Practical comparison of 5 differential high-voltage probes

[hanakp]

I did some SMPS development lately and I needed a reliable differential high-voltage probe for that. Reliable in the sense it would represent the input signal without additional overshoot, ringing or offset. As I found out the hard way, that's not a given. In the end, I scoured my workplace for all HV probes I could borrow and made a practical comparison. Some of them are not produced anymore, but people visiting this forums usually buy used equipment anyway, so I decided to share the results here. Keep in mind that all the probes were at least 10 years old and long out of calibration, but I don't think new ones would fare much differently. I tested these 5 types:

Tektronix P5205
Pintek DP50
Metrix MX9030
Metrix MTX1032-C
Agilent N2772A

I used Agilent 81150A generator as signal source, because it has differential output. I set the output voltage as high as I could (I switched the output amplifier to "high voltage" mode). In practice it was around 40 Vpp, becasue the probes have high input imedance. I always tried to use as short leads as possible, I twisted them together etc. The results are in the images below, test conditions are in file names.

-Tektronix P5205 worked the best, so I used it as "the standard" for all other probes (dark blue trace). Although all probes have different attenuation, I always set scope channels so the traces would overlay.

-Pintek DP50 doesn't have an output cable, so I tried BNC and SMA cables of different lengths. As expected, that caused various ringing, because it has 1 Mohm impedance instead of 50 ohm. But I think the overshoot would stay even if used no cable at all.

-Both Metrix probes are absolutely useless as they produce terrible overshoots, no matter how you connect them. I think they use the same sensing circuit internally.

-Agilent N2772A is fine as long as you don't measure signals with fast edges. But even if you don't, it always introduces  persistent ringing at 30 MHz and significant delay (18 ns). I think it's because the input leads are so long.

Well, maybe someone else will find this info useful (or even add more results).

[PartialDischarge]

Now add a common mode and get ready to panic

[tautech]

-Pintek DP50 doesn't have an output cable, so I tried BNC and SMA cables of different lengths. As expected, that caused various ringing, because it has 1 Mohm impedance instead of 50 ohm. But I think the overshoot would stay even if used no cable at all.

It comes with one:
http://www.pintek.com.tw/customer/pintek/upload/DP-50-DP-100-back.pdf

[alm]

Common mode would indeed be interesting. Particularly, although this probably requires specialized equipment, with large voltage swings (that's why you use a HV differential probe, right?), so the voltage coefficient of the input attenuators.

[David Hess]

Now add a common mode and get ready to panic

I agree.  Connect both inputs to the same fast source and measure the common mode response.

High voltage probes are a place where a high voltage charge line avalanche pulse generator is useful for testing.

[hanakp]

Oops, I didn't think of common mode because the step response was the most pressing problem. I did these tests a month ago and I already returned most of the probes. Well, I have no easy (and repeatable) way to add significant common mode voltage to the input signal, anyway.

[scoper007]

You can take a look at the Micsig's differential probe :