Want to build HV HF oscilloscope probe.

[Domo]

Hello all..

I would like to build a HV and HF probe, that can measure up to around 30kv (TV flyback output).
Since the output  (by calculation) of the flyback is around 3-4 mA, i didn't want the load of probe to draw much current, so I decided to use 20 100MOhm resistors (total of 2 G Ohm). This also was influenced by use of 1/2 watt resistors. Output would be 1000:1, so measuring 10kV would show up as 10V.  By calculations, probe would draw around 15 uA.


I have not yet build it, it is only in simulation, but i am already running into a problem. From the probe to oscilloscope, I would be using coax cable, and I am aware that it has average parasitic capacitance of 66.7pF per meter. When i add that to the simulation, sine wave does not rise more than 3Vpp. Changing values between output of probe and ground does not increase the sinewave amplitude, How can I compensate this?

[Twospoons]

Short answer is you need to add capacitance across your 2Gohm resistor to form a capacitive divider than matches your resistive divider, which in your example will be 66fF, or 1.3pF across each 100Mohm resistor . 

Also I suggest you run your sim as an AC analysis - then you will see the frequency response of your probe, and why the capacitive compensation is necessary.

Here's a couple of resources you should find useful :

Oscilloscope Probe Circuits (tektronix) https://w140.com/tekwiki/images/6/62/062-1146-00.pdf   (I would attach the file but it exceeds the file size limit.)
100kv_probe.pdf

[Domo]

Very good information on attachment.. Thank you.

[klugesmith]

Also be sure to mind the voltage rating of your resistors.   Look for ones that are called "high voltage resistors".

Most ordinary half-watt axial lead resistors have maximum rated voltage of 350 V.
https://www.seielect.com/catalog/sei-cf_cfm.pdf
A typical resistor product series has a wide range of ohm values, all with the same max power and max voltage.
For R values near the high end,  the voltage limit gets you before the power limit.
Large voltage overloads don't typically cause sparks or smoke, but they can cause resistance change (reversible or not) and long term degradation of the ohmic material.

[edit] your low side resistor is shown as 1Meg + 15.5k, for DC attenuation of about 1/2000 instead of 1/1000.
I guess that's what you want.   The simulation without capacitors shows RMS voltage of 10 kV in, 5 V out.
Lets see if your AC analysis, with 66.7 pF capacitor in parallel with low side R, gives you a 2.35 kHz lowpass filter.
At 10 kHz most of the current goes through the capacitor.

As twospoons said, to get flat frequency response you need a capacitive divider that matches the resistive divider.
Then at 10 kHz, the probe's input capacitance will draw a lot more current than its input resistance.

[Twospoons]

There is an argument to be made for putting a cable-driving buffer in your probe. That way you can reduce the capacitance of your probe and extend the frequency range without unduly loading whatever you are testing.  It also means you can correctly terminate your cable to avoid ringing.

What upper frequency are you aiming for?

[Domo]

Also be sure to mind the voltage rating of your resistors.   Look for ones that are called "high voltage resistors".

The resistors I want to get are from digikey, HVA05FA100M. Datasheet says max voltage is 3500V. They are 1/2 watt, so it should withstand around 30KV input.

What upper frequency are you aiming for?

This is a project to measure output of TV flyback transformers, so average around 30KV output at 15.7 kHz. In simulation I tested with 10kv RMS @ 10 kHz, and those were arbitrary numbers for easier calculations.

[Domo]

So it appears that AC probe would be not easy using coax cable... Can I just use banana-to-bnc cables? It would eliminate parasitic capacitance. Besides, tv flybacks have pulsed DC, and that is the only thing I would be using it for, to measure 30kV, with minimal current draw.

Now, a new dilemma appeared.. Oscilloscope input impedance is 15pF and 1M Ohm (both to ground), and the current coming out of the voltage divider is so low, that the 15pF is not charging fast enough, and simulation show an RC charge time instead of DC pulses... dang, back to drawing board..

I have tested few circuits in simulator, and I cannot find any work properly..

Edit:
If I am to only use  for flyback transformers, then I guess i will have to compromise any only use for one purpose. I won't be using coax, just banana-to-coax conector.

Signal input to probe is 30kV, 15.7 kHz.


If the output of transformer is 30 kV pulsed DC (0-30 kv), then the output will be 3Vpp (+1.5 to -1.5).



Before I build this i would welcome comments and suggestions.

[Twospoons]

No-one suggested an AC probe.  This was about the simulation you are running - do an AC analysis instead of a transient analysis. The result should be a plot of gain vs frequency. Ideally you want the gain curve to be flat over the frequency range of interest. In your case I'd suggest a bandwidth of about 100kHz, which would cover the 5th harmonic of your TV flyback 15kHz.

Can I just use banana-to-bnc cables? It would eliminate parasitic capacitance.

No. It wont eliminate parasitic C, only change it in an unpredictable way. It would also increase your parasitic inductance.  You'd do better to add a buffer amp, if driving the cable and scope input is  too hard.

What simulator are you using?

[klugesmith]

If you just want to know the peak voltage, why not get a HV rectifier and charge a small HV capacitor,
to a voltage that can be measured with ordinary kilovoltmeter probe?


If you want to see the waveshape on oscilloscope, can you fit a test winding of a few turns onto the core of your flyback?

[Domo]

No-one suggested an AC probe.  This was about the simulation you are running - do an AC analysis instead of a transient analysis. The result should be a plot of gain vs frequency. Ideally you want the gain curve to be flat over the frequency range of interest. In your case I'd suggest a bandwidth of about 100kHz, which would cover the 5th harmonic of your TV flyback 15kHz.

This is taken from output of the probe / input to oscilloscope.

What simulator are you using?

NI Miltisim 13

Sorry for late reply.. busy at home..

[Twospoons]

Can I suggest replotting using dB on the Y axis, and only plot the voltage. Putting V and I on the same graph creates a scaling issue that makes it hard to interpret the results.  We're interested in seeing the -3dB point.  You should also plot the X- axis  using a log scale, covering 10Hz to 1MHz.

[Domo]

Can I suggest replotting using dB on the Y axis, and only plot the voltage. Putting V and I on the same graph creates a scaling issue that makes it hard to interpret the results.  We're interested in seeing the -3dB point.  You should also plot the X- axis  using a log scale, covering 10Hz to 1MHz.

Frequency in LOG,
Amplitude in dB,
10 Hz - 1 MHz.

[Twospoons]

Much easier to interpret!
So you currently have a bandwidth that covers 400Hz to ~30kHz.  Which is why your square waves look messed up.
Now if you get rid of C1 and instead put a capacitor across each 5Meg resistor to match the division ratio you have with your resistors, you can greatly extend and flatten the response of your probe.  In practice there will be stray capacitance from your divider resistors to ground that you will have to account for.   Achieving 1MHz bandwidth on a real life probe shouldn't be to hard with a bit of care in construction.

Simple diagram of the end goal, where the ratio of the capacitors matches the ratio of the resistors :