Jump to content

SGI Origin 350 Capacitor Replacement Guide

From RetroTechCollection
SGI Origin 350 power supply removal
Taking out an Origin 350 power supply: (A) two screws and the screen cover, (B) the cover swung clear of the fan ribbon cable, (C) release button and green handle, (D) the supply sliding out (SGI User's Guide, Figure 6-22)

Capacitor notes for the SGI Origin 350. SGI's user's guide treats the power supplies as sealed customer-replaceable units and gives no component lists for them or for the compute module's boards.[1] No owner has published a capacitor list for the Origin 350 either, so the values in your module have to be read from the parts themselves. In the owners' repairs reported on the IRIX Network forum, the parts found faulty were a regulator chip, monitoring chips and a supply diode; new output capacitors, where they were tried, did not cure the fault.[2][3]

Warning: the supply runs from mains and its input capacitors hold a dangerous charge after the cord is pulled. Unplug the module, let the supply stand, and confirm with a meter that the input capacitors have discharged before touching anything inside it.

The power supply

[edit | edit source]

SGI rates each supply at 500 W from 110 or 220 V AC, with outputs at 12 V, 5 V and −12 V.[4] jan-jaap, a collector on the IRIX Network forum, identifies it as a rebadged Delta DPS-500EB with these outputs:[3]

Origin 350 supply outputs, as given by jan-jaap
Output Current
15 V bias 50 mA
−12 V 0.5 A
5 V 2.0 A
12 V 40 A

He takes the bias output to be the always-on supply for the L1 controller, and the module's other voltages to be made from the 12 V rail. Older revisions of the supply carry a manufacturing fault, the subject of Intel Technical Advisory TA-0674-5, in which the primary power factor correction diode fails; his own Origin 350's supply failed this way, and after he repaired it and others the fault did not come back.[3] Check that diode as well as the capacitors in any supply you open.

Regulators on the boards

[edit | edit source]

The interface board carries switching regulators for the module's lower voltages, and the IP53 node board carries three VRMs for the processor core, 1.5 V and 2.5 V supplies.[4]

  • weblacky, who repairs these boards, describes the output capacitors in this generation's regulators as surface-mount solid aluminium or polymer types. On the Tezro main board, which he says uses the same regulator design, he found 470 µF Panasonic solid polymer capacitors of about 25 mΩ ESR, and pointed to a newer series offered at 560 µF and about 14 mΩ.[5][2]
  • He expects the ESR of these parts to rise with age before they would ever leak, and suggested a raised ESR as one cause of rails that drift low after a few minutes. That remained his theory; he had not confirmed it.[5][2]
  • aSpartan111 put new Panasonic 220 µF 6.3 V capacitors on the outputs of the PCI 5V and 3.3 V regulators of his interface board. That did not cure his fault; a new LTC1629 regulator chip did.[2]
  • aSpartan111 also found a DS1780 monitoring chip on the PCI riser and an LTC3728 regulator making 1.8 V and 2.5 V.[2]
  • One owner's module had no processor VRM fitted; the VRM is a standard VR091B080CS part.[3]

See SGI Origin 350 Troubleshooting Guide for the full reports.

Recording your own module

[edit | edit source]
  1. Shut IRIX down, power the module down from the controller and unplug it. The base module has no power switch, and its L1 controller stays live while it is plugged in.[6]
  2. Take the supplies out from the front (see SGI Origin 350 General Maintenance); to reach the boards, pull the module forward to its safety latches and remove the ten cover screws.[1]
  3. Photograph both sides of every board before desoldering anything, with designators readable.
  4. For each electrolytic or polymer capacitor, write down the designator, capacitance, voltage rating, temperature rating, maker and series, size and polarity.
  5. Look for domed tops, leakage at the base, brown or crystalline residue, and corroded traces near capacitors.
  6. Measure ESR with the part out of circuit and compare it with the maker's figure for that series.
Record sheet
Board Designator Capacitance Voltage Temperature Series Size Condition
Power supply, primary
Power supply, secondary
2U interface board, regulators
IP53 node board VRMs
PCI riser

Replacing

[edit | edit source]
  • Fit the same capacitance and at least the same voltage rating, 105 °C parts, and low-ESR series wherever the original was a low-ESR type.
  • Replace solid polymer or solid aluminium parts with the same technology, at an equal or lower ESR.
  • Keep the original size where a heat sink, duct or neighbouring part leaves no room for a larger one.
  • Observe polarity; mark it on your photographs before removal.

After the repair

[edit | edit source]
  1. Refit the supply and close the screen cover without pinching the supply fan's ribbon cable.[1]
  2. Plug the module in. The supply LED should blink green, then turn steady green once the module is powered on; amber means a fault.[1]
  3. At the L1 prompt power the module up and run env. Every rail should be inside its warning limits; SGI counts 10% from nominal as a warning and 20% as a fault.[7]
  4. Measure the rails you worked on with a meter as well, to confirm that the monitoring chips are reading them correctly.
Selected rails from jan-jaap's working Origin 350 (L1 env output)
Rail Warning limits Fault limits Reading
1.8V 1.62 to 1.98 V 1.44 to 2.16 V 1.791 V
12V #2 10.80 to 13.20 V 9.60 to 14.40 V 12.000 V
3.3V 2.97 to 3.63 V 2.64 to 3.96 V 3.302 V
5V AUX 4.50 to 5.50 V 4.00 to 6.00 V 5.044 V
PCI 5V 4.50 to 5.50 V 4.00 to 6.00 V 4.966 V
PCI 3.3V 2.97 to 3.63 V 2.64 to 3.96 V 3.302 V
PCI 2.5V 2.25 to 2.75 V 2.00 to 3.00 V 2.509 V
IP53 VCPU 1.13 to 1.38 V 1.00 to 1.50 V 1.241 V
IP53 SRAM 2.25 to 2.75 V 2.00 to 3.00 V 2.470 V
IP53 1.5V 1.35 to 1.65 V 1.20 to 1.80 V 1.495 V

[3]

[edit | edit source]

References

[edit | edit source]
  1. ↑ 1.0 1.1 1.2 1.3 SGI, 007-4566-001, chapter 6, pp. 155-201. Customer-replaceable units, power supply LED (Table 6-1) and replacement, opening the module.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 "Origin 350 - PCI 5V missing", IRIX Network forum, posts by aSpartan111 and weblacky, September to December 2023, forums.irixnet.org/thread-4064.html. Interface board regulator repair, output capacitors, PCI riser parts.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 "o350 env chip locations", IRIX Network forum, posts by jan-jaap and vvuk, 28 February 2022, forums.irixnet.org/thread-3424.html. Supply identity and outputs, TA-0674-5, VRM part number, env output.
  4. ↑ 4.0 4.1 SGI, 007-4566-001, chapter 3, pp. 114-122. IP53 node board VRMs, interface board switching regulators, power supplies.
  5. ↑ 5.0 5.1 "O350 Voltage Fault:ATTN: 1.8V low fault limit reached @ 1.396V.", IRIX Network forum, posts by weblacky, November 2020 to March 2022, forums.irixnet.org/thread-2563.html. Regulator output capacitors on Tezro and Origin 350 boards.
  6. ↑ Silicon Graphics, SGI Origin 350 Server System User's Guide, 007-4566-001, June 2003, chapter 1, pp. 53-59 (File:SGI Origin 350 Server System User's Guide (007-4566-001).pdf). Powering off; the base module has no power switch.
  7. ↑ SGI, 007-4566-001, chapter 7, p. 206, Table 7-2. Voltage warning and fault thresholds.