First steps testing and troubleshooting bench power supply

[guymo]

Hello,

I'm a relative newbie and have just got hold of my first bench power supply, an old HP 6205C which should be more than enough for my interests for now. I got it cheap and sold "as is" so I'm just starting to play around with it and see how well it works. I have a few questions!

At first sight it seems the output voltage drifts down quite a bit. This is very apparent on start up, but settles down after 30 minutes or so, though still noticeable even on my low res multimeter: maybe 10mV every few minutes. I have not tested under load, though -- would that make a difference? And is a rapid drift expected as the supply warms up? (The service manual does suggest a 30 minute warm up before making stability tests.)

The manual suggest using a 133 Ohm load resistor to test the supply at 20V/0.6A. Is it safe to use e.g. 12 x 2k 0.25W resistors in parallel? Each one should then be dissipating 0.2W. Do I need more margin than that? Or should I try something else entirely?

I have more questions but I'll leave it here for now. Thanks for any help!

[alm]

My basic tests would be:
- With knobs set to max, does it output close to its maximum voltage or somewhat higher? Check the output for ripple.
- If you turn down the voltage knob, does the voltage go down? Go all the way down to zero.
- Load it with something that will draw close to the maximum current. Ideally something that can draw 20 V / 0.6 A (= a fairly beefy 33 Ohm resistor or electronic load) and 40 V / 0.3 A (= the 133 Ohm resistor). If that is not available, then testing the maximum current at a lower voltage should also give you a decent indication. Measure ripple again under load (will likely be higher than ripple with no load).
- Test the current limit. Do not short the power supply before doing this, because it could blow the pass transistor if the current limit is faulty. For a 0.6 A current limit, I would use something like a 5 Ohm power resistor and set the supply to 2.5 V. 0.5 A should now flow. Increase the output voltage, and the current limit should kick in sometime before reaching 5 V. For 0.3 A you would use a 2.2 Ohm resistor.

Obviously for a dual power supply you would perform those tests twice.

I am sure in the manual you will find a much more complete performance verification that will verify every spec in the datasheet. This would be my off-the-cuff test. I would use an electronic load instead of messing with different power resistors, but resistors will work just as well.

Paralleling multiple resistors is fine, but try to have some air between them. Running them close to the maximum is also fine, but make sure you do not touch them or put them on something that might melt. They can get quite hot, in excess of 100°C.

Can you quantify 'quite a bit'? I turned on a fairly similar HP 6200B, and it drifted from 15.123 V to 15.067V in thirty minutes (most of it in the first ten minutes). Fifteen minutes later it is down another 6 mV (ambient temperature also went down). Drift like that is normal during warm-up. And even then, these are not precision voltage sources. 0.1 % + 5 mV drift at 20 V is 25 mV. Add to that changes in load (if any), ambient temperature or line voltage. Is your DMM stable enough to measure 10 mV change on a 20 V signal? What happens if you warm up, measure it, and measure it again after a couple of hours without significant fluctuations in ambient temperature, is it then more than 25 mV off?

[guymo]

Thanks for your reply.

The output voltage falls by over 1V in the first five minutes before levelling off. After that it varies +/-15mV or so over time. These measurements were made with no load.

On the 20V setting the output voltage goes full range with no load (40mV at the low end, up to just over 20v). It was only after reading your post that I realised I hadn't tested on the 40V setting yet. Bad news here: setting the switch for 40V output makes very little difference to the actual output and the maximum voltage with no load is 20V on one channel and 22V on the other. Should the output voltage jump up immediately when the 20V / 40V switch is pressed?

It gets worse with a load connected. I wired up 15 2k resistors in parallel to make a 133R load. On the 20V setting, the maximum outputs on the two channels were 10V and 12V, and on 40V setting, 12V and 14V. So, overall I conclude that it's not working very well.

The manual has a fairly detailed procedure for checking out the regulation circuits. I've made a start and the first batch of reference voltages are all ok. Going further will involve some fairly detailed work shorting out various transistors in the feedback loop.  However, the fact that both channels are defective in the same way seems suspicious to me. Could the transformer be to blame? That's all that the two channels have in common. Or is it reasonable to imagine that the regulation circuitry has gone wrong on both sides in the same way?

