Compaq PDP-121P - Baked MOV

[momaka]

Re: Compaq PDP-121P - Baked MOV

A year to fix a PSU - LOL

Hey, better late than never!

I have some projects that have been open/in-a-fix for years now.
The point is you got it done and possibly learned something new.

Also the 12.8V rail doesn't put out 12.8V - it's only 0.1V more than the 12V rail (which is around 12V) - maybe it was never actually 12.8V

Quite possible... or at least when loaded down with the CPU it might not be 12.8V. IDK, it's a strange design indeed with the Compaq/HP PSUs, but I do recall seeing a few of them back in HP DC5000 SFF computers and similar.

Also reconnected the passive PFC, after soldering new (and thicker) wires to it - your mention of efficiency made me want to put it back in

Well, PFC (be it APFC or PPFC) doesn't actually increase the efficiency of the PSU. If anything, it decreases the efficiency. In the case of APFC, heat from the APFC MOSFETs, diode, and inductor are essentially those losses. And in PPFC, it's heat generated in the PPFC choke/inductor's core and windings.

However, PFC does reduce the reactive power from the PSU, which in turn may make the delivery from the power co. to your house a little less lossy (though that depends.) So if you consider just the PC/PSU as one system, PFC decreases efficiency. But if you consider the PC/PSU + power co. delivery system as one, overall that may increase the efficiency or at least increase stability in the system (highly inductive or capacitative loads... i.e. loads with high reactive power... can be problematic for the power company.) So all in all, it's probably worthwhile to keep it.

Anyways, congrats on the repair!

[socketa]

Re: Compaq PDP-121P - Baked MOV

Thanks for that additional info - that's quite interesting - the sort of thing that i reread quite a few times.

Yep, i should have said "power" instead of "current" - old habits dying hard

I did test the resistor properly;
but i put another resistor (same size and colour, but the wrong value) back in,
and i put a heatshrink around it - so i didn't notice.

Reconnected with dim bulb tester to make sure all was good enough to turn the PSU on - Yep
I reconnected the hard drive, and this time, it didn't fully start up (just spun for a second).
Added some 12V bulbs to the 5v and/or 12 rails, and it still didn't start (fan just briefly spun, then PSU shut down)
Maybe it needed a higher load to start now, since the current sensing resistor is now not 10X what it's supposed to be.
Feeling confident that i had fixed it, i put it back into the case, connected it all up, and it started up, and all is good.
Also the 12.8V rail doesn't put out 12.8V - it's only 0.1V more than the 12V rail (which is around 12V) - maybe it was never actually 12.8V


So, in summary, i replaced:
The fuse - with another of equal or less value than the original
The shorted MOV (there's two of them - one across each primary cap)
The current sensing resistor (that was burned open) that's connected to the primary controller chip
The shorted Primary controller chip
The shorted primary MOSFET, with a slightly different one, from a parts PSU.
I left the three optocouplers in, that i changed over from a parts PSU, while looking for the fault.

Also reconnected the passive PFC, after soldering new (and thicker) wires to it - your mention of efficiency made me want to put it back in
There is an unoccupied space for another MOV (MOV3) that is in parallel with the passive PFC - that maybe protects the PFC coil?

I put some UHU glue on the bases of the primary caps, and between them, to keep them stable

And, of course, didn't touch the primary heatskink when the PSU was switched on, because it's live,
and i reconnected the earth terminal to the case.

A year to fix a PSU - LOL

[momaka]

Re: Compaq PDP-121P - Baked MOV

So the PSU shut down point is somewhere between 8W and a 23W, if a load is added on either the 12V or the 12.8V rail
And the PSU shuts down if there is between 9.68W and 13.7W load added on the 5V rail
The common ground here is a range between 9.68W to 13.7W
So i need to try to configure the bulbs to some value between 9.68W and 13.7W and hopefully find a combination that will shut down the PSU on the 12V rail.
I added two 23W/12V bulbs in series, to create a total of 11.5W load (since there is a 6V voltage drop across each 12V bulb, and therefore wattage will decrease by a factor of 4, for each bulb) and the PSU shut down.
Yay?
So that means that PSU is shutting down when a load is added that is somewhere between 9.68W and 11.5W
Looking more like the fault is on the primary side now?

