HV probe not reading right

[Qmavam]

I purchased a used Heathkit HV probe, it is a high value resistor that uses the meter as part of a voltage divider. The manual says it's a 990MΩ and a X100 with an 11 MΩ meter, but the resistor is labeled 1090MΩ. This is what works out to a X100 with an 11MΩ meter.
 The probe seemed fine the first few times I used it and I compared it to another probe that had a meter on it and it was fine, then it wasn't. This was on DC, but now I'm testing on AC. (I'll set up a bridge to test DC later)
  I have a 9kV 60hz source and two meters, a Fluke 79 III and Brymen BM235. The Fluke reads 238V and the Brymen reads 269V. A X100 Probe should read 90 Volts on the probe.
 The highest voltage I have measured was 12kV and it's made for much higher. I was very careful when I took it apart, didn't touch it with my fingers, I have since tried wiping down the 1090MΩ resistor with Acetone and alcohol, it's glass, so don't expect any harm. The Fluke rose to 259v and the Bryman to 299V. This would lead me to thing the 1,090MΩ resistor is acting like a 300MΩ resistor. I also measured the input impedance of the Brymen with the Fluke at 10.00MΩ, but I get no reading when trying to measure the Fluke with the Brymen.
 Any ideas what is going on with this HV probe?

[bdunham7]

Any ideas what is going on with this HV probe?

Is the probe specified for AC use?  What is likely going on is that the capacitances don't match.  HV dividers have to match the capacitance ratios as well as the resistance ratios and that really wouldn't be possible if your HV probe depends entirely on the meter for one section of the divider.  The common Fluke and BK/CalTest 40kV probes that are rated for DC and AC (typically 50/60Hz only) use a 1000X divider ratio with a ~1.1M section in parallel with the DMM so that the capacitance of the DMM has less effect.  They're still only accurate to 5-10% on AC.

[Qmavam]

Any ideas what is going on with this HV probe?

Is the probe specified for AC use?

I think that's it. I just finished putting full wave bridge on my 9kV AC source (a neon sign transformer) and now both meters read 97 Volts when measuring the rectified (non filtered ) AC. I don't know my 9kV source is really 9kV, in fact the other meter I have says 10.1kV.

What is likely going on is that the capacitances don't match.  HV dividers have to match the capacitance ratios as well as the resistance ratios and that really wouldn't be possible if your HV probe depends entirely on the meter for one section of the divider.  The common Fluke and BK/CalTest 40kV probes that are rated for DC and AC (typically 50/60Hz only) use a 1000X divider ratio with a ~1.1M section in parallel with the DMM so that the capacitance of the DMM has less effect.  They're still only accurate to 5-10% on AC.

Hmm, well my next question, I'll be trying to measure a variable frequency HV AC signal,
Say 4Hz to 50 Hz and maybe 2kV to 8kV. Is there any way to modify this to get a resonable 5% to 10% accuracy?

I'm starting to believe I was fooling myself, I built a probe using ten-10MΩ resistors and connected a meter or scope to it thinking it was 10 to 1 divider. Hmm, maybe it was OK, because I did measure DC and then turned on the switching and still got the same peak voltage. Simplified circuit below. When I turn of the grid drive the (actually I have to turn off the tube heaters) I get my full DC voltage when tuned back on, I have the full pkpk voltage. I suspect the lower impedance has some effect on that.
 

[bdunham7]

Hmm, well my next question, I'll be trying to measure a variable frequency HV AC signal,
Say 4Hz to 50 Hz and maybe 2kV to 8kV. Is there any way to modify this to get a resonable 5% to 10% accuracy?

It's going to be quite difficult to do that with a ~1G input impedance probe, although at those low frequencies and a 10% accuracy limit might be possible.  The real issue is that you'd need some way to calibrate it to be sure it performs acceptably.  The more accurate options that aren't as frequency dependent are typically lower input impedance.  Would a 75M load be too much?  Keep in mind that your actual impedance with AC ends up well under 1G due to capacitance anyway. 