The first thing I had to do when receiving the unit was re-jumper the transformer for 240V operation so that the two primary windings are in series rather than parallel. Could I have damaged something in doing this? It was a very quick and easy job. 

[bd139]

Nice power supplies. Setting to 240 shouldn't have harmed anything.

I had a 6236B front the same series for a number of years. The main filter caps were shot. This was causing ripple bad enough that the voltage reference wasn't working properly. This got worse as the supply warmed up. This is probably the first thing to check. Either an ESR meter or observing voltage across them with a scope will do the trick. You can probably get away with a DMM on AC settling to get an idea of the ripple as well.

[guymo]

I had a 6236B front the same series for a number of years. The main filter caps were shot. This was causing ripple bad enough that the voltage reference wasn't working properly. This got worse as the supply warmed up. This is probably the first thing to check.

Thanks. For the avoidance of doubt, you mean the large capacitor which is connected across the output of the bridge rectifier, right? There's a resistor in parallel which is pretty accessible for my scope leads so I will give that a look later today.

[bd139]

Yep that's the one. Check the rail voltages against the service manual as well while you are there. That's usually a big indicator of where failures are.

[guymo]

What do you mean by rail voltages here? Sorry for being dim. Feeling very glad I posted in the beginners' forum...

[bd139]

No worries. We've all been there

There are a number of voltages defined in the service manual. These are usually the reference voltage which the output is compared against, the positive and negative supplies for the opamps etc and the main regulator input supply. These are generally referred to as "rails" as in power distribution rails because they are constant voltages which are used to supply power to utility functions in the device. They usually have a defined voltage to expect with a tolerance. Best to look at the service manual for what these should be. You can mark up the schematic in the manual with "expected" and "actual measured" voltages. This tends to lead to clues when things are wonky. I've fixed many a thing over the years purely from inferring the failure from these measures voltages.

Edit: before I forget, I'd just like to add that a broken bit of kit is the best education you can get

[alm]

Sounds definitely like a bad filter cap to me. The ripple gives it insufficient headroom to go beyond 20 V. As you increase the current draw, ripple will increase. It would probably be more stable at very low output voltages. Measuring output ripple with the supply set for 40 V should show a very large ripple.

If this is not the case, then I would suspect a bad range switch contact or rectifier diode before suspecting the transformer.

[guymo]

Thanks again.

That was my guess as to what you meant The manual has a good table of reference voltages and test points. I've checked them and they all seem good.

I'm with you on the education point, and enjoying the process despite feeling a bit in the dark for now.

@alm: I've just had a quick look at the voltage going into the bridge rectifier and the range switch has a convincing effect on that side, more or less doubling the voltage as expected. So I think the transformer is likely ok. Will get the scope on the filter caps later.

[guymo]

Hmmm, so I scoped the output ripple of the filter capacitors, and I think they look good.

With the 133R load connected, the power supply was able to deliver 10V, and the scope showed a skewed triangle wave (steeper rise than fall) of 2V amplitude at 100Hz, riding on a DC offset of 35V or 50V depending on the range switch.

I think that is what would be expected, isn't it? The mains power is 50Hz so 100Hz ripple makes sense. The filter cap is 490uF so I think the 2V ripple is the right order of magnitude. So I'm thinking that the bridge rectifier and filter cap are ok. Does that sound right?

The service manual's clues for troubleshooting low output voltage are, first the rectifiers and filter cap, second the front panel meter (are they just suggesting that the readings are wrong, or that the mete itself is somehow spoiling things?), and then the regulator feedback loop. Does this mean it's time to get stuck in to the loop circuitry?

[bd139]

That's a good set of measurements. Transformer, rectifiers and caps are likely good then.

Id ignore the meters for now and check with an external DMM.

Regulator/feedback it is. Start with voltage reference. Monitor that over time. That determines half of the output stability over time. Other half is output sample voltage. Monitor that over time too. I don't have the schematic here but there's usually a voltage divider across the output. Might be dodgy pot as well you see.

[alm]

That sounds fine. One thing to check that I believe was not mentioned is the strapping on the terminals on the back. Especially the sense and programming terminals. The manual should have the correct strapping for front-panel operation with local sensing.

Did you also measure the output ripple with a scope on both the 10 V and the highest voltage setting? Will probably be negligible since the cap looks fine, but it would also show if something is oscillating.

