High Voltage Differential Probe Design Help

[atomic antics]

Hi everyone,

I'm working on a high voltage differential probe to help troubleshoot the primary of my tesla coil.  This is my first PCB design project that requires some serious thought about design and layout to maximize bandwidth, so I'd like some feedback. 

I have been using this page as a bit of a guide, but have also simplified it down some to have fewer parts and also worked on the CMRR.
https://www.eevblog.com/forum/projects/oshw-diy-1kv-100mhz-differential-probe-(dilemma-vs-hope)/

Requirements:
+-2kV input transient, +-1kv min
flat response out to 20MHz
6-10Meg input impedance


Short summary of design:
I am using a high impedance voltage divider to divide the input signal down to tolerable levels.  I can select between 1/1000 and 1/100.  The total impedance is around 7Meg.  I am aware that this will introduce nyquist noise, but I do not really see any better options.  I
This is fed into an instrumentation amplifier.  The input buffers are made from jfet op amps (opa659) and the subtracting amplifier is the LM7171.
The output is BNC through coax to my oscilloscope.

The op-amp power supply uses 6.5V switching converters, some filters and then 5V linear regulators to hopefully bring the ripple down around 1mV.

I attached a PDF document with screenshots of the schematic and layout.  I would appreciate any feedback on that. 

Currently, one of my largest concerns is reflections.  I tried to keep trace lengths as short as I could and maintain a continuous ground plane beneath signal traces, but I do not know much about reflections other than sudden changes in impedance can cause them.  Is this something I need to worry about?  I am specifically concerned about the output of the probe (low impedance diff amp output) to my scope input at 1Meg 20pF.

Thanks in advance for any help!

[David Hess]

The high impedance attenuators are going to require frequency compensation.  High voltage ones typically use a distributed capacitor on the input resistor and an adjustable capacitor on the shunt.

Instead of making the input attenuator x1000/x100, make it x500/x50 so that a 50 ohm series termination on the output can drive a 50 ohm oscilloscope input for an additional x2 factor making x1000/x100 total.  That gets rid of any issues with driving the oscilloscope.

Or just make the cable between the instrumentation amplifier output and the oscilloscope input short.

The instrumentation amplifier should have adjustments for DC and AC common mode rejection.

Instead of building your own probes, I would consider using a pair of existing x100 or x1000 oscilloscope probes with a home built differential amplifier to drive the oscilloscope.  You still need the buffers and instrumentation amplifier but it saves trying to figure out how to get the high input impedance attenuator to perform well.