Faulty SMPSU.

You're running at 220V? OK

Make sure the choke ends aren't touching as they leave the toroid center. Wires not shorting to core. No sign of cracked cores.

Check big primary bulk electros - voltages should be roughly equal with small low-frequency ripple.

Replace small electros on primary side, then optocoupler.

A basic forward converter. Integrated primary switch should be replaceable, but check transformer Lp first.

RL

Check.

Do you mean the opposite ends RL, in as it passes past out? If so, no they seem to be a good 2mm apart.

Measurable just with my DMM? (if so, I can't measure anything).

Do I only have one on the board (the big one in the final stage (is it?)) and if so, not that I can easily see?

I was inspecting those earlier and noted the caps were spanning an 'island' and each cap only 200V (so in series where the working voltages add)?

Would AC volts on my DMM be ok for that RL (I do have a USB scope but I wouldn't want to use it on mains and I haven't modded it to be AC coupled yet).

The ones near C18? (just for reference, C16 is 100uF, 25V and C17 is

4.7uF, 50V). I have one of those automatic component testers. Would an 'ok' from one of those be ok?

An EL817. [1] (there is a TL431A (voltage reg?) beside the opto FWIW).

I agree it doesn't look 'overpopulated' as such (thank goodness!). ;-)

Is that 5 legged device on the underside?

Sorry, you have lost me on that one. ;-(

Cheers, T i m

[1] When I was a datacomms guy we used to make and sell a Current loop 'Line driver' that used an opto for the receiver. They would often be sent back marked as faulty, we would bench test them and loop the output back to the input and they would often pass (BERT). Heat, cool, tapping, still ok. Then I designed and built a small box where we could vary the input current to better represent true line (loss) conditions and lo-and-behold, some of those that were NFF now failed because their sensitivity had dropped, putting the sense threshold right near the TX current level. We found a new opto would trigger at around 10mA, the max current from the TX was 20mA and some of these returned 'faulty' ones were nearer 18. I had to build testers for all the techs. ;-)

166V across each, no real measurable ripple (off load of course).

Cheers, T i m

To test C16 and 17 you would need an ESR tester. It is a pretty good bet that both are bad. Make sure that you use 105C caps.

TNo, that has to be the low-voltage rectifier (if it really IS rated for 20A, it has to be heatsinked); if it's leaky it would be a quasi-short-circuit, and that will deliver AC to the capacitors you replaced, which would kill 'em soonish.

That's the likely culprit, probably will need to be desoldered to find the part number, and will be easy-ish to test. Sometimes there's two diodes, center pin common, in this form factor.

Yes, I have one. Of the 4 x 1000uF smoothing caps across the output (where one had blown) 3 had an ESR around 0.1 and the blown one a touch higher? Is that right / normal (FWIW they also showed a Vloss of ~2% and were all around 1100uF)? [1]

Ok, I'll pull test / replace those next.

Check (I did so with the 4 smoothing caps, but didn't check the lead spacing on my order (doh)).

Cheers, T i m

[1] Could it be bulging because it just happened to be the closest to the output choke and that had been running fairly hot, rather than it had failed by drying out or whatever? (they were nichicon 105 1000uF 16V YV(M) H1305)

The two TO(xxx?) packaged devices under the board were bolted to the square U section case with insulators. eg, the whole ally plate chassis and perforated ally 'lid' would be a heatsink.

OOI, as it seems this PCB was originally designed to be 5V / 40A, so is it likely they would have used the same component there for 12V /

20A (I know it has to dissipate the same number of watts but wondered if it might be able to handle 20A easier if it was originally specced to take 40A)?

Ah. ;-(

Understood.

Yes, I think I've seen such on the various SMPSU diagrams I've been checking. ;-)

As advised elsewhere, I think I'll pull and test (and possibly replace

*anyway*) the two small electrolytics first and then move to the rectifier.

FWIW, I do read around 10V DC on the output but it's fluctuating. I could put my PC digiscope on it (after checking the output for AC on my DMM)?

The wires without magnet wire enamel look pretty close to me. You should be concerned that wires exiting core in different directions are not visibly or physically separated. Force these apart with a flat screw driver and insert biz card thick paper stock.

You're looking for discoloration, crumbling of core where magnet wires are forced against core edges due to winding force. Discoloration is sign of point-contact heating.

If you don't hear grinding noise or see shifting, when manipulated, then there's probably no issue.

Your later description sound like these are OK. Large voltage imbalance or large ripple would indicate failure of one or both.

Main transformer primary inductance of a forward converter running off mains will be more than 400uH - typically 2mH.

Lower values suggest shorting of windings or terminal connections/traces.

RL

I think it may be my bad picture, IF we are talking about either of the pairs of wires as they pass in and out of the ring choke?

Ah, right, then yes, there is a good 2mm between them.

A stack of 5 biz cars would fall out. ;-)

Can I just confirm something please? The output choke seems to be bi-wound, two wires wound side-by-side so in parallel, both going to the same pad at each end so carrying the output current between them?

So, it wouldn't matter if the 'pairs' of wires touched as such, as long as the input and output wires didn't as they will be at a different potential / phase?

Check.

Well I think the output choke windings have been running hot as the enamel looks darkened compared to those bits that were likely cooler as they were in the air and had heatsinks in the form of the soldered connections to the board but I can see no obvious signs of failure of the enamel.

