Computer power supply-mother board intermediary tester?

[Dawn]

This has been a chicken or egg problem I've run into several times. The classic computer won't start symptom with the telltale p.s.fan that that just quivers. Ninety plus % chance there's a dead supply and the user has the computer overloaded with peripherals that tax a stock supply. In most all cases, a new supply is all it takes to bring it back to life.

Other times it's been a bad board or peripheral. Sometimes noticeable bulging caps/leakage on a 7+ y/o board. Something I'd take a shot at changing out and ESR'ing suspected others if it were my personal machine. These are obvious or easily traced, but occasionally you get a weird one like I did a few months ago.

Seven y/o machine whith the above classic fan twitch, 200W stock supply with 3 IDE drives, a racehorse video card, obsolete, analog tv tuner card, original modem, not to mention choked with dust and under the CPU heatsink packed with dog hair and dust mat. This supply must have been gasping to keep up. The user reports that it's been taking several tries to get the computer to turn on recently. Now it's dead.


Pull all cards, dust/blow out the machine, unhook everything except board video and add new 480W supply. All is well, add back only video card and drives one at a time, no problems. Two weeks later, computer is dead. pull supply and have a go/no go tester/load, won't start and dead.
Swap supply and all is good again. About a month later the same. Put in another new one and ask them to run it w/ primary drive and board video for a while. Kid's at school, so no gaming and they only use it on the net 24/7. Less then a few days I get a frantic call that smoke is pouring out of the supply. Sure enough, and electrolyte leaking everywhere. Obviously a board problem that took out several supplies I have to eat.


So you know the scenario. You can't be sure in this case intially which was the problem and not much you can do otherwise. Board shows no visible traces of overheating, bad caps, all fans working, and a supply large enough for the job. Home audit shows a perfectly working UPS/Surge supressor and small multipurpose scanner/printer working perfectly on the same ups. Transients on the power line are out of the question. This leaves the board the major suspect.

For supply side testing, I have both a field go/no go load and led idiot box and a more elaborate dynamic LCD metered, load tester to check supplies independently in the shop.
Does anyone know of any diagnostic ATX connector interface that can be put between a good supply and a board that can flag excessive power demands and perhaps help isolating the board? This would be nice tool in the days of aging desktops.

[Monkeh]

So what spec and brand PSU did you fit?

And what components are in the machine?

[pickle9000]

I haven't seen a tester as you described. I have however seen a few instances of a failed electrolytic close to the real time clock battery that has caused a similar problem, in fact when tested they where passable but in reality bad. No boot and no POST indication. Like you say ESR meters are handy in this kind of situation you can easily check all the caps (psu and motherboard) in under a half hour.

My situation was that these where old SBC's used for industrial control. A small form factor PC P2 running DOS. Replacement was not possible so they needed to be repaired.

[Dawn]

Local sourced one hung low 480w p4/SATA compatible supplies from a trade supplier that I've never had problems with and has taken each back for exchange. I've swapped out many using the same guy. Brands in stock vary, but no problems to date with any supply. In this last case, the PC is a packaged Emachines with a non-intel sourced board. no clue to the mnfr. 3 different vendors in the series. '754 sempron 1.8g IIRC. system reduced as close to bare bones during last shot. Just on board memory and original seagate 100g hdd, 2x1g memory. Other drives are pretty old and came from old machines,so pulled just in case, pulled the stock modem and nvida card and some tv card they don't use to further isolate possible faults. Original supply as above was 200W. fine for a bare bones of that time. Just to make sure all base covered, even the cpu & heatsink were pulled,cleaned, and used arctic silver over the white silicone compound used back then just to be thorough when I cleaned up everything and hit all bases. Many of the packaged machines really skimped on the compound with coverage just over the core and I've seen some machines having heat issues after the heat sink is choked with dust or have less then optimum ventilation. In this case, the board boots and initalizes bios sequence, so no post test is needed and fault doesn't occur within a reasonable amount of bench burn in.

This is water under the bridge and I'm not going to be doing board level rework even if there were obvious bad caps on something that's not worth repairing and I don't personally own or a worthwhile labor rate to ring out the caps with an ESR meter. That's enough to declare the board dead unless this was a unreplacable or mission critical machine and I have the paid liberty of pursuing an unguaranteed resolution.

