How to test a power supply

I've got a 430W NMB power supply that I need to test. I've already checked it with a multimeter and all the rails are fine. I've used it in a rig for several months and the rails were solid even under full load.

However, I recently killed 3 motherboards and 5 processors. The motherboards were all refurbs from NewEgg, and I can't be sure why they died. It's possible that I just got a string of bad boards, and that they then killed the processors. I know that electrical components fried on each of them. Two of them had mosfets pop, the third had some tiny little thing fry (I couldn't tell what it was after the fact). All of these boards were initially powered by the NMB, though some of them actually died while being powered by an Antec TruePower 430. THe boards would either not post at all, or would post one or two times and then not at all. All of them had at least some difficulty right from the first power on. Most of the chips would never post in any of the boards. Some of them posted once or twice and then never again. I have since tried all of them in a good board, and theyre all dead, 4 of them don't post at all, the 5th generates a memory error when it tries to post. After I had switched to the Antec PSU, I also blew an IC on a floppy drive. The computer wouldn't post with the floppy drive attached. I reconnected all the cables and got a pop and a puff of smoke from the bottom of the floppy. A rather large IC had brust.

Testing the rails with a digital multimeter at no load with the black probe on a ground wire generates solid readings, a bit on the high side, but within reason. However, there are some oddities. First, the power supply case is not grounded. If the black probe is placed on the case, and the red probe on a hot wire, then I get a reading of -00.05v on the 20v scale. The truly odd thing is that I get the same reading if I reverse the probes. Even weirder, no matter what scale and polarity I use, I always get a reading of all 0s and a 5 at the end, negative. So if I use the 200v scale I get 000.5, the 2v scale reads -0.005. If the psu is installed in the computer case, these readings can also be obtained from the computer case. The nature of the readings leads me to suspect I'm actually picking up some EMI. Is this a likely interpretation? I've checkec continuity with the swith on but the unit unplugged, and non of the rails are continuous with the case, not even the ground wires.

I need to know whether this supply is safe. I'd like to continue using it as it's a really nice power supply. However, I don't want to blow up another motherboard, even if it is still under warranty.
How can I test this psu to make sure it's safe? I don't have any other ATX systems to test it on, though I do have some AT systems.

Any input would be greatly appreciated. Including leads on some place that could (quickly) test it for me. Thanks

nihili

you running ok now on a different psu? i have no idea what's going on. definately something. i don't think you should be getting any readings from the psu housing.

Yes, right now I'm running fine onthe Antec 430. The Antec is grounded to the PSU case, so touching the black probe to the case and the red probe to the hot wires yields the same readings as running the black probe to a ground wire.

The thing that puzzles me on the NMB is that the readings I pick up are the same no matter what scale I use. THat's what has me thinking that maybe I'm just picking up EMI. Do you happen to know how I could test the psu without threatening my current setup?

first, glad you got it running!. you know you can short the green wire in the psu's mobo plug over to a ground wire and get it running right? then just monitor the voltages and check the psu housing for positive voltage levels. besides that i don't know how to do any more extensive testing. might be a good idea to check all the wires in the mobo plug for good readings.

This is my personal opinion after 30 years in electronics:

Never use a PSU that does not hold the return for its outputs at safety ground (chassis) since your motherboard and components do. It is a high voltage switching power supply and it will capacitively couple the return to safety ground from the capacitance between all the blcak wires that snake around the chassis. Individually, each black wires insulation does not prodide much capacitance, but together the sum of all of them does. floating the DC ground (black wires) above safety ground blocks DC current from flowing between the two as long as the line voltage and output voltages do not change suddenly, such as occurs when the processor kicks into a high load or a drive spins up, etc. When you get a line or load side transient, for a brief moment there is an additional differential between the chassis and the black wires. It sums with the DC output voltage momentarily, spiking the actual rails way above the values they are supposed to be. The refresh rate of most digital multimeters is too slow to show those transients, but a floating oscilloscope will catch them. If you were only powering a light bulb, the short duration of those transients would not effect it, but semiconductors are fast responding and hence fast damaging devices. That's all the deeper I care to go into the theory behind why it is a dangerous scenario right now, but suffice it to say, stay away from PSUs with a floating return rail.

As to why some of the failures occurred with your trusty Antec, that's harder to speculate upon. In that (those) case(s) it could have been a shortcoming in the device that caused the failure, that devices time had come, or in the case of some motherboards, the metal standoffs that it sets upon which hold it to the case were shorting to a conductor that runs close to the mounting hole on the foil side of the board. Most manufacturers are careful to keep those paths away from the plated-through holes, but some come darn close to them. If you look at the holes for the northbridge heatsink, there are signal traces that come extremely close to those holes and care has to be exercised when putting a heatsink or waterblock on the northbridge which uses metal mounting hardware (screws, flatwashers, nuts, etc). I smoked a motherboard when the nut on the foil side of the board under one of those northbridge holes shorted some nearby traces to the plated-through hole and heatsink. That's why you see most, if not all, northbridge heatsinks retained with plastic snaps or plastic screws with plastic nuts. If you have an aftermarket heatsink or waterblock that you intend to use on your northbridge, inspect the trace clearance on the foil side of the holes before you put metal hardware on them.

Whew....

