ATX recommended ripple measurement methods

[Agent24]

So, I just built myself a little differential probe kitset to do "proper' ripple measurements with, and decided to test it on an Antec Basiq BP350 that I have just recapped.

I started with the 5vSB rail, and was getting some 200mV p-p spikes.

That's not good, obviously. But then I remembered the ATX guidelines about testing ripple... and that it requires capacitors (100nF and 10uF) across the probe.. so I added them.

That reduced the spikes to about 20mV which is obviously within the spec of 50mV.

Does this waveform look normal for a DM311 based 5vSB rail? Sorry for the photo quality, the camera refused to focus on the trace.



Ignoring the recommended probing setup, this means the 5vSB is genuinely generating 200mV p-p spikes when running. Is this normal? It is a flyback design. Do they just not care and assume that the motherboard filtering will take care of it? (which I suppose it must do, if the test set does.)

Or have I done something stupid? I did not use the isolation transformer on the PSU as I am using the differential probe, and the ATX spec diagram shows the PSU to be grounded, which it is.

There is no indication in the diagram as to whether the oscilloscope should be isolated or not. It recommends a Tek TDS460 which is a standard mains-powered bench-top CRO, so I assume not, and I did not isolate mine.

[budm]

Re: ATX recommended ripple measurement methods

Ripple spec with Load, no load?
https://www.youtube.com/watch?v=Edel3eduRj4

[Agent24]

Re: ATX recommended ripple measurement methods

Ripple spec with Load, no load?
https://www.youtube.com/watch?v=Edel3eduRj4

If you mean ATX spec, the ripple & noise is defined as 50mVp-p maximum over 20MHz for 5vSB, 5v and 3.3v rails, which I assume is under any load condition from zero to full load. The ATX test procedure just says to test the rail at both no load and full load. The first quick test I did was at 300mA with a resistor.

At no load, I get a very different ripple waveform, like a sawtooth, 10mVp-p, at 333Hz. This DM311 has a burst mode operation at no load so I suppose that's what it's doing.

I hooked up an adjustable load and it goes out of burst mode at around 130mA.

The overall ripple & noise does not really go over 20mVp-p through the whole range from 0-2.5A. The waveform just changes shape. The noise spikes/bursts actually reduce as the load goes up, but the ripple increases and the waveform is always at around 20mVp-p overall.

That is of course with the 100nF and 10uF capacitor in circuit as shown in the ATX test spec diagram.


He does not show the concept of adding a filtering capacitor(s) when probing. So I assume that those 200mVp-p noise bursts I measured are real. However, when measuring with the ATX recommended capacitors attached, it is only 20mVp-p.
Seems a bit bizarre maybe but that's what the specification says to do.

[Agent24]

Re: ATX recommended ripple measurement methods

I found a similar PSU, FSP300-60THA(1). The waveform from the 5vSB is exactly the same, though the ripple & noise is a little lower amplitude (maybe 15mVp-p).
But also very high noise spikes (150mVp-p) like the Antec, without the ATX test spec capacitors.

Comparing that, with an AcBel ATX-400P-DNSS, which had a completely different waveform. Barely any noise spikes at all on 5vSB.

So the noise is likely not from an external source, and perhaps it is normal for that FSP based design with the DM311, and they just rely on motherboard filtering capacitors to remove the spikes and also to pass ATX test.

That's all I can think of anyway.