testing power supply ripple?

[Pentium]

C_hegge, I am curious what your method is for testing power supply ripple? I just got a rigol digital oscilloscope, and I have never used one before. I understand they have to be tuned. All I really know is that you want to test peak to peak voltages in mV. I would really love some insight because your method seems to be extremely accurate. Thank you!

[c_hegge]

There's not actually much tuning involved. What is important is to use the 1X probe, not the 10X. Most probes have a switch on them allowing you to set whether it's 1X or 10X. You also need two small capacitors in parallel with the probe - a 10uF electrolytic and a 1uF Ceramic. I use Nichicon PW for the lytic and whatever I can find for the Ceramic. I think Wolf uses a Panny FC (same specs as PW) for the lytic. Note too, that, when testing the -12V rail, you need to reverse the capacitors (so you don't make the lytic explode), but not the probe. That grounding wire must be connected to ground at all times, even when testing a negative voltage. The way I have it set up on the tester is to have a rotary switch to select the rail to monitor, and a set of caps for each rail, like illustrated here. I couldn't come up with a way to use the one set of caps for all rails without reverse polarising them for the -12V rail.

[Pentium]

Thank you so much for the info and diagram, I thought it would have to be on 10x. I assume a 10uF 50V cap will do? Do you only need to use one probe? Do you think you could take a picture of yours ready to go just so I can make sure I'm not doing it wrong? I really appreicaite it!

[c_hegge]

Yes, a 10uF 50V cap would work (although 16V would probably be preferable - it's supposed to be said capacitance at the testing voltage. I doubt it would make much difference, though), and yes, you only have to use one probe. I'll get a picture of the back of the load tester's switch later today, once I get around to taking the cover off. The wires and joints are all covered by tape, though, so I'm not too sure how useful it will be.

One small test you can do to make sure your wiring is good and not picking up interference is to first test the rails directly at the output by placing the capacitors legs in the connector and testing them with the scope. The numbers from the switch should be the same as those taken from directly at the output.

[Pentium]

I feel really stupid asking all these questions, but I'm a little confused, thank you for your patience. So is the rotary switch a separate circuit that you made yourself and you have to use eight 10uF + 0.1uF caps in series, or does that just make it easier? Do you just twist together the leads of the caps? I'm having a hard time picturing it. Is measuring the ripple something you can't do in a running computer because of interference? Like for instance what would happen if you just hooked the ground alligator clip to the ground on the scope and stuck the probe into say the 12V wire of a molex plug, would it just give you an inaccurate reading or would it damage the scope? You pros make it look so easy!

[c_hegge]

I feel really stupid asking all these questions, but I'm a little confused, thank you for your patience.

No Problem

So is the rotary switch a separate circuit that you made yourself and you have to use eight 10uF + 0.1uF caps in series, or does that just make it easier? Do you just twist together the leads of the caps?

The two caps go in parallel with each other, not series. The rotary switch is indeed a separate circuit that I built myself - I used the 8 sets of caps because I couldn't find a way to use the one set without reverse-polarising them when testing the -12V rail. The legs are both twisted together and soldered.

Is measuring the ripple something you can't do in a running computer because of interference? Like for instance what would happen if you just hooked the ground alligator clip to the ground on the scope and stuck the probe into say the 12V wire of a molex plug, would it just give you an inaccurate reading or would it damage the scope?

You can measure the ripple with the PSU on a running PC. There's probably a few more sources of EMI, but, I doubt if it would seriously throw the results out. Maybe you could give it an extra 20% headroom (ie, if its out of spec by 20%, don't worry about it). It won't throw it out by hundreds of mV, though.

I have added a picture of the setup with the switch. It was hard to get a decent shot, though. The switch is in a bit of an awkward position just in front of the reinforcement strut and there are wires everywhere. I've taken off some of the tape just for the shot, though, so you'd want more insulation on there than what you see there..

[Pentium]

Thank you very much for the picture, you have quite the setup. Waiting for those 10uF caps to arrive. I guess my last few questions, is that directly implemented into your load tester? Because I don't have a load tester yet so would it be more difficult to test the ripple without one?

[c_hegge]

In my case, yes, the switch with the caps is installed inside the load tester. However, you don't need to have one in order to do it. I would just modify a 24 pin ATX extension cable and splice some extra wires in. If you were doing so, you would only have one 12V position (My tester has four - one connected to the ATX connector, one to the 12V CPU connector, and one on each of the two PCI-E Connectors), but I rarely need to use the last three. Even most multi-rail units share the same filtering, and have very similar waveforms and ripple amounts.

EDIT: I probably should mention one other thing to be careful of when building this circuit. Don't hold the soldering iron to the switches terminals for very long, as the switch will melt, and the terminal will come out. I've destroyed a few switches from that before.

[Pentium]

Parts came in So when testing the ripple, what should the alligator ground clip be connected to when testing, to the case of the PSU or to the chassis ground of the scope? I was just going to for instance stick my soldered 10uF lytic + 0.1uF ceramic into the ground and 12V wire of a molex cable and test with the probe on the leads of the lytic

[c_hegge]

Connect it to the ground - leg (or + leg in the case of the -12v rail) of the capacitor.