Excellent work there!

Now the question is, what if you distribute this 8 to 11.5 Watt load between the 5V and 12V rails evenly? Will it still shut down?

My proposal is to put maybe 5 Watts on the 12.8V rail and 1-2 Watts on the 12V rail - i.e. keep the total power on these two under 8 Watts... but make sure the PSU can still run with whatever load you put on these. Once the PSU is running with that load on the 12.8V and 12V rails, add a small load on the 5V rail as well - 1W, then 2W, then 3W, and so on. See if/when the PSU shuts down. The idea here is to see if the PSU will still shut down somewhere in the 8 to 11.5 Watt range if the power is distributed more evenly between the 5V and 12V/12.8V rails. Maybe even divert some of that load on the 3.3V rail, just so that everything is balanced. If the PSU still appears to shut down somewhere in the 8-11.5 Watt range like you found earlier, then this is probably a primary-side issue then.

Yep, the resistor measures about two ohms
I've ordered a 0.15 ohm one

Just to make sure you didn't get a false positive - did you measure this with your multimeter? And if so, did you first try measuring resistance with the two probes on the multimeter shorted together? Reason I ask is because some multimeters have a pretty high offset (especially the cheaper ones when they have sat for a while without use, which makes their contacts oxidize and give higher readings.) If the resistance you measured is not due to multimeter offset... then yeah, that 0.15 Ohm resistor is likely the culprit if it measure 2 Ohms (almost 10x higher resistance.)

Let's propose that there is an issue with the U3845 chip's current sensing (or it's associated circuit).
I'm assuming that the current that it senses is the total current of all rails. (is that correct? - since all the secondary current, except for the auxiliary, is passing through the main MOSFET)

Kind of but not exactly.

Basically, the current sense resistor for the MOSFET senses the current on the primary side. Since the primary side of the transformer usually operates with 340V pulses of DC (square-wave AC), the current though the sense resistor will not be the sum of the currents of the secondary side. Rather, each current on each rail on the secondary side makes up a power load. The load from each rail adds up to one total power load... which the primary side must supply.
In simpler terms:
primary side power (primary side current x primary voltage) = secondary side power (secondary side current of each rail multiplied by its respective rail's voltage and all of them added together)

So if you pull 2 Amps from the 12V rail (24 Watts) and 2 Amps from the 5V rail (10 Watts), the total power on the secondary side is 24 + 10 = 34 Watts. For the PSU to work, the primary side has to supply that amount of power (not accounting for inefficiencies here, just for simplicity.) Thus, if the primary side operates at 340V, then it must supply 0.1 Amps of current continuously to keep the voltage regulation of the secondary in check. However, we also know that the primary side cannot supply a continuous (DC) current through the transformer, because transformers only work with AC. Therefore, that 340V supply is actually cycled On and Off through the transformer. The average current will still be 0.1 Amps over the course of any one complete On-Off cycle, but the peak current will not. Let's assume 50% duty cycle (denoted by D) here, just for simplicity. That means during the On period in the duty cycle, the current will have to be 0.2 Amps (equating to 68 Watts of power). But in the Off period, the current will be 0 Amps (0 Watts). Since D=50%, that means the On and Off periods will be equal, and so you have 68 Watts for one period and 0 Watts for the other period. The total of the two periods makes one cycle, and the average power of those two periods is 68+0 / 2 = 34 Watts, which is what the secondary side is pulling. So that's how PSUs work.

Now in reality, the duty cycle (D) for a low load will be much much smaller - typically in the single % digits for a single-transistor forward converter PSU like this. Actually, we can calculate that! So I'll do that below.

The Compaq PDP-121P is rated for 220 Watts... but more than likely, the primary side has some over-head built into it to account for low AC line voltage, inefficiencies, and whatnot. If the PSU has worst-case efficiency of 70%, then the primary must be able to do at least 315 Watts. And with low like AC input, let's bump that up a notch to 350 Watts.

