Noise with Hall sensor at high speed

Have you tried full sensorless to be sure the noise don't come from vibrations at high speed?!

I have tried several motors and some very light outrunners are vibrating.

Bearings can also produce noise.

(edit)

The noise is coming from interpolation limit in my opinion. As long as I remember from practice it should be around 13k EERPM.

Interpolation limit is written in code and can't be changed from VESC Tool. It can be found in mcpwm_foc.c file. 

if (rpm_abs < 100) {
                // Don't interpolate on very low speed, just use the closest hall sensor
                ang_hall = ang_hall_now;
            } else {
                // Interpolate
                float diff = utils_angle_difference_rad(ang_hall, ang_hall_now);
                if (fabsf(diff) < ((2.0 * M_PI) / 12.0)) {
                    // Do interpolation
                    ang_hall += speed * dt;
                } else {
                    // We are too far away with the interpolation
                    ang_hall -= diff / 100.0;
                }
            } 

If you want to increase EERPM of hall sensors just change the line :

 if (fabsf(diff) < ((2.0 * M_PI) / 12.0))

to:

 if (fabsf(diff) < ((2.0 * M_PI) / 6.0))

And your max supported hall speed should be around 26k EERPM. Just bare in mind that hall detection should be done really precisely (use higher current when detecting hall).

@e-biker: If you want to use hall sensors at a higher RPM, there is a setting for that in VESC Tool. (It changes the m_conf->foc_sl_erpm setting used further up in the function you quoted, to set using_hall true or false.)

If I understand this code correctly, the

if (fabsf(diff) < ((2.0 * M_PI) / 12.0))

line compares the actual, read hall position to the interpolated one. If they are less than 1/12 of an electrical revolution apart, the interpolated estimate is adjusted according to the current speed. Otherwise, the estimate is considered faulty and backed off(?) a bit - I think.

There are three hall sensors, and they go through two values each per electrical revolution, for 6 hall states/rev. The estimated value should never pass outside the "zone" of the current hall state - if it does, then we know the estimate is wrong. The "zone" stretches "half a zone" in width on each side of the hall position, i.e. ±1/12 of a turn.

So changing this to 1/6 of a turn like you are suggesting should allow the estimate to fall far outside of the correct zone. Not sure why you'd want to do that?

@arvidb: You are absolutely right, but I had no choice.

I have made my own high power VESC based on HW 4.12 and found that in FOC it can't run properly on high currents (400A). So I changed interpolation limit that I can run it only on hall even that it is not accurate as senseless observer.

And it runs pretty smoothly, with my hubzila i can hit 16k EEPRM @ 87V with no problem (with no load you can hear some desync, but riding in real life I haven't noticed anything - maybe some efficiency loss). Increasing switching frequency to 30kHz due to better timing helped a bit. 

Thanks for explanation.

Martin.

@yosoufe: Did you try different sensorless erpm settings in hybrid mode? Especially very low ones? With such a low K_v your motor should generate lots of voltage at low rpm which I guess should make the observer able to track OK even at low (e)rpms. Maybe you can find a setting where the hall sensors and observer "agrees" about the position of the rotor, for minimal jerk at the switch-over?