Troubleshooting a ripple voltage.

[npelov]

Hi,

I was about to ask for help, but I figured it out. I thought it might be helpful to share.

I was going crazy while trying to reduce the ripple voltage of an ebay DC-DC buck converter. I got 50-60 mV no matter how many inductors and capacitors I put. I changed all the caps with good low esr ones. I added 2 stage LC filter - still the same. I used 1cm spring on the scope probe and I used only resistive load. Then I remembered the Lights in Dave's lab. So I started switching off stuff. It was the fume extractor that I made from two 120mm PC fans. I switched it of and I got 5-10 mV ripple (20MHz bandwidth).

I started suspecting what's going on when I saw the same waveform when the buck converter was turned off. The buck converter was powered from a battery and the fans - from a low quality chinese AC-DC SMPS - the "industrial" type. So I guess the inductors picked up the noise from the fans that were 10-15 cm away.

I also have conductive rubber mat (similar to the one dave uses) that was connected to mains neutral (it's 2 wire installation). Does it matter it's not connected to earth? I guess neutral carries current, so it might not be ideal for earth connection. What else can I do to reduce noise with such measurements? Maybe turning off everything is the way to go.

[tatus1969]

could be RFI pickup from your scope probe / ground lead. Try using low impedance grounding to reduce that and get closer to the true voltage.

[npelov]

I already said I'm using a spring for low impedance ground path - I highlighted it.

[tatus1969]

sorry, didn't read carefully enough.

Maybe you can post some screenshots showing the ripple. Which frequencies does it contain? Is it also present at the buck converter's input? What happens if you remove the ESD mat? PCB layout will also play an important role here. Is the buck converter's power inductor shielded? It may also be coupled into the converter's feedback path.

I recommend not to add LC filters after the converter output, as this will create a pole at its resonance frequency, which will amplify every load modulation at this specific frequency, instead of damping it. If you want to add further filtering, I would add a ferrite instead of an inductor. You only need to care about frequencies that are beyond the converter's bandwidth. Everything below that is handled by the control loop. Depending on the buck's design this is in the order from kHz to tens of kHz.

What I would do next is take the clip of my probe and remove the ground lead, turn up vertical gain on the scope, and use it as E field sniffer. Not as good as a spectrum analyzer but in most cases I can find possible sources of radiation. You can also probe H fields to some degree by connecting the ground clip to the tip, but that is not very sensitive.

Next would be adding copper foils soldered to circuit ground.