DECstation 5000 Troubleshooting Guide

An H7878 power supply opened up; the three fans that cool the whole system unit sit in the removable panel (Rob Jarratt)

Fault-finding on the DECstation 5000 Model 200 series, from DEC's Model 200 Maintenance Guide (EK-370AA-MG-002) and Model 200 Series Maintenance Guide (EK-PM38C-MG-002), with power supply repair notes from Rob Jarratt's H7878 and from owners of the Model 100 series H7826. DEC's procedures stop at the field-replaceable unit: system module and chassis, CPU module, memory module, power supply. Below that level the notes are from community repairs.

Order of work

DEC's troubleshooting overview works in this order:

1. Power. If the green LED on the power supply does not light, get power to the unit before anything else.
2. Self-test. Read the code on the diagnostic LED array if nothing reaches the screen, or the error message if one does.
3. Configuration. At the console prompt (>>) type cnfg to see whether every module, the expected memory and the right SCSI IDs appear; cnfg followed by a slot number gives detail for one module.
4. Environment variables. printenv shows whether boot points at a working drive and how haltaction is set.
5. Tests and scripts. t slot/? lists the tests for a module and ls slot lists its scripts.

On the Model 200 the system module answers as slot 7, so its tests are run as t 7/test; TURBOchannel options are slots 0, 1 and 2.

Diagnostic LEDs

The eight red LEDs on the rear panel show the progress of the power-up self-test. On the Model 200 the four left-hand LEDs, read as you face the back of the machine, give the slot number of the unit that stopped the test.

Model 200: self-test does not complete (DEC Table 4-2)

Left LEDs (1 = on)	DEC's corrective action
None; test does not start	Check for loose or damaged connectors and reseat loose modules
1111	A hardware fault is stopping the diagnostics from running
0111	Replace the memory module in slot 0 and run the test again; if it still stops at 0111, replace the system module and chassis
0110 or 0101	Replace the system module and chassis
0010 or 0001	Replace the option module in the slot shown; for a 2D graphics accelerator, try a new VSIMM first
0000	If the right LEDs also show 0000 the test passed; otherwise replace the option module in slot 0

The later guide for the Model 200 series gives the whole eight-bit pattern as a hexadecimal code. The left digit is the slot number in codes of the form x3, x2, x1 and xB.

Model 200 series LED codes (DEC Table 9-1)

LEDs	Hex	Meaning
1111 1111	FF	Initial power-on and hardware initialisation
0011 1111	3F	Firmware initialisation
0011 0101	35	I/O ASIC initialisation
0011 0110	36	Firmware memory test of the first 256 KB
0011 0111	37	Firmware calculating the cache size
0011 1110	3E	Calibrating the millisecond delay loop
0011 1101	3D	Power-up or reset set-up code; memory and modules being configured
xxxx 1011	xB	Loading the console from module x
xxxx 0011	x3	Module x reported an error in the power-up self-test
xxxx 0010	x2	Firmware in module x started
xxxx 0001	x1	System software loaded from module x started
0000 0000	00	No errors during power-up

DEC's flowcharts for the stuck codes: at 36, put a known-good memory module in slot 0; if the test still stops there, replace the system module, otherwise test the remaining modules one at a time. At 3B, 2B, 1B or 0B the console cannot start on the graphics module: remove the graphics modules, attach a terminal to the communications port, and type setenv console s to run the console on the terminal; setenv console * restores the monitor.

The Model 240 and 260 CPU modules have two LEDs of their own. If neither lights when the self-test fails, the CPU module is faulty; if only one lights, the base system module is faulty; if both light, the two modules have communicated.

Test error messages

Self-test failures appear on the screen as ?TFL messages naming the test that failed. DEC's table for the Model 200:

Failed test and the unit to check (from DEC Table 4-1)

Test named	Unit tested	DEC's action
cache, dz, fpu, tlb, misc/wbpart	System module	Replace the system module and chassis
mem, ecc	Memory modules	Reseat, then run the RAM board test
misc/kbd, misc/mouse	Keyboard or mouse	Swap the keyboard, mouse or cable
misc/pstemp	Power supply	Check the power supply fans
cntl, target	SCSI controller or bus	See the TURBOchannel maintenance guide
esar, ext, int, ram, reg	Ethernet controller in the slot named	See the TURBOchannel maintenance guide
box, curs, fb, vdac, vram, stamp, stic	Graphics module in the slot named	See the TURBOchannel maintenance guide

DEC's guide lists misc/wbpart against the system module in Table 4-1 and against the memory modules in Table 5-1.

Power faults

No power

DEC's procedure for a system unit with no power:

1. Check the green LED on the power supply. If it is off, check the switch, the power cord and the outlet.
2. Switch off, unplug the cord, remove the cover and front plate, and check that the red 5 V and black ground cords from the supply are in the right connectors on the system module.
3. If there is still no power, replace the power supply.

The 12 V system module cord carries the ±12 V rails and two status lines. DEC's pin assignment for the Model 200 series:

Power supply connector (DEC Table F-5)

Pin	Signal	Wire gauge
1	+12 V	18
2	Ground	18
3	−12 V	22
4	Ground	22
5	POK (power OK)	22
6	Warning	22

The KN02 is specified for +5.1 V, +12.1 V and −12.1 V, each ±4%, drawing up to 20 A, 2.7 A and 0.22 A. A reading outside ±4% at the system module connector points at the supply or its wiring.

