Since I asked the question first I have gone through several tests. Working together with a motor manufacturer some samples were created including an additional linear hall sensor to the usual three digital sensors used for commutation. I cannot recommend that setup, as there is always to be one part of each turn where the linear sensor signal will be non-monotonic which will create a difficult to solve problem.
So the next stop is what TD-Linux suggests, to use one magnetic rotary sensor that can be had cheaper than an optical encoder with the added benefit of providing an absolute angle measurement (instead of the relative one provided by an incremental optical encoder).
I reckon I read that these magnetic encoders (by AMS) have already been used successfully by openservo project.
You can see a short video of what I am getting now using a brushless motor with an optical encoder [youtu.be]
So the next stop is what TD-Linux suggests, to use one magnetic rotary sensor that can be had cheaper than an optical encoder with the added benefit of providing an absolute angle measurement (instead of the relative one provided by an incremental optical encoder).
I reckon I read that these magnetic encoders (by AMS) have already been used successfully by openservo project.
You can see a short video of what I am getting now using a brushless motor with an optical encoder [youtu.be]
