Oscilloscope power supply burnt and strange clicking noise

[CodingMarco]

Hello guys,

--- Summary ---

--> Something burnt down in my oscilloscope power supply from ~1994 (-->psu.jpeg)
--> I can't see any obviously burnt components.
--> Now a strange clicking noise comes from it (mp3 as attachment).
--> What could cause this noise?
--> Voltage spikes compared to earth but not compared to ground pin on power supply (--> spikes_turned-on.png)
--> Spikes continue a little while after power is switched off
--> Sometimes everything seems to work (without load), and sometimes not --> None of the "essential" components burnt?
--> I can observe many things (see detailed desctiption) but don't have enough experience and knowlage to combine them into a conclusion
--> Any help would be so nice!

--- Detailed description ---

A few weeks ago, a very strong smell like burnt semiconductor came from my HP 54540A oscilloscope and all the lights and the display were flashing randomly.
I removed the cover and the smell seemed to come from the power supply (HP part no. 0950-2369), from the region marked red in psu.jpeg. I looked over it and saw no obviously burnt transistors, MOSFETs etc.
I probed the transistors, MOSFETs and diodes with a multimeter and none of them seem shorted.

I removed the connector going from the power supply to the mainboard and turned the supply on.
Now a pretty loud and high-pitched clicking noise comes from it around every 25 ms. I recorded it with my laptop microphone (mp3 as attachment to this post).
From what kind of compoment could this come from?

I also looked at the outputs with another oscilloscope. What is really strange is that I can turn the power switch on and it works just normal (--> normal.png) and sometimes there are these spikes, around every 25 ms like the noise (--> spikes_turned-on.png). One spike can be seen in detail on spike.png.
I don't know if I'm wrong, but I conclude from that that it's none of the really "essential" components that is burnt. Am I wrong with that assumption?

I noticed a few other things:
    --> Another thing I noticed is that the spikes and the noise continue a little while when I turn the power switch off (--> spikes_turned-off.png).
    --> There is always one of these spikes when I turn the supply on. Sometimes this is the only spike and the power supply seems to continue normally (--> powerup-normal.png)
        and sometimes the spikes continue and the offset of the output approaches zero (--> powerup-fail.png)
    --> Connecting a 100 ohm resistor across a 5V pin and ground seems to trigger the spikes (also turning them off while the power switch stays on), but not always...

I notice that I observe many different hints, but since I study computer science with a lot of EE lectures since only half a year, I don't have enough experience and knowlage to combine them into a conclusion which part(s) might have failed.

So any help and hints which components to look at closer would be very much appreciated!

[Wallace Gasiewicz]

In some power supplies there are circuits to shut down the supply if an overload is sensed. Then the PS starts up again only to repeat and repeat. If this is the case and you have disconnected all the leads to other components, the protection circuit could be the problem. Maybe the caps are enough to keep power going into the other components.  This is just a guess in your case. You have a switching PS.
I cannot find any real schematics on your PS. I think that you should not run the scope because if something more fails it may put way too much volts on the other components.
You said "Voltage spikes compared to earth but not compared to ground pin on power supply" Therefore:
Might also be the power supply connection block, it is "filtered" and the filter can be a problem
Here is a link to another thread discussing your problem:

https://vintage-radio.net/forum/showthread.php?t=123197

Apparently the PS has known problems

[JKKDev]

It's either a pretty big coincidence that the frequency is somewhat close to the mains one or there is a problem with one of the caps close to the mains side of things. The only noise I would expect from a PSU is either a relay gone mad (which it's not the case here) or one of the caps has an internal short that is sparking. At lease that's my opinion

[Wallace Gasiewicz]

Yea, JKKDev has a point, I agree

On further thought, if the clicking is caused by the PS overload protection, it could easily be one of the caps in the PS. You cannot test "leaky" caps with an ohmeter. You need  to remove them and test them with a voltage supplied across the leads. You will notice that the cap is drawing a lot of ma through it.

Probably easier to replace the likely suspects.

PS: I really do not like switched mode power supplies but they apparently are a part of modern life.

[CodingMarco]

Thanks for the replies!

