Compaq PDP-121P - Baked MOV

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

Thanks for your reply I sort-a lost track of this post and found it again

[socketa]

Re: Compaq PDP-121P - Baked MOV

The fact that you can get it to run when you play around with the PS-ON signal for a while tells me either something on that circuit is dodgy, or perhaps one (or both) of the optocouplers are not 100% good after the PWM IC blew up. So if you have any spare parts from other PSUs, perhaps start to swap these around.

Thanks,
i do have such spare parts; so will try that when i get back home in a week or two.

Also, when you are able to get the PSU to start, what voltage does the PG (Power Good) signal measure? If it's logic level "low", that means the PSU is not "happy" about the state of the output voltages and so won't signal the PC to turn ON yet. On the other hand, if PG does go high, then that means the PSU thinks its output voltages are OK and "ready".

Pretty sure that PG logic level goes high when the main rails are up and running - as i took note of that when i tried it on that test motherboard in post #38
I did notice that the rail voltages started to come up on the compaq motherboard; but i didn't check what the PG voltage was doing - i'm assuming that it would briefly be high before the main rail voltages were shut down by whatever was shutting them down (maybe some PSU controller IC?, or whatever is on the compaq motherboard that is recognizing "overvoltage")
The voltages didn't reach their normal spec voltages, so it seems odd that the LED sequence on the front panel indicated "over voltage"

I'll also try this PSU on a good motherboard, and see if it will start up.
And i'll record the voltages if it does.

[momaka]

Re: Compaq PDP-121P - Baked MOV

With the 23W bulb on the 3.3 rail, before i ground PS-ON there is 159.3V and 153.8V across the primary caps
When i ground PS-ON and the PSU is running, there is 156V and 152V across the primarys
If i then add the hard drive while it's running (easier than adding it before grounding PS-ON, then having to jiggle the paper clip for several minutes to get it to fully start), there is 155V and 151V when it's running

23W bulb on 3.3V, and 5W bulb on the 12.8 rail : there is 155V and 151V
Ground PS-ON and PSU running: 155V and 153V
Add a hard drive hard drive to the above : 159V and 156V (but, as before, the PSU doesn't start when PS-ON is grounded, and still 159V and 156V)

OK, that's good. Looks like the primary caps are holding up voltage alright.

The NTC thermistor seems OK too. Just had to ask, in case it had gone higher than normal resistance. After all, sometimes that can happen when an MOV blows, causing the fuse to blow too (and sometimes the fuse might not blow, but the NTC might instead.)

So as for the issue at hand... I'm still not sure what's causing this. The fact that you can get it to run when you play around with the PS-ON signal for a while tells me either something on that circuit is dodgy, or perhaps one (or both) of the optocouplers are not 100% good after the PWM IC blew up. So if you have any spare parts from other PSUs, perhaps start to swap these around.

Also, when you are able to get the PSU to start, what voltage does the PG (Power Good) signal measure? If it's logic level "low", that means the PSU is not "happy" about the state of the output voltages and so won't signal the PC to turn ON yet. On the other hand, if PG does go high, then that means the PSU thinks its output voltages are OK and "ready".

I have to ask the question isn’t the voltage of 12.8 a little bit high for a computer switching power supply

Yes... but in the case of this power supply... let's just call it an HP/Compaq peculiarity. IIRC, the HP DC5000 computers also had something like with their power supplies when I recapped a bunch of them for an office about a decade ago (or thereabouts.)

The 12.8V rail is only used for the CPU VRM high side and nothing else. I suspect HP did this so that they could feed this rail directly to the CPU VRM high side (for which 0.8V higher voltage probably won't make much of a difference, if not help a tiny bit in terms of drawing less current) while also being able to use this rail to linearly regulate it down to 12V (so only 0.8V maximum Vdrop in the worst case) for the HDDs and the rest of the components on the board. I'm not sure if they did this to get a more stable 12V rail (since many PSUs on which they did this were from the Pentium 4 era) or just to save on filtering components. But either way, it's an HP/Compaq -specific thing.

Because the computer switching power supply tester I have if the voltage is over about 12.2 it starts flashing over voltage

Yeah, that's not really a good tester, then - those testers should only warn of over/under-voltage when the 12V rail goes above 12.6V or below 11.4V, respectively. Both of these values represent 5% deviation, which is the max allowable for ATX spec on the 12V, 5V, 3.3V, 5VSB, and (I think) -5V rails. For the -12V rail, it's 10%, so regulation can be quite loose there.

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

I have to ask the question isn’t the voltage of 12.8 a little bit high for a computer switching power supply

Most of the time I have seen 11.8 to maybe 12.2 but this is if you are lucky

Because the computer switching power supply tester I have if the voltage is over about 12.2 it starts flashing over voltage

[socketa]

Re: Compaq PDP-121P - Baked MOV

Removed the NTC thermistor, and it measures 4 ohms
There is a 330Kohm resistor across the live and neutral which is good, and is what is measured across live and neutral

With the 23W bulb on the 3.3 rail, before i ground PS-ON there is 159.3V and 153.8V across the primary caps
When i ground PS-ON and the PSU is running, there is 156V and 152V across the primarys
If i then add the hard drive while it's running (easier than adding it before grounding PS-ON, then having to jiggle the paper clip for several minutes to get it to fully start), there is 155V and 151V when it's running

23W bulb on 3.3V, and 5W bulb on the 12.8 rail : there is 155V and 151V
Ground PS-ON and PSU running: 155V and 153V
Add a hard drive hard drive to the above : 159V and 156V (but, as before, the PSU doesn't start when PS-ON is grounded, and still 159V and 156V)

And this PSU has balancing resistors across the primary caps - which is a good feature

So that all looks OK, right?

