PMP9256: Output Ripple is too high

Hello Isabelle,

Sorry about the link, I will have to see if there is any reason you cannot access it. From the presentation on measuring output ripple:

So on this one, we've got two output voltage measurements. On this left one, there's a lot of noise, and it's very dirty. It's very ugly-- this output voltage measurement. And on the right, it's much cleaner, and the output voltage ripple is very visible. So which a voltage measurement is right?

So what happened and what's causing such a noisy measurement on the first one is measuring the switching node-- there's a couple of things actually. First, we're measuring the switching node at the same time as the output voltage ripple, and what this causes is coupling through the oscilloscope back from channel 2, which is the switching node, back to channel 1.

The other thing with this measurement is no bandwidth limiting is used. This causes you to pick a very high frequency noise, which may really not be there. Typically, when you're measuring the output voltage ripple, you're more concerned about the conducted noise and not the radiated noise. The noise at these high frequencies typically will not propagate down the PCB due to the parasitics of the PCB.

So the best way to take a measurement is to use bandwidth limiting and remove the switch node probe. This gives a very clean switch node output voltage ripple measurement, and it looks more like what you would expect. So there are some cases where you may want to do a high bandwidth measurement, but in those cases, you just be careful to know that you're measuring what you really want to measure.

So just a few more tips when you're measuring the output voltage ripple-- so first, it's good to measure right across the output capacitor. And when you measure it, you should use a tip and barrel approach as shown in this middle image-- tip and barrel so that the probe is right across the upper capacitor, and there's a wire wrapped around the ground of the probe that's connected to the ground side of the probe.

Another thing you can do is use probe sockets, and then you can just plug it in and get a really good measurement. One thing you definitely don't want to do is have a flying ground lead like the image on the right. This just creates a loop that can pick up a ton of noise and cause a lot of inaccuracy in the measurements.

Another couple of things with a setup-- as we said before, it's good to limit the bandwidth to 20 megahertz just to make sure you're picking up the noise that you really want to measure. You want to use the full y-axis scale, so use the smallest voltage scale possible just to include-- get the best resolution out of the oscilloscope.

Lastly, if you're measuring a very low output voltage ripple, consider using a one-to-one probe or an active probe. This helps to limit the noise in the measurement, again, to get a more accurate measurement

Regards,

Bob Sheehan

Hi Bob,

The tip and barrel technique improved the ripple seen to about 600 mV, which isn't too bad compared to TI's reference test report.

Attached is a screenshot of the new waveforms: primary and secondary switching, and output ripple.

The ringing seen on the secondary switching waveform seem to have gone away after upgrading a transistor. The ringing on the primary side remains (CH4). We are still investigating why that is.

Also the by-pass caps are layed out exactly like in the reference design, and shorting them did not make any difference.

Thank you for your help!