The low frequency (20k and 700k) limit violations indicate to my opinion, that motor wiring and shielding/grounding is inappropriate. Depending on the application (cable length, motor mounting) this may acutally be difficult to achieve.
The 700k component should be reducable with LC filters at the controller, 190 MHz(?) with small RF chokes. Most likely the 190 MHz isn't generated by the switching transistors istself but some (too?) fast freewheeling diodes in a bad designed PCB.
The freewheeling diodes I am using are those internal to the FETs - FETs are IRF 7341Q (
http://www.irf.com/product-info/datasheets/data/irf7341q.pdf).
The traces between the power supply (12 v car battery) to the FET drain to the FET source to the current shunt to the GND are as short and wide as they can be (they are restricted by the connector position).
So the freewheeling current wouldn't need to make a big loop. The PCB is a disk with 2 inches diameter.
So I hope the PCB is not that bad - could you give more details about what you mean by bad designed PCB - maybe it didn't cross my mind. Otherwise I think our PCB designer did the best he could.
Oh, and that frequency was not 200 MHz, but rather 150 MHz - where we are over the limit.
Interesting is that the duty cycle of the PWM command for the FETs has some importance - for example at 80% DC the results are not so bad as at 30% DC.
Could that be because the rippple current would be lower in case of the higher DC?
Let me know if you have some time for me - I could pm you with more details.
I greatly appreciate your help,
nike