NEC APC III Power Supply Repair (FIXED)

[sean0118]

So I'm trying to get an old (1985) NEC APC III computer up and running as a project.

The power supply (YEC PU506) seems to be broken, the secondary isn't putting out any voltage. Rectified mains is present but the primary side transistor is not being switched.

I think I should be okay with the repair, but the primary side controller is a green zip package with "YEC YD-010 4Y2" printed on it. This seems like some very old obscure part that I can't find any information on... does anyone have a datasheet or recognise it as a generic smps controller?

[TheMG]

These are, as far as I can tell, custom hybrid circuits made specifically for these power supplies. They were pretty common at a time when fully integrated SMPS controller ICs were not commonly available. It allowed power supply manufacturers to cut down on the size and cost, while also avoiding the processes of having both SMD and through-hole parts on the same PCB, which was much less common at the time.

First things first, I would check all electrolytic capacitors. Pretty common for a bad cap in the primary side to cause a SMPS not to switch. Also check the power resistors, in this older design they do run quite hot and it would be possible for one to go open after years of thermal cycling. In these designs, they typically used large resistors in combination with a zener diode (could be internal to hybrid or external) to power the control circuit direct from the rectified voltage full time (instead of the more common arrangement these days of a feedback winding providing power to the controller IC after the power supply has started switching). Similarly, check for cold/cracked solder joints, especially around the power resistors.

You can also check that the hybrid is getting power to operate. You can find it by locating and measuring voltage across any small electrolytics that have the positive leading to a pin on the hybrid, and the negative to rectified common. The positive side of the cap should have one or more power resistors from it to the rectified positive. If you find this in the circuit, then you've found where the power comes to the control circuit. It will normally be in the range of 10-30V give or take, and that depends on the design. But if you see no or very little voltage there, you know that's your problem.

[sean0118]

Thanks, that makes sense, hopefully it isn't the controller which is damaged then.

I have already replaced the smaller electrolytic caps on the primary side, but just with ones I had in my junk bin, good quality brands but one was slightly down from its rated capacitance. I also don't have an ESR meter. Eventually I'll replace all the smaller electrolytics with new ones.

In these designs, they typically used large resistors in combination with a zener diode (could be internal to hybrid or external) to power the control circuit direct

That's interesting, I didn't know that, there are some zener diodes and very large power resistors so I'll look into that. I guess I could remove the zener diodes and test their reverse voltage too.

You can also check that the hybrid is getting power to operate. You can find it by locating and measuring voltage across any small electrolytics

That's another good idea, I think I'll do that first. 


I'll upload some more pics soon too, I need to resize them because most of them are above the 5MB attachment size limit.

[sean0118]

It's still broken, but I did a bit more investigating. 


I found the power rail to the primary side controller, it's only 4.3V.

Seems kind of low, but the capacitor is only rated to 10V so 4.3V can't be too far off. The rail itself is feed directly from mains (before the main rectifier), it's rectified by a diode (D2) and a series 18kohm resistor (R3) and then filtered by a 100uF 10V capacitor (C9). R3 measures correct, so I'm not sure why the voltage would be low.

The rail is also attached to the transistor side of an optocoupler. With the collector attached to the power rail and the emitter goes back to another pin on the control IC. I need to investigate this a bit more, I'm not sure what it's function is. The LED side of the optocoupler goes to the secondary side somewhere.


I also found some solder joints that were possibly cracked on some power resistors (R18, R19). Both of these are 100k and measured correct out of circuit. I'm not sure of their function but they are the blue ones with the sleeves around the main primary switching transistor and transformer.

[sean0118]

I did a bit of tracing out (schematic attached)

Still trying to work out what's happening, but looks like some sort of over voltage protection circuit?

[sean0118]

I did something bad...   

I measured the optocouplers in circuit and found their LEDs had 0V across them, and was measuring the same voltages on both collector and emitter of the transistor side.

So I decided to power it up with the optocouplers removed to see if the supply to the IC remained at 4.3V, in case there was leakage through the opto.

When I plugged it in the mains fuse blew instantly, seems the optocouplers were doing something. The switching transistor on the primary is now completely shorted between base, collector and emitter...
that means the controller IC (which drives the base directly) was exposed to the full rectified mains, so there's a chance that's completely fried now.

 

[sean0118]

Got the main switching transistor desoldered, had to take out the transformer first.

It's a C3507 NPN transistor, all pins are definitely shorted together.

[shakalnokturn]

Basics of playing with SMPS: Do not open the regulation loop. On most designs it leads to smoke, either the switching transistor survives and pops all the secondary or it fails first from being driven at a duty-cycle that forces too much current through it.

The transistor would be a 2SC3507, if you're lucky the C-E short will have taken the fuse out before anything got through to the YEC hybrid.
Before replacing transistor, fuse and maybe resistors check hybrid base drive pin resistance relative to supply and ground pins. If you get low values it's fried then no point replacing the transistor.

[TheMG]

I did something bad...   

I measured the optocouplers in circuit and found their LEDs had 0V across them, and was measuring the same voltages on both collector and emitter of the transistor side.

So I decided to power it up with the optocouplers removed to see if the supply to the IC remained at 4.3V, in case there was leakage through the opto.

NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO!!!!!!!

The optocoupler is for negative feedback from the secondary side. As the secondary voltage reaches the desired level (set by either a voltage reference IC or zener diode) the optocoupler LED begins to turn on, causing the transistor side to begin to conduct. This causes the control circuit to reduce the switching duty cycle.

