How to test power supplies

[iloveelectronics]

At about 2:20 it is mentioned that you would need a scope capable of 500Mhz to accurately measure noise. Does that mean it's pretty much impossible for a hobbyist (with typical hobbyist equipment) to measure PSU noise?

[c4757p]

Access to expensive equipment breeds dependence on expensive equipment. There was a time when almost nobody had 500 MHz oscilloscopes (we'll forget the time when nobody had them), and trust me, there were power supplies back then. Low noise ones, even. All depends on how accurately you need/"need" to measure the noise.

[c4757p]

Also, there were stable feedback loops before network analyzers.

[Harvs]

I can only speak for myself, but when looking at power supply noise, I'm typically just looking at switching supplies (or combination of switching and linear post regulator.)  In that case, with the majority of switching supplies operating in the 100k-1MHz region, the majority of significant noise of interest is easily viewed on a 100MHz scope.

Obviously if you've got specific requirements well above that frequency range then you've probably also got access to high performance equipment.

[vk6zgo]

I think he has been within the cloistered halls of Texas Instruments way too long.
Out in the real,big,scary,world,you consider yourself lucky if you have a reasonable DMM & a 100MHz Oscilloscope!

[fpliuzzi]

Check out the other short youtube video in this power supply series by Texas Instruments...

"Engineer It - How to test power supplies - Measuring Noise"

In this video he mentions that when measuring the supply's ripple noise (across the output capacitor) he sets the bandwidth limiting feature of his oscilloscope to 20 MHz. Later, when he goes to measure the transient response of the power supply he switches the scope back to its full 500MHz bandwidth (frequency of the supply's  transient noise was only 41 MHz in his particular case).

Nice little series of videos.

[c4757p]

It's one hell of a power supply if you need a 500 MHz scope to measure its transient response. The guy's got serious ivory tower syndrome.

[robrenz]

Download This older HP dc power supply handbook and start at page 32 for proper ripple and spike measurement techniques.  The whole thing is excellent and covers all aspects of power supply designs and testing.

This is a later version of the same document with some variations.

[c4757p]

That HP/Agilent app note is very good. Hadn't noticed the revised version before, and the revisions look decent to me. (Note that even the revised power supply handbook from Agilent in 2000 (obviously they made higher bandwidth scopes and had an interest to push them) suggests a 20 MHz bandwidth for this measurement.

[robrenz]

This post shows a differential noise measurement of a very low noise power supply using the methods shown in the HP handbook.  This also shows that with a little luck and patience it is possible to get high performance gear for very little money.

[Christopher]

I work in a company which designs military switch mode power supplies, costing upwards of 5k each...

The production test dept which hand tests each power supply before going out the door use analogue scopes (The newish fancy ones with digital coursers and measurements), 100Meg. Obviously 6.5 digit HP meters etc are used. Engineering use recent Tek scopes approx 100Meg 2GS/s...

Normally the test specs require a 20Meg bandwidth limit, and with a bandwidth limit, surely (for a dig scope) the accuracy of measurement would be with the sample rate and not the bandwidth...?