[bd139]

Good ideas!

[guymo]

Thanks again. I'd checked the strapping before but I had another look just in case and it is correct for ordinary front-panel controlled operation.

I scoped the filter cap on both the 20V and 40V settings, and I varied the control pots a bit to see if there were any differences. The overall ripple decreased when I lowered the output voltage, but overall looked the same.

It seems to me that the low level of the output is more of a worry than stability at this point, and presumably there's a good chance that whatever is wrong that keeps the output voltage low might also be causing the drift. So is there anything to look at first to diagnose the output voltage?

The front panel pots measure ok as far as I can tell, though I suppose I could eliminate them from the picture by setting the unit up with a programming resistor instead.

To check my understanding of the operation:  is the regulation supposed to deliver a certain voltage regardless of the 20V / 40V setting, so that on 20V the output would max out (and presumably stop being very well regulated) as you turn the control pot up about half way? Or should switching from 20V to 40V double the output voltage for a given setting? It's easier for me to figure out how circuits work when I know what they're meant to do

[alm]

With output ripple, I meant ripple on the output connector (i.e. after the regulator). It is an easy thing to check and could give you some information.

On my 6200B, which is fairly similar, the output voltage stays the same regardless of the range setting, unless it is close to 40 V (on the 20 V setting it goes up to about 38 V without load, less under load). So basically the 20 V mode is only using half of the range of the pot.

I would start with the steps in table 5-5 of the manual.

[guymo]

Ahh, sorry for my confusion and thanks for your patience. I'll have a look for oscillations etc, probably tomorrow. And thank you for explaining the expected operation.

I've been eyeing Table 5-5 with a little apprehension all day, and was hoping you wise folk might have another test or two to suggest first, but ok, I'll jump in. This probably means a rapid reduction in the number of questions I ask because I imagine it'll take me a while, so that's good news for you at least

[alm]

Not sure if you already did it, but measuring the reference voltages and ripple (table 5-2) would probably be my first step (thou shalt measure voltages) after a visual inspection for burnt components, loose connections, blown traces, etc. Measuring the supply voltages looks easier than table 5-5, and an unstable +12.4 V rail could easily cause issues with regulation, since the reference voltage is derived from that.

[JXL]

Your no-load voltages before for the regulating transistor measure OK.  I suspect the supply may be current limiting.  Did you check the current limiting circuit as suggested by @alm in reply #1?  Turn the current limit pot to both extremes and check the loaded output voltage.

[guymo]

Did you check the current limiting circuit as suggested by @alm in reply #1?  Turn the current limit pot to both extremes and check the loaded output voltage.

I haven't done that, because I don't have an appropriate load resistor. But it does make sense to see what current it can deliver. I'll get some appropriate resistors, and also have a poke around the current limit circuitry for voltage measurements. Thanks!

[bd139]

I tend to just stick my DMM across the output in mA range with current limit set to around 100mA and compare the supply readout / metering to the DMM output as you ramp the limit up and down. Saves smoking resistors then. Occasionally costs a fuse however. Basically the shunt in your meter is a very small resistor. The power supply voltage readout will drop line a tone to under 1v in this circumstance.

[Vtile]

I'm reading this from phone so I'm not 100% sure if I have seen everything related. If the voltage controller circuitry is negative feedback type using integrator the offset error might come from blown integrator cap if the parasitic capacitance is not enough to pull it to set value. But this pretty far fetched I think.

Edit. C5 in Driver & Error amp. section is I guess it is also integrator of the feedback loop. I'm not particularly good with analog implementations though so I might be totally wrong. Look at it when all other thing is checked and if the offset error is still there.

[guymo]

The current limiting seems to be ok: I got hold of two 2.2Ohm 10W resistors and hooked them up in series to the supply. It was able to deliver 360mA on the 40V setting and 720mA on the 20V setting, just a little bit above the advertised .3 and .6A, which is what the manual told me to expect. I confirmed the current with my DMM as well as the panel meter. So that looks ok.

It seems there is nothing else for it but to get stuck in to the feedback loop. Thanks for your replies so far.

[bd139]

Good luck!

[guymo]

Progress of a sort.

