Oscilloscope SMPS repair - where to look?

[c4757p]

OK, time to pick the EEV brains again. I've got yet another scope to repair - a Hitachi V-665. The PSU squeals loudly at about 10kHz. I've checked all the capacitors and diodes, and as far as I can tell, they're good. None of the rails are shorted, and unplugging either one of the two power output cables doesn't make it happy (so I'm assuming the fault isn't on one of the other boards). Nothing overheats, even if I allow it to run like this for a minute or so, but nothing works correctly either.

+110V, +55V, -12V rails are reasonably close to regulation (I don't have the exact numbers with me, but about 120V, 58V, -12V). The user manual doesn't specify what the acceptable tolerance is, and I haven't been able to find a service manual. However, the +12V and +5V rails are down to 9V and 4V, and as mentioned above, oscillating angrily. Tweaking the "+12V adjust" pot just makes it louder.

Attached are scope captures of the +5V rail before and after the output LC filter, and the relevant schematic page. The other rails look exactly the same on the scope, so I didn't bother capturing them.

Unfortunately, all the interesting functions are mains-side (and also it's a massive pain in the ass to disassemble), so before I start dicking around aimlessly with an isolation transformer and my safety, I thought I'd ask if anyone has any suggestions of exactly what to suspect. I'm not 100% sure I understand at a glance how the controller chip works, though I'll dig into the schematic a bit more before doing any real work. I haven't been able to find a datasheet that isn't in Japanese.

A few thoughts and comments:
- I'd assume the STK7308 controller/switch chip is good, since it is running, if poorly.
- Since only two of the rails are out of regulation, even though the only thing linking them is the turns ratio of the transformer, I'd normally suspect the transformer. However, the actual scope circuitry doesn't appear to be running properly, so I think it's safe to assume that the loads on the rails aren't distributed the way they're supposed to be.
- Complete shot in the dark: Maybe the soft start bypass TRIAC is fried and isn't closing? Would an SMPS stuck in soft-start behave anything like this? And do they often fail open?
- Would it be risking damage to disconnect the power supply and externally supply power to just the +/-12V and 5V rails (not the 55V and 110V), to see how they behave when supplied with a known good PSU?

[SeanB]

C1507, C1508 and C1512 would be good bets to change in the first place. The higher voltage loads probably have enough capacitance not to drop when the supply is oscillating, but the low voltage rails have a larger load. The triac will be working, as otherwise the smoke would be coming out of R1501. With these Sanken supplies I am always suspicious of small electrolytics on the primary if they are not working well and the outputs are not dead shorted.

[Paul Price]

I'd put my money on C1561 and C1551 and possibly C1552 but I can't tell because you've not presented a scope trace here.

You should scope the capacitors with a ground reference on the left side of the small L1510, L15aa chokes that connect all the filter caps to ground, but that would mean you would isolate the ground lead of your scope and you wont do that..I would.

At telltale sign of a diseased cap is any bulging of the vent area on top of the filter caps, a visual check can be quicker than a scope to find the problem.

You can make a good check of the inrush triac circuit by simply unplugging the power supply and checking (oops don't burn yourself) the temp of the resistor it shunts. The resistor should be reasonably cool if the triac is doing its job.

I would use an AC multimeter to check the AC ripple across the caps in question, it is a safe and isolated way to spot an open or opening capacitor.

I think measuring the DC voltage and AC ripple just post input bridge filter cap point,  this is also an important place to look  at. The voltage should be in the 350V range and ripple should be <10V RMS.

[c4757p]

Thanks for responding quickly as always!

I'd put my money on C1561

I'll check it more carefully. It didn't stand out with a visual check and in-circuit ESR probe (admittedly probably inaccurate because of the filter chokes), and it's a top quality Nippon, so I didn't suspect it, but I will. I've got an appropriate replacement on hand, so I can just test by replacing it.

[Paul Price]

Measuring ripple with a MTM and/or shunting a new capacitor across a suspected cap is much faster, easier trouble-shooting technique that results in quickly finding problems due to capacitors.

