Encoders are not linear devices!!!!!!!
Peter
give me one or two machine examples where you would need this level of precision please.
Dan Bentler
Any encoder that is going to be used as a master or reference should have the highest resolution possible. In my example above I showed that even with over 500,000 counts per revolution that the velocity resolution is only 0.1 inches per second. How does one gear to that?
Positioning systems should use the derivate gain which is a gain that is multiplied by the error between the target velocity and actual velocity. If you can't compute a decent actual velocity then the control output will be very 'noisy' looking. This really isn't noise. It is the effects of feedback quantizing or simply put, ENCODERS ARE NOT LINEAR DEVICES ARE THEY?
Ideally resolution would approach infinitely small. Then the non-linear encoders will appear to be linear. The goal of a good digital design it to approach that of an ideal analog design.
See the difference between a 4000 counts per revolution and 524288 counts per revolution
ftp://www.deltamotion.