I worked through the feedback loop troubleshooting procedure from the manual, saturating and shorting various transistors, and everything checked out: each step sent the output voltage to maximum (38V on the 20V setting, higher on the 40V setting). I skipped the step of disconnecting the current limiting section because it seemed like hard work for a long shot. Hmm.

I started to wonder about the output sampling section. After poking about a bit and getting suspicious of some of the voltages, I decided to try disconnecting the panel control and using a resistor to program the output voltage. According to the manual, each 200 Ohms should give 1V of output. I tried a 4k7 and a 6k8 that I had on hand, and lo and behold, 22V and 32V output -- not exactly what is expected but close, and I think I had been tinkering with the programming trimmer, so no big deal -- and stable with my 133Ohm load.

With the panel controls connected, that programming resistor is replaced by the panel pot and a 5uF capacitor in parallel. This reminded me of the suggestion to check for oscillation, and sure enough, the output is oscillating: there's a downward dip of about 4V every microsecond. (It is possible this is caused by capacitance in my scope probe but it was the same on 1x and 10x setting.)

So I am wondering if this oscillation is messing up the output sampling because it passes through the 5uF cap. Does that sound plausible?

If so, there's a capacitor to check out and a trimmer to adjust, though I lack the equipment to test as suggested in the manual (para 5-72 if anyone is reading!). But any other thoughts you might have would be most welcome.

[guymo]

I am now convinced that the oscillation is the main issue with this power supply.

I adjusted the resistor in the compensation loop and was able to reduce the oscillation on both channels to about 1V amplitude. This done, the output voltage can now be adjusted across the full range and is stable under load; "stable" apart from the 1V 1MHz oscillation.

Is there anything else that can be done about the oscillation? The compensation is a trimmer and a 1nF cap in series. Is it worth tinkering with the capacitance? What should I try?

[bd139]

Is the output electrolytic capacitor duff? They tend to provide a big chunk of the stability.

[guymo]

Is there a way to test in-circuit? So far I've got away without desoldering anything.

[bd139]

You can usually spot them with one of three methods in circuit:

1. Temperature. Knackered ones tend to get quite warm if you run them hard. Fire up the supply for 5/10 minutes with no load, then shut it down and feel the capacitor and see if it's warm.
2. Connecting another one across it. At least one order of magnitude larger than the existing one. If it becomes stable then the cap is shot or the design is bad. HP designs usually aren't bad so by elimination, it's the cap. This doesn't work at all if the cap is shorted for obvious reasons. One on the end of a couple of croc clip/minigrabber leads is fine - doesn't have to be soldered in.
3. Ohmmeter across it. Any DVM should be fine. The current source in the meter will charge the cap up which will increase the voltage across it and read a higher resistance over time. If it doesn't go up the cap is shot. check the polarity of the DVM first - some of them have a positive voltage on the negative terminal in ohms mode.

You can trace out shorted ones with a 4.5 digit or better DVM in low ohms mode based on the trace resistance.

[guymo]

Thanks again -- I'll give all of those a go later today .

[alm]

And make sure what you are measuring is really conducted from the power supply and not common mode noise. Connecting both probe tip and ground lead to the same output terminal (shorting the probe) and checking if you still see the signal is a way to detect this. Dave also discusses it in this video:

[guymo]

Thanks for the pointer. I am concerned that my measurement equipment and skills aren't quite up to the task, but when trimming the compensation network I could clearly see the ripple changing smoothly as I turned the trimmer, first reducing in amplitude and then suddenly jumping up to a full rail-to-rail clipping oscillation when I took the trimmer too far, so I am pretty sure this ripple is coming from the supply. I'll try it on a different mains circuit and try shorting the probe as you suggest to see if anything looks suspicious, though.

(The schematic of the error amp section is attached. R30 is the trimmer that I'm talking about. I thought I'd better include this in case it gives someone an idea or you notice that I have done something appallingly wrong!)

[Vtile]

Try to DVM method for the C5 and try to put the another capacitor in parallel. It somewhat sounds that the controller circuit is oscillating?? A couble oscillator screenshots could say something to people who have lots of experience with these.

[guymo]

Progress!

I checked the output cap using an Ohmmeter as suggested, and it appeared to be ok, reading a rising resistance. It topped out around 11k, but I guess the actual reading is kind of meaningless, isn't it?