Hoot

Hoot,

Would it be sufficient to ground the case? I'm guessing that should be done to the third prong of the 120 and not by just hooking a black wire to it. In fact. I think I'll open up the PSU (again) and see if I can figure out why the case isn't grounded. That just seems too weird to me.

And what's your opinion on using plastic standoffs to isolate the motherboard from the case? I checked all the motherboards btw, none of them have traces that would have grounded out with the standoffs.

nihili

Last first. I'm a big fan of not insulating the motherboard from the standoffs despite some argument for only allowing one standoff to conduct to the motherboard and no others, preferably one near the ATX connector. That's based upon the single-point grounding sceme to avoid loops. Problem is if that sole conducting standoff gets corroded or goes to a high-z state for some other reason, no more grounded motherboard, other than through the black leads to the psu. Yes, I definitely prefer to have the motherboard grounded to as many points on the case as possible. These paths are short enough that loops should not be a problem.

The case and PSU housing should always be at the same potential as the safety ground, AKA third prong. Otherwise, when nearby lightning causes ground bounce, you can fry other peripherals connected to your PC. This is a common cause for devices on a serial port to lose an RS-232 driver or receiver, especially modems.

You are correct in pursuing opening up the PSU and figuring out why the output returns (black leads) are floating above safety ground. Usually you see all of them staked to a single point inside the PSU and that point connected to the groundplane of the circuit card which eventually finds its way to the safety ground. Some people will disassemble a power supply to adjust an output pot, clean the fan, who knows why and when they put everything back together, omit some critical screw that just so happens to be the one which bonds the PSU case and/or the black returns to the safety ground. Don't laugh, it happens. Murphy's Law you know.

Hoot

Ok, I've verified that the case is continuous with the safety ground of the outlet. However, the black leads are not continuous with the ground plane of the circuit card (I'm assuming that's the part that connects to the case. Thus the black leads are not continuous with the safety ground.

However, this may be an issue with some resistors. In trying to trace my way from the known ground plane back to the black leads, the first thing I come across is a capacitor (little brown disk with two leads coming out). My meter does not read the two legs of the capacitor as continuous. When I try to measure the resistance on a 2M scale, I get very close to 2M or sometimes over depending on how good a connection I get. On the capacitor it says:

F
104
KCK

There are several of these capacitors ont the boards of the PSU. Some of them measure almost no resistance, some measure at or above 2M.

Does this indicate a problem? I'm going to carefully reassemble the PSU and see how it does on the multimeter. I'll let you know.

nihili

Ok, after reassembly it is essentially the same as before. I have good solid readings on all rails when measure between hot and black leads. But measuring to the case instead of the black leads yields a minimal negative reading regardless of polarity or scale. This time it was 0.001 regardless of scale. After turning the supply off I checked again and the case is indeed connected to safety ground.

I'll have another look inside tomorrow. Any hints about what I'm looking for?

nihili

I have a Delta DPS-300BB where the earth ground is connected to case but the DC ground (black wires on the output side) is not, except at high frequency, through 4700pF capacitors. I also have an Antec 300W SmartPower where the DC ground is connected to the case only through a jumper wire. I read that only the better quality power supplies keep the AC ground separated from the DC ground and that it's actually desirable to keep the AC and DC grounds separate and to not have current flowing through the chassis, although I don't understand why.

nihili said:

Ok, I've verified that the case is continuous with the safety ground of the outlet. However, the black leads are not continuous with the ground plane of the circuit card (I'm assuming that's the part that connects to the case. Thus the black leads are not continuous with the safety ground.

However, this may be an issue with some resistors. In trying to trace my way from the known ground plane back to the black leads, the first thing I come across is a capacitor (little brown disk with two leads coming out). My meter does not read the two legs of the capacitor as continuous. When I try to measure the resistance on a 2M scale, I get very close to 2M or sometimes over depending on how good a connection I get. On the capacitor it says:

F
104
KCK

There are several of these capacitors ont the boards of the PSU. Some of them measure almost no resistance, some measure at or above 2M.

Does this indicate a problem? I'm going to carefully reassemble the PSU and see how it does on the multimeter. I'll let you know.

nihili

Click to expand...

Those are .1 microfarad disk ceramic capacitors. They are quite common in bypassing.

I didn't want to start down this slippery slope because there are equally good arguments for floating a power supplys output above safety ground and grounding it. Each camp has supporters and detractors depending upon the particular application (does not mean software). All I am saying is that in the case of all the PCs I have had success with, the PSU, motherboard, components, et-al, have had their return rail held at a DC ground (meaning dead short) common to safety ground, at every point where that can be accomplished. So, my opinion on this issue is based more upon anecdotal experiences than pure electronic theory. "If it works, don't fix it." If I had a PSU that for whatever reason, trouble seemed to follow, I'd toss it.

Hoot

Hoot said:

If I had a PSU that for whatever reason, trouble seemed to follow, I'd toss it.

Click to expand...

Have to agree.

It sounds to me like the supply is probably putting out large voltage spikes. Maybe something as simple as an output filter cap that has failed open. (Although they usually fail with a short and a bang and a puff of smoke.) Without a scope it is going to be near impossible to tell what the problem is. And without a lot of experience repairing supplies, it would be easy to replace a bad component that is only a symptom of the underlying problem. (Fairly likely this is what happened when it was "refurbished".)