This is a catch-22 bind that's going to be increasingly more common as the desktop market ages. I'm hoping to find some sort of diagnostic ATX connector supply to board test device to prevent having to deal with this kind of problem again. It's not something that I've seen and don't know if anyone sells such a device. Lots of folks don't seem to want to give up functional desktops and the machines I see are getting older. 

[Monkeh]

Local sourced one hung low 480w p4/SATA compatible supplies from a trade supplier that I've never had problems with

Yargh.

Original supply as above was 200W.

There's a lot more to it than the advertised wattage.

[notsob]

From what I have observed, most or the generally available ATX style power supply testers, are pretty much just glorified volt meters.
 I don't think any of them measure actual load [ie current - Amps]on for each Voltage input or connected device.

[alm]

Some reviewers test under load (eg. Hardware secrets). This involves more equipment than a cheap power supply tester with a few LEDs, however. The best test in my opinion is an actual computer. Stress the various parts (CPU, RAM, GPU, disks) and see if it remains stable. You can even measure voltages and noise under this load.

I find it very surprising that a bad mainboard takes out a decent power supply. Good power supplies are protected against over current and over heating. My guess would be that you used two crap power supplies that died because the load was not very well behaved.

[M. András]

we had at home a brand new 200usd+ asus mb for the newer core i7 intels. it took out a 600watts cheiftec ps completly it wasnt older than a year, the mb fried the whole thing after the first restart

[Mint.]

Instead of using cheap power supplies you should seriously just invest to buy a quality one from trstable names such as corsair, ocz, antec, etc.

[NiHaoMike]

If you can find a recent 4U rackmount server for cheap (one that still runs but is obsolete), it's pretty easy to swap out the existing motherboard with a new one. It's a cheap way to get a very well built power supply and a sturdy case designed for excellent cooling.

Beware that "variable speed" is generally not a good feature. All of the good power supplies seem to have it nowadays (server units generally don't but those are really expensive), so try to pick one that has a switch or other adjustment to disable that feature. (It gives them extra marketing material and gains them a tiny bit on the efficiency tests, at the expense of lowered reliability and lifetime.)

Also, for home use, PFC does little or no good while adding points of failure and causing some efficiency loss (especially on 120V). But since all good power supplies seem to have it, there's not much you can do about it unless you're willing to make extensive mods to disable or bypass it.

[Monkeh]

Beware that "variable speed" is generally not a good feature. All of the good power supplies seem to have it nowadays (server units generally don't but those are really expensive), so try to pick one that has a switch or other adjustment to disable that feature. (It gives them extra marketing material and gains them a tiny bit on the efficiency tests, at the expense of lowered reliability and lifetime.)

You mean the fan? It's called noise control, and no, it doesn't make any difference to efficiency or reliability.

Also, for home use, PFC does little or no good

Take PFC off a couple hundred thousand PCs, see how much difference it makes to your grid.

while adding points of failure and causing some efficiency loss (especially on 120V).

No, it doesn't cost efficiency. Using 120V costs efficiency.

I think you might be confusing efficiency with low power consumption. They are two drastically different things, and in any ATX machine, the PSU is the least of your worries for the latter.

[NiHaoMike]

I know that at least Antec is pretty bad at implementing variable speed fans. The power supply would get really hot (to the extent that the coils start giving off a "hot varnish' smell) before the fans start to wind up, which causes the caps to fail early. Also, most modern CPUs (and the latest Nvidia GPUs) have Turbo Boost, where the clock frequency is boosted as long as the core temperature is low enough. Poorly implemented variable speed fans would let the chip heat up and drop out of turbo early, causing a performance loss.

And don't forget that the variable speed circuit itself introduces some more failure points. I managed to get a 550W Antec Trio for free because someone threw it away. When I tried to start it up, the fan did not turn, but all the output voltages checked fine. Turns out a failed transistor disabled the fan, so I just rewired the fan directly to 12V and it worked like a charm.

PCs are such a small percentage of residential load (in most homes anyways) that it really doesn't have much effect. (For businesses, PFC does make sense.)