There are just two reasons for the +5 and +12V supplies to sag, they are either not supplying enough power because of power supply problems or the circuits they connect too are greedily asking for too much. Disconnecting the load, if practically possible, and shunting the gap with a .1 or .01 ohm resistor will reveal the current drawn by the respective supplies with ohms law and a low-voltage reading across the resistor. I sometimes do this using an razor to cut a tiny paper-thin break in the PCB copper (that can be easily repaired with small wire bridge after tests) and insert the current-sense resistor across the gap.

[c4757p]

Good point.

[Paul Price]

Connecting an external power supply to surrogate the existing is a great idea only if the circuits they are supplying are not in a fault condition and currently asking for too much while offering too little.

[c4757p]

Nope, the output caps are fine. I tried putting known-good, low ESR capacitors across each of the output caps, including on the 55V and 110V rails, with no change in behavior at all. Took one out and tested it out of circuit as well, and it's fine.

I doubt it's the load causing the problem - as I said, there are two output ports on the power supply, and I tried disconnecting them one at a time with no luck. I'd be surprised if both boards failed on the same rail. I'll try inserting current shunts, but I'll replace Sean's caps first (honestly, if he thinks they're likely suspects, I'd want to replace them anyway, in case they go bad in the future).

1) Since the PSU schematic lists the output currents, perhaps I will try disconnecting the loads and temporarily placing dummy load resistors.
2) If I were to use an external power supply, I'd make sure to set a current limit at the PSU's rated output current.

[Paul Price]

I was referring to the electrolytic caps on the +5V and +15V to the left of, prior to the 5V output caps..did you bridge those? You should not see the waveform you posted for the +5V at the left of the two decoupling chokes connecting to the common ground.

Replacing capacitors that check out to ensure they won't fail in the future is IMHO a complete waste of time. Most capacitors will never fail, some most stressed will likely fail, and life is too short to spend any time trying to predict their future. I would liken it to going to a doctor with a broken leg and then having your tonsils taken out, and unless your real name is Michael Douglas, it would not be a good idea.

[c4757p]

Yes, that's what I meant. I bridged exactly the one you said.

[Paul Price]

I would solder a wire across the two decoupling ground choke and re-scope the +5V and +12V output with the ground lead of the scope at the wire bridging these chokes. I would also do a digital (finger) temp test of D1561, it may be breaking down or just overheating because of an overload condition.

I would really like to see a scope of the +5V  at diode cathode of D1561 and filter cap C1561 under these conditions.

L1510-11 chokes are only a noise filter and would not endanger any circuit component  by this temporary test modification.

[Paul Price]

If the caps and diode checks ok, and output currents are ~ correct, then I'll move my bet to  c1512 and check R1522, and then  my new bet on C1512 as the cause, this side of next really suspecting the STK7308 is gone sour.

But this iSTK7308 dea contradicts the fact the HV supplies are near their correct  output voltage, so an overload condition on the +5V supply and +12V supply must be checked first.  Since you say they are reading slightly higher, this quickly suggests that there is an overload condition on the +5 or +12V supplies or the diode I suggested is at fault.

[c4757p]

Erm... well, new development. In the process of checking the temperature of the diode, I let it sit and scream its heart out for about a minute. Naturally, it went from "God I hate faulty power supplies" to "what the hell?" when it suddenly started working...! I turned it off and back on a few times and it booted up just fine every time (the supply lets out a tiny squeak and then shuts up). All is well, except I don't think a single trimpot is in the correct position.  I'm letting it "cool" (yeah, right, it's 90F in the house...) for a while to see if the bad behavior returns.

There's no digital readout, but that could be operator error, or for all I know there's a trimmer out and it's being displayed off screen. The trace is fine.

By the way, when it's oscillating, the currents on the 5V and 12V rails, respectively, are 200mA and 340mA, definitely within limits. They increase just slightly when the scope starts working properly. The STK7308 is comfortably warm.

Bypassing the ground filter chokes and probing from there only makes the peaks of the output voltage spikes slightly taller, predictably. The waveform is identical whether I probe on the capacitor or the diode.

So... what failing components work once they warm up?

[bilko]

Dry joint ?

