Exploding bypass caps

[rexxar]

I just built up a bench power supply from a kit (this one), and when I was testing it, I shorted the output, which caused one of the bypass caps (C3 in the attached schematic) to blow up. Okay, so much for "short circuit protection". I replaced it with one I had on hand of the same value, and carried on. I then tried pulling 1A at the rated 30V, and the other bypass cap (C2) exploded as well. 

Judging by the pcb footprints, a much larger cap was designed in here, but was changed out for something smaller. Same thing happened with one of the main filter caps. I wrote to customer support, but I don't really expect anything useful, so I'll ask my question here.

Is there another value cap that's supposed to be here, or just one rated at higher current? The ones they gave me are puny little 16v 10uf caps.

[ejeffrey]

Did you put tantalum capacitors in backwards?

[rexxar]

Did you put tantalum capacitors in backwards?

No, they're electrolytics, and I put them in the right way round both times.

[c4757p]

Got a DSO? Stick it across a capacitor, then make it blow up and see what the capacitor sees.

On a hunch, stick diodes antiparallel to the outputs and try again.

[TerminalJack505]

They didn't spare any expenses there, did they?  16V caps for a 15V rail.  I'd put some 25V or 35V caps in instead. 

10uF should be fine.  You usually don't want to put more capacitance on a power supply's output than is necessary--which is basically enough to keep the regulator happy.

[ejeffrey]

I wouldn't use 16 V caps in a 15 volt supply, but it should still work, even if the longevity is compromised.  Could be the caps are fakes.  In principle the output could be oscillating, and the ripple current is blowing up the output caps, but I don't see any reason that would happen in this circuit.

[rexxar]

They didn't spare any expenses there, did they?  16V caps for a 15V rail.  I'd put some 25V or 35V caps in instead. 

10uF should be fine.  You usually don't want to put more capacitance on a power supply's output than is necessary--which is basically enough to keep the regulator happy.

Well, I had some 25V caps in my junk bin, but when I put them in, I accidentally put the one for the positive rail in backwards, and the magic smoke got released from something. 

Now the positive rail is dead, and the negative seems to be stuck at about -25V regardless of what I turn the pot to.

Did I somehow kill both the regulators? Nothing inside looks burnt, so I don't know where the smoke came from, but I definitely smell burnt silicon.

[megajocke]

If you short from negative to positive output, the strongest regulator will "win" and reverse the polarity of the other output, for example driving +15V into the -15V output which would blow up C3. Diodes to prevent reverse-biasing of those capacitors would be well advised.

The designer seems to have only considered shorts from either or both polarities to the output common terminal.

Your new problem could also be caused by an open potentiometer or some bad connection, for example.

[ElectroIrradiator]

You probably blew or weakened the LM317 on the positive supply rail, when you shorted the positive rail. Both regulators are missing a protection diode across R1 and R2, respectively. These are needed when C1 and C4 are used. Check the datasheet for the LM317.

The open ends of VR1 and VR2 is a bad design decision. If the wiper arm goes open circuit, then the output voltage will soar to the full, unregulated input voltage across C5/C6, probably killing C2/C3 and perhaps even the load. The wiper arm should always be connected to one end of the potentiometer. This way the circuit will - worst case - see the full resistor value of the potentiometer, if there is a bad connection at the wiper.

A preferred solution would be to connect the pots in series with R1 and R2, with a fixed resistor located where the potentiometers are now. This requires calculating new resistor and potentiometer values, to ensure the regulator pin current stays within spec, yet would result in the output going down/toward common if there is a contact problem in either potentiometer.

[cube1us]

Interesting.  I built a dual output power supply from a Radio Shack "curkit" back in the day.  I have shorted its output lots of times, had it fight with batteries, etc. etc. and it just works.  It also has built-in switchable current limiting, and dual meters.   (No +5 fixed supply, though).  The thing looks kind of awful, but just works.   

[rexxar]

They're sending me a replacement kit. To beef up the circuit, I should replace the 16v caps with 25V, and if I'm reading megajocke correctly, I should put a diode in series with the caps? Does it matter what type, or just anything I find in my junk bin?

[c4757p]

Just standard 1N400x. Not in series - across each output in antiparallel, cathode to positive.

[rexxar]

Just standard 1N400x. Not in series - across each output in antiparallel, cathode to positive.

Why would we want to let a reverse voltage pass through, rather than just block it?

[David_AVD]

Just standard 1N400x. Not in series - across each output in antiparallel, cathode to positive.

Why would we want to let a reverse voltage pass through, rather than just block it?

You don't want to add voltage drop in the output lines.  Even if you did, the Vf of the diodes will vary (with temp and load) and lead to less accurate outputs.  You could sense the voltage after the outputs, but then you've just come full circle and defeated the protection for the feedback.

[ElectroIrradiator]

Why would we want to let a reverse voltage pass through, rather than just block it?

When you disconnect an inductive load, there will be a potentially very powerful back-EMF voltage spike. This is how the inductor attempts to keep the current flowing at its previous value. A diode in series with the output will do nothing to help in this situation.

With a 'catch diode' connected in reverse across the output, you will still get a small spark at the contact point, but the reverse voltage spike within the PSU is kept to safe levels.

[rexxar]

Okay, so replacement came in the mail, and it's not blown up yet.

Interestingly, in this one they gave me 25V caps instead of 16. I'd already bought 36V replacements for the ones in question, so I used those, and stuck a 1n4007 in parallel with them, all good so far. I also changed R1 and R2 to 1/2W resistors because the 1/4W ones looked a bit singed from overheating 

Anyway, I wanted to look at the output on my scope, and I noticed that there's still voltage on the output for a long time after I turn the supply off. Since the power switch is on the primary side of the transformer, it has to be charge stored in the caps.

My guess is that discharge resistors across C5 and C6 (and C7) would be in order, but I don't know what value to put in. Second, what's the purpose of these caps? Is it just filtering?

Also, the output of this power supply is really clean, but I guess that's to be expected of a linear supply?

[mariush]

A 1 kohm or something a bit higher would probably be a reasonable discharge resistor.

As for why that capacitor is required, have a look at the LM317 datasheet.

[MrAureliusR]

I'd already bought 36V replacements for the ones in question, so I used those, and stuck a 1n4007 in parallel with them, all good so far.

Did you make sure to put the diodes in antiparallel as c4757p mentioned?

[rexxar]

I'd already bought 36V replacements for the ones in question, so I used those, and stuck a 1n4007 in parallel with them, all good so far.

Did you make sure to put the diodes in antiparallel as c4757p mentioned?

Yes, and 1k resistors across the caps seem to be working just fine to discharge them.

Thanks for the help, all!