[jr2]

Hi guys, just dealt with this problem and wanted to know what you thought:

Customer has PS/2 mouse that won't work (vertical won't work).

Sell customer used PS/2 mouse - 5 bucks

Customer's new mouse (that you tested in the store in front of him) doesn't work

Customer (meanwhile) buys new memory on your advice (just for performance, not to fix the mouse) updating from 128MB to 1GB, running XP and installs it himself.

Customer troubleshoots mouse over phone - it won't even work in safe mode, it refuses to work - XP drivers uninstalled / reinstalled, sfc /scannow run, still no dice.

Customer takes computer to you to fix troubled mouse.

You try to power on the computer - no video

Unplug all peripherals, add-on cards, and disk drives, still no video

Turns out to be:   Bad power supply.  Mouse works fine now.



EDIT: And all voltages on the old P/S were within tolerance and the power good signal is fine.  Tested with a power supply tester.  I just threw it on a new P/S for the heck of it and everything started working again.

[Iss Mneur]

That doesn't surprise me, bad power supplies can cause any number of problems.

As for the voltages on the old power supply, did you test while the power supply was under load or just shorted the powergood with the power supply on the bench?  Also, even tiny out of range events can cause problems, because not everything will have a capacitor to protect it or be correctly spec'd to work within the input tolerance.

[jr2]

P/S was under load from disk drives only - I had it plugged into a Coolmax PS-224 PSU tester.

[Tyrian]

Electrical engineering student here, so I feel compelled to post.  

I've had known bad power supplies test normally when their enable/sense line was shorted to ground.  (Which is essentially what the mobo does to tell the PSU to start pushing current.)  If you take a multimeter and probe the voltages on the 20/24 pin connector they'll usually come back normal when open circuited.  However, when you put the supply under load, the voltages at the pins will actually drop, a phenomenon known as "pull down".  Normally the supply with compensate for this.  However, if the compensator (which is sometimes just a resistor) is blown, which it sounds like it was here, then those voltages can drop substantially under load.  Even with a fairly small load, you can see drops of as much as 2V on a +12V DC line, which is about 17% error.  Who knows what else gets undervolted as a result; that depends on the mobo configuration.  

EDIT:  I'm going to take a closer look at that tester later...

[S-99]

So, if the power supply actually works, that means take a look at the motherboard next (after testing power supply in known working computer (you got one around, it's that loaner/doesn't really matter what happens to it computer that in general works pretty ok, but since it has no purpose, and it's an extra, you normally use it for frankenstein experiments). But, if you're still certain the psu is the problem, plug a known working psu in the customer's computer and see if you can turn it on with that. If not, that means take a look at the motherboard.

So what if they end up needing a new motherboard, just get them one they can plug all of their existing hardware into.
I recently had a problem similar to this last week. The computer would actually turn on, but as far as i could tell nothing was happening except the cooler fan for the cpu and the cooler fan for psu were working. Surely enough every component in that computer was just fine except for the motherboard. The problem was those really cheap business grade desktops from dell have sub quality proprietary motherboards (made by dell of course). They die out in like a couple years, most of the time for no apparent reason than that it decided to give up the ghost and not turn on one morning at work.

[Kosh]

PS/2 mouse? 128 MB of installed RAM? What kind of dinosaur did he have? 

[chief1983]

At least he was on an NT-based OS, unlike some people around this forum...

[Tyrian]

So, if the power supply actually works, that means take a look at the motherboard next (after testing power supply in known working computer (you got one around, it's that loaner/doesn't really matter what happens to it computer that in general works pretty ok, but since it has no purpose, and it's an extra, you normally use it for frankenstein experiments). But, if you're still certain the psu is the problem, plug a known working psu in the customer's computer and see if you can turn it on with that. If not, that means take a look at the motherboard.

