ATX Power supply in High accuracy measurements

[grimmjaw]

Hi eevbloger,
Has anyone used an ATX Power Supply (PC power supply) to power a precision DAQ circuit  . Since an ATX power supply has 5V,+-12V, and sometime 24V rails
it would be great for powering  a circuit.

Say a 24bit- DeltaSigma ADC with a external 1.024V Voltage reference with fontend instrumentation amplifier (like PGA281 from TI).The meausrement is slow around  100SPS- Would the ripple from the power supply effect the measurements?the ripple/noise for a typical ATX can be ca. 120mV.

Regards,
Grimmjaw

[Monkeh]

ATX supplies do not have 24V.

Precision and a butchered PC PSU generally don't mix..

[ejeffrey]

ATX power supplies are generally pretty noisy.  They also have extremely high current output on the primary rails (3.3, 5, 12), which can make them a fire hazard if used improperly.  You can certainly use them to power circuits, but you need to use fuses to protect your wiring, and add additional filtering and post-regulation to get something quiet enough to use for analog electronics.

[dannyf]

Say a 24bit- DeltaSigma ADC with...

As long as your target circuit doesn't demand lots of current (it doesn't sound like in your case), you can use the atx to power a linear regulator which then powers your circuitry.

[dannyf]

Another good / inexpensive source to consider is laptop power supplies. They are typically 19-20v and can supply more power than 99% of what your projects will need.

[dannyf]

For example, I have built quite a few Hakko clones using HP / IBM laptop power supplies. Couldn't be happier.

[SArepairman]

thats like tying to do surgery in a jeep

get a linear supply!!!

If you do make some ATX>Linear monstrosity be sure to fuse it you don't want 20 amps going into something if something shorts out.

[grimmjaw]

Hi everyone.

I'm planning to use polyfuses on all the rails to protect the circuit.Since the ATX power supply is use in powering CPU chip and other
sensitive chip should't it be stabil enough?

[casinada]

As they're switching power supplies, they're expecting some load at least on the 3.3V, 5V, and 12V rails. Some have separate rails for the same voltages. You have to provide a load for every rail otherwise the power supply might not work as expected. This power supplies are designed to have good efficiency under certain load conditions, if you don't load each output they might even have high ripple. Probably a very cheap one with very low power one < 300W might work. On the dangerous prototypes web site they have a board that you can connect directly to the power supply connector:
http://www.seeedstudio.com/depot/ATX-breakout-board-bench-power-supply-p-1222.html
As you can see there is a power resistor used to load the output of the power supply.
I would go for a linear power supply but if your budget is limited and already have the parts..........

[Mechatrommer]

For example, I have built quite a few Hakko clones using HP / IBM laptop power supplies. Couldn't be happier.

it doesnt sound like High accuracy measurements

[chickenHeadKnob]

@grimmjaw:

There is a way you can utilize cheap switchers in a volt-nut lab, use them in a battery charging capacity. I would design the power supply system backwards from the outputs. First figure out the range of voltage/current you need. Design your linear voltage/current regulator(s), they can be as simple as lm317 based for low noise. Then given the dropout voltages of the regulators design a battery bank to give some drop out headroom. Finally design a battery charging circuit appropriate for the chemistry of your batteries and pick a switcher to power that. You turn off the charger when making measurements (obviously).

[grimmjaw]

@chickenHeadKnob

Interesting idea.But your method needs batery banks, which makes the design a bit complicated compare than designing a linear power supply.

[jmaja]

Any kind of switching supply is going to suck for precision anything, unless you have so much filtering you might as well build a linear one. Just make one, it isn't hard.

I have used succesfully switching supplies with 24 bit Sigma Delta ADC's, but I have always had a linear regulator after the switching one (in one case integrated to the switching mode one). Noise has been at 18-23 bits (or 0.5-20 uV), which I would call precision something. I don't have any special filtering, just the lineasr regulator (+ a ferrite bed in one case). These haven't been from mains, but from 12/24V down to 6 V and then with a linear regulator to 5/3.3V or from USB 5 V first down to 3.3 V then up to 6 V and finally back to more accurate 5 V with linear regulator.

[grimmjaw]

Any kind of switching supply is going to suck for precision anything, unless you have so much filtering you might as well build a linear one. Just make one, it isn't hard.

I have used succesfully switching supplies with 24 bit Sigma Delta ADC's, but I have always had a linear regulator after the switching one (in one case integrated to the switching mode one). Noise has been at 18-23 bits (or 0.5-20 uV), which I would call precision something. I don't have any special filtering, just the lineasr regulator (+ a ferrite bed in one case). These haven't been from mains, but from 12/24V down to 6 V and then with a linear regulator to 5/3.3V or from USB 5 V first down to 3.3 V then up to 6 V and finally back to more accurate 5 V with linear regulator.

Just to make sure I understand correctly jmaja..The supply you used is not ATX power supply?

[jmaja]

Just to make sure I understand correctly jmaja..The supply you used is not ATX power supply?

No, it's not. I have no experience using ATX power supply outside computers. I have used e.g. old mobile phone chargers (12 and 6 V) to supply linear regulator.