Gigabyte PSU as a high current source

[enut11]

I started to mod this PC PSU for low voltage/high current testing of shunts. According to the label, it is supposed to work up to 36A at 3.3v or 30A at 5v. More than enough for my tests.

Starting with the 3.3v, I paralleled about half of the black wires and all of the orange wires including the brown voltage sense wire. At a 15A load the voltage dropped initially to 2.6v and slowly went up to settle on 2.9v. As this load was less than half the rated capacity it seems like very poor regulation. Maybe computers are tolerant of this drop?

Also, does anyone know if the 3.3v is derived from the 5v as I noticed a drop in the 5v rail as I loaded the 3.3v rail.
enut11

[jonpaul]

PC ATX PSU are regulated SMPS at primary side of SMPS xformer, and all Vout will track ONE sensed voltage eg 5V or 3.3V.

The designs assume a normal PC  load with most of the outputs loaded and at least +5/+3.3/+12 with some amps.

Your attempted use cannot expect normal regulation.

Finally the PSU may be defective.

See links for dummy load tests of ATX /PC PSU.

First verify it is OK under a normal load.

Jon

[Terry Bites]

Its unlikely. They're both high current outputs. If the 3V3 would be a heavy the burden on the 5V circuit. It would be a switcher running from a switcher, nah...
You could always open it up (power off) and see if there's buck converter for 3V3 pulling power from 5V. It seems unlilkely.

[coromonadalix]

sure  it could be a possibility

but is it worthwile ? 

you have server grade psus for cheap prices

even theses https://www.ebay.com/itm/193878694060?     ok its an 30 amp   but there is other models  etc ...

https://www.ebay.com/itm/264375040903?     seems a 60 amp model ??

[enut11]

PC ATX PSU are regulated SMPS at primary side of SMPS xformer, and all Vout will track ONE sensed voltage eg 5V or 3.3V.

The designs assume a normal PC  load with most of the outputs loaded and at least +5/+3.3/+12 with some amps.

Your attempted use cannot expect normal regulation.

Finally the PSU may be defective.

See links for dummy load tests of ATX /PC PSU.

First verify it is OK under a normal load.

Jon

So what would a 'normal' load be for such a supply?

[enut11]

sure  it could be a possibility

but is it worthwile ? 

you have server grade psus for cheap prices

even theses https://www.ebay.com/itm/193878694060?     ok its an 30 amp   but there is other models  etc ...

https://www.ebay.com/itm/264375040903?     seems a 60 amp model ??

Yes, only cheap if you live in the USA. Most are well over $200AU by the time they are shipped to Australia.

I am wondering now if the PC PSU can be modded for constant current operation?

[MathWizard]

IDK that PSU model but it's not uncommon to have only 12V output from the main transformer, then have 2 step down converters to make 3.3V and 5V. The typical output supervisor chip, of the overall PSU, would be monitoring the voltage and current of all 3 main rails and not be happy with UV/OV/OC.

IDK what happens when 1 rail uses way more power than the others. The PFC/PWM chip that runs the main transformer, usually only gets feedback from a circuit on the 12V rail, at least the ones I've seen.

IDK the controller chips near good enough to get the small signal details right, or what the transformer and any buck converters would really do, when just 3v3 or 5V is loaded, but here's some datasheets for some really common controller chips. Most computer PSU's use them or clones of them, with the same pin out too.

If I read some of these design guide PDFs, I'd do better at repairing these, but I have other stuff to learn 1st.

[M0HZH]

Add some load on the other rails and you should be fine.

I found 1A load on the 5V rail is more than enough to get maximum current out of the 12V rail on most ATX PSUs.

[enut11]

Modified Gigabyte power supply with 3.3v, 5v and 12v outputs. The 12v rail was loaded with an internal 47R 5W resistor.

I tested the 5v rail at ~18A with the result shown below. The initial voltage dropped from about 5.1v to 4.66v under load which does not look good. However, the output was then stable to within ~2mV during the 1.5min test. This was 0.043% short term regulation under load and good enough for my shunt cal tests.

[Vovk_Z]

So what would a 'normal' load be for such a supply?

Add about 5-10% of rated load (or at least 1-2 Amps of load) to unused channels.