[socketa]

Re: Compaq PDP-121P - Baked MOV

So with all of the above tests... did you still have the series incandescent bulb(s)?

Nope - i'm quite aware that doing so will leave less voltage for any downstream load.

I'll check the NTC thermistor, and the voltages across primary caps, and report back

[momaka]

Re: Compaq PDP-121P - Baked MOV

So with all of the above tests... did you still have the series incandescent bulb(s)? If yes, remove it and try the PSU without. I think the fact that the PSU can turn on (sometimes) and not blow up means it's ready to test without a series current limiting. Then let's see how the PSU behaves then. If it still trips, check the NTC thermistor, as perhaps that could have gone high impedance?? When the PSU does not start (but also when it starts), can you measure the DC voltage of the caps on the primary side and post what values you get here? (Basically, I'm looking for voltage before the PS-ON signal is jumpered to ground and then when the PS-ON is jumpered to ground.) I'm still trying to determine if this regulation thing is a primary-side issue or something else.

if R is constant, then P will be directly proportional to V²
Then, if voltage is halved, what will the power be?
V X V = P
0.5(V X V) = 0.5P
V² = 0.25P

Yup, good old power/square law, basically.

So those caps would just go from the probe ground, to the +ve part of the probe?

Yup, right at the connector where you are testing the voltages.

When using incandescent lamp as a load, it will have high inrush current until the filament heats up.
The cold resistance of the incandescent lamp will be about 10 ~15 lower than Hot resistance (when filament heats up the resistance will rise).

Indeed.
Halogen bulbs are even worse in that regard - a high-power one (35W+) can downright trip circuits at nominal rated voltage applied.

On the other hand, high-power halogen bulbs are great for series power limiting (or lack thereof), particularly on PSUs with APFC, because the low cold resistance will make it easier for the APFC to start up.

[budm]

Re: Compaq PDP-121P - Baked MOV

When using incandescent lamp as a load, it will have high inrush current that can trigger power supply over current protection circuit, as filament heats up the resistance will rise. An incandescent light bulb with a tungsten filament has a positive temperature coefficient.
The cold resistance of the incandescent lamp will be about 10 ~15 lower than Hot resistance (when filament heats up the resistance will rise).
You can easily test cold resistance with Ohm meter, the hot resistance can be calculate using Ohms laws.
30W 12V will have Hot resistance of 4.8 Ohms, so cold resistance will be < 1 Ohms.

https://www.fluke.com/en-us/learn/bl...cent-lightbulb

https://www.pathwayz.org/Tree/Plain/LAMPS

[socketa]

Re: Compaq PDP-121P - Baked MOV

Thanks for the corrections. Appreciated.

In case i (or someone else) look back on these threads (which i do), i'm just putting this in as a reminder, about what happens when you use 12V bulbs on a lower voltage rail:
P = V²/R
if R is constant, then P will be directly proportional to V²
Then, if voltage is halved, what will the power be?
V X V = P
0.5(V X V) = 0.5P
V² = 0.25P

Seems that there is a problem with 12.8V rail supply:

I plugged the 24 pin in into a good motherboard, and the board starts using the power button, and runs (12.8V rail = 12.22V); but it wont start if i plug in the 4 pin CPU connector.
If i connect a hard drive (but not the 4 pin CPU connector) the board starts, but then shuts down as the drive starts to spin up.
Also checked the Taicon (probably a PW, as it's the same colour as the other Taicon on the 12v rail) 1000uf, 12.8V PSU rail cap, and it's not bad - 872uf 0.044 ohms (i just de-soldered the +ve leg, and tested it, because it's got tough glue on it and it's in a tricky spot)

Unplugged from motherboard, and then loaded the PSU rails:
3.3V rail_____21W 12V bulb
12.8 rail_____5W 12V bulb
Hard Drive
12V_____ Unloaded

Result:

Starts, but shuts down when HD is spinning up. (and shuts down faster if there is a 73ohm resistor on the 12V rail)
if 12.8V rail is then unloaded, it starts and doesn't shut down : 12.8V rail = 11.86V
if 12.8V rail is then loaded with the 5W 12V bulb, and HD is removed, it starts and runs : 12.8V rail = 12.06V
if 12.8V rail is then only loaded with the 5W 12V bulb, and HD is removed, and the 12V rail has 74 ohm resistor connected. 12.8V rail = 11.82V

Most of the time that the 12.8V rail is loaded with the 12v 5W bulb, it can take quite a few trys at shorting the PS-ON pin to ground untill the PSU starts up

If it's loaded with the 23W 12V bulb, it won't even start, nor does the fan twitch, maybe just a slight, but brief, glow on the bulb.