We also know this is a single-transistor forward-converter PSU, so the primary side MOSFET's duty cycle will probably be limited to about 40% (that is, the On period in any one cycle cannot be On for more than 40% of the time in that one cycle.) This means that at full power output of the PSU (220 Watts) and 70% efficiency (typical for these PSUs at max load), the primary side will be pulling about 315 Watts... but again, we rounded that up to 350 Watts, just for some headroom. And since the duty cycle is limited to 40%, that means the MOSFET will have to supply "short" bursts of really high power, so that the total average power is equal to 350 Watts. These short busts will be 350W / 40% = 350 / 0.4 = 875 Watts. Might seem like a lot... but remember the primary side has 340V to work with (though in reality, that will dip a little)... So the maximum current that will pass through the MOSFET will be 875 / 340 = 2.57 Amps <-- this is for when the PSU is at max load.

Now we want to know what happens at just 34 Watts of output power (again, just to continue with the same example above and not that this number has anything to do with your PSU.) To find the duty cycle for 34 Watts, simply take that and divide by the "burst" power, then multiply by 100 to get the percentage.
D_34watts = ( 34 / 875 ) x 100 = 3.89%

Note that the average peak current on the primary side (in a properly working PSU) will still be about 2.57 Amps for the above example, regardless if you're pulling a little or maximum power. For the case of 34 Watts, we ca prove that this is true, since the primary voltage (340V) times the peak primary current (2.57A) times the duty cycle (3.89%) = 340 x 2.57 x 0.0389 = ~34 Watts

How does any of that help us here? IDK But figured I'd post the theory of operation on SMPS anyways.

Also, it's always nice to have a thread that doesn't have a dead end

Indeed.
This is why I like BCN! - you can literally revive a thread whenever you have relevant information for it, regardless if that's in a day, a week, or even a few years down the road.

[socketa]

Re: Compaq PDP-121P - Baked MOV

Yep, the resistor measures about two ohms
I've ordered a 0.15 ohm one
Thanks, Momaka, for that suggestion; for without that, i might not have figured this out, and learned a thing or two along the way.
Also, it's always nice to have a thread that doesn't have a dead end

[socketa]

Re: Compaq PDP-121P - Baked MOV

Subsequently, i thought to check the resistor that's going from source of the primary MOSFET to ground.
and it tests 1.8 ohms in circuit
I will desolder it tomorrow, and measure it's resistance again, as that doesn't seem to be right when i compare it to another schematic of an ATX PSU that has a UC3843 chip.
In post #28 and #31, i checked this replacement resistor
And the picture shows it to be either 0.15 or 0.25 ohms
But maybe it actually measures around 2 ohms, and i put the decimal point in the wrong place despite what i thought was careful observation in post #28.
Will find out tomorrow, and i will post the result.

From what was said in post #31, if the resistor is larger by a factor of ten, then that would shut down the PSU at way too low power (a tenth less?), which is roughly what is happening

[socketa]

Re: Compaq PDP-121P - Baked MOV

So the PSU shut down point is somewhere between 8W and a 23W, if a load is added on either the 12V or the 12.8V rail
And the PSU shuts down if there is between 9.68W and 13.7W load added on the 5V rail
The common ground here is a range between 9.68W to 13.7W
So i need to try to configure the bulbs to some value between 9.68W and 13.7W and hopefully find a combination that will shut down the PSU on the 12V rail.
I added two 23W/12V bulbs in series, to create a total of 11.5W load (since there is a 6V voltage drop across each 12V bulb, and therefore wattage will decrease by a factor of 4, for each bulb) and the PSU shut down.
Yay?
So that means that PSU is shutting down when a load is added that is somewhere between 9.68W and 11.5W
Looking more like the fault is on the primary side now?

[socketa]

Re: Compaq PDP-121P - Baked MOV

Yes, I think the issue is indeed more likely to be on the secondary side at this point... unless somehow the UC3845 is incorrectly sensing the current going through the MOSFET on the primary side and shutting it down (thinking it's an over-power protection issue.) However, you can rule this out by loading the 5V rail or the regulated 12V rail with 25 Watts or more (as it seems that the 23W bulb on the 12.8V rail is able to make the PSU shut down, but the 8W bulb isn't.)