Supply cuts out or will not start

Rob Jarratt's DECstation 5000/240 powered off after running for a while, then its H7878 stopped altogether, with no fan movement and no LED. His findings:

• The rectified mains across the two input smoothing capacitors (C10 and C11) read about 170 V DC instead of the 325–330 V expected; one capacitor was faulty. Both are glued to the board with epoxy.
• With new input capacitors the supply produced output but could not hold a load: the fans twitched and the +12 V output wandered between about 4 V and 8 V.
• Replacing output capacitors, including the large 3300 µF and 8200 µF parts on the +5 V output, made little difference. The fault was on the output riser board, which Jarratt took to be doing the regulation; there C407 (120 µF 25 V) had leaked and C408 (120 µF 16 V) showed heat marks. With both replaced the supply ran the machine for over an hour.
• The whole supply, primary and secondary, is referenced to mains earth when fitted in the enclosure, so connect an oscilloscope ground only to mains earth, such as the chassis.
• The supply's LED is soldered to wires, and a heat sink has to be partly removed to take the board out.

Replacement values are on DECstation 5000 Capacitor Replacement Guide.

Model 100 series (H7826)

The 196 W H7826-AA has three fans and a green LED like the larger supply, and DEC's procedure is the same: check the LED, the cord and the outlet, then replace the supply. Owners who have repaired it report:

• A DECstation 5000/133 supply stopped with its fans and no output and no blown fuse. One of the three 1800 µF capacitors on the 5.1 V output was very leaky and had vented from the bottom, where it could not be seen. Three 120 Ω 2 W resistors beside it, the minimum load on that rail, had darkened the board.
• In the same thread the poster paul reported a shorted tantalum capacitor on the 12.1 V output of his Personal DECstation 5000/25, a different machine with its own supply; he suggested checking for the same fault.
• In a 5000/125 supply, the start-up capacitor C24 near the big input capacitors was bad. Two SCRs on the small primary-side daughterboard act as crowbars and latch the supply off until its capacitors drain; one is fired by a shutdown signal from the secondary side through an optocoupler, the other by primary overcurrent. The negative side of C24 sits at a dangerous negative voltage relative to neutral.
• Rob Jarratt's H7826, from a TURBOchannel extender box, had corroded heat sinks shedding flakes across the board and a corroded track between the 5019574 shutdown daughterboard and the main board; a wire link replacing the track brought the supply back.

Warning: these are mains switch-mode supplies. The input capacitors charge to over 300 V DC and the primary-side control circuit floats at mains potential. Unplug the supply and confirm the input capacitors have discharged before touching anything.

Overheating

The message 7/misc/pstemp (system is *HOT*!) means the system unit is overheating. DEC's procedure:

1. Remove the system unit cover and switch on.
2. If all three fans turn and the fault persists, DEC's guide refers the machine to Digital's support centre.
3. If only one or two fans turn, replace the power supply.
4. If no fan turns, switch off, remove the eight screws holding the fan assembly to the supply base and check the fan cable connection. If it is connected, replace the power supply.

The overheat detect test can be run on its own as t 7/misc/pstemp.

Memory faults

A mem or ecc error names the slot of the memory module that reported it. DEC's procedure on the Model 200:

1. Make sure every module is seated.
2. Run t 7/mem * to test every module in turn, or t 7/mem n for the module in slot n. Each module takes several minutes.
3. If every module fails, remove them all and fit one known-good module. If that fails too, replace the system module and chassis.
4. If only some fail, replace those.

Failures appear as ?TFL 7/mem/code: board (x: description), where board is the slot number. The Model 200 slots run from a0000000 (slot 0) to a77FFFFF (slot 14) in 8 MB steps, so an address in an error log identifies the slot.

Serial ports, printer and terminal

If the tests pass but a printer, modem or terminal does not work, fit communications loopback connectors to both ports and the keyboard and mouse (or a mouse loopback) to the keyboard-mouse cable, then run t dz/modem/ line cts-lbk, with line as comm (right-hand port) or port (left-hand port) and cts-lbk as n for data lines only or y to include clear-to-send. An error means the serial controller on the system module; a pass means the external device is at fault.

Error logs and addresses

Under ULTRIX, memory and error-and-status-register error logs give a physical address. DEC's table maps it to the failing unit:

Model 200 physical addresses (DEC Table 4-4)

Address range	Hardware
00000000–1DFFFFFF	Main memory
1E000000–1E3FFFFF	Option slot 0
1E400000–1E7FFFFF	Option slot 1
1E800000–1EBFFFFF	Option slot 2
1EC00000–1F3FFFFF	Reserved; DEC treats errors logged here as system module errors
1F400000–1F7FFFFF	SCSI controller on the system module
1F800000–1FBF0000	Ethernet controller on the system module
1FC00000–1FFFFFFF	System module interface

Console locked or settings lost

A forgotten console password is cleared by fitting the clear-NVR jumper to the two-pin header on the system module, switching on until the console prompt appears, switching off and removing the jumper. All environment variables then need setting again with setenv. Console status code 0439 reports that the time-of-day clock values indicate a dead battery.

Blank screen

DEC's owner's guide checks, in order: press a key in case the screen saver is on; check the system unit and monitor are on; turn up brightness and contrast; with both off, check the power cords and the monitor-to-system-unit cable and the video cable. If the green power supply LED is lit but the screen stays blank, follow the LED code, or run the console on a terminal as described above.

Related pages

• DECstation 5000
• DECstation 5000 General Maintenance
• DECstation 5000 Capacitor Replacement Guide
• DECstation 5000 Documentation
• Capacitor Failure Symptoms