All replacement capacitors in iMac G5 PSU burned up

[gee]

Re: All replacement capacitors in iMac G5 PSU burned up

UC3845 is dumb old SMPS IC and not fast enough to sense trouble. And came from same company who designed original UC3482.

This type of crappy IC will keep trying (ticking) and blow up (not fast enough on some situations especially with bad caps) and do strange things too. There is electrolytic caps that is very important for proper function of the UC3845

I've designed and worked on several UC384x based supplies. Behind the TL494, it's probably the most common, well understood SMPS controller on the planet. It's definitely not crap if it's implemented correctly.

Toasty: I'll look into that when I get home later today. Do these supplies have a habit of burning vias between top/bottom?

[Toasty]

Re: All replacement capacitors in iMac G5 PSU burned up

Not that I'm aware. Found 2 of 3 units here with cold joints there. Since I am not working on them at this time, I'm trying to offer info as I find it.

I am intrigued by the crowbar idea on the mobo. I am wondering if the 3.3v is rising and tripping the protection. That would start to make sense of the coil burnout.

Updated (Values & designations) JPEG of the 3.3v output with burnt coil attached.

Toast

[gee]

Re: All replacement capacitors in iMac G5 PSU burned up

I just had a chat with a SMPS guy here at work, who had some insight. L6 is probably a saturable inductor, and is used as a mag-amp circuit. There's a bit of circuitry somewhere here which tags on between L6 and D33 and varies the saturation of L6 to regulate the +3.3V output.

Searching a bit more on Google, check out this book (click 'preview') pages 800-804. From the schematic on 804, it looks like this circuit could be hiding amongst the 'jellybean' SMT bits on the supply that I haven't considered yet.

http://books.google.ca/books?id=-R2vt2L03l0C

Anyhow, a failure/instability of this circuit could cause the +3.3V rail to go high, and perhaps that crowbar circuit does exist on the G5. More digging to do when I get home...

[Toasty]

Re: All replacement capacitors in iMac G5 PSU burned up

The other point that was raised by PCBONEZ several pages ago (#54 LOL) is that L8 is feeding the pi filter (C52, L10, C56) here:
>>It's not part (o)f the PI filter, it's a current surge limiter before the PI filter.<<
>>So, if the TOROID was marginal and then you put in nice STRONG caps [that are capable of charging much 'harder' than than the old ones] that may have been enough to push it over the cliff.<<

Initially, nwd's recap resulted in a hot PSU and the smoking coil. Prior to recap it is not known if that coil was damaged or not. Its black paint likely hid any prior overheating. Yet, the boards in the ones I have are not discolored and all seem to be post recap overheats. Is there a pattern here?

TOast

[gee]

Re: All replacement capacitors in iMac G5 PSU burned up

Confirmed, L6 is a saturable and there's a mag-amp circuit; Q14 controls L6's saturation through D24. Ultimately, regulation is performed by TL431 part M9. I haven't traced the whole circuit out yet, the layout of the board is horrible for tracing out and I'm too busy doing other stuff.

If the circuit fails such that L6 is almost always saturated, the +3.3V output will rise to +5V. Also, I don't know if this magamp controller circuit is sensitive to the amount of output capacitance, output capacitor ESR or anything of that sort such that recapping the supply causes it to blow...