Power supply repair

[rodrigopires]

Hi everyone! I have a HP PS-3701-1 that is not working right. When i turn it on it outputs the 5v and 12v but it instantly cuts off the 12v power just leaving the 5v line on. The fans work as well. I think the overcurrent protection circuit is being activated and thus cutting the power. What can i do? What to look for? I check the traces and there is no short, if there was it would explode cuz it outputs 60A max and it cuts off silently. No burned component also. Could it be the OCP chip the gone bad? I realy want to fix it...

Thanks! 

[singapol]

Hi everyone! I have a HP PS-3701-1 that is not working right. When i turn it on it outputs the 5v and 12v but it instantly cuts off the 12v power just leaving the 5v line on. The fans work as well. I think the overcurrent protection circuit is being activated and thus cutting the power. What can i do? What to look for? I check the traces and there is no short, if there was it would explode cuz it outputs 60A max and it cuts off silently. No burned component also. Could it be the OCP chip the gone bad? I realy want to fix it...

Thanks! 

You have post same problem in June: https://www.eevblog.com/forum/repair/hp-ps-3701-1-failure/
Sadly no one replied.

It's best to learn and know how switchmode power supply work while you must know your 3701 has high current and high voltages.Know the safety aspects of electricity before you do anything. Try measuring voltages at the proper places of the circuit..read the basics below.

Switchmode basics: http://www.onsemi.com/pub_link/Collateral/SMPSRM-D.PDF

http://www.power-analysis.org/files/The_Voltech_Handbook_Of_Power_Supplies.pdf

[rodrigopires]

Thanks for the support! I will look into it and try to fix it. I really hope i can do it...

[rodrigopires]

Done some poking around and the two main mosfets (20N60) are at 391V on source and 112V on drain, one with 200mV gate and other with 250mV gate. I think that the gate is too low and not opening the mosfets totaly. What could be the cause? What is afected by the overcurrent protection circuit? i cant see where the traces go since some parts have silicone and the unit is very compact.

[amspire]

Is it failing in a computer, or are you running it on the bench? If you are running it on the bench, there are probably supply control lines that you need to activate to get the supply out of standby mode.

[rodrigopires]

yes i'm running it on the bench and i have already activated the ps, otherwise it wouldn't even turn on, but thanks!

[janoc]

yes i'm running it on the bench and i have already activated the ps, otherwise it wouldn't even turn on, but thanks!

Does the PSU cut out even with some load connected to the 12V rail? These computer supplies sometimes don't like to run unloaded, because then regulation isn't working OK in such case. So it could be detecting that and turning the 12V rail off.

I am not familiar with this particular supply, but usually when the OC trips, it shuts the entire supply off, not only a single rail. So it is probably something else, unless the supply has separate OC protection on each rail - uncommon on ATX supplies, but this is a server supply, so no idea.

[rodrigopires]

Yes i have tried it with a load and it does the same thing. I have had that problem with a atx power supply and i was going crazy why it wont work until i tryed it with a load...

The supply has seperate shunts and OC circuits for the 12V and 5V rail, thats why i think the 12V OC circuit is being activated or not funcionig as properly. Also there is a LED on the back of the supply that shuts off immediately after turn on. I have another supply just like this one and the LED turs off only when the OC is triggered.

[janoc]

Well, we are shooting in the dark here. The LED could go off also simply because the regulator or the switcher IC detects a fault (not necessarily a short/OC condition) and shuts the power off.

I think you will need to find/post the schematics of the thing or at least detailed photos of the circuitry inside, including both sides of the PCB if you want someone to help you.

[CJay]

It's a server PSU, have you followed one of the conversion guides?

Like this one:

[rodrigopires]

Yes I have done all that, this power supply was working fine until this happened. Also searched for the schematic but was unable to find one, that would really help. I'm afraid that photos will not help, the board has a double layout and it is realy packed and some parts like the high current shunts have silastic on the traces so i cant even see what traces are what. But heres are some photos:

[singapol]

Let's put it this way, if you know electronics..ie. the functions of components, how to test them like using diode test,continuity and measuring voltages I don't think you stand a chance of repair. From your comments YOU seem to know... so don't be lazy and start doing the usual drill and check the suspects (components) rather than waiting for the magic bullet to appear. The links I posted show different switchmode topologies (ways) supplies.

Apart from diodes,transistors,pwm controller ic and opto-isolators/couplers, the most useful test is capacitor esr...if you have esr meter.Test the small electrolytic caps..it should be an eye opener for you. Good luck and don't get zapped by high voltage...discharge the caps when you want to dive in with your hands.