I can confirm that the PS has a protection circuit for all the outputs because I have access to the document with the specs that HP sent to the external supplier (Computer Products Inc.). But otherwise there is no useful information nor schematics in it.

I can measure the spikes on all outputs, does this mean that the fault lies on the primary side or the filter? Because according to this reverse-engineered schematic http://www.radiocollection.be/images/restaurations_img/HP54503/supplyHP54503.pdf the outputs seem quite isolated from the large transformer on...

[CodingMarco]

But the smell really seems to come from the secondary side

[JKKDev]

There's a power supply voltage check in the service manual. If you follow the isolated test and check the output voltages do all of them have spikes? Are all of them present and around their nominal level?

[CodingMarco]

I have ordered a 25 watt, 2 ohm resistor on ebay and I'll try it out. I'm also going to need it when I replaced some parts to see if it works again.

Without load, the spikes are present on all outputs. When there are spikes, the voltage difference between a ground pin and another pin is pretty much unpredictable, but always less than the desired output. When it suddenly works, all volatages are ok (without load).

[Wallace Gasiewicz]

They also talk about the AC connector as a problem, It has a filter in it that goes bad,
and smells like cat pee when it does

[shakalnokturn]

Where are your waveforms from and what is their reference point?

Considering you have a smell and what could look like a random overload symptom try removing the clumps of larger secondary side electrolytics, for all you know the electrolyte may have leaked and caused the PCB to char between tracks / pads.

[vk6zgo]

Some older SMPS don't like operating without a load & will go into "hiccupping mode".
This is a Philips thing in particular.

Every 25ms does see very fast, though.--it would almost appear as a continuous tone.
It also doesn't relate to any commonly used Mains frequency.

We had 40Hz Mains in Western Australia years ago, but that is long gone, & I haven't ever heard of anywhere else that used it.

If a DC relay is trying to operate at 50Hz or 60Hz, from unfiltered halfwave rectified DC, it may have enough "lag" that its "clicking" occurrs at a lower frequency.

[CodingMarco]

Where are your waveforms from and what is their reference point?

I measured all the waveforms at the output pins that go directly to the oscilloscope mainboard and I didn't connect the ground leads at all so they are floating measurments.

Considering you have a smell and what could look like a random overload symptom try removing the clumps of larger secondary side electrolytics, for all you know the electrolyte may have leaked and caused the PCB to char between tracks / pads.

Yeah I think I'll desolder a few caps and test them for leakage and capacity.

[wn1fju]

I have fixed a few of these same power supplies - they are used in several HP models.  In all cases, the solution was to replace ALL the electrolytic capacitors since they are quite prone to leakage.  Remove them all, scrub the board with alcohol, scrub it again, scrub it again, then replace the caps with new (non-Chinese) ones, say from Nichicon, Panasonic, etc.  If you find electrolyte on the board, also check the trim pots as I had electrolyte get into some of them.  Unfortunately, you will need to spend about $25 (US) to buy the couple of dozen new caps.

Even if the problem with your power supply remains and requires further diagnoses, it is still worth the effort to replace the caps because if they aren't leaking now, they soon will be.

[CodingMarco]

I have fixed a few of these same power supplies - they are used in several HP models.  In all cases, the solution was to replace ALL the electrolytic capacitors since they are quite prone to leakage.  Remove them all, scrub the board with alcohol, scrub it again, scrub it again, then replace the caps with new (non-Chinese) ones, say from Nichicon, Panasonic, etc.

Thank you very much for your reply! I'm going to replace the caps. Is it also necessary to replace the very small caps on the control / overvoltage protection circuit on the side?
The smell definetly comes from the 6 caps in the middle. But I'm still wondering if it's only the capacitors because there really was a really, really strong smell like a burnt transistor very suddenly in my room. But I could see the white hole in a previously burnt transistor very clearly but now there are no signs at all...

With a better google search (site:eevblog.com + googles similarity algoritms) I found another thread referring to the exact same clicking noise:
https://www.eevblog.com/forum/repair/hp-54502a-repair-once-again/ It links to a very helpful pdf from Tektronix which describes the clicking noise exactly on page 29: https://www.sphere.bc.ca/test/tek-parts/troubleshooting-scopes.pdf.