I also checked the resistance of the rails after about 10 seconds of connecting the probes:
3.3V___8 Kohms
5V_____1.1 Kohms
12V____11 Kohms
12.8___233 ohms (so that doesn't look like a problem - only 0.7W)
5VSB___265 ohms

Did you have the proper bypass caps on your probe when you did that test? (IIRC, it's in Intel's ATX spec sheet, in the section about power supplies and PSU voltage/ripple specs.)
And were both PSUs' covers on and secured properly? A PSU with the cover off may induce a lot of noise on the scope, so that's something to always watch out for.

No, and no.
And thanks for that - indeed, you do recall correctly

Measurements shall be made with an oscilloscope with 20 MHz bandwidth.
Outputs should be bypassed at the connector with a 0.1 μ F ceramic disk capacitor and a10 μ F electrolytic capacitor to simulate system loading

So those caps would just go from the probe ground, to the +ve part of the probe?
And the one that i have is 60MHz.
I'll try it again when i put it back together

If the PSU still refuses to regular, remove both primary caps and test them. Perhaps one of them went open or partially open when that MOV blew. That is the only thing I can think of right now as to why the PSU is not regulating, aside from possibly more/other bad caps on the output filter.

The primarys are good; even replaced them with others that tested good - still no change

Plugged it back into the Compaq computer, and still the "over voltage condition" LED sequence is displayed, and it wont start regardless of whether i plug in the 4 pin CPU connector or the hard drive
I can see the voltages start to come up when i press the power button, and none of them is over-voltage, but they get quickly shut down by the motherboard.

[momaka]

Re: Compaq PDP-121P - Baked MOV

[Edit]
i just realized, and figured out, that 30W@12V is the same as 12.5W@5V, or 2.5A

Close...
Dropping the voltage in half will drop the power by 4x. So a 40W bulb running on half of its rated voltage will draw only about 10 Watts.

- so that 23W 12V bulb, when placed on the 5V rail, should have worked if it required more loading for the PSU to start;
but, as i've discovered, the problem appears to be with the 12V rail, not the 5V rail
I didn't think to try it on the 12V rail at that time

Well, that's an interesting result then.

Generally, when a group-regulated PSU (like this one) has a low-voltage reading on one rail (in this case the 12V rail), then loading the other rails (5V and 3.3V) should work to raise the voltage on the rail that was reading low. Hence why I suggested you put a load on the 5V rail only.

When i plugged the PSU ATX into the test motherboard, i got 11.8V out of it; so i know that it can do it, so it's odd that the max voltage that i can draw out of the 12V rail is 10.64V.
...
It's still a mystery to me as to why i can get 11.8V out of it when plugged into a motherboard ATX, but not when loading the 12V rail with any resistor. Obvioulsy something's happing within the motherboard that's causing the 12R to come into spec, that i can't replicate.

If that's the test motherboard where you had the 4-pin CPU connector unplugged, then you probably got a higher voltage on the 12V rail out of it, because the chipset, RAM, and standby circuits on the motherboard are all powered from the 3.3V and 5V rails... thus drawing more power from those rails only. The 3.3V/5V-"heavy" cross-regulation is likely what's bringing the 12V rail up higher.

I removed the 25V 470uf Taicon PW cap that was on the 12V rail
It measures , according to the 328 transistor tester, 454uf and 0.42 ESR (or 0.83 ESR using a dedicated ESR tester)
The datasheet says that it should have an impedance of 0.273, so if "impedance" = "ESR" then it's out of spec.

Yes, impedance spec @ 100 KHz is equivalent to ESR.
Cap may not necessarily be out of spec, though - you also need to consider the ambient temperature too.

And isn't there something else filtering the 12V rail too? 470 uF seems way too low. I think you might be looking at the -12V rail cap. 12V rail should have at least 1000 uF (but almost always more for PSUs that have a 4-pin 12V CPU connector.)

And when i compared the 12V rail with a good PSU 12V rail, on an old oscilloscope (probably not doing it right though), without changing the previous settings, i could notice significant ripple compared to the other PSU.

Did you have the proper bypass caps on your probe when you did that test? (IIRC, it's in Intel's ATX spec sheet, in the section about power supplies and PSU voltage/ripple specs.)
And were both PSUs' covers on and secured properly? A PSU with the cover off may induce a lot of noise on the scope, so that's something to always watch out for.

Even though the other rails are good, might it be that they need extra loads to push up the 12V?

That's what I'm thinking and that's generally how group-regulated PSUs work too.

This PSU had an 8A fuse in it. P=IV = 1840W. That seems to be a bit high for a PSU that's rated at 220W.

Indeed it is.
Fuse is original?
Normally, a PSU like that should only come with a 5 to 6.35 Amp fuse.

