5170 power supply (ASTEC AA13240) power-on spike bug

[stepleton]

The rapid, helpful response to my 5154 question (thanks @modem7 !) has encouraged me to ask about the other power supply thing vexing my 5170 these days

It's not too pressing because I can use an adapted ATX PSU, but I'd like to restore the original PSU that came in my IBM 5170. There were several different kinds, but this one is an ASTEC AA13240, made in the UK and switchable between 115VAC and 230VAC. I'm running it on the 230VAC mains here, which usually run hot at 245VAC or so. Unfortunately, there appear to be no schematics for this specific unit that I can find.

The PSU was working fine for a little while after I bought the machine, but something has failed and it immediately shuts down when turned on. This occurs while it is loaded with the (large) correct load specified in the 5170 manual --- it also happens when installed in the computer itself.

The overvoltage circuit is responsible for the shutdown: removing the optoisolator prevents the shutdown from occurring. I believe the circuit is operating correctly, because if the overvoltage detection is defeated temporarily, you can observe a sizeable transient on the DC power rails when the power supply starts up. The +12V rail briefly jumps up to around +18V if memory serves; +5V goes to around +9V I think. So the PSU is correct to shut itself down. However, if the PSU is allowed to continue running, then regulation looks good and the unit can supply power at the correct voltage under a range of appropriate loads.

The PSU has been recapped with new, quality aluminium electrolytic capacitors, except for the two big smoothing capacitors on the rectifier PCB. The new capacitors have not helped. I've also tried replacing the opto and all of the commodity ICs, which include a quad comparator and a TL431 adjustable voltage reference. This didn't help either. There is one custom ASTEC IC in the PSU marked "AS4301", but I haven't been able to find out much about it.

I expect that the real answer will come from reverse-engineering the PSU. (I've already done this for the riser PCB that contains the voltage monitoring circuit and some control circuitry, but there is much more to understand on the "main" card in order to deduce the source of the fault.)

But perhaps this failure is familiar to someone out there who can offer a shortcut to the solution?

 

[NeXT]

Did you recap it before or after the PSU started acting up?
A few times now I've caught people using low-ESR capacitors in place of regular grade in these power supplies because they think that will be a good measure since Low-ESR caps were expensive and premium when the PSU was new but the excessively low ripple can confuse the controller which expects to see ripple but doesn't as the rails come up. Depending on how the rest of the switching supply is designed I can theoretically see the controller keeps driving the rails harder and harder until the over-voltage shuts it down.
Yes the rails stabillize but that 18V spike would be a no-no.

 

[stepleton]

The recap came after the failure. No worries, I'm not putting this PSU in service, it's been sitting open on my bench for weeks

 

[modem7]

A few times now I've caught people using low-ESR capacitors in place of regular grade in these power supplies because they think that will be a good measure since Low-ESR caps were expensive and premium when the PSU was new but the excessively low ripple can confuse the controller which expects to see ripple but doesn't as the rails come up.

Example follows.