I have a MicroVAX 3100 Model 95 that I acquired some years ago. At 32 VUPs it is one of the faster MicroVAXen.
From the front with RX33 and TZ30
Showing 32MB of installed memory
Some time ago I stopped using it because memory modules would suddenly stop working. At the time, I checked the power supply and some of the voltages seemed to be a little out of tolerance, in particular the output marked as 5.1V seemed to be producing 5.3V. I put the machine to one side because at the time I had little knowledge of repairing power supplies.
Recently I decided to take another look at the power supply with a view to getting the machine up and running again, given it is such a comparatively fast machine. I checked the ripple on the outputs but what I saw were spikes. The spikes seemed to reduce in amplitude after a few seconds. It turns out that I was measuring the ripple with poor technique, but I don’t know what the original ripple was now, because I replaced the majority of the electrolytic capacitors before realising my error.
I decided to replace any vaguely suspect electrolytic capacitors, which included the two big smoothing capacitors as one had a high ESR and the other appeared to be bulging slightly. I discovered that one of these would not fully discharge, which I found when I tried to measure its ESR after having had it powered on. Thankfully the meter had some protective diodes, which now need replacing.
After replacing the suspect capacitors I excitedly put it all back together, only to find that it would not power on. It turned out, that following some other sparks I managed to get from the not fully discharged smoothing capacitor I had ended up frying the UC3842N pulse width modulator. Fortunately I had a spare on hand (UC3842AN), replacing it fixed the power supply, and also seemed to resolve the problem of the capacitor not discharging.
Sadly, I also managed to break the connection to the power LED out to the front of the PSU, so now I won’t have that working.
After resolving my measurement mistakes the ripple on the 5V and 12V outputs was about 20mV and the voltages seemed to be right. So I put the machine back together again.
Unfortunately it seems that one of the memory modules is not quite right because the firmware reports an error:
KA51-A V2.6, VMB 2.1 Performing normal system tests. 74..73..72..71..70..69..68..67..66..65..64..63..62..61..60..59.. 58..57..56..55..54..53..52..51..50..49..48..47..46..45..44..43.. 42..41..40..39..38..37..36..35..34..33..32..31.. ? Test_Subtest_40_06 Loop_Subtest=00 Err_Type=FF DE_Memory_count_pages.lis 30..29..28..27..26..25..24..23..22..21..20..19..18..17..16..15.. 14..13..12..11..10..09..08..07..06..05..04..03.. 16 MB RAM, SIMM Set (0A,0B,0C,0D) present Memory Set 0: 00000000 to 00FFFFFF, 16MB, 32768 good pages, 0 bad pages Error: SIMM Set 1 (1E,1F,1G,1H) SIMM_1E = 16MB ?? SIMM_1F = 16MB SIMM_1G = 16MB SIMM_1H = 16MB Memory Set 1: 01000000 to 01FFFFFF, 16MB, 0 good pages, 32768 bad pages Total of 32MB, 32768 good pages, 32768 bad pages, 112 reserved pages Tests completed.
However, VMS seems to think everything is OK.
$ sh mem System Memory Resources on 10-MAY-2020 13:24:53.80 Physical Memory Usage (pages): Total Free In Use Modified Main Memory (32.00Mb) 65536 21316 42491 1729
So I am hoping the machine will now work reliably.
Rob's Old Computers 
Seems like you are having fun Rob.
I don’t have any old iron here except for a Sinclair Spektrum and a Z80 based system we built for teaching here way way back post Manchester. I need to follow your example and dust them off. Particularly the latter to see just how much slower subjectively it is compared to the stuff I have on my desk. The modern systems seem to spend an inordinate amount of time running around fixing software “engineering” deficiencies on the fly.
Big caps can be a hazard and often it is worh haveing a bleed resistor across them to make sure they are fully discharged in a reasonable time. Particularly important if they had in my case 600V stored on them.
Stay safe.
My experience of the PSUs used in DEC machines is that they tend to have bleed resistors, but without a schematic it is hard to know for sure if this particular PSU doesn’t (it is made by a different third party to most of the others I have seen). It is odd that fixing the PWM seems to have improved the discharge of the capacitor (it was only one of the pair that would not discharge).