Ok, let's assuime we have the ideal machine first. Some problems
still left with the slow speed. As to noise, it will have place,
more or less. However, most of precision system could not be built
without special measures such as signal filtering, detection and
capture. If there is nothing clear to measure, we need to try to
separate a signal of noise.
Also, it may be appeared other problems - sampling rate, interrupted
rotor movement, etc. on the slow speed.

Sum, could you describe your system a little more details please.
What's the slowest linear speed on Rmax. If it's going to be rotated
something big like telescope, its speed measurement would be easyer
than a small steer motor of orbital space station or capstan motor
in SONY's HandyCam.
I have found one more AP of Intel, related to your problem. Probably
you are familiar it already, it's AP-428 "Distributed Motor Control
Using the 80C196KB" that discribes small speed machine control as
well as others. Houever, we need look at fundamental science of
automatic control first, a lot of such resources I have left and
forgotten in Dnepr. I am sure, most of the question are under our
descussion, have been discovered already at the science. Just find
it, and take time to learn this nice technic field.
Any control system depends very much on its target and required
control parameters such as amplitude-phase-frequency response,
reaction/stabilization time, controlled parameters allowed error.
Since PWM and uCU is going to be used, this system is nonlinear,
and such system are a little more complicated to calculate it in
front of linear, but we need basic calculation of the system, as
far as we ned to know CPU instruction set to build an asm code.

This question is my nastalgy. I will be glad to see if it'll be
continued more.

Thanks.
Best regards,
Oleg