E3631A hangs on power up

[knotlogic]

My E3631A power supply just failed on me yesterday.  It was running a light load on my desk when it beeped, and appeared to reset itself, setting outputs back to the power on default of 0.  I disconnected my load, turned it off, then on again, and it hung with the display all on, no beeps on startup.  Subsequent attempts to turn it on have the same result, no beeps and the display hangs at all on.

I've had a quick look around the insides.  Nothing obviously wrong, though there's a slight buildup of fine dust in some areas.  My plan is:

1) Give the insides a quick dust off with a static safe brush.
2) Do a check of the internal voltages.  I've been able to find a copy of the service guide online with schematics and a list of bias voltages and test points for them.
3) Maybe do a quick thermal check with a thermocouple probe.  A thermal camera would be great but unfortunately I don't have one.

Once that's done I'll have a better idea of what I need to do.  In the meantime, I'd be open to any advice anyone has about repairing one of these units.  Especially if you've come across the same problem before.  Is there anything in particular I should or shouldn't do at this stage?

[z01z]

I repaired one, sadly with different symptoms, so I doubt it would help you.
There are several topics about E3631A repair in this foirum, I'd suggest to read through them, if not done yet.

[RFMWdude]

I experienced a similar problem with the exact same model. It turned out to be a defective SRAM chip (U15 on the 'Top PCB Assembly'; 32kbx8).

[knotlogic]

Managed to get it nicely dusted off.  Turns out to have been a bit more dust that I initially thought in some areas.  What I noticed was:

1) The top PCB has two black wires permanently soldered in, I think they run to the front panel, probably common?  The bottom PCB has one red wire, probably the +6 V line, also soldered in.  Because of this neither board can be fully removed, and it took a bit of careful positioning to dust them off.

2) One of the wires on the J3 top to J3 bottom wiring loom wasn't crimped properly!  I'd disconnected it from the top board, and then noticed the end at the bottom board was wiggling free.  Luckily it appears to be a ground line - there are multiple on that loom.  I forgot to check the bottom board, but that end might not even be connected.

3) Rifa caps.  The X and Y capacitors are Rifas.  Should I replace these?

Cleaning took longer than I expected, but I did manage to probe all the voltage points listed in the service guide.  All were within specs.  And halfway though that the unit started working again.  I turned it off after probing the points on the top board, and it powered up fine when I was about to start on the bottom.

If it's not a power supply problem, maybe it's logic?

[knotlogic]

I repaired one, sadly with different symptoms, so I doubt it would help you.
There are several topics about E3631A repair in this foirum, I'd suggest to read through them, if not done yet.

I've started reading a few of them.  Given I have no idea what's wrong with mine, there's no telling what info will come in useful.

I experienced a similar problem with the exact same model. It turned out to be a defective SRAM chip (U15 on the 'Top PCB Assembly'; 32kbx8).

How did you manage to narrow it down to the SRAM?

One thing I wanted to do but didn't have time for was to probe the power lines to the CPU and SRAM with an oscilloscope.  I'm wondering if there's anything I should look at now while it appears to be working.

[dominitech]

"The Signal Path" has a YouTube video covering the repair of an E3632A.  It might be of some help to you.  He uses an Isolation transformer for safely troubleshooting.  I usually use a variac transformer myself.  I provided the link below to his video.

[RFMWdude]

How did you manage to narrow it down to the SRAM?

One thing I wanted to do but didn't have time for was to probe the power lines to the CPU and SRAM with an oscilloscope.  I'm wondering if there's anything I should look at now while it appears to be working.

Good question. As I recall, it was more intuition than actual troubleshooting. The supply voltages were OK, the CPU clock was running and there was activity on the data and address lines, but the program crashed somewhere during startup. I could've been a faulty CPU, ROM or RAM chip (among other things). I replaced the SRAM chip and got lucky

[WattsThat]

Yes, the black wires soldered on the top board are the common wires for the +/-25 volt supplies. It could have been some conductive crud on the logic as the fan deposits dirt on both boards, more so on the bottom which contains the logic portion of the supply.

My recent E3631A repair, illustrating that factory mistakes can take years to reveal themselves:

https://www.eevblog.com/forum/repair/hpagilent-quality-not-a-repair-but-a-rant/

An isolation transformer is not required to work on these beasts but a strong back and bench are!

[knotlogic]

Well it's been a busy week and I haven't had time to look at the unit further.

An isolation transformer is not required to work on these beasts but a strong back and bench are!

I noticed!    What makes it worse is trying to carefully maneuver the boards around to hold them at an angle while trying to dust them off.  I'm probably going to have to replace the Rifas at some point, and I haven't figured out how to get at them yet.

It's possible I had some conductive enough crap on some logic somewhere.  I'd cleaned out the interior and was testing the voltages on the top board.  At this point it was still failing on power up.  Shut it down to reposition it so I could access the bottom voltage test points, and it started up fine.  It's been OK since, though I haven't really used it.  There was a particularly hard to get clump of dust under the SRAM pins, I had to use a cut down toothpick to get it out.  But that was before I started testing.

Good question. As I recall, it was more intuition than actual troubleshooting. The supply voltages were OK, the CPU clock was running and there was activity on the data and address lines, but the program crashed somewhere during startup. I could've been a faulty CPU, ROM or RAM chip (among other things). I replaced the SRAM chip and got lucky

Can the CPU clock be checked by probing the CLKOUT pin (65) indicated on the schematic?  If not should I just monitor the SRAM address lines on startup?