Measuring 12V on a dual rail psu

[rscl]

Hello,

I was trying to measure my +12V but ended up uneasy since I'm not sure if I'm doing it correctly. I would greatly appreciate if you could help.


Background: Fan broke down on my PSU, I was way too scared to change it so I bought a slightly worse PSU (17A/17A instead of 18A/18A) from FSP.

I measured how much I was actually using from the 500W supply (300W) so I thought a 470W couldn't be too bad.

Apparently I was wrong. I started playing a low-load game and I suddenly heard my GPU fan spin furiously. It would seem as if this PSU change has increased GPU heat from 55C to 88C which made my GPU fan run at almost full speed (checked @ MSI Afterburner)

Not that I know much, but I realized that the PSU must be overloaded. So I realized I better take a shot at changing that PSU fan even if it means I'll die touching those capacitors.

I had to cut the wires to the fan-connector on the PSU and connect the new fan using electrical tape since the connector was located in a very weird place I couldn't reach. I realized that I changed from a Yate Loon temp sensitive 0.30A fan to a 0.35A Zalman fan as well.

Everything seems to be working well, bios reads 12.40V which should be no biggie.

Speedfan does not seem to be able to measure too well, it shows 0,63-1,48V which made me unsure so I got a instrument to measure, which brings us to the question:


Is it possible to measure 12V by reading a molex connector? Or do I ultimately just read the 5V? I measured the 7V and the 5V which together read 12,38V

Do I have to slaughter a fan to measure the 12V or does this way work too?

[Shadowww]

Is it possible to measure 12V by reading a molex connector? Or do I ultimately just read the 5V? I measured the 7V and the 5V which together read 12,38VOn molex connector:

Black wires are common ground
Red wire is +5V
Yellow wire is +12V.

On PCIe, CPU power and other wires it should be pretty much same, except for few more colors:
orange = +3.3V,
purple = +5VSB (one that can provide current even when your PC is powered down),
green = PS_ON (connect it to any common ground to 'jumpstart' your PSU without actually having motherboard or anything like that connected to it),
blue = -12V (it's fine for it to be out of spec, like -13V or so, it's unused in modern systems),
white = -5VDC (may even not be present),
gray = PWR_OK (should be floating around 5V).

[rscl]

Oh, ok. I tried measuring red+black and then red+yellow added those together.
It should be the same though, right? Or do you think I should do it again, correctly? (yellow+black) or perhaps use the built-in gpu and measure through a PCI-E connector?

EDIT:
I measured it correctly now and it reads 12,40 idle and 12,34 on load. I suppose I'm all good.


I am still curious to for what reason a psu can make a GPU 30C hotter while not adding any stability issues in general.

[Shadowww]

12.4 seems a bit on a high side - it's not far away from maximum of 12.6.
Mine PSU has around 12.1 on idle and 11.9 on load.

Try changing your multimeter's battery and see if readings change - usually multimeters show too high voltages when their batteries are near to empty.

[Digerati]

Try changing your multimeter's battery and see if readings change - usually multimeters show too high voltages when their batteries are near to empty. Ummm, no, I don't think so - multimeters typically don't need batteries to measure current or voltage. Batteries in multimeters are used for continuity testing and resistance measuring, and to illuminate the display on digital meters - not for measuring voltage.

Also, the ATX12V Form Factor PSU Design Guide, Ver 2.2, March 2005 (http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf), Table 2 on page 13 states the +12V tolerance can go to ± 10% at peak loading. So that would mean voltages up to 13.2VDC are still acceptable. Therefore, 12.4VDC (12.34VDC under load in your case), which is still well within the ±5% tolerance, is nothing to worry about as the motherboard and graphics card regulator circuits can easily compensate for that minor variance.

That said, a PSU cannot be conclusively tested with a common multimeter because most simply are not capable of measuring ripple (http://en.wikipedia.org/wiki/Ripple_(electrical)) and other anomalies. For that, you need an oscilloscope or power analyzer - sophisticated and expensive test equipment that takes special training to use accurately. Therefore, if you suspect your PSU is giving problems, the best solution for most users is to swap in a known good PSU.

[Shadowww]

Ummm, no, I don't think so - multimeters typically don't need batteries to measure current or voltage. Batteries in multimeters are used for continuity testing and resistance measuring, and to illuminate the display on digital meters - not for measuring voltage.From manual of my cheapo multimeter:

"Low battery indication: 7V typ, with voltage/current accuracy lost at 6.3"
:P

[Digerati]

Interesting. What is the brand and model number? Again, I see no reason a meter needs a battery to read voltage or current in a circuit - other than to power the display on a digital meter.

Multimeters (http://www.kpsec.freeuk.com/multimtr.htm)All digital meters contain a battery to power the display

[cypherpunks]

Ummm, no, I don't think so - multimeters typically don't need batteries to measure current or voltage. Batteries in multimeters are used for continuity testing and resistance measuring, and to illuminate the display on digital meters - not for measuring voltage.

This is utterly false. A decent multimeter pulls very little current when measuring voltages, certainly too little to power its workings. Heck, good-quality old-fashioned VTVM or FET-input voltmeters needed power for their high-impedance input amplifier stage.

My shitty little Harbour Freight multimeter (http://tubesound.com/2011/05/13/limitations-of-cheap-meters/) has an input impedance of 1 Mohm, meaning that it draws 12 μA from the +12V rail. Do you think that 144 μW is enough to run the multimeter? What about 25 μW (the +5V rail)? 10 μW (3.3V rail)? What about 0V that's displayed when nothing is connected?

The only multimeters that don't need power are your basic cheap analog multimeters, a rarity these days.

Likewise, when measuring currents, the voltage drop is less than 100 mV, again too low to power anything.

And Shadowww is right; often when the batteries are low, the first thing to suffer is the voltage reference circuit, so the input voltage is compared against a falsely low reference, thus producing an erroneously high reading.

[Digerati]

A decent multimeter pulls very little current when measuring voltagesNow wait! That's how any meter works - even a $5 meter. certainly too little to power its workingsNo. Sorry. So you think a modern digital multimeter requires more power to run than an old analog meter (discounting the power required to illuminate the LCD display)?

And yes, I think a good meter can easily be powered by 10 μW - since the voltage is the potential, not the current. Note my old analog Simpson 260 had a 50µA scale, meaning it could measure current levels under 10µA - and that was 40 years ago!

100mV to low to power anything??? Sorry, but 1/10 of a volt can power lots of things. The purpose of that 1MΩ resistor is so it will not load down the circuit (though I suspect it is really a 10MΩ resistor - a more commonly used resistance in today's digital meters).

At any rate, I concede there are many multimeters that use batteries to measure and display voltage and current in a circuit though I note there are many that don't too. And many that do require batteries do so simply to operate the digital display (and resistance/continuity) because in practice, you do not need power to read power.

So, to prevent further argument, I bow out and say if your meter is one that requires batteries for voltage and current measurements, then by all means, make sure they are good. That said, I think most meters that use batteries have a battery strength indicator too. And if you can't trust that, then can you trust the meter?

[rscl]

Thanks for your help guys. For the record, the first multimeter I used had a depleted battery which we swapped out for a new one. (The one I used to measure correctly but using the wrong method (added 7V+5V).

And the second measuring I did (with the correct method) was with a brand new multimeter I bought for myself. So I have no doubts regarding this :D

I understand 12.40V isn't optimal but as long my HDDs wont instantly die I'm not complaining :d

[jonnyGURU]

12.4V is fine. Anything higher than 12.6V is over spec.