How to measure current of pulsed power, what's wrong with my current shunt?

Question

I'm trying to measure the peak current of a pulse generator, I've been using a current-transformer type current probe (TEKTRONIX A621 AC) which can only measure up to 2000amps. I've also tried a few shunt resistors (500A, and 1500A continuous rated. The measurements included below are measuring the 500A shunt. The measurements I'm getting on the shunt are not what I'm expecting (they don't match the signal of the tek current clamp). The signal looks like inverted voltage across the IGBT (light-blue trace), and above about 8kA (measured from the tek current clamp(which is only rated for 2kA, it shows a very strange spike which is confusing me. Am I missing any filtering or anything here for the shunt mesurement? Should it match the trace of the tek current clamp? Thank you all for any help/advice as always!

Here are some traces showing the various signals at ~9kA, 8kA, 6kA and 4kA. Ch1(Yellow) = IGBT gate signal Ch2(Light-Blue) = voltage across the IGBT Ch3(Magenta) = Current measured from Tektronix current clamp) Ch4(Dark-Blue) = Voltage across shunt resistor

Below I've included a few photos of the setup of the current shunt and the setup. I've switched to using shielded wire from the shunt to the differential probe leads.

Answer 1

The current probe has a low pass characteristic. According the datasheet, it's upper cut-off frequency is 50kHz. Hence any current components with frequencies greater than this will be under represented in the output. Note the negative transition in Ch3 just after the middle of the screen occurs over 10us.

Using BW = 0.35/tr = 35 kHz. So that edge alone is already pushing the limits of the probe.

The shunt effectively has the opposite behaviour as it not just a resistance but also has series inductance. It will over represent any frequency components when the reactance of the inductance is greater than the resistance. Taking a guess as some values R=1m L=10nH the frequency will be on the order of 15kHz. The excessively high current measurements from the shunt are all on fast edges.

This can be corrected by adding a 1st order low pass between the shunt and differential probe to correct this up-turn in sensitivity. The hard part is working out the values. The following needs to be met: Lshunt/Rshunt = R1*C1. In concept it could be done with a DSO and signal generator but given the current levels you're playing with the shunt will be 1mOhm or less. Any sensible signal generator will not be able to drive a voltage the DSO can measure into that impedance. If you access to a Network Analyser finding the frequency would not be too difficult as they will work happily with the very low signal levels.

The following simulation show the uncompensated and compensated outputs for the example values I suggested

The absolute values of R1 and C1 are not important, just the product of the two values. Given R1 and C1 are across a shunt, the voltage across these parts is minimal. The lead lengths to the series RC should be short - right on the shunt.