Then I tried a 470uF cap in parallel with the output cap (80uF). Bingo! With just that capacitor as load, the scope reads a dead flat line with no noise or ripple visible.

Testing with a resistor in parallel, things looked ok but as I turned up the supply beyond about 10V  it began current limiting and then blew the top off the capacitor, which I guess was a bit underspecified at 35V    Would things have been better if I had turned up the voltage more slowly?

Brushing aside my poor test procedure (ahem! beginners' forum!) does this mean that the output cap needs replacing? I suppose it is still working as a capacitor but no longer the right value to keep the circuit stable. If so, what should I look for in a replacement? Clearly it should be 8uF 300V as specified for the original, or at least close to that, but is there any other spec to bear in mind or any recommended brand or series for such parts?

Many many thanks to you all for getting me this far. I think we are almost there  -- let me know if you disagree!

[alm]

A 35 V cap should be fine up to at least 35 V . Not sure why it died around 10 V. If it was an old cap (as in decades old), it might have been leaky. Did you monitor the voltage and do you know if the supply was oscillating?

The 300 V cap sounds way overrated to me. I would not hesitate to replace it with anything over 60 V. But if I was ordering, I might look for something close to 300 V.

Do you have anything in the range 80-100 µF with a voltage rating of at least 40 V that you could put in?

[bd139]

This cap is overrated so that if you're floating the supply at +250v, no horrible accidents take place.  10uF oxide layer may have been popped by the sudden potential difference between ground / negative rail (I think there's a small cap between there) and after that, kaboom on lower voltages! I've done that.

WEAR SAFETY GLASSES!

That's hopeful

You probably won't get one close to original spec. Choosing capacitor:

1. Measure pin pitch and note if it's radial / axial.
2. Hit RS: http://uk.rs-online.com/web/c/passive-components/capacitors/aluminium-capacitors/
2. 80uF doesn't exist any more. Try 82uF or 100uF (these things are usually +/- a ton so it usually doesn't matter terribly).
3. Try closest voltage. 250-350 is probably fine.
4. Narrow to lead spacing required.

Good brands: Vishay (BC / Sprague come under this brand), Rubycon, Nichicon.

This isn't necessarily a particularly special capacitor but try and look out for one with a long life rating (2000h+) and 105oC temperature rating instead of 85oC. Hour rating is at full temperature, which it won't get near unless something is seriously wrong. It'll last a lot longer at lower temperatures.

Side note: if it's got three pins you will require a bodge most likely. Pick the nearest thing and fudge it in with some thick wire. There's not much else you can do now about this.

[guymo]

A 35 V cap should be fine up to at least 35 V . Not sure why it died around 10 V. If it was an old cap (as in decades old), it might have been leaky. Did you monitor the voltage and do you know if the supply was oscillating?

When I was looking at the scope I didn't see any oscillation. Then I turned my attention to the voltmeter and turned up the supply. Once I noticed it topped out at 10V I took a look at the ammeter and could see about 750mA there, slowly falling. I experimented with this a bit, turning the voltage up and down, and then the cap popped. It wasn't old.

I wonder if my tinkering with the error amplifier compensation has slowed down the transient response leading to these current spikes? I'll try putting the trimmer back to roughly where it was and play around some more (with appropriate safety gear).

And then it seems like the next move is to get the board out so I can remove the output caps and try out some replacements. Great -- I'll get on with it.

[bitseeker]

Hey guymo. For a beginner you're doing really well. These are good little supplies and it sounds like yours is well on its way to being healthy again.

[guymo]

Hey guymo. For a beginner you're doing really well. These are good little supplies and it sounds like yours is well on its way to being healthy again.

Thanks! I'm enjoying and finding it interesting too. The real tricky part is yet to come, though, because I am much better at thinking and learning than I am at physically interacting with things, so my removing those output caps poses a significant risk to the circuit. I'll be ok soldering new ones in, though.

[alm]

Connecting a cap to the output that does not have six times as much capacitance as the original cap and that does not die when exposed to > 10 V might still be a useful test.

[guymo]

Connecting a cap to the output that does not have six times as much capacitance as the original cap and that does not die when exposed to > 10 V might still be a useful test.

My choice of capacitance was driven by bd139's suggestion to use a cap an order of magnitude larger than the existing one. Well, I got close.

