Finding a dead short component

[Beton]

Hi guys!
I searched for the term "dead short" in the forum but no specific thread came up however if I'm just blind and couldn't see the thread talking about this please forward me to that direction.
So my problem is I don't know how to find the source of a dead shorted voltage rail. Dead short = ~0 ohm resistance.
I have two VGAs (HD 4850 and HD 4870), both dead shorted to ground on the memory rail. I have the boardview file for the 4870 and the memory rail consists around ~7892346 capacitors...
I have a bench PSU and if I try to inject voltage the voltage drops to ~0.1V (if I remember correctly). I can go up to 6-7 amps but nothing gets hot at least not in a way I can feel it with my hand/fingers or can see it with ipa.
Obviously taking into consideration Ohm's law this is expected.
I suspect I killed these cards a long time ago by applying a lot of current (if they were not dead before) nevertheless I don't want to make the same mistake in the future.
So I tried measuring the resistance at different points on that rail but my multimeter only shows four digits in ohms mode. The lowest on the rail was 00.21 ohm but that's the resistance of the leads...
I tried measuring voltage at different points (while injecting voltage), I can measure 4 digits in millivolts but that didn't show me anything meaningful (or I'm just stupid).

So what do you suggest? I've seen a Northridgefix video in which the guy pinpointed the location of the short on a dead shorted board with a Flir cam, the heat spot (mosfet) was barely noticable on the cam (only a few degree difference vs the board) so even with a Flir cam it's not easy but I don't have a Flir cam.

How do you deal with this situation?
Thank you in advance.

[foshland]

Re: Finding a dead short component

Hi, first of all, how many volts did you inject? You should never inject more than the rail is suppose to have in working condition, for example, in Vcore, never go up to 0.8 or more, on memory work with 1V or near it.

Have you tried with alcohol? it disipates faster in the heated zone, apply in the suspected zone and inject, maybe you find it.

Some people uses flux atomizers, others thermal cams. I invested in a Seek thermal cam for mobile phones, 250-300$ and is a good invest for this and other things

[Beton]

Re: Finding a dead short component

Hi, first of all, how many volts did you inject? You should never inject more than the rail is suppose to have in working condition, for example, in Vcore, never go up to 0.8 or more, on memory work with 1V or near it.

Have you tried with alcohol? it disipates faster in the heated zone, apply in the suspected zone and inject, maybe you find it.

Some people uses flux atomizers, others thermal cams. I invested in a Seek thermal cam for mobile phones, 250-300$ and is a good invest for this and other things

I tried with 1V, I know it has to be equal or lower than the rail voltage itself. But it doesn't matter because the short protection lowers the voltage on the psu.
I tried the alcohol method, ipa is isopropyl alcohol. Btw a 0.1 ohm resistor was the hottest (barely noticable hot) by my observation with ipa but the resistor is fine.
I haven't tried the flux atomizer yet.
I'm asking this because as I said a 0 ohm dead short not gonna get hot, that's the whole point.

[DynaxSC]

Re: Finding a dead short component

U can try to examine the MLCC caps with a magnifier glass, sometimes the shorted caps have some visible irregularities. Other way is apply more amps, with 15-20 amps, you schould be able to observe faster disappearing of alcohol, however it is not easy to get 15-20 amps with a normal psu. I put two 10 amps psu's in paralel. You might also see the area where most current is flowing narrowing the search. Cables must be thick too. The not easy way is desolder the rams and gpu, then you must not care so much of voltage, as caps have usually higher tolerances. But you need then to reball the devices and solder back after shortage found. Do it only if you know how to. Another method is to measure when injecting the voltage on the caps in a sensitive mv range with a high resolution meter, the voltage near the short will be with good chance dependent on the board topolgy and injection point lowest.

[eccerr0r]

Re: Finding a dead short component

getting two PSUs in parallel is also a feat...not all can be connected in parallel.

A low resistance device will indeed get hot if you force it...

[Beton]

Re: Finding a dead short component

Wouldn't applying such high amperage destroy the traces? I know these VGAs in question might be already totally dead but in case I run into something else with similar problem.
Btw it just occured to me that the testing probes were hot when I injected voltage, I mean I could feel it but nothing serious... but not the VGA.
My psu on paper is capable of 30V/10A.

[DynaxSC]

Re: Finding a dead short component

Hi, I did go to ca 15A on different GTX1070, you need to do it step by step, not 20A at once, this can lead to track burnout, but if you do it with care and slowly, you should be safe on the safe side. Also it is not easy to get 20A with 0,8V, due to the supply cable resistance. If you have luck, 10A might be enough, depends on "how good" the short is. The alternative is to desolder all caps one by another, could be time consuming and destroy the board too from heat.

[DynaxSC]

Re: Finding a dead short component

... and solder the supply cables to the board, the positive lead after the coils, eg. on the memory secondary coil side, and the negative lead somewhere to the video output sockets on the back side of the card, and observe what is happening when increasing the amperes, I use my hand to do it, the best tool to feel the heat, however do it with care, at least the first touch should be very short. Isopropanol is also OK, and has the protective advantage, that it takes heat when vaporating, at least for a short time when something becomes really hot.