Hello all,

I wanted to get your thoughts on some techniques that might be used to measure noise and ripple on the output of a high voltage DC to DC converter.

Additional details:

1. DC to DC converter has variable output from 0v-4kV and the output is specified to 0.25mA max. This would be connected to a metal plate to hold it at that voltage potential. Switching Frequency range is 75kHz to 150kHz and ripple is specified to <0.1%
2. DC to DC converter GND is common with earth ground, as is the oscilloscope.
3. O-Scope (Tektronix TBS2204B) inputs rated for <= 350V

I searched around the internet and found that some people use an external AC coupler (looks to me like a high-pass filter) with a high-voltage capacitor in series with the HV line to block the DC from the output. Could I use a Cera-Mite 4700pF 6kV cap, and a 470 Ohm resistor to give me a cutoff of about 72kHz? This would block the DC voltage from the scope input while still allowing me to measure the ripple and noise, right? (granted that the noise frequencies would have to fall within what the filter allows). Would this method alter measurements and if yes, is there a way mathematically to compensate for that loss?

Am I on the right path or is there an easier way to achieve this? I have a few high voltage resistor-divider probes but it's hard to get a meter to pick up ripple in the mV/sub-mV range when voltages are divided down by 100 or 1000. What would you suggest?

 

Could I use a Cera-Mite 4700pF 6kV cap, and a 470 Ohm resistor to give me a cutoff of about 72kHz?

Yes, that would work, but why are you rolling off the response at that frequency?

 

Yes, that would work, but why are you rolling off the response at that frequency?

Well...No real reason other than that I wanted to be sure to capture the frequency range of the DC/DC converter. I'm very open to suggestions lol. How would you go about it?

 

You could significantly increase the value of the 470Ω resistor to increase the cut-off frequency, but remember that with a x1 scope probe the input impedance (to the scope) will be 1MΩ. A further measurement error will be due to the scope probe capacitance, but as this should be of the order of 10s of pFs it should be insignificant when using a 4700pF blocking capacitor.

 

The primary purpose of the resistor is to assure that the voltage applied to the scope probe does not exceed the probes voltage rating.

 

At these levels one is always wise to protect their instrument and themselves with the acquisition of a 1000x HV probe.

 

I searched around the internet and found that some people use an external AC coupler (looks to me like a high-pass filter) with a high-voltage capacitor in series with the HV line to block the DC from the output. Could I use a Cera-Mite 4700pF 6kV cap, and a 470 Ohm resistor to give me a cutoff of about 72kHz? This would block the DC voltage from the scope input while still allowing me to measure the ripple and noise, right? (granted that the noise frequencies would have to fall within what the filter allows). Would this method alter measurements and if yes, is there a way mathematically to compensate for that loss?

Am I on the right path or is there an easier way to achieve this?

Yes. That would work, but I would suggest some back-to-back zeners across the 470Ω resistor to prevent high voltage spikes getting to your scope input when you first connect the capacitor to the circuit.
You might want to lower the cutoff frequency with a larger capacitor, because there might be ripple at lower frequencies that the operating frequency. Switched-mode circuits are prone to sub-harmonic oscillation if the feedback is wrong, so there might be something interesting at 36kHz that you wouldn't want to miss.
The lower you go, the bigger the capacitor and the bigger the spark when you connect it!
Or you could use a smaller capacitor and a bigger resistor. The scope input itself is 1M

 

You could significantly increase the value of the 470Ω resistor to increase the cut-off frequency, but remember that with a x1 scope probe the input impedance (to the scope) will be 1MΩ. A further measurement error will be due to the scope probe capacitance, but as this should be of the order of 10s of pFs it should be insignificant when using a 4700pF blocking capacitor.

The primary purpose of the resistor is to assure that the voltage applied to the scope probe does not exceed the probes voltage rating.

I really appreciate the clarification on these, thanks!

At these levels one is always wise to protect their instrument and themselves with the acquisition of a 1000x HV probe.

Are you saying this in a general sense or are you saying I should use the 1/1000 divider probe along with the HV External AC coupler? I would think the addition of the probe might complicate things a bit depending on how it's designed?

Yes. That would work, but I would suggest some back-to-back Zeners across the 470Ω resistor to prevent high voltage spikes getting to your scope input when you first connect the capacitor to the circuit.
You might want to lower the cutoff frequency with a larger capacitor, because there might be ripple at lower frequencies that the operating frequency. Switched-mode circuits are prone to sub-harmonic oscillation if the feedback is wrong, so there might be something interesting at 36kHz that you wouldn't want to miss.
The lower you go, the bigger the capacitor and the bigger the spark when you connect it!
Or you could use a smaller capacitor and a bigger resistor. The scope input itself is 1M

Thank you for your suggestions. I will plug some different component values into the calculations and see what I get. As far as feedback is concerned, the DC/DC converter internals (feedback loop included) are all contained inside a metal cup filled with epoxy. Should I devise some method to discharge the cap by shorting the input side of the capacitor to GND before connecting/disconnecting? The input will be connected via an Amphenol SHV connector. I attached an image to verify what you meant by a "back to back zener" arrangement. I believe that should clip peaks off the AC waves as the voltage climbs past a certain amplitude in either direction though I am not sure which zener values would be ideal for this. I don't see ripple voltages going past a few volts so maybe some 12V or 15V zeners?

 

tautech said:
At these levels one is always wise to protect their instrument and themselves with the acquisition of a 1000x HV probe.

Are you saying this in a general sense or are you saying I should use the 1/1000 divider probe along with the HV External AC coupler? I would think the addition of the probe might complicate things a bit depending on how it's designed?

A 1000x probe is a divider like any other.
With a DSO you just specify the probe attenuation and set the appropriate input attenuation and AC coupling or use an appropriately rated capacitor.
An example:
https://www.pintek.com.tw/productDetail/land-ctop-2/index/pscsn/17071/psn/127265

 

4kVdc with <0.1% ripple is a peak-peak voltage of <4V.

Using a 1000x probe this would be <4mV; even using the blocking capacitor/resistor with a 10x probe this would be reduced to <400mV (I’d stick with a x1 probe).

Don’t forget to load the dc output at its rated current (0.25mA) with a 16MΩ resistor.

 

I replaced the 4700pF capacitor with a 0.01uF capacitor and as an added safety against input pulses, the voltage was incremented in steps of a few hundred volts at a time. With no load and half the voltage applied (2kV) just under 100mV was observed in the frequency ranges expected. From here I plan to remove the 5.1V Zeners and replace them with faster responding SA5.0A TVS diodes and call it a day.

Thanks everyone for all your help. Much appreciated!