[Rasz]

1 smps tutorial series:

I have another supply just like this one and the LED turs off only when the OC is triggered.

'like this' as in same model? well we are in business
you can use monkey repair method by comparing two boards, run two next to each other and measure voltages, alternatively measure resistances on powered down boards.

5V is 100% independent circuit providing standby voltage, its presence doesnt mean a thing for the main supply. How long is the 12V present? do you have a scope?

my HP ESP115 has everything (under/over/fan) controlled by oldschool comparators, I would suspect its the same here, except more modern so smd The board you call 'led and fan controller' has a bunch of chips that look like lm324, start there.

[janoc]

Just a hunch, perhaps it is a non-issue because of the angle of the photos, but better verify:

The board next to the large heatsink and the big inductor/choke with the brown electrolytic caps stuck down looks weird with the caps bent and leaning all over the place. It is difficult to see from the photos, but did you check that there is no leaking capacitor on that board? The rest of the PSU looks decent, so these "drunk" caps look really odd.

[R005T3r]

If you are working on a defective PSU, be very careful as the capacitors may take some time to discharge. I'd exclude that the OCP chip is gone.

Your best bet may be following the 12V pcb path until you find the faulty item from the mains to the rail (don't do the opposite since you will loose it). Needless to say: working on live equipment is very dangerous, so use proper probes and gears. The path will lead you to a point where there's no voltage, and just before there's the faulty item.

If the item is a transistor/mosfet/ regulator, there might be a problem on the control circuit.

[Nermash]

Check output rectification diode(s) for the 12V rail.

[rodrigopires]

The board next to the large heatsink and the big inductor/choke with the brown electrolytic caps stuck down looks weird with the caps bent and leaning all over the place. It is difficult to see from the photos, but did you check that there is no leaking capacitor on that board? The rest of the PSU looks decent, so these "drunk" caps look really odd.

I have visually inspected every cap, those caps your are referring are bent as they came from the factory because they have silastic underneath.
I wish it was just a bad cap... Could be...

5V is 100% independent circuit providing standby voltage, its presence doesnt mean a thing for the main supply. How long is the 12V present? do you have a scope?

12v is non existent. I think the problem is on the primary side. No, i dont have a scope

If you are working on a defective PSU, be very careful as the capacitors may take some time to discharge.

Actually when i disconnect the plug the fan keeps working and it discharges the capacitor from 390V to round 16-12V, but i always discharge then anyway. Learned it the hard way 

Check output rectification diode(s) for the 12V rail.

One of the first things i checked but they are ok and there is no output on the transformer, so the problem must be somewhere in the primary side.

Let's put it this way, if you know electronics..ie. the functions of components, how to test them like using diode test,continuity and measuring voltages I don't think you stand a chance of repair. From your comments YOU seem to know... so don't be lazy and start doing the usual drill and check the suspects (components) rather than waiting for the magic bullet to appear. The links I posted show different switchmode topologies (ways) supplies.

you are right, but i only know the basics and without a schematic and cant even see the traces its hard to poke aroud the board... i wish it was just a cap... The thing i like the most about eletronics is big power supplies and high power stuff, the higher the power the better and this is one of my most powerfull supplies so i really want to have it working...

So, i redid some poking around and found that the 2 big mosfets (20N60) have 390V on the source and 115V on the drain so i presume they are good. There are 4 more mosfets (STP20NM60) that connect to the 12v transformer primary. The 2 top ones have 390V on gate and source, no drain output. Those drains go to the source of other 2 that have no source voltage. What is happening here? 390V on the gates is not normal  And why does the drains of 2 go to the source of other 2 and then drain to positive?

Also tested all the diodes i could see and they all seem ok

[singapol]

No worries..Think out of the box, life is like that...be creative and don't say there is no schematic..scrap this power supply or learn the basics and go with that. Links below teach how to test mosfet but do know there are N and P channel mosfet . Links for testing mosfet below :
Also take note of the test voltage that your multimeter uses for diode test mode.

http://www.androiderode.com/how-to-test-mosfet/

http://www.4qdtec.com/mostest.html

http://www.electronicrepairguide.com/test-mosfets.html

youtube:

[R005T3r]

You don't need a scope for following the 12V path. Use a multimeter.

[Rasz]

12v is non existent.

erm, you liar! you said:
>When i turn it on it outputs the 5v and 12v but it instantly cuts off

have you checked primary mosfets  for shorts?