I thought more about this:
Let's propose that there is an issue with the U3845 chip's current sensing (or it's associated circuit).
I'm assuming that the current that it senses is the total current of all rails. (is that correct? - since all the secondary current, except for the auxiliary, is passing through the main MOSFET)
The 12v/23W bulb is causing the PSU to shut down when it's connected to either the 12.8V or the 12V rail.
To test the above proposal using the 5V rail, i would require the same amount of watts that triggered the PSU to shut down when loading either the 12V or 12.8V rail
After applying the power/square law for conversion of a 12V bulb that is fed with 5V, (so i multiplied power by a factor of (0.42)², or 0.176), this equates to, at most, a 12v/132w bulb.
So i turned on the PSU with just the hard drive attached again, then put the 55W bulb across the 5V rail,
and then also added the 23W bulb in parallel,
then the PSU shut down
The total additional power that the 5v rail is using is 13.7W, which equates to 2.74A@5V
The specification of the PSU is 11A@5V - so 2.74A, plus the hard drive current of 0.85A, is well within that boundary.
So, is it looking the the fault is on the primary side U3845 circuitry?

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

Because some switching power supply for computers requires a load on 3.3 or 5.0 and 12.0 and sometimes all three power supply rails the other power supply like - 5 and the - 12 volt power supply rails most of the time do not need a load on to keep the switching power supply on

[socketa]

Re: Compaq PDP-121P - Baked MOV

typo
should read as 12V/23W, not 2V/23W

[socketa]

Re: Compaq PDP-121P - Baked MOV

Thanks for that

Yes, I think the issue is indeed more likely to be on the secondary side at this point... unless somehow the UC3845 is incorrectly sensing the current going through the MOSFET on the primary side and shutting it down (thinking it's an over-power protection issue.) However, you can rule this out by loading the 5V rail or the regulated 12V rail with 25 Watts or more (as it seems that the 23W bulb on the 12.8V rail is able to make the PSU shut down, but the 8W bulb isn't.)

With the PSU working, (with fan and hard drive spinning) when i put either a 2V/23W bulb, or a 12V/55W bulb, across the 5V rail, the bulb lights up, and the PSU doesn't shut down
When i put either of the same bulbs across the 12V regulated rail, then the PSU shuts down right away

What, if anything, can be ascertained from those results?

[momaka]

Re: Compaq PDP-121P - Baked MOV

Would the fault more likely be on secondary side, since the PSU is not shutting down when i attach the 23W bulb to the 12V rail?

Yes, I think the issue is indeed more likely to be on the secondary side at this point... unless somehow the UC3845 is incorrectly sensing the current going through the MOSFET on the primary side and shutting it down (thinking it's an over-power protection issue.) However, you can rule this out by loading the 5V rail or the regulated 12V rail with 25 Watts or more (as it seems that the 23W bulb on the 12.8V rail is able to make the PSU shut down, but the 8W bulb isn't.)

Looking that the datasheet schematic of LM339N, it doesn't look like there should be any resistive path between the positive input and VCC, on a comparator.

Correct.
Well, technically speaking, the transistor internal to the LM339 might show a very very high resistance between VCC and any of the input pins, as well as ground and any of the input pins... but it will probably be in the high 100's of KOhms if not several MOhms.

On the daughterboard PCB (voltage protection board?) there is a 10K resistor across VCC and positive input pin 5
But 5K is measured across it's leads, in-circuit

Well, what's the resistance of VCC/Aux rail to ground?
If it's anything less than 1-2 KOhms, then have a look at the circuit in the picture you attached: you will have a 10-KOhm resistor going to VCC and another 10-KOhm resistor going to ground. If VCC and ground have less than 1-2 KOhms of resistance to each other (perhaps due to a minimum load resistor somewhere or other circuits in the PSU), then essentially the two 10-KOhm resistors on pin 5 will be pretty much in parallel (but with a 1-2 KOhm or whatever other resistance you get between VCC and ground)... so that would exactly explain why you are getting only about 5 KOhms across each.