Since you live in a country with 220/230/240V AC, even a 1.5 Amp fuse would technically work for you, as 1.5A x 230V = 345 Watts. Of course, you do have to factor in PSU inefficiencies. So let's say the PSU is doing 70% eff. when loaded to the max at 200 Watts. Then that means it will be drawing about 290 Watts from the wall... hence the need for slight over-rating (in addition to also have a bit of over-head so that it doesn't blow randomly from surge current from the mains caps.) But when the PSU is moved over to a country with 115V, the current will be twice as high and the PSU will be slightly less efficient too. In such a case, a 3 Amp fuse would be the minimum.

As these PSUs are manufactured for both 115V and 230V AC operation in the same factories, the manufacturer generally just installs the fuse that will work in both occasions - hence going with the bigger fuse.

That said, I agree with you that I too don't see the reason why they went with an 8 Amp fuse. 5-6 Amp would be more than reasonably adequate with enough headroom for the PSU's inefficiency. So probably worthwhile to downgrade that fuse, IMO.

I have a 5W ceramic fuse, so i think that there would be no problem replacing the original fuse with that. Correct?

You mean 5 Amp fuse?
If yes, correct.

In case anyone's interested, this PSU's max power ratings are
+3.3V _____ 15A
+5V_____ 11A
+12V _____ 5A
+12.8V_____ 7.5A
-12V_____ 0.15A
5.05V Aux_____ 3A

Should have put that in the beginning so we wouldn't have to guess if it's an older 5V-heavy design or not.

And from that label, I think I know now what's happening here...

That "12.8V" rail is actually the 12V rail that is produced by the PSU and is connected to the 4-pin 12V CPU connector. Meanwhile, the "regular" 12V rail is just a regulated rail off of the 12.8V rail... so the lone 470 uF cap you saw on the 12V rail may be normal. And the low voltage may or may not be normal. Better to just check the voltage on the 12.8V rail. If that one checks out, then we need to worry about why the "regular" 12V rail is reading low.

I trust the MESR ESR meter more, as it lines up better with the datasheets and other newer capacitors that i brought

Yeah, the MESR meter should be better.

My GM328 tester consistently reads much higher ESR on caps when doing the testing between pins 1 and 3. So that's why I always test my caps between test pins 1 and 2... which also isn't 100% accurate and seems to under-report ESR. Probably something to do with the firmware, but I haven't felt like messing with it. In any case, it's still more than helpful for finding bad caps. I rarely use it to check caps against their datasheets. Rather, I use it to just see if a cap is starting to fail (will read an Ohm to a few Ohms of ESR... which is what it usually takes to make PSU circuits misbehave.)

So i replaced it with an OST RLP 1000uf (actually 860uf) that has ESR 0.027
But now the PSU wont start unless i lower the 12V load resistance (i used 47ohms instead of 83 ohms)
And when it starts, the 12V rail voltage is still too low, as before

OK, seems like we are not getting too far with these load tests.
And it seems the PSU is no longer blowing fuses.

So perhaps try connecting the PSU without the series bulb(s) (i.e. directly in the wall) and give it a good load on the 12.8V rail, as this appears to be a more modern PSU meant for 12V-based PCs. And to make sure the other rails are loaded too, perhaps now try the HDD on there too (that is, in addition to whatever load you have put on the 12.8V rail, which I would suggest something in the range of 10-20W 12V car/auto bulb.)

If the PSU still refuses to regular, remove both primary caps and test them. Perhaps one of them went open or partially open when that MOV blew. That is the only thing I can think of right now as to why the PSU is not regulating, aside from possibly more/other bad caps on the output filter.

[socketa]

Re: Compaq PDP-121P - Baked MOV

I removed the 25V 470uf Taicon PW cap that was on the 12V rail
It measures , according to the 328 transistor tester, 454uf and 0.42 ESR (or 0.83 ESR using a dedicated ESR tester)
The datasheet says that it should have an impedance of 0.273, so if "impedance" = "ESR" then it's out of spec.

Actually, the datasheet says 0.065 (looked at the wrong column), and i meant 0.083 instead of 0.83
and it now measures , according to the 328 transistor tester, 454uf and 0.2 ESR (or 0.080 ESR using a MESR-100 ESR tester)
(Possibly due to a change in temperature)
I trust the MESR ESR meter more, as it lines up better with the datasheets and other newer capacitors that i brought

So i replaced it with an OST RLP 1000uf (actually 860uf) that has ESR 0.027
But now the PSU wont start unless i lower the 12V load resistance (i used 47ohms instead of 83 ohms)
And when it starts, the 12V rail voltage is still too low, as before

[socketa]

Re: Compaq PDP-121P - Baked MOV

I just missed out on editing my previous post. So ignore that one, and use this one instead; as i had further thoughts, connected more dots, and corrected and tidied it up.

Getting something like a 12V car light bulb or 12V halogen light bulb (anywhere from 10 to 35 Watts will do) and put that on the 5V rail only.

Seems like this PSU might not not 5V heavy (as previously thought)?
It's rating are:

+3.3V _____ 15A
+5V_____ 11A
+12V _____ 5A
+12.8V_____ 7.5A
-12V_____ 0.15A
5.05V Aux_____ 3A

Hence that's why adding the car bulb didn't have any effect, because i'm fairly sure that the problem is with the 12V rail (see below).