[c4757p]

That was my first thought. In general, the soldering looks fantastic, though I may go over a few of the joints again just to be sure.

[Paul Price]

Old men and oscilloscope power supplies, that's what.

The god of troubleshooting invented the hair dryer, freez mist and the plastic handle of a screwdriver to bang around with to help diagnose this kind of fault.

Even old men sometimes get lucky.

[c4757p]

Well, I think I'm onto something here. As it warms up, the voltage output drifts upward slightly. If I tweak the trimmer, the highest I can make the 5V rail go is 4.1V, with the 12V following at around 10V, and that's when it's working properly. So I suspect what's really happening here is that the scope malfunctions because the output voltage is too low, and eventually comes into proper function as the voltage reference warms up and the output climbs. The oscillation may be due to one of the loads malfunctioning during the undervolt conditions (I haven't probed the current on the 55V and 110V lines (the 55V line actually supplies significant power, including the entire CRT+heater), and also, the CRT flyback driver and horizontal amplifier are on the same board as the PSU and cannot be easily disconnected, so god only knows what they're doing. I also haven't probed the current with an oscilloscope, only a multimeter, so there may be something interacting with the PSU at its switching frequency and upsetting the loop.)

The whole thing is packed with somewhat precariously biased discrete transistor circuits, so I wouldn't be surprised to find it does bad things when the power supply is too far out.

I've got to take a break now (time to start making dinner), but after the break I think I'll go probing some resistors and diodes in the feedback loop.

[Paul Price]

The power supply is a self-oscillating regulator that works best at a certain load, so adj. the output to the correct voltage brings the circuit into a comfortable operating state.I am thinking of some bad solder joint, esp. around the STK chip and the power rectifier diodes.

The hardest things to fix are them that ain't quite broke. If it is working, then bang the PCB about a little to shake up the cookies, it that fails, try hot/cold slow torture. If that fails call it fixed.

Check R1504.

[c4757p]

R1504 and its companion R1503 are fine.

[c4757p]

After giving it some time to cool off I'll report back on whether fixing a few solder joints has fixed it. I think it may have, but I want to wait until the chickens hatch before I start counting... I still want to investigate and repair the undervoltage - I currently have the voltage trimmer at maximum.

[c4757p]

And the chicken never hatched. But after playing with it a bit, I am almost certain that the problem is caused by the bad voltage control.

[SeanB]

Change the little caps then, they do have a bad habit of going high ESR or low in value ( sometimes very leaky with no other symptoms) and i am always wary of them in that area. Use higher voltage ones that will still physically fit.

[c4757p]

Yep. Paul not wanting to preemtively replace aside, I want to be able to sell this when I am done and want to know it will continue working for a while. Just have to grab some good ones, the ones in there are very good quality and I refuse to downgrade.

Think I've found a leaky transistor, so more on that later.

[c4757p]

The transistor was indeed failing, so I replaced it, but it did work, and replacing hasn't improved the function of the power supply. But....

Notice above where I said the readout wasn't displaying either? I figured that was an unrelated problem. Well, as I've said many times, things just don't have multiple faults. When something fails, people stop using it, and it doesn't develop another. And, of course, I finally figured out the rather bizarre arrangement of the readout "system", and noticed something I overlooked: obvious thermal damage on a section of a board responsible for the readout! So once I figure out the bad part (lots of SMD stuff on this board...) I'll replace it, and hopefully that was the problem!

[c4757p]

I shelved this one for a while out of frustration, and finally came back to it, and as usual, SeanB was 100% correct. C1507 had over 60 ohms ESR, and C1508 was a complete oddball - couldn't even get it to balance on my LCR bridge. Replaced both of them and the squealing was much quieter. Swapped another near the STK7308 (can't be bothered to look up the designator) and it became completely silent. Now I just have to grab a couple replacements of a quality I can actually trust.

The low output voltage appears to be a separate issue, as I still can't get the rails up. (In fact, with the caps replaced the rails are even lower, and the +5V won't even hit 4.3V, so the thing doesn't boot up.) But that shouldn't be hard, just a toasted resistor or Zener diode or something. Nothing is overloading the supply and it's not straining.