Measuring small differential voltages with high common voltage applied

[Momchilo]

Hello,
I try to find a way to measure "small" differential voltages across two transistors/capacitor with high common voltage applied.
I know there are some differential op-amps like AD629 or INA149 which tolerate quite high voltages up to \$ \pm 275V \$ but it's just not enough.
In the picture you see one half bridge of a three phase multi-level converter (MMC). In this example it's a six level converter.
Basically I have to measure the voltages Vm1-Vm5 in the attached picture. The voltage between the "ports" 1 and 2 is 50V. So it would be possible to measure the differential voltage with the named op-amps.
But the goal is to increase the numbers of transistors and capacitors in series to get more cells built by two transistors and a capacitor(-bank) instead of the five cells in the picture. In the end the differential voltage is around \$ \pm 10V \$ and the common voltage will be around \$ \pm 500V \$ with many cells in series (50). Or less cells with higher differential voltages, but the common voltage will be 500V. The switching frequency will be around 10kHz.
If you have any idea, tips, publications or books with this topic, I would be very grateful.

Best Regards
Momchilo

[Momchilo]

I don't need new probes blueskull. Sorry if I wasn't clear about this. The CMRR of a differential probe is good enough to verify the measurements at those low frequencies.
I can't use probes, I need all the differential voltages as inputs for the control. For every of the six half bridges there are 5 different differential voltages, at all 30 voltages have to be measured and processed.

[MasterT]

Isolation Amplifier:
https://www.digikey.ca/product-detail/en/silicon-labs/SI8920BC-IP/336-3329-5-ND/5358956
or this:
https://www.digikey.ca/product-detail/en/broadcom-limited/HCPL-7520-000E/516-1684-5-ND/825320

[floobydust]

I doubt you can look at mV due to circuit stray inductance on the IGBT sections, it's going to be pretty noisy.

Look at Pintek. I've used DP-15K High Voltage Differential Probe.
It's overkill at 15kVpp common-mode, 35MHz, 100mV but their product line is low cost ~$400 and they have other models. This is if you want to buy off the shelf.

If you want to build them, I drew a DP-15K schematic and it's straight forward - a instrumentation amp from three op-amps.

Pintek Differential Probe Comparison Chart

[Momchilo]

Thanks MasterT and blueskull for the help I will look at your suggestions.
Yes I'm using a C2000 (TMS320F28335) for the control, at least for the most part. For all the signals I need a bus system, because there will be 30 voltages, 30 PWM signals, which will be increased in the future and some other measurements like temperatures, phase currents and so on. So every cell gets its own µC which is connected via a bus to the F28355. Do you think CAN bus will be the best option? At the moment the signals are generated by one FPGA. But with a bus system the scalability would be much better.
I just started with the project and have to do a lot more research.

I still don't need any probes floobydust.

[duak]

There is an old technique that is quite sensitive and has excellent common mode rejection.  It uses varactor (voltage variable capacitance) diodes in a simple bridge circuit connected to the circuit nodes to be measured.  A transformer couples a drive signal to the varactor bridge and a synchronous demodulator measures the imbalance caused by the change in varactor capacitance which is a direct function of the voltage being measured.  Here are links to info on the technique: http://protorit.blogspot.com/2011/11/varactor-bridge-op-amps.html
https://www.electronicdesign.com/analog/what-s-all-varactor-input-amplifier-stuff-anyway
To make it a true isolating amplifiter, another transformer would be required to bring the AC signal back from the bridge to the AC amplifier.

There are more modern techniques and products available, but if they are ruled out for whatever reasons, this technique should be considered.  With modern components a design could be quite simple, although it will require a good knowledge of discrete circuit design.

Cheers,
 

[David Hess]

Self powered isolation amplifiers could work if their bandwidth or settling time is suitable but they are pretty expensive at $30 to $50 each.  The more modern approach is to use an ADC or ADC and microcontroller for each channel with a galvanically isolated serial link.  You did not give any details about how the drive signals are applied to the power transistors but if power for the ADC or ADC and microcontroller is not available from the drive circuits, then provisions like a floating supply will need to be made.