I'm still not sure what happened to that capacitor. If the ammeter's readings were anything like correct then something weird was going on. The voltmeter read ~10V, the ammeter 750mA, and the load was 470uF in parallel with 133R. You can't put that much current through that for any period of time without considerably more voltage, can you?

But yes, before I start taking things apart, I'm going to get hold of a 100uF cap with a decent voltage rating, and see what happens when I connect it across the output after also returning the error amplifier feedback trimmers to their original positions.

[bd139]

Probably a silly question but did you put it across it the right way?

Reason to use a much larger cap is that if the current one has a high leakage then it usually compensates for the reduction in time constant for the dead one. It's a diagnostic tool, not a solution. If it does something, suspect that cap and replace it with a closer to reality one then measure it again.

[guymo]

Wrong polarity was my first thought (after "what the...?") but a post mortem revealed that I had not been as careless as I thought after all.

I have some 100uF 65V caps in hand now. I will try them out this evening.

[alm]

I was not saying using a much bigger cap was a bad diagnostic test, but just that killing completely the bandwidth with a huge cap is not a 100% proof to me that the cap is indeed the problem.

[bitseeker]

The real tricky part is yet to come, though, because I am much better at thinking and learning than I am at physically interacting with things, so my removing those output caps poses a significant risk to the circuit. I'll be ok soldering new ones in, though.

Be sure to check out some videos on desoldering and practice on boards that you don't care about before tackling your power supply. You don't have to buy a vacuum desoldering station to do a good job, solder braid is fine, but if you find yourself doing more and more of it, vacuum makes life easier. I use braid, sucker, or vacuum depending on the situation. They all have their pros and cons.

[guymo]

I was not saying using a much bigger cap was a bad diagnostic test, but just that killing completely the bandwidth with a huge cap is not a 100% proof to me that the cap is indeed the problem.

No, indeed -- understood, and thank you.

I've just tried with a 100uF cap and 133R resistor. There seems to be about 30mV peak to peak noise: it looks like a damped wave, not really sine-ish, every 7 microseconds. I wasn't able to get a convincing photograph of my scope -- all reflections.

I haven't been able to reproduce yesterday's flat lines at all, even using the larger capacitor. But then, that cap did pop, so I suspect I was doing something appallingly wrong. I haven't seen any issues with high current flow in today's testing and I haven't exploded anything.

I wouldn't be at all convinced that my measurement setup and skills are up to the job of this high frequency low voltage stuff. Under some conditions I could see some noise when the power supply was not switched on, but the frequency was different.

Is this as good as I'm going to get, do you think, or is there more to look into?

[alm]

30 mV is way worse than it should be. But I am skeptical (as are you) about the result. Firstly, did you exclude common mode noise? Did the noise disappear with probe tip and ground connected to either output? When you left the probe connected but turned the supply off (sounds like sometimes it did and sometimes not)?

Are both output terminals of the output you are testing floating (not connected to the ground terminal)?

You should be able to get better pictures of the scope screen if you turn down the lighting on whatever is in front of the scope (e.g. you). If less light falls on you, then less light will reflect of you, via the scope screen, to your camera.

[guymo]

I set the power supply up in what the manual calls the normal way: with the positive output connected to ground. I measured at the negative output pin, obviously. I've just tried the other way around with the same results. I've managed to take a photo, not great, but attached. The vertical scale is 10mV per division, horizontal is 2us. The shape and amplitude of this wave changes as the feedback trimmer in the error amplifier is adjusted -- this was the smoothest I could get it.

With the supply floating, the results are much worse. At first I couldn't get the scope to show anything sensible, but after much tweaking of the trigger settings, the second image attached came up. Here the vertical scale is 1V. The trace moved up and down the screen as I changed the trigger level. I have no idea what is going on here.

Yes, I could sometimes see much lower noise levels, 5mV or so, with the supply switched off, but not all the time.

I am becoming deeply confused but still very grateful for all the help!

[guymo]

I had the idea to take a look at the internal voltage references on the scope. They show similar looking noise when I turn up the sensitivity, with spikes at a similar frequency. The peak-to-peak was pretty large, 300mV or so, which I think is way more than zener noise is meant to be.

Obviously this doesn't tell me whether the supply or my measurement setup is at fault. Not sure what to do next really.

[bd139]

It's possible whatever is giving current to the zener ref is shot or something upstream from that.

