BLDC motor - FOC Iq sign reversal

Question

I operate BLDC motor with FOC controller, and attempt to characterize Iq as a function of load/torque.

From FOC theory, I would expect a linear relationship between the torque and the Iq current component.

What I observe is a sign reversal, where Iq goes towards zero as the load (friction) increases, and then changes sign.

In this test, the torque never changes direction, as the speed is held mostly constant by the speed PID loop, and the inertia is negligible.

So what could possibly cause an effect like this ?

(the control is well-tuned, and follows desired speed for a wide range of RPM setpoints and loads)

Edit: There is a load value where the controller reports zero current which is nonphysical (the motor spins, overcomes friction and consumes energy). So I don't see how it can be anything else but an artifact/bug of the FOC algorithm.

Answer 1

To answer my own question, the root cause of the effect was a phase current measurement error (as mkeith suggested), due to insufficient filtering of ADC input.