So I think the issue is likely elsewhere. But where? IDK.
Perhaps try removing various resistors (or at least lift one of their legs from the PCB) and see if they measure in spec. In particular, go for the high-resistance resistors first (100 KOhms or more), as these are the more likely ones to go open-circuit. If all are OK, check all electrolytic and ceramic caps on the supervisor board (preferably out of circuit.) If all of these are good too... maybe just time to swap the LM339 and LM324 ICs in a "shotgunning" approach and hope for the best (not my favorite way, but sometimes that's faster than troubleshooting.)

[socketa]

Re: Compaq PDP-121P - Baked MOV

Looking that the datasheet schematic of LM339N, it doesn't look like there should be any resistive path between the positive input and VCC, on a comparator.
On the daughterboard PCB (voltage protection board?) there is a 10K resistor across VCC and positive input pin 5
But 5K is measured across it's leads, in-circuit
Looking at my photo, the only path that could provide this lower resistance is through the chip.
So, if my thinking is correct, either there is a problem with the chip, or the resistor
The fact that it measures 5K (a too coincidental value) makes me suspect that i'm missing something, and that it's meant to be like that
There is also the same thing happening across another resistor in the photo (just underneath it).

Also, the 12.8V rail doesn't go to the daughterboard

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

This is also a possibility as well but the question is why

[socketa]

Re: Compaq PDP-121P - Baked MOV

i get the feeling that it's more like an incorrect voltage (that's somehow derived from the 12.8V rail) that's being applied to one of the comparator's inputs, and hence, not creating the required output

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

That might be a possibility that something is wrong with the LM339N if it used voltage monitoring

[socketa]

Re: Compaq PDP-121P - Baked MOV

on the secondary side, this only has two 14-pin chips:
LM324 Opamp
LM339N Comparator
i don't know what these do, except that the LM339 might be used for OVP
Any suggestion as to what to look for, in oredr to find out what is causing this PSU to latch?
Would the fault more likely be on secondary side, since the PSU is not shutting down when i attach the 23W bulb to the 12V rail?

[socketa]

Re: Compaq PDP-121P - Baked MOV

It's definitely a problem with the 12.8V rail on this PSU
That other PSU, that i found, starts up the computer no problem

When i connect a 23W bulb to the 12V, rail the bulb lights up and the PSU keeps running fine
But when i connect it to the 12.8V rail, which is supposed to be able to supply 7.5A (which means 96W), the PSU shuts down and won't start again until i unground and reground the PS-ON pin
But it runs OK if i use an 8W bulb instead

So it seems like a protective latch is being activated that is incorrectly sensing the 12.8V rail current

Does that sound right?

[socketa]

Re: Compaq PDP-121P - Baked MOV

I'm getting desperate!
I checked the rail pins of the motherboard ATX connector, in case the motherboard is faulty, but couldn't find any shorts -
just 11ohms on the 3.3 rail, and 15ohms on the 5v rail, so that shouldn't trigger over-current protection, because it's not even 1A
Then the thought occurred to me that there may be a slight possibility that i may have a similar computer lying around, so i noticed one that has a similar front bezel - its a HP/Compaq and it has a PSU with a 12.8v rail - nice!
So now i will be able to narrow down where the problem is (assuming that this other PSU is working)
I can now test the PSU safely also, because i finally made an enclosed dim bulb tester
I'll post my findings
And it's being so long, that i forgot that i tested the start-up cap (10uf rubycon YXF, which tested good) but i might try replacing it with a lower impedance/higher ripple one (33uf Samxon GF)

[sam_sam_sam]

Re: Compaq PDP-121P - Baked MOV

I lost track of this post again and I just found it again and I am hoping that you figure out what is really wrong with this power supply issue is and it you can fix it or not

[socketa]

Re: Compaq PDP-121P - Baked MOV

It would be the UC3845 main supervisory chip that is shutting this down, right?
probably because it senses a change to an undervoltage condition at Vcc, that is reset when the power plug is removed from the wall.
There is a 10uf electrolytic cap that is attached to Vcc, so maybe this this a starting cap and it provides enough energy when not connected to a motherboard, but fails to provide enough energy when connected to a motherboard?
Maybe it's capacitance/ESR was degraded when the surge took out the fuse and thermistor?