[bitseeker]

If you have a DMM with sufficient bandwidth on AC, you can use that to measure the amplitude of the noise as confirmation.

[guymo]

I don't think my DMM is up to the task. Haha, I should probably buy a really good DMM to help me fix this ancient and cheap PSU... and on down the rabbit hole.

Meanwhile I found this thread https://www.eevblog.com/forum/repair/linear-psu-where-does-these-spikes-come-from/  with similar looking spikes on a PSU output. The conclusion there was not 100% convincing but that it came from noise on the mains. Hmmmm.

[bd139]

This is how it always starts and the first thing you after this is buy a bust power supply to fix the bust power supply with and then your scope will blow up so you'll buy another one to fix it

Looks to be around 166KHz which is in the reasonable range for switching supplies and lots of interference. Might be picking up French TDF time signal https://en.wikipedia.org/wiki/TDF_time_signal

Ergo, can you stuff the case back on it if it's open and check the noise on the output. If it's still noisy, I'd go around and see if you can find any wall warts and appliances and try and turn them off. I forever end up chasing odd noise problems inside analogue scopes only to find that it's picking up my lamp's power supply transients etc. If that makes no difference then we're almost certainly dealing with oscillation.

I would replace C5 and possibly the trimmer in that case as in the manual that says it's critical for the purpose of oscillation control.

[guymo]

Law of unforeseen consequences:

My in-laws are coming today so I have to put all the electronic stuff away. While packing up I came across a Pomona shorting bar. It occurred to me that this could be used  to connect one end of the output to ground. Yes, it fits, so I connected it and set up the scope for "one last look" before the weekend. Hooked up a 100uF cap across the output and measured, and what do you know, no spikes. Both channels of the supply are behaving themselves, reading a flat line on the scope (a bit "fuzzy", +/-1mV at most).

Obviously it could be that there was some external cause of the spikes that is not there this morning. But it seems more likely that it was my fault, and that lazily grounding one end of the supply with a 50cm test lead caused problems. Could that be down to inductance in the cable? I don't understand inductance.

So I think I am back to where I was the other night, in need of replacing the output caps.

I'm really grateful for the help and support I've received here. Just not being completely alone in puzzling over this is a great help, but the specific advice and navigation has been fabulous. Thank you! No doubt something else will go wrong and I'll need to revive the thread later.

[bd139]

Damn those in-laws. Mine have to put up with it

That might actually have been the issue. If it is, phew! Good find! 1mV or so of noise is pretty good. My Thurlby TS3022 is around 2.5mV noise. I suspect that needs refurb as well but I can't be bothered at the moment

Inductance is hard to visualise and think of really. There are two domains to think of it in; the time and frequency domain. In both domains it is exactly the opposite of the capacitor. Fundamentally in the time domain, if you stick a voltage across it, the current rises from 0 to infinity exponentially at a rate defined by the inductance. When you remove the voltage source, the magnetic field collapses and generates a high voltage to try and force the current to flow again (this is incidentally basically how switching power supplies work). In the frequency domain it's a frequency dependent resistor. The higher the frequency, the higher resistance (generalised as reactance in these things). Typically outside of RF, which is purely concerned with high frequencies, they are used to get rid of high frequencies like oscillations which is why you see a ferrite bead on a transistor leg occasionally. That's a little inductor that stops high frequency oscillation.

You have three modes that could have caused noise which I will detail:

1. The loop acted as an antenna (electric field coupling). This could be a resonant length of some RF flying around in your house. This would develop a voltage across it. At that size this is possible but unlikely I would suggest. Anything resonant at 166KHz is usually pretty big as the wavelength is 1.8Km!
2. The loop acted as an inductor (magnetic field coupling). Try this as a demonstration. Take your oscilloscope's probe, stick it on 10X if it's switchable and clip the ground wire to the tip creating a loop. Set it on the lowest volts/div setting and go and wave it around laptop chargers, televisions etc and see what happens. Basically the magnetic field lines of force will cut the loop and induce some current. Because the probe impedance is so high it'll result in a visible voltage. Some EMC testing tools actually use things that do this on purpose.
3. The additional inductance of the wire created an accidental resonant oscillator somewhere with a capacitance inside the device.

Have fun and let us know if anything else goes wonky