And there are low power ATX platforms, and not just Atoms either. There's an Ivy Bridge system that idles at just under 6W. (Granted, it has a lot of mods as well as a SSD.)
http://ssj3gohan.tweakblogs.net/blog/8217/fluffy2-59-watt-high-end-desktop-computer.html
At such low load, the magnetizing loss in the PFC inductor can be very significant. (Some power supplies automatically disable the PFC or start period skipping at low load.)

[Monkeh]

I know that at least Antec is pretty bad at implementing variable speed fans. The power supply would get really hot (to the extent that the coils start giving off a "hot varnish' smell) before the fans start to wind up, which causes the caps to fail early.

That would be because 'Antec' PSUs are often just cheap chinese things whacked in a fancy box.

Also, most modern CPUs (and the latest Nvidia GPUs) have Turbo Boost, where the clock frequency is boosted as long as the core temperature is low enough. Poorly implemented variable speed fans would let the chip heat up and drop out of turbo early, causing a performance loss.

PSU fans have nothing to do with this in a modern design, and poor implementations of fan control are not an argument against proper fan control.

And don't forget that the variable speed circuit itself introduces some more failure points. I managed to get a 550W Antec Trio for free because someone threw it away. When I tried to start it up, the fan did not turn, but all the output voltages checked fine. Turns out a failed transistor disabled the fan, so I just rewired the fan directly to 12V and it worked like a charm.

Whoopee, one more transistor in a machine with hundreds of millions of them. See above point about cheap Antec PSUs.

PCs are such a small percentage of residential load (in most homes anyways) that it really doesn't have much effect. (For businesses, PFC does make sense.)

It has an effect when you multiply by millions.

And there are low power ATX platforms, and not just Atoms either. There's an Ivy Bridge system that idles at just under 6W. (Granted, it has a lot of mods as well as a SSD.)

That is nothing like a standard ATX machine, and I'll believe the 6W number when he gets some more reliable equipment to measure it.

At such low load, the magnetizing loss in the PFC inductor can be very significant. (Some power supplies automatically disable the PFC or start period skipping at low load.)

At such low load you don't use a normal ATX supply.

[M. András]

Instead of using cheap power supplies you should seriously just invest to buy a quality one from trstable names such as corsair, ocz, antec, etc.

yeah... its a much older brand then those 3 btw i have 1 older 400watts more than 15 years old and still going without a problem,  the current ones on the 2 desktop are a 550 watts on mine, and a 600 watts ony my fathers desktop, mine is a 8 year old unit. i cant really remember when did corsair start to rebrand atx psus myabe 4-5 years ago? ocz the same i guess. btw these arent cheap most of their higher end models were all delta made

[tom66]

PFC is a waste of time for an average computer power supply.

Resistive loss in UK grid = approx 8.5%
Efficiency of PFC circuit = 85-92%
Uncorrected power factor = ~0.6
Corrected power factor = 0.99

So you lose up to 15% of power in the PFC circuit, to save less than 8.5% (because you still have to draw power... but at least your power is in phase!)

Now there's a good reason manufacturers use it: you can get away with a smaller mains bulk cap, and cheaper main transformer (smaller operating range means it's easier to design), allowing other parts of the power supply to be smaller or cheaper. But it doesn't save energy at all. It costs the customer more than it saves the electricity company.

[T4P]

PFC is a waste of time for an average computer power supply.

Resistive loss in UK grid = approx 8.5%
Efficiency of PFC circuit = 85-92%
Uncorrected power factor = ~0.6
Corrected power factor = 0.99

So you lose up to 15% of power in the PFC circuit, to save less than 8.5% (because you still have to draw power... but at least your power is in phase!)

Now there's a good reason manufacturers use it: you can get away with a smaller mains bulk cap, and cheaper main transformer (smaller operating range means it's easier to design), allowing other parts of the power supply to be smaller or cheaper. But it doesn't save energy at all. It costs the customer more than it saves the electricity company.

Certainly, but only the caps on Active PFCs are alot smaller BUT again, Passive PFC can only be found in cheap arse power supplies, that's the power hungry one and Active PFC is not.