There doesn't seem to be?

Ok, I just pulled these had found:

The 100uF, 25V: VLoss= 1.8% Capacity= 90uF ESR= initially 2.1 ohm then dropped to .69 when repeating the test.

The 4.7uF, 50V: VLoss= 3.0% Capacity 4755nF ESR= 5.5 Ohm and stayed around that after several repeats of the test.

So, is that our 'bad' cap?

Ok, I'll try that now ('strike whilst the iron is hot'). ;-)

Understood.

What I think is the primary (marked in red) was: .4 ohms, 17.4mH.

Another winding on the primary side (marked yellow and going to the small caps etc) showed: .3 ohms, .08mH

FWIW, the output seems to be made of 3 windings in parallel (marked in green and tested wired that way) measured: .02 ohm, .25mH.

Does any of that help?

Cheers, T i m

They're shorted on the printed wiring, so just simple 2wires in parallel.

On closer examination of choke, I see it's been impregnated, so there's little likelihood of fault due to enamel failure or vibration.

Anything serious would have showed up measuring Lp.

These are often part of feedback network for simplest integrated switcher ICs. Probably not the culprit, but low side of tolerance on an electrolytic is indication of end-of-life. Changing esr can indicated corroded/intermitent inner contact.

Small 4u7 can have esr in the mid-ohms range. Probably not an issue on housekeeping supply, but parts with >Main transformer primary inductance of a forward converter

That's ok as long as measuring units are correct (common mistake). Fairly high value for this power level and topology, but only low side limit is important.

No surprises.

Integrated switchers can be finicky when it comes to self-generated noise. This makes snubbers and clamps pretty important. As a last resort, check primary parts wired across main primary winding and secondary side RC before replacing the big integrated switch.

RL

Ok, so just to increase the current carrying capacity or something more subtle? I'm more of a DC / LV / serial datastream sorta guy and know 'frequencies' can do weird stuff sometimes (skin effect etc). ;-)

Good news, thanks.

Noted.

Ok, and cheap / easy enough to replace so ...

Ok.

D7 checks out as a std rectifier on my DMM (ohms Diode test).

My tester actually displays the values as quoted so I'm happy there. ;-)

Ok.

Good news. The txfmr also 'looks' ok FWIW, no burn marks etc.

So that's the 5 pin TOxxx case device?

Do you mean the small electrolytic's, diode and other caps (is that what they are / doing)?

So are we still on for replacing the opto device RL? I don't think my tester can do that, other than the opto side as a diode possibly and if it wasn't so cheap and already fairly old, I'd set up a test for it myself on some breadboard. ;-)

Just to give me a better visualisation on all this, could you point me to a schematic of something similar perchance please?

Cheers, T i m

Optos get leaky, light seals break. - in most feedback loops that means it's turning the circuit off without assistance from the TL431.

You'd probably get better app notes chasing down the actual part number in the unit.

RL

Update: Whilst waiting for the smaller components to turn up I've de soldered the two devices on the rear of the board to check the part numbers and (as you say), the three legged one is a rectifier, a '30H150C' and I've tested it with my DMM (diode test) and it checks out ok on that (centre pin to both outsides, both ways round etc). I hope it's good under load as they don't look as easy to find as the switcher, a KA1M0800?

Cheers, T i m

Any change in performance with ANY of these part subs?

RL

I'm still waiting for the small caps (might be here today, whilst I have some stock of caps, I never seem to have the spec I want), and I intend, assuming this wouldn't risk any other stuff, to replace the two caps, then opto, voltage reg then switcher in that order (as I believe that was the order you mentioned the likely cause)?

The idea being that it might be interesting to see what it actually was?

As an aside, I replaced the 4 x 1000uF, 105, LESR output smoothing caps with new caps but with the wrong wire spacing (I was jumping about on eBay and clicked the wrong one) but have some good s/h 1500uF (16V 105, LESR) caps with the right pin spacing that would fit better. Would using higher capacity be 'better', given it's soft start etc or would it negatively influence the L-C of the output and so end up being 'worse'?

Cheers, T i m

You are, occasionally, reapplying power to check function?

RL

Update: The caps arrived this morning so I re-fitted the txfmr, the switcher and rectifier, then fitted the new caps and tested it. DMM shows around 10V on the DC range (as before off load) but it's jumping about a lot and about .5V on AC.

I then replaced the opto and then the voltage reg with the output looking the same each time (allowing for the voltage on the main caps to bleed down each time etc).

So are we now looking at the switcher itself and / or is there anything else I can check in the meantime please?

Cheers, T i m

I think we just keyed up at the same time? ;-)

After re-assembling and between replacing the components, yes.

Cheers, T i m

Coming back to this before buying / replacing the switcher, I tested it out of circuit on my DMM (diode test, all ways etc) and it seemed to check out ok?

Whilst rummaging though my spares last night I came across a similar looking device (with the two diodes facing the centre pin marked on the case) so assuming the spec checks out similar (at least 12V etc), should I try swapping that in and see what happens, in case the other is breaking down in use?

FWIW I measure about 10V DC on the output (no load) and only about .5V on the AC range (ripple / noise)?

Cheers, T i m