It was after that, that i plugged the 20pin ATX connector into the test motherboard, just to see what would happen.
One useful thing that came out of that, was seeing correct voltage on the 12V rail

So, in retrospect, it's no surprise that 23W bulb (and only the 23W bulb), had no effect on starting the PSU when connected to the 5V rail

I just realized, and figured out, that 30W@12V is the same as 12.5W@5V, or 2.5A - so that 23W 12V bulb, when placed on the 5V rail, should have worked if it required more loading for the PSU to start;
but, as i've discovered, the problem appears to be with the 12V rail, not the 5V rail
I didn't think to try it on the 12V rail at that time

After getting no joy when adding a wide range of resistors to the 5V rail, the PSU now starts if i attach a 3W 74ohm load resistor to the 12V rail to draw close to 2W, and i arrived at that value by trying decreasing resistances until the PSU started and the voltage hit it's highest value, which is 10.64V
The other rails are well with spec and stable
When i plugged the PSU ATX into the test motherboard, i got 11.8V out of it; so i know that it can do it, so it's odd that the max voltage that i can draw out of the 12V rail is 10.64V.

I removed the 25V 470uf Taicon PW cap that was on the 12V rail
It measures , according to the 328 transistor tester, 454uf and 0.42 ESR (or 0.83 ESR using a dedicated ESR tester)
The datasheet says that it should have an impedance of 0.273, so if "impedance" = "ESR" then it's out of spec.
And when i compared the 12V rail with a good PSU 12V rail, on an old oscilloscope (probably not doing it right though), without changing the previous settings, i could notice significant ripple compared to the other PSU.
It's still a mystery to me as to why i can get 11.8V out of it when plugged into that test motherboard ATX, but not when loading the 12V rail with any resistor. Obviously something's happening within the motherboard that's causing the 12V rail to come into spec, that i can't yet replicate.
Even though the other rails are good, might it be that they need extra loads to push up the 12V? - which is a bit different (voltage is too low) from the scenario that you mentioned:

If 12V or 3V3 are too high, you may have to put a small load on those as well

Anyway, before removing the cap, i plugged it into the Compaq computer with some 5A fuse wire and got an LED code of over voltage condition, and it wouldn't start
That was no surprise

While removing the cap, i noticed that this has a +12.8 rail as well, for the 4pin CPU plug.

This PSU had an 8A fuse in it. P=IV = 1840W. That seems to be a bit high for a PSU that's rated at 220W.
I have a 5W ceramic fuse, so i think that there would be no problem replacing the original fuse with that. Correct?

[socketa]

Re: Compaq PDP-121P - Baked MOV

(Please read all first)

Getting something like a 12V car light bulb or 12V halogen light bulb (anywhere from 10 to 35 Watts will do) and put that on the 5V rail only.

So for a 60W bulb, don't put more than 6W of load on PSU output

Something to the equivalent of 120-150W for the series device might be the minimum you'll need if you want to run just an HDD or the bulbs I suggested.

I 'only' have about 180W total of bulbs connected, so i suspect that that's why the PSU didn't start when i connected only the 23W bulb - i had a feeling that the 35 watts that you suggested was too much. Feel free to correct me if i'm not seeing clearly enough here, because if ever there was an age of folk not seeing clearly, then we are in it right now.
It was after that that i just plugged the 10pin atx connector into the test motherboard, just to see what would happen - as you can see, with only the ATX 20 pin connected, the board's pulling 15W max, which is half the amount of a 30W car bulb.
According to this additional info that you gave, "10 to 35 Watts" will not necessarily work, as it would have to be "10W to 18W"

So, accordingly, it no surprise that 23W bulb (and only the 23W bulb), has no effect on starting the PSU.

[Edit]
i just realized, and figured out, that 30W@12V is the same as 12.5W@5V, or 2.5A - so that 23W 12V bulb, when placed on the 5V rail, should have worked if it required more loading for the PSU to start;
but, as i've discovered, the problem appears to be with the 12V rail, not the 5V rail
I didn't think to try it on the 12V rail at that time
[end Edit]


After getting no joy when adding a wide range of resistors to the 5V rail, the PSU now starts if i attach a 3W 74ohm load resistor to the 12V rail to draw close to 2W, and i arrived at that value by trying decreasing resistances until the PSU started and the voltage hit it's highest value, which is 10.64V
The other rails are well with spec and stable
When i plugged the PSU ATX into the test motherboard, i got 11.8V out of it; so i know that it can do it, so it's odd that the max voltage that i can draw out of the 12V rail is 10.64V.