Actually, if the power supply works, I'd start with the RAM first.  That can be easily wrung out with a tool like MemTest.  Next I'd do the processor, which can be tested with Prime95's torture test feature.  (Just make sure you run SpeenFan or another temperature monitoring program along side it, so you don't burn out the chip.  This will also tell you if you have a thermal problem.)  The peripheral cards that draw power from the PSU I'd test next, if only for the reason that they can be quickly plugged into another system and benched there.  If all those pass, then I'd take a look at the motherboard, because that's hardest to test for functionality.  Granted this is prioritized based on ease of testing.  If you have some behavioral evidence that a particular part is bad, like the mouse in the above case, and replacing the mouse doesn't work, then I'd start there.  Given that the mouse didn't work and he stated that the problem was the PSU, I just jumped to probing the 20/24 pin connector.

One question about that PSU tester: Do you just plug your supply in to it and hit "Go" or does it sit between the PSU and the components it powers?

[jr2]

PS/2 mouse? 128 MB of installed RAM? What kind of dinosaur did he have?  

Elitegroup K7VMM+ rev 3.1 with a 1GHz Athlon.

EDIT:

One question about that PSU tester: Do you just plug your supply in to it and hit "Go" or does it sit between the PSU and the components it powers?

Unfortunately you plug it in and hit go... so there is no load on the PSU when it's tested.  xP bummer.  Anyone know where I can get a Y-splitter for a 20/24 pin power connector so I can put a load on it at the same time?

[S-99]

Actually, if the power supply works, I'd start with the RAM first.  That can be easily wrung out with a tool like MemTest.  Next I'd do the processor, which can be tested with Prime95's torture test feature.  (Just make sure you run SpeenFan or another temperature monitoring program along side it, so you don't burn out the chip.  This will also tell you if you have a thermal problem.)  The peripheral cards that draw power from the PSU I'd test next, if only for the reason that they can be quickly plugged into another system and benched there.  If all those pass, then I'd take a look at the motherboard, because that's hardest to test for functionality.  Granted this is prioritized based on ease of testing.  If you have some behavioral evidence that a particular part is bad, like the mouse in the above case, and replacing the mouse doesn't work, then I'd start there.  Given that the mouse didn't work and he stated that the problem was the PSU, I just jumped to probing the 20/24 pin connector.

In your case your problem was a little more different than the one i mentioned, they shared similarities though. Albeit your problem requires all of this extra testing because the cause is a little more mysterious compared to mine where the computer just plain old wouldn't turn on and ultimately was just plain old easier to diagnose which also equalled less hardware for me to test (mobo and psu).

Ultimately, if a computer doesn't want to turn on, as a start going straight for the mobo and the psu to check them out would be warranted.

[Kosh]

At least he was on an NT-based OS, unlike some people around this forum...

That's what he upgraded to, not what he upgraded from.

[chief1983]

At least he was is on an NT-based OS, unlike some people around this forum...

Ok fine, fixed.

[Tyrian]

Unfortunately you plug it in and hit go... so there is no load on the PSU when it's tested.  xP bummer.  Anyone know where I can get a Y-splitter for a 20/24 pin power connector so I can put a load on it at the same time?

Just using a Y-splitter, if my mental picture is correct, won't actually help.  Some of the power supply's behavior will actually be determined by the load attached to it.  (See Ohm's Law.)  For you to manually load the supply would mean that you'd have to know the equivalent input resistance for that port.  And getting that is impossible due to all the (millions of) transistors in the circuit.  A transistor's resistance will actually change depending on the voltages you apply to certain parts of it. 

I've dealt with similar problems to what you've described before.  I currently have on my to-do list to design a pass-through connector that will allow you to probe voltages and currents while the supply is running your system.

[jr2]

I've dealt with similar problems to what you've described before.  I currently have on my to-do list to design a pass-through connector that will allow you to probe voltages and currents while the supply is running your system.

That would be sweet... sell it on eBay, you might make some dough.  I'll buy it anyways as long as the price is reasonable.