I have to admit that this is starting to get funny because of the challenge it is.
Maybe this is beginning of a long TEA journey...?? I think working at Keysight doesn't make this any better

[wn1fju]

I wouldn't worry too much about the control board capacitors just yet.  Besides, it is not easy to get to the back side of that daughterboard without removing it, which is a pain.

The two large caps on the primary side are most likely OK.  But you could desolder, measure them and reinstall (assuming there is no sign of wet electrolyte).

For amusement, you might take an ohmmeter and check the various output pins to ground and see if any are very low resistance (before you attempt any repairs).

Sometimes, when there is a short, or close to a short, a switching supply will fold back due to current limiting and attempt to restart over and over.  That's a very common occurrence (especially with vintage Tektronix stuff).

[TheDefpom]

If you cannot see anything obvious then check the IEC mains socket closely, the old ones can blow the filter caps that are inside them, they usually smell incredibly bad, you may even see residue by the socket where it has vented.

You can source replacement IEC sockets from various suppliers, just find one that fits.

[CodingMarco]

I removed the 6 capacitors in the middle and found that one of them was leaking and the PCB underneath it is more or less fried.
Without the caps the output is pretty noisy (the voltage difference has a pk-pk voltage of 400 mV --> or is that fine?) but the clicking noise is gone apart from a few spikes when it is turned on which which I think should be normal. 
But I will still replace all the caps on the main pcb and post an update here.

EDIT: The "noisiness" is actually a good sign. This means, that the inverter is working and active.

[TheDefpom]

Yep that would do it.

[CodingMarco]

Hello y'all, here is my update:

I've desoldered all caps and only the one already mentoined was leaking. I've checked the large caps on the primary side and they seemed to be fine.
After soldering new caps to the board and turning it on I was disappointed at first because the clicking was still there. But the final solution was suggested by vk6zgo:

Some older SMPS don't like operating without a load & will go into "hiccupping mode".

After connecting a 25W, 2 Ohm resistor across a 5V pin as described in the service manual, the clicking was gone and I could trigger it by (dis-) connecting it again.
So it seems like the controll circuit sees no load as a faulty condition and tries to restart the inverter and "the clicking noise is the transformer core responding to those short bursts of energy from the inverter" (from the Tekrtonix troubleshooting guide)
I checked all output voltages and they were fine. The reassembled scope also works perfect and all selftests are successful!
If anyone needs a list and the positions of the capacitors, I've included them as attachment.

We have a room full of old HP and Agilent equipment at my Keysight site so maybe I can get something else to repair at some time because afterwards it was quite exciting

Thanks to everyone who helped!

[radicalrendell]

I'm glad I found this thread... Same symptoms and same smells.  I looked at the photos and saw the burn mark at C12, hmmm, time to double check... I checked my PS earlier and saw with no visible burns or popped caps, BUT, with a much more careful look and better lighting, I saw a small burnt blob of electrolyte(?) near C12.   

I have some caps on order and plan to swap out one... just to test and do a ripple voltage check.  I'll replace the others as long as they are 35V 1000uF.

The problem isn't solved yet, but I think I am close to having it working again.

I'll attach my photo of C12 which may help others....

[james_s]

That's not a burn mark. That appears to be leaking electrolyte. Feel it to see if it's wet, if it is, wash it off ASAP because it will corrode the heck out of traces and any other metal pieces.

[radicalrendell]

So, I replaced 10 of the 35V 1000uF (since I bought 10 of them), cleaned up the flux residue and electrolyte.  I resealed the a 4 square mm area where the conformal coating was blasted off, with a dab of green Tremclad oil based paint (it just happened to be green)

I then precariously balanced the power supply, 2 ohm output resistor and o-scope chassis so that I could get the power input connected.  Power on.... everything powered up nicely, low ripple. Accurate voltages.  No smells. No sparks.  I put some globs of silicone as the factory had done and reassembled the power supply.

I reassembled the scope and powered it up and we are back in business.  Thanks everyone for their contributions to this thread.  I am sure the one cap was all that was needed to fix this, but feel better knowing that I have bought hopefully another 5-10 years of life before the next cap goes south.

FYI the old caps measured in the range of 968uF-946uF (including the bad one)

Attached: pic of new nichicon caps...