[TimFox]

A typical way to measure HV AC is to use a capacitive divider instead of a resistive one.
The lower capacitor (to ground) needs to be large enough to get a reasonable high-pass response with the input resistance of the meter, and the actual input capacitance of the meter and cable is in parallel with it, so it should be reasonably high capacitance.
Therefore, if the meter input resistance is 10 M\$\Omega\$ and the lowest frequency is 4 Hz, the lower capacitor must be >> 4 nF.
If you use 20 nF and want 100:1 division ratio, then the upper capacitor is about 200 pF.
If that be too high, then you need a higher division ratio (smaller upper capacitor).

[Qmavam]

A typical way to measure HV AC is to use a capacitive divider instead of a resistive one.

Therefore, if the meter input resistance is 10 MΩ and the lowest frequency is 4 Hz, the lower capacitor must be >> 4 nF.

OK, I see 4nf at 4Hz is about 10MΩ and it needs to me much, much higher.

If you use 20 nF and want 100:1 division ratio, then the upper capacitor is about 200 pF.
If that be too high, then you need a higher division ratio (smaller upper capacitor).

 I can come up with a 20nF low voltage cap, but all I have in HV caps are 3kV 5pf and
3kV 40nf.  I'd like to build it good for 20kV for later use, so I guess I'm shopping for HV caps, again.
 

[Qmavam]

Oh, I found a 391pf 20kV doorknob cap, that with  a 40nF is a 103 to 1 divider.
 Close enough.
 

[bdunham7]

Oh, I found a 391pf 20kV doorknob cap, that with  a 40nF is a 103 to 1 divider.
 Close enough.

Can your source deal with that load?  You'll have about 6.7M @ 60Hz.  Something like a P6015 probe will give you 100M resistance and ~3pF capacitance on the input.

[Qmavam]

Oh, I found a 391pf 20kV doorknob cap, that with  a 40nF is a 103 to 1 divider.
 Close enough.

Can your source deal with that load?  You'll have about 6.7M @ 60Hz.  Something like a P6015 probe will give you 100M resistance and ~3pF capacitance on the input.

Uhm yes, for my initial testing the neon sign transformer has a 30ma output, when I put this in practice, I'll need to be watchful of the frequency 4Hz is fine with 100MΩ, looks like I will need to develop a chart to account for the loading of the probe capacitance as I change frequency.

[1audio]

With that cap you will be running the sign transformer in current limited mode with reduced output + its all very dangerious. You can make a cap divider with a piece of glass and some foil. 5 pF to 5000pF is X1000. Your scope/meter will be less than 5000pF. so all you need is to add cap at the input. You can buy scope probes that are literally spring clips to attach to HV wires for HV AC. Now getting into HV AC at frequencies above 20 KHz a lot of other aspects will come to make life worse.

Here is a tutorial on calculating the cap. https://study.com/skill/learn/how-to-calculate-the-capacitance-of-a-parallel-plate-capacitor-of-given-its-geometry-explanation.html#:~:text=For%20parallel%20plate%20capacitors%2C%20the,plates%20of%20the%20parallel%20capacitor.  Permmativity of Glass is around 8.

[kleiner Rainer]

If you need a high voltage capacitor for a capacitive divider, you can use a HV rectifier tube for colour TVs - they are cheap and nothing insulates as well as a high vacuum enclosed in glass. I can get NOS GY501 (european equivalent of 3BH2) for 60ct, doorknob caps are much more expensive. I learned this trick from an oldtimer during my internship in a company that built gas lasers in the kW range for welding and annealing - measuring real HV was daily business. Due to the small capacitance, this method also is useful to measure high power RF.

The 3BH2 datasheet specifies a capacitance of 1.5pF and a PIV rating of 35kV - that should be sufficient for most hobbyist applications. BTW since we only use the capacitance, we dont need to power the heater...

https://frank.pocnet.net/sheets/035/3/3BH2.pdf

You have to calibrate the divider, preferably by adjusting the value of the lower divider cap, due to the tolerance of the plate-to-cathode capacitance and stray capacitances of the divider, but you have to do that too when using resistors for a divider. Careful parts placement is required to keep stray capacitance low - avoid sharp edges (corona discharge!).

Greetings,

Rainer

[TimFox]

The handy thing about such tubes for this purpose is the plate cap for the top end of the divider.
Be sure to connect the (cold) heater to the cathode to avoid floating bits of metal inside the tube.