Electronic – How to Filter Very High Voltage Power Supply Noise

filterpower supplyRF

Any suggestions of how to filter out this noise? Basic filter math is not working out.

I am adapting a cheap commercial HV supply (variable DC voltage up to 40kV at 10mA) for physics experiments. The supply works adequately, but it has output noise I can’t seem to filter away.

The noise comes as regular high frequency pings at a significant percentage of total voltage. For example (at 20kV into a 4 MOhm load), there is a ping each 18us with a maximum amplitude both up and down 6kv from the base 20kV. The frequency of the ping is about 36 MHz and it mostly fades away after 500ns.
An RC filter (10k Ohm / 4nF) cuts this noise in half, but the results are far short of theory, and this still leaves a big spike. The capacitors are high voltage polyester & foil caps. LTSpice modeling with ideal components suggests that I should have a clean output, so I assume that real capacitors, resistors & wires are not ideal at this high frequency and high voltage.

Adding ferrite rings between the R and C cut it about 10% more.

Another thought I have not yet tried is to loop the HV wire a few times around a ferrite torroid and then add a separate single short-circuit loop to act as a short circuited transformer secondary. My thinking is that this would pass the DC but waste the AC part into heat.

Best Answer

It may be that there is noise being coupled in from whatever is driving the system, I'm going to assume this is either a HV flyback or a voltage multiplier (and from the some od it, the multiplier seems mpre likely). Now when the switcher in the base turns on, it dumps all it's energy onto the stack in short sharp bursts (you get the same problems when rectifying AC feeding a big smoo5hing capacitor). Now at 10k ohms, your filter will allow A LOT of current to flow into your smoothing capacitor. Try either adding an extra cap directly to the output before the filter and/or increase the value of the resistor substantially (try 100k) and if that's still not enough, try (as Eric suggested) adding some large inductors to the output (start with something in the millihenry range) and see what happens.