Autotune syncofs?

[slow67]

Johannes,

Have you thought about creating an autotune feature for this? Have it set iq to 0 and have it step through offset until it wont move (with big jumps). Then maybe have it step through a finer range, and find the one in the middle?

[slow67]

Just thinking out loud here. If you command positive current to phase a, and negative current to phases b and c, the motor will align itself (if it can freely spin) to a known position. If inductance is used in the future, Ld can be measured also (once its aligned). This *should* be the same as 0 degrees (on a 0-360 degrees scale)? And therefore the resolver position should equal syncofs

[slow67]

More thinking out loud. Assuming motor is on the car, and has a working parking brake. If syncofs is now known, we can lock the parking brake, and command something like 10% rated current, and measure the d axis inductance (Ld). Unlock the parking brake, move the rotor 90 electrical degrees, lock the parking brake again, and run the inductance test again. This will be q axis inductance.
If we have the rotor flux (volts per radians per second), we also now have the ld-lq difference, and know what the MTPA setting should be without having to tune with trial and error. Also combined with known phase resistence, the FOC PI controllers kp and ki settings should derived from these.

[johu]

Dima just implemented a graphical tuning page for the esp8266
https://github.com/jsphuebner/esp8266-web-interface
Your idea is probably good too
Find the inverter software here to have a play: https://github.com/jsphuebner/stm32-sine

[dima]

I have a small "tweak" and idea! But I cannot visualize it.

Maybe someone can explain in a "round" perspective how angle "looks like" relative to all those magnets.

I know that it is calculated by:

Code: Select all

angle = (syncofs * 360 / 65536)

but how to visualize that? Is this diagram way off or am I on the right track?

[slow67]

I think maybe showing all the poles makes it too busy. Maybe only display it as a 2 pole motor and only 3 windings like this

[slow67]

More thinking out loud. For finding the correct syncofs, if in manual mode, we can ignore the resolver position, tell the FOC process that we are at position 0. Command id current, and the motor will align itself to position 0 (assuming it can mechanically). The resulting position should be syncofs.

[sfk]

More thinking out loud. Assuming motor is on the car, and has a working parking brake. If syncofs is now known, we can lock the parking brake, and command something like 10% rated current, and measure the d axis inductance (Ld). Unlock the parking brake, move the rotor 90 electrical degrees, lock the parking brake again, and run the inductance test again. This will be q axis inductance.

Do you mean the parking brake on the transmission? Typically these are mechanical devices that rely on a pawl locking into a fairly coarsely toothed gear. Which means it won't engage unless the teeth happen to be lined up at the time. What usually happens is the car (and the transmission) will shift a bit before the pawl locks.

If you were depending on this to positively lock the trans at a specific known location, it probably won't work as expected. Not like a brake rotor at least.