I am developing a relatively small motor (12V, 3A) which should run sensorless at very low speeds (0 - 70 RPM) with a certain and almost constant torque (around 6 or 7 Nm, max) and as smoothly as possible (speed error below 5 or 10%). I know this is very challenging, and we have not achieved appropriate results yet.
Right now we are running tests and the first prototypes with an encoder, but that should definitely not be present in the final product. We have spent a lot of development time in an HFI implementation in an IPMSM motor (with saliency approx 0.7). I suspect it sometimes estimates the angle with a considerable error or offset, which causes the motor to run in the wrong direction, both from the start and when it is already running and a certain (not even too big) change in the load is applied. I understand from your information here http://www.ti.com/ww/en/mcu/instaspin/instaspin-zero-and-slow-speed.shtml?DCMP=ep-mcu-c2x-gen&HQS=ipd that you have a similar solution to offer. And some of the drawbacks you address might be similar to the ones I am finding.
One of the problems I found in the solution I am trying to implement now is that the lib is prepared to run mostly big motors and at much higher speeds (their speed control has a resolution of 6 RPM, their HFI Amplitude value has 1-Volt steps, etc). I managed to find some workarounds for these issues, but it would be ideal to find a platform which does not need this. Do you have any papers or have you run tests that might fit in my application? I would eventually analyze moving the whole project to your MCUs if a proper and promising solution was available. Even the motor design could be changed accordingly if there was a strong reason for it, not implying disadvantages in other aspects.
Thank you in advance.
