, and tantalum chips are a reasonable substitute elsewhere. Through-hole parts soldered to the surface-mount pads work if nothing else is to hand, at the cost of appearance and of standing proud enough to be sheared off later.

Removal is the awkward part: the pads are small, and leaking parts contaminate their own joints so they refuse to reflow and the leg pulls free before the solder melts. Cutting the can in two with flush cutters, removing the rubber seal and then working on the exposed stubs from above is far less likely to tear a pad than pulling on an intact part.

Mitsubishi drives in these machines are specifically noted for capacitors that produce 6xx electronic errors at POST.

Faults that look like capacitor faults and are not

• 162 at POST with the drive otherwise dead. During POST the heads are driven off track 0 and back within a fixed window. A sticking head carriage, a dirty track-0 photo-sensor or a detached head damper misses it and the drive is dropped from the configuration. ALPS drives lose the damper pad glued to the head assembly and have a much higher error rate than the Mitsubishi drives; the Mitsubishi 355 collects dirt around the read/write head actuator spindle.
• Speed or timing complaints blamed on a belt. There is no belt. Mitsubishi's own specification for the MF355C states a DC brushless direct-drive motor coupled straight to the spindle at 300 rpm.
• Dirt on the planar's diskette controller. The controller sits directly under the power-supply fan, collects dirt and leaks current between pins, producing 6xx errors that look exactly like a dead drive. Clean that area before condemning a drive.

Fixed Disk Subsystem

The Model 80's fixed disks are conventional ST-506, ESDI or SCSI units on a separate adapter card, not the integrated drive-plus-controller assemblies used in the desktops. No capacitor list has been published for the drives or for the adapters.

Check the obvious before suspecting electronics:

• 1780 on an 8580-041 — IBM's own retain tip, record H064300, says to verify that the fixed disk cables are installed properly. Cable orientation is the documented first cause of this code on this machine.
• A 314 MB drive on a -311 that will not come ready is covered by IBM ECA 040, which attributes it to a fault in the drive logic board.
• A drive that will not spin up after long storage — stiction. Unplug the unit, shake it vertically, then rotate it sharply a few times and try again. Whatever frees it, back up immediately.
• 104xx codes. 010455 and 010456 are controller errors, 010462 a controller seek error, 010464 a drive read error and 010454 a sector-buffer test error.

The code often quoted alongside 10455 as an "ESDI controller error" — 10463 — does not exist in IBM's code tables. The range runs 010462, then 010464.

Memory Cards

The 8580's planar memory sits on IBM's own memory cards in two connectors on the board, and no capacitor data has been published for them. Memory errors here are far more often a card, a seating or an ECA problem than a capacitor problem:

• 00011000 is a system-board parity check: remove the card in J15 or J8; if the problem goes, replace that card, otherwise try the one in J16 or J9.
• 00021500 indicts the card in J15 or J8; 00021600 and 00022100 the card in J16 or J9.
• Intermittent 110 during POST, diagnostics or applications — IBM retain tip H006554 notes that Advanced Diagnostics loads into the first 256 KB and therefore cannot stress-test it. Swap the two cards and re-run the memory test.
• OS/2 Trap 0002 or NMI on a 25 MHz machine with 8 MB fitted — ECA 051. Examine the modules at U24 and U34 near J8 and J9; if they are marked FCT841 the board qualifies for replacement. Any other marking is good, and anything printed on the board near the module should be ignored.

After Working on the Machine

1. Bench-test the supply on its own, under load, against IBM's service limits.
2. Refit and start from the Reference Diskette, with the diskette in the drive at power-on.
3. Run Automatic Configuration if any adapter changed, then Advanced Diagnostics (Ctrl-A).
4. Check that the power light comes on — it is driven by Power Good.
5. Run a surface scan if you worked on the fixed-disk subsystem, and format and verify a diskette if you worked on the drive.
6. Leave a long memory test running.

If a fault appears immediately after a recap, check the polarity of every part you fitted before suspecting anything else.

When Not to Recap

A Model 80 that starts reliably, POSTs clean, holds its configuration and reads and writes diskettes does not need its planar touched — there is nothing on it to recap.

Do act if you see:

• Fluid, crust or crystalline residue near a capacitor on a drive or supply board, or discoloured, grey or rough solder at its pads.
• A rail outside IBM's service limits under load.
• The power light staying off, or the supply latching off and needing the switch cycled.
• Diskette read/write failures or 6xx errors on a mechanically sound drive.
• A supply that whines, chirps, smells, or will not start when warm.
• A decoupling capacitor confirmed shorted out of circuit.

Related Pages

• IBM PS/2 Model 80
• IBM PS/2 Model 80 Maintenance Guide
• IBM PS/2 Model 80 Troubleshooting Guide
• IBM PS/2 Model 60 Capacitor Replacement Guide — same chassis and same supply family
• IBM PS/2 Model 70 Capacitor Replacement Guide — the desktop sibling, with a different and undocumented supply
• Capacitor Failure Symptoms
• Battery Explosion, Capacitor or Corrosion Damage
• Floppy Drive Repair
• Recommended Tools