No voltage across the diode is normal if the secondary hasn't reached proper output voltage.

Without a functioning feedback path, either the secondary voltage will rise until something blows, or the duty cycle could become too great and blow up the switching transistors.

Never mess with the feedback loop with power supply energized.

As for the 4.3V to the hybrid, without a datasheet it's not possible to really tell if that is within normal range, but being as you said that the capacitor is rated at only 10V, that could very well be its normal operating voltage. The supply voltage for the control IC in most SMPS is not tightly regulated and has a fairly wide operating range.

[sean0118]

With hindsight it's pretty obvious that was going to happen. I was thinking the duty cycle would be limited to a safe value, but I guess with no load attached even a small duty cycle could cause a massive voltage on the secondary. 

I think there's a decent chance the controller is okay, its gnd is referenced to the emitter of the transistor, so the controller shouldn't have had a large voltage across it. Measuring the pads where the transistor was I'm getting open circuit on CE and BC, with 100R between BE, which is the same as before.

I'm not sure why it decided to start switching with the optocouplers removed, because earlier I had scoped the transistor and found it was not switching at all. I think either resoldering the joints had fixed something or it has some protection circuit which causes it to latch off (but without the optocouplers this didn't happen).

Looking at the schematic I traced out before, I can see that Z4 might be part of the feedback circuit, assuming D18 is a zener diode, but what is the function of the second optocoupler Z5?


The 2SC3507 is obsolete, but looking around it seems the MD2310FX might be a good replacement?


edit: Forgot to mention, I tested the controllers base drive to its supply, it's 4kohm or so which seems okay. Base to the controllers gnd is 100R due to the resistor between base and emitter.

[TheMG]

I'd say the second opto could very well be a shutdown latch signal. The circuitry would have to be analyzed to say for sure.

It is possible the power supply is starting briefly, then the protection circuitry on the secondary side sees something it doesn't like (most likely either over-current or over-voltage) and triggers the opto to shut the thing down. Could be there is a fault in the protection circuitry too.

Pretty typical for PC power supplies to latch off on a fault condition (as opposed to current limiting, foldback, or hiccuping).

A trick I use when I suspect a fault with protection and/or feedback loop in a SMPS on the secondary side, is (with power supply disconnected from mains) I slowly apply external voltage to the power supply outputs with a bench power supply. You'll immediately know if there's a short on the secondary side.  Also, as the voltage is increased just slightly above the regulation point, the opto in the feedback loop should now have forward current through the LED, and drop back down close to nothing when the voltage is below the regulation point. If this happens, then you know the feedback loop (at least the part of which is on the secondary side) is intact. You'll also be able to determine if there's a significant problem with the protection circuitry (ie: triggering a shutdown signal when it shouldn't be).

[amyk]

Basics of playing with SMPS: Do not open the regulation loop. On most designs it leads to smoke, either the switching transistor survives and pops all the secondary or it fails first from being driven at a duty-cycle that forces too much current through it.

No kidding. As another datapoint, a regular PC ATX PSU can easily produce >30V on the 12V output if the regulation loop breaks.

[sean0118]

Good news, it's all fixed. But I'm not sure if it was broken originally. 


So I replaced the switching transistor I blew up with ST BU508AW. I also replaced the 0.47ohm primary side current sense resistor, mains fuse and replaced all the electrolytic capacitors...

But it still wasn't starting up, so I put a scope on its output and found the rails did go to +5, +12 and -12, but only briefly before turning off. It also latches off, so unplugging isn't enough, the primary side capacitors have to be discharged before power is applied again.

I found optocoupler Z4 had over a volt across it, so I pulled zener diode D18, which was marked B3 12, and I measured it's reverse voltage as 12.8V. Which seems okay, except looking at the schematic I posted earlier (that I reversed engineered) it seems like Z4 might always have voltage over it? R27 is 100ohms which pulls one side of the zener to +4.3V, while the other side of the opto is at -12V, so it seems like the opto and zener would always be conducting.

I have no idea what that Z4 circuit is for or how it works, unless voltage over the opto is normal. I should go back and remeasure it now that it's working.


Anyway, none of that matters, because I decided to attach a HDD as a load, and the power supply worked! It must have been detecting a no load condition and latching off.

So I'm not sure why it didn't work originally, I think either one of the other components in the computer has a fault or the electrolytics were bad. I don't have an ESR meter, but they all tested roughly the right capacitance.

 


edit: I also have a second NEC APC III which doesn't turn on, it will be interesting to see if the power supply in that is faulty or not.

[sean0118]

I put this back in the NEC and it turns on now, so there was definitely an issue with the power supply initially. Maybe the capacitors were bad, I'm not sure what else it could have been.


Thanks everyone. 

[TheMG]

Yes, very likely it could have had bad caps initially.

As for the power supply requiring a load attached to work, that was quite common with older SMPS. The control circuits were not as sophisticated as modern SMPS control ICs we have available to us nowadays.

With these older SMPS designs, the switching control circuit has a minimum switching duty cycle that it is able to achieve, but even at this minimum duty cycle, if there is not at least a certain minimum load on the output of the power supply, the output voltage can climb out of regulation. If the power supply is equipped with OVP (virtually all PC power supplies are), it shuts off. With power supplies that do not have OVP, most of the time the manufacturer would include load resistors on the output to ensure a minimum load even with the output disconnected, however, this obviously wastes energy and increases heat output.

Modern SMPS control ICs typically handle this situation by skipping pulses entirely, enough to ensure the output voltage remains in regulation, and therefore have no minimum load requirement.