High frequency Current sensor

Allen,

Thanks for using E2E!

I can recommend a few design paths here:

For the conditions you describe, the challenge here is the 300V-400V working voltage, as our current shunt based solutions are typically unable to handle a common mode voltage of this magnitude. One easy option you could examine is to design the sensor on the low side of the heater coil. If you implement the sensor between the coil and ground, this would eliminate the common mode voltage, as the common mode voltage produced on the shunt would be close to that of GND. There are tradeoffs to a low side design, though, which you can learn about here. 

Second, to get your design to this level of voltage, additional components can be used, such as we show in this cookbook circuit based around the INA138 for 400V operation. However, this design only works for our current output parts, so only either INA138 or INA168 can be used here.

Finally, I would also recommend checking out the TMCS1101, which is a Hall-effect based part that can withstand voltages of up 600V, and constant currents up to 15A. This part may be ideal for this scenario, but comes with additional layout considerations due to the Hall-effect nature of its operation. 

Also in all of these design paths, the bandwidths of these devices are not meant for high speed operation (INA138 BW = 800kHz and TMCS1101 = 80kHz), so regardless of which design path you go, if you are planning to sample the output at a 500ksps frequency, you will most likely need to buffer the input into your ADC with a high speed buffer for optimal settling. 

Please let me know if you have any questions. 

Allen,

Yes, but specifically for Vsense = 100mV.

At Vsense less than this, the error may be greater due to increased error contribution from the offset. For total error, you may use the root sum square approximations discussed in our video series if you wish to theoretically calculate total error, but this is the typical total error you would see across temperature for the INA138.

The 2% limitation at room temperature is actually a tested limit, and any devices exhibiting >2% error at room temperature are removed from the yield. 

Allen,

Unfortunately, I don't think are very many options for this level of working voltage with a BW this high to get it done in a single IC. You could run a high speed amplifier on the low side of the load, effectively eliminating the common mode problem, as the VCM would be roughly zero, but that comes with its own tradeoffs, namely losing the ability to detect faults to GND. 

Another option you could look at is the AMC1300 series to take the design isolated, but I believe these devices also top out around 250kHz output BW or so. To drive a 500ksps rate, I would think you would need an amplifier with a BW around 2MHz or so.