How to check for a short before powering it on

[sourcerer]

I did some research into GPU repair lately, and there is one thing that I definitely learnt from that field: Always check for shorts before turning it on. How to check for shorts? Well, that is surprisingly simple: You take a multi-meter, put it into resistance mode, and measure the resistance of the device that you want to test on the power pins / power connector. If the resistance is too low, then you have a short and you will most likely get the magic smoke when you turn it on.
For SATA disks you check the SATA power pins (one pin of your multimeter to the SATA ground pin, and another pin of your multimeter to SATA +5V pin. SATA provides 3 different power levels, +3V, +5V, +12V so you should check all 3 different power level pins.
As a reference, on the Samsung EVO840, the +3.3V on pin 1 gives 7.9 MegaOhm, the +5V on pin 8 gives 9MegaOhm and the +12V on pin 15 gives infinite resistance, which means that it is not connected.
If you measure a resistance of less than 10 Ohm, you most likely have a short and should fix the short before applying power to it.
You can apply the same procedure to nearly every device, from hair driers to refridgerators to mainboards, ... which you get and you aren't sure whether they are OK or not. Just measure the resistance.
The only exception I have come across yet are GPUs , the GPU chips actually have a very low resistance so for them, measuring 0.9 Ohm of resistance on the power rails can be OK, so you need for them you need more knowledge on the expected resistances of the GPU chips to decide whether they are OK or not. For everything else, <10 Ohm should be a usable rule.
Since I learnt about the resistance measurement a few months ago, I applied to to many different devices and successfully identified shorts in devices before they went up in smoke. Therefore I highly recommend that you check the resistances on every customer disk you get before powering it up. This way you can save them from bigger damage.

[grevan]

That´s a must tip, i was just wondering why +12v is not connected, i think it may not share the same ground.

[pepe]

It is less likely to smoke if it is shorted, coz it shuts down the PSU in that case. Problems come when it is not shorted enough to shut it down and some ampers can happily smoke it as you said.
However, i wonder how - on any power rail of a digital circuit - can the static, off state resistance be 0.9 ohms. It just doesn't fit. Have you checked if it is actually the GPU chip exhibiting this resistance or some other component on the board?
I can only think of some protection resistor put in paralel with the GPU chip on some low voltage power input (1.2V for example) which would still burn around 1W for no reason, but it sounds abnormal for the chip having so low off-state resistance.

Anyway, yes, checking it is useful, (albeit i usually do it in diode mode).
The worst things can probably happen if one has a cheap PSU that regulates 5V only, and by overloading it without reaching the shutdown current, the other voltages can rise significantly, burning stuff on the 12V line for example.

pepe

[pepe]

That´s a must tip, i was just wondering why +12v is not connected, i think it may not share the same ground.

12V is not used in an SSD (usually). It might be used in enterprise SSDs though. And it does share the same GND.
3.5" HDDs use it to feed the spindle/vcm controller only.

pepe

Last edited by pepe on September 28th, 2021, 17:27, edited 1 time in total.

[fzabkar]

That´s a must tip, i was just wondering why +12v is not connected, i think it may not share the same ground.

Consumer grade SSDs generally don't use the SATA +12V supply. Some enterprise SSDs do, though. I've also seen one 2.5" enterprise HDD which uses +12V.

The 840 Evo in fact generates its own +12V supply from the SATA +5V supply via a switchmode boost converter. I believe this voltage is used by the NANDs during writing.

[fzabkar]

However, i wonder how - on any power rail of a digital circuit - can the static, off state resistance be 0.9 ohms. It just doesn't fit. Have you checked if it is actually the GPU chip exhibiting this resistance or some other component on the board?

I believe this is normal, even though it looks strange to me, too. I've seen it several times at eevblog.com.

[pepe]

I am curious what voltage level is the affected power line...

[fzabkar]

I am curious what voltage level is the affected power line...

Vcore, 0.9V or thereabouts, IIRC.

[pepe]

not quoting myself, i thought about the same, but the question was rather like what's the real-life value there. Measured on an actual card.

[fzabkar]

not quoting myself, i thought about the same, but the question was rather like what's the real-life value there. Measured on an actual card.

eevblog.com has plenty of real measurements. I assume the OP is quoting figures from personal experience.

[sin]

I did some research into GPU repair lately, and there is one thing that I definitely learnt from that field: Always check for shorts before turning it on. How to check for shorts? Well, that is surprisingly simple: You take a multi-meter, put it into resistance mode, and measure the resistance of the device that you want to test on the power pins / power connector. If the resistance is too low, then you have a short and you will most likely get the magic smoke when you turn it on.
For SATA disks you check the SATA power pins (one pin of your multimeter to the SATA ground pin, and another pin of your multimeter to SATA +5V pin. SATA provides 3 different power levels, +3V, +5V, +12V so you should check all 3 different power level pins.
As a reference, on the Samsung EVO840, the +3.3V on pin 1 gives 7.9 MegaOhm, the +5V on pin 8 gives 9MegaOhm and the +12V on pin 15 gives infinite resistance, which means that it is not connected.
If you measure a resistance of less than 10 Ohm, you most likely have a short and should fix the short before applying power to it.
You can apply the same procedure to nearly every device, from hair driers to refridgerators to mainboards, ... which you get and you aren't sure whether they are OK or not. Just measure the resistance.
The only exception I have come across yet are GPUs , the GPU chips actually have a very low resistance so for them, measuring 0.9 Ohm of resistance on the power rails can be OK, so you need for them you need more knowledge on the expected resistances of the GPU chips to decide whether they are OK or not. For everything else, <10 Ohm should be a usable rule.
Since I learnt about the resistance measurement a few months ago, I applied to to many different devices and successfully identified shorts in devices before they went up in smoke. Therefore I highly recommend that you check the resistances on every customer disk you get before powering it up. This way you can save them from bigger damage.

What you have highlighted makes sense.

Infact even while repairing audio power amps, studio gear (analog compressors, eqs, preamps, audio monitors etc etc) this trick is always handy
1)a transformer at primary in such cases will always be DC shorted
2)Post rectifier, pre regulators and post regulators this trick works neatly even in the above-mentioned cases of audio gear.



--

[sourcerer]

I made a table of various HDDs and SSDs and their resistances on the SATA/Molex connectors:
https://docs.google.com/spreadsheets/d/ ... mwHdpErRI/
If anyone wants to contribute to the table, just let me know.

[pepe]

Since these are semiconductors, resistance is not a linear thing, so it may vary with the measuring current, thus with type of multimeter used.
my oppinion.
Btw, have you compared multiple samples of the same device?

pepe

[michael chiklis]

A short usually develops heat if the shorted component has pretty low resistance (if is crossed by a large amount of current).
You can see the heat with a thermocam (which is very expensive) and thus locate the shorted component.

If you can't afford a thermocam you can use freezing can, but that's not very useful in summer because the freezed surface might dry too fast and you'll not have enough time to spot the point where it dries first (expecially if you're working on large PCBs, or if a large portion of pcb surface is suspected).
The same can happen by using IPA alcohol instead freeze.

If you are interested you might use this "new" technology that covers large portion of pcb surface with white powder created from solid Rosin flux
https://aliexpress.com/item/1005002997973015.html

It will stay on pcb longer than freeze or alcohol so it will be easier to find the point where the heat is coming from.