Backdriving against large torque causes jerky torque reduction

[dzid26]

With high torque (high current) and backdriving against it, there are two effects contributing to the jerky movement.

1. In the case of the present current control, PWM to DC filters cause DC offset and more crucially phase delay on the current request. Partially reduced by this.
2. High motor inductance causes phase delay on the current request

When backdriving under smaller currents, both effects are not as visible as the current can quickly change direction. For larger currents, it takes time for it to change - the phase delay.

When phase delay is large enough more than 90deg, the driving current stop adding to torque generation, and causes rapid torque decline.

The phase delay can be compensated using speed calculation by additionally taking into account motor inductance.

This is mostly an issue in voltage-control mode.

[dzid26]

Currents data during the snap event:
Snap on overload.xlsx

Current changes fast during the snap, because the shaft accelerates rapidly (jerks, snaps).

But why does it snap? Either current is not commanded right or it has something to do with large torque ripple under load. (#11) To be investigated...

[dzid26]

I think I mixed two issues/scenarios here. :(

1. One could be related to decay modes and distortion - which at higher speed opposing the driving torque, the torque value drops. But it is normal scenario of driving the motor when bemf makes it harder to maintain the high current. Here we are braking, so bemf helps to have high braking current.

2. At much higher requested torque (say 3Amp) somewhat similar effect can be visible at 0 speed! If the load torque increases, eventually the motor is "giving up" and the torque drops. It's a snap. But this only happens in voltage-control mode - not used by default.

However I think underlying causes are different. Although eventually they may get combined once the speed increases.

For the 2. scenario, (which is more disturbing), the decay mode doesn't play a role. - initially the speed is close to zero (before it snaps and looses torque), and so bemf is close to zero and slow decay will do just fine. It is confirmed by the circle plot above.

Realistically, I think the seconds scenario is caused by nonlinearity of the torque and some processing delay. That is the holding torque is probably only high where teeth are aligned and relatively small in-between. And once it goes pass the tooth, it looses torque and motor speed goes up due to inertia and springiness of the system. Once it starts rotating, perhaps the torque never comes back due to combination of the 1. scenario.