Another point I'd like to make:

In my experience, resolving a position to an encoder can be simple and straight forward, if you apply a little common sense in the beginning.

1. When deciding the number of encoder quadrature pulses, decide how accurate your position needs to be FIRST. For instance, in a linear application, I want to resolve +/- 0.00025", so my window is 0.0005" - I need at least one quadrature pulse every 0.0005", or 2000 quadrature pulses per inch on my linear encoder.

2. Always allow for at least 2 motor step pulses per encoder pulse. That way you can resolve ambiguity in position by adjusting the device to match the encoder value. Let the encoder be the final say for any position.

3. Whether you are using a trapazoidal motion profile or a parabolic profile, keep track of where you are during the move, and begin ramping down early enough to avoid overshooting. I employ a box-car filter which works even when there is 2" of slop in my motor's belt drive (very loose belt).

4. Use an external flag mounted to the device your are moving, along with an optical interrupter light switch for a reference home. Never rely on the encoder. Reading the interrupter (from the same direction everytime) can yield repeatability measurements on the order of 1/10 mil, if done properly.

5. And last, don't forget about things like gear ratio, torque, speed, inertial moments, dah, dah, dah... They really do matter.

Good luck,

D Brooks