[rodrigopires]

Thanks Singapol, i have desoldered all primary mosfets the tested them with the multimeter and with the LED circuit on your link ( http://www.androiderode.com/how-to-test-mosfet/). The two big 20n60 worked fine with the multimeter, opening and closing the gate, The other 4 stp20nm60 could not open the gate with the multimeter so i used that circuit and they seem to work fine. All mosfets are good so the problem must be on the triggering because there was voltage on the source right? The main chip is a UCC3818N, The TO-92 are MPS 651 and MPS751 transistors. Could the failure be on the main chip? The 5V mosfet is working fine. Could it be a transistor? Can/should i turn the unit on without the mosfets?

12v is non existent.

erm, you liar! you said:
>When i turn it on it outputs the 5v and 12v but it instantly cuts off

have you checked primary mosfets  for shorts?

Well my fault, in the beginning yes, that was true, the voltage was very wierd it started low on 2v then went up to 4-6V and the it jumped to 12v. After that it just gave a 12V pulse when turned on and shuted off. Now, if that pulse is still existent, i dont know because my multimeter is not fast enough to get it, but i think not.


Thanks everyone for all the help, i realy apreciate it 

[janoc]

Could it be a transistor?

It could be literally anything, even a blown resistor somewhere. You need to start checking things systematically, not grasp at straws. This is a relatively complex device with lethal voltages in it, not really a good thing to poke around at random.

Can/should i turn the unit on without the mosfets?

And what do you expect to happen, exactly? 

Well my fault, in the beginning yes, that was true, the voltage was very wierd it started low on 2v then went up to 4-6V and the it jumped to 12v. After that it just gave a 12V pulse when turned on and shuted off. Now, if that pulse is still existent, i dont know because my multimeter is not fast enough to get it, but i think not.

If all the FETs are good, then this behaviour would indicate that it is detecting a fault and shutting off. Turn everything off, discharge all output caps and check with the ohmmeter whether you don't have a short to ground at the output. If you measure a lower resistance than few hundred ohms, then start checking the components on the secondary for shorts - bad capacitors, diodes, any transistors or voltage regulators that could be on the 12V rail, etc.

Another thing worth checking is the optocoupler - there is very likely a feedback opto somewhere across the isolation slot between the secondary and primary side. They don't fail often, but I have seen cases where the opto got damaged or the LED became so weak due to age that the phototransistor wasn't opening properly anymore. That would directly affect the startup and the regulation of the supply.

BTW, that UCC3818N is a power factor correction chip, along with the circuitry around it. That is not the part that actually does the voltage regulation.

[Rasz]

If all the FETs are good, then this behaviour would indicate that it is detecting a fault and shutting off. Turn everything off, discharge all output caps and check with the ohmmeter whether you don't have a short to ground at the output. If you measure a lower resistance than few hundred ohms, then start checking the components on the secondary for shorts - bad capacitors, diodes, any transistors or voltage regulators that could be on the 12V rail, etc.

better yet, he said he has two units, one working fine. Nothing easier than laying two pcbs next to each other on the desk and comparing resistances systematically 

btw desoldering primary side was a TOTAL waste of time if you had this 12V even for a fraction of a second.

[singapol]

I think it's best to study and understand how ucc3818 is used in a switchmode supply by reading datasheet and design guide for reference. Is the 5V only for auxillary start up or main supply aslo? Anyway there should be a 12V regualtor ic or a discrete transistor version or it just use 12V rectified by schottkey diode only? PS there is a typical circuit in datasheet on page 7 for primary side. No point shooting in the dark.

ucc3818:

http://www.farnell.com/datasheets/1861018.pdf?_ga=1.257954319.423484779.1454294857

Design guide using ucc3818:

https://www.fairchildsemi.com/application-notes/AN/AN-9041.pdf

[CJay]

better yet, he said he has two units, one working fine. Nothing easier than laying two pcbs next to each other on the desk and comparing resistances systematically 

btw desoldering primary side was a TOTAL waste of time if you had this 12V even for a fraction of a second.

This is where a component tester/octopus is ideal, you can whizz through a power supply or other analogue device in no time with no waiting for the meter to settle, if you've got an identical working unit it's even simpler and quicker.

In principle I would tend to agree that the primary side is unlikely to be at fault but still worth investigation as it's not unknown to have a fault there causing a PSU trip.

Server supplies are usually engineered to fail safe, those supplies would have been a minor cost item in a high ticket price server, having one of them fail and toast the server is not an attractive option so they are usually designed to trip out at the first hint of a failure.