I removed the 25V 470uf Taicon PW cap that was on the 12V rail
It measures , according to the 328 transistor tester, 454uf and 0.42 ESR (or 0.83 ESR using a dedicated ESR tester)
The datasheet says that it should have an impedance of 0.273, so if "impedance" = "ESR" then it's out of spec.
And when i compared the 12V rail with a good PSU 12V rail, on an old oscilloscope (probably not doing it right though), without changing the previous settings, i could notice significant ripple compared to the other PSU.
It's still a mystery to me as to why i can get 11.8V out of it when plugged into a motherboard ATX, but not when loading the 12V rail with any resistor. Obvioulsy something's happing within the motherboard that's causing the 12R to come into spec, that i can't replicate.
Even though the other rails are good, might it be that they need extra loads to push up the 12V? - which is a bit different (voltage is too low) from the scenario that you mentioned:

If 12V or 3V3 are too high, you may have to put a small load on those as well

Anyway, before removing the cap, i plugged it into the Compaq computer with some 5A fuse wire and got an LED code of over voltage condition, and it wouldn't start
That was no surprise

While removing the cap, i noticed that this has a +12.8 rail as well, for the 4pin CPU plug.

This PSU had an 8A fuse in it. P=IV = 1840W. That seems to be a bit high for a PSU that's rated at 220W.
I have a 5W ceramic fuse, so i think that there would be no problem replacing the original fuse with that. Correct?

In case anyone's interested, this PSU's max power ratings are
+3.3V _____ 15A
+5V_____ 11A
+12V _____ 5A
+12.8V_____ 7.5A
-12V_____ 0.15A
5.05V Aux_____ 3A

[momaka]

Re: Compaq PDP-121P - Baked MOV

I will repeat...

You CANNOT test motherboards with series power/current limiting device. Period. There is no point in doing this test.

You can use series power limiting only with very low and known loads: i.e. 12V car bulbs, resistors, and etc., but the total load from these must NOT exceed 1/6-1/8 of the series device's power rating. And that doesn't even factor in PSU inefficiencies. So a safer bet is 1/10 of series limiting device's power rating, MAX. So for a 60W bulb, don't put more than 6W of load on PSU output. Anything more, and no guarantee how the PSU will behave.

That said, you can also monitor the PSU's AC input after the series limiting device. If the voltage drops below 180-190V AC, there's a good chance it will not work with that low of an input.

All that i can think of is maybe the computer is sensing high ripple on one of the rails because of a bad output cap, despite good and stable voltages.
Is that a known possibility?

Yes, that's a possibility.
Which is why I suggested to check output caps with ESR meter.

[socketa]

Re: Compaq PDP-121P - Baked MOV

The ESR of the small caps is not too important - so long as it doesn't go sky-high, of course... or at least to unreasonably high levels. Just check the datasheeet. It's not

uncommon at all for small 5x11 mm caps to measure over 1 Ohm ESR... but also OK if they measure less too.

Thanks, i measured the ESR of that cap that i put in backwards, and it's now 4 ohms, whereas it was initially 1.2 ohms, and the capacitance was about the same as before; so now i really know how to turn a low ESR capacitor into one that's not low ESR.

That 1hz revving is the Hard drive trying to spin up

I have 3 parallel series limiting bulbs 2X60W 1X75W.
When i plug it into the original compaq computer, with the 3 series limiting bulbs, the front led blinks and it means "Power supply overload" according to the documentation.
When i plug it into my test board it wont power up via the header pins, but will power up with another couple PSUs via the header pins. (so either the PSU still has some issue,

or more (or greater wattage) paralell "series limiting" bulbs are required.
The test motherboard powers up with other PSU's via the header pins, and there is a MOSFET on the motherboard that gets super hot and smells
The test motherboard doesn't power up with this compaq PSU via the header pins. It only powers up via the paperclip, and the MOSFET doesn't get hot like it does when i use a good PSU

Anyway, using the test motherboard, the Compaq PSU, and 3 series limiting bulbs ( 70W 60W 60W):
With the 3 series limiting bulbs, the motherboard doesn't start easilly if 4pin CPU connector is plugged in (have to jiggle the paperclip lots, until it fully starts)

ATX 20 pin connected. :
Bulbs-----Total power-----3.3V-----5V-----12V-----Neg12v-----PG
3 bulbs---15.0W-----------3.27-----5.05---11.47---11.79------4.39
2 X 60W---15.1W-----------3.27-----5.05---11.47---11.79------4.39
1 X 60W---15.6W-----------3.27-----5.05---11.47---11.79------4.39

ATX 20 pin connected. CPU 4-pin connected :

3 bulbs : 15.1W-----3.27-----5.05-----11.46----11.23 to 11.80-----4.39
2 X 60W : 15.2W-----3.27-----5.05-----11.46----11.26 to 11.80-----4.39
1 X 60W : 15.7Wh----3.27-----5.05-----11.46----11.21 to 11.80-----4.39 Bulb glowing

ATX 20 pin connected, hard drive connected (and trying to spin up) :

3 bulbs : 18.5W-19.9W-----3.27-----5.05-----11.46----11.23 to 11.80-----4.39
2 X 60W : 15.5W-19.9W-----3.27-----5.05-----11.46----11.26 to 11.80-----4.39
1 X 60W : 13.3W-----------5.05-----11.46----11.21 to 11.80-----4.39 Bulb glowing and PSU shuts down
1 X 75W : 20W - Shuts down

ATX 20 pin connected. CPU 4-pin connected. Hard drive connected (but not trying to spin up) :

3 bulbs : 12.6W-----3.29-----4.47 to 4.45-----10.98 to 10.92-----11.30 to 11.8-----62mV
2 X 60W : 12.7W-----3.29-----4.47 to 5.51-----10.88 to 10.92-----11.26 to 11.8-----62mV
1 X 60W : 13.3W-----3.29-----4.47 to 4.49-----10.83 to 10.90 >>> PSU shuts down

Adding a 23W 12V bulb to the 5V rail, in any of the instances above, immediately shuts down the PSU

Adding a stove element to the paralell bulbs, doesn't make the PSU any easier to start when the 4 pin CPU connector is plugged in (sill have to jiggle the paperclip, and get lucky, until it fully starts up),
neither does it help to start the Compaq computer, with or with the 4 pin CPU connector (the front panel flashing LED still indicates "Power supply overload") - i only used the front panel button because the other PSU starts it up using that button.
And connecting a molex to the hard drive makes no difference either

Each time that i push the front panel power button, the fans twitch, and i can hear that the speaker gets power. The 5VSB rail is good and at least 5V, 1V, and 0.5V appear on the 12V, 5V, and 3.3V rails when i push the power button, but they instantly get cut off.

Any more ideas?
All that i can think of is maybe the computer is sensing high ripple on one of the rails because of a bad output cap, despite good and stable voltages.
Is that a known possibility?

[momaka]

Re: Compaq PDP-121P - Baked MOV

Replaced the cap
I plugged in a hard drive for the easiest option first.
Grounded PS-ON, and the fan, and the hard drive were spinning good for about 2 seconds - then the bulb came on and glowed brightly and the PSU shut down.
...
The PSU still starts if i short the optocoupler secondary side pins; but the 60W bulb starts to glow after a second or so if a hard drive is plugged in - so now i immediately turn it off.

Like I mentioned, this looks like an older PSU, so it may need more load on the 5V rail to regulate properly. A hard drive *won't* do in such cases, as HDDs typically use more power from the 12V rail. Getting something like a 12V car light bulb or 12V halogen light bulb (anywhere from 10 to 35 Watts will do) and put that on the 5V rail only. This will ensure that the 5V rail has enough load to run well. If the 12V rail starts looking too high (over 12.6V), then you can add a few resistor to it too, to balance out the load.

Also, you *can't* load the PSU with just the 60W series bulb, much less a motherboard (though unplugging the 12V CPU power connector may work in some cases.) Something to the equivalent of 120-150W for the series device might be the minimum you'll need if you want to run just an HDD or the bulbs I suggested.

1. Yes
2) Visually look OK
3) Yes
4) As above

For 2), you can't rely on visual check. That's why I mentioned check ESR and capacitance (if you have one of those GM328 or similar testers.) I've encountered enough Ltec caps that looked "fine" but measured out of spec to know a visual inspection is not good enough with these. Same with Teapo... and many of the other crap cap brands. The UCC KZE are probably the only ones that don't need to be checked, because they are known reliable brand and series.

OK, i plugged in a motherboard that i don't care about too much (as it seems to produce a smell in the vicinity of the northbridge) and also connected a hard drive - didn't plug in the 4pin CPU connector
After a few touches of the PS-ON to Ground (where the fans just twitch), the PSU fully starts up and the CPU fan spins but revs at a rate of about 1hz, voltages are good.
The PSU wont start up (it's fan just twitches once) if i also connect the 4-pin CPU connector (but that could be due to an issue with the motherboard)

So without the 12V CPU power connector, the PSU starts and the voltages are good? If so, then you were impedance-limited by the series bulb on the AC input on the other tests for sure.

And again, with the series bulb(s), you just *cannot* put too much load on the PSU and expect it to work. From what I've been experimenting on lately with low AC input / series power limiting and ATX PSUs, the general tendency appears to be that you can pull only about 1/6 to 1/8 of the power rating of the series limiting device and still have the PSU work. So with a 60W bulb, you might be able to pull about 10 Watts from the PSU, tops.

As for the fan on the motherboard revving with 1 Hz... that's probably the motherboard boot-looping and restarting, because the CPU power is not connected, so that may be normal. And of course it not starting with the series bulb in place when the 12V CPU connector is inserted is expected.

Was it OK that i replaced the 10uf 50v Teapo SC with a low impedance (ESR measures about 0.7 ohms) Rubycon YXF?

100%
The ESR of the small caps is not too important - so long as it doesn't go sky-high, of course... or at least to unreasonably high levels. Just check the datasheeet. It's not uncommon at all for small 5x11 mm caps to measure over 1 Ohm ESR... but also OK if they measure less too.

[socketa]

Re: Compaq PDP-121P - Baked MOV

Hmmmm... Maybe not enough load on the output of the PSU, then? 5V looks normal, but 12V and 3V3 are low.

OK, i plugged in a motherboard that i don't care about too much (as it seems to produce a smell in the vicinity of the northbridge) and also connected a hard drive - didn't plug in the 4pin CPU connector
After a few touches of the PS-ON to Ground (where the fans just twitch), the PSU fully starts up and the CPU fan spins but revs at a rate of about 1hz, voltages are good.
The PSU wont start up (it's fan just twitches once) if i also connect the 4-pin CPU connector (but that could be due to an issue with the motherboard)

Was it OK that i replaced the 10uf 50v Teapo SC with a low impedance (ESR measures about 0.7 ohms) Rubycon YXF?

So try running the PSU with two parallel bulbs as the series limiting device and put a larger load on the 5V rail (perhaps a 12V car bulb around 20-30 Watts.) Then try powering on the PSU normally through PS-ON an observe what happens. If nothing, try the shorting of the opto again and see what the output voltages look like with the now extra load on the 5V rail. If 12V or 3V3 are too high, you may have to put a small load on those as well. Try going for resistive loads (or 12V incandescent bulbs) as HDDs and fans may not present a load quick enough to allow the PSU to regulate properly.

i'll try that in the next day or so.

[socketa]

Re: Compaq PDP-121P - Baked MOV

Replaced the cap
I plugged in a hard drive for the easiest option first.
Grounded PS-ON, and the fan, and the hard drive were spinning good for about 2 seconds - then the bulb came on and glowed brightly and the PSU shut down.
Have tried it again several times, but now i'm back to just the twitching fan.

1. Yes
2) Visually look OK
3) Yes
4) As above

The PSU still starts if i short the optocoupler secondary side pins; but the 60W bulb starts to glow after a second or so if a hard drive is plugged in - so now i immediately turn it off.

[momaka]

Re: Compaq PDP-121P - Baked MOV

Turned on the PSU, with PS-ON grounded, and the fan twitched slightly.
Looks like some protection is kicking in?

Maybe.
Did you have the series bulb on the AC input? If so, what power rating? Go with 100 Watts (or thereabouts). And if the PSU still doesn't turn on, try about 150-200 Watts (two or more 40-100W bulbs in parallel.)

I have to wait about 10s, in between removing the paper clip and reinserting it, to get a response

That almost looks like short-circuit protection is triggering, but since you do get output voltages on all of the rails, I don't think that can be the case.

It wouldn't be because of the replacement MOSFET that i used in post #16, would it?

No, the replacement you have in there is reasonably close enough.

When i ground PS-ON, about 9V appears on Vcc, then it immediately drops off.
And i get about 10V on the 12V rail (and fan spins), 5V on the 5V rail, and 1.5V on the 3.3V rail, if i accidentally short the optocoupler pins on the primary side.

Hmmmm... Maybe not enough load on the output of the PSU, then? 5V looks normal, but 12V and 3V3 are low.

So try running the PSU with two parallel bulbs as the series limiting device and put a larger load on the 5V rail (perhaps a 12V car bulb around 20-30 Watts.) Then try powering on the PSU normally through PS-ON an observe what happens. If nothing, try the shorting of the opto again and see what the output voltages look like with the now extra load on the 5V rail. If 12V or 3V3 are too high, you may have to put a small load on those as well. Try going for resistive loads (or 12V incandescent bulbs) as HDDs and fans may not present a load quick enough to allow the PSU to regulate properly.

Could it be base voltage is initially driving a transistor that's supplying Vcc, but then it quickly falls, and so the transistor switches off?

Maybe.
It's hard to say, because feedback from the secondary gets sent out so quickly. But the test you did with shorting out the optocoupler for supplying power to the UC3845 IC shows that the PSU can work... just that voltage are a bit low. That voltage on the 3.3V rail is probably what's tripping it.

I'm going to retest the startup(?) cap, and look for a replacement; as that might be the cause of this after all.

Yes, do check all of the caps in this PSU. It appears to be an older HiPro unit - old enough that it has green UCC KZE caps (among lower-tier brands like Ltec and whatnot), looks like (so possibly Pentium 3 / early P4 era.) I'm not 100% familiar with this particular version, though... and it may be a custom unit too, judging by the layout.

In any case, check the basics first, as always:
1) Voltage switch in the correct position and primary caps getting about 320V DC across them (about 160V DC across each.)
2) All large output filter caps are good in terms of ESR and capacity
3) 5VSB always present when the PSU is plugged in, even when shorting the PS-ON signal to ground and afterwards?
4) Enough load on the output when testing the PSU? This being an older PSU, probably will be happier with more load on the 5V rail (so an old HDD may not be enough to "do the trick".)

[Edit] after previously testing that cap, i have noticed that i put i back in with the wrong polarity

Might be worthwhile to change that cap then, as it could have been damaged. Or at least pull out and reform for 30 minutes, then re-test before re-using.

[socketa]

Re: Compaq PDP-121P - Baked MOV

When i ground PS-ON, about 9V appears on Vcc, then it immediately drops off.
And i get about 10V on the 12V rail (and fan spins), 5V on the 5V rail, and 1.5V on the 3.3V rail, if i accidentally short the optocoupler pins on the primary side.
Could it be base voltage is initially driving a transistor that's supplying Vcc, but then it quickly falls, and so the transistor switches off?

I'm going to retest the startup(?) cap, and look for a replacement; as that might be the cause of this after all.
It's connected to a PNP transistor that's connected to Vcc

[Edit] after previously testing that cap, i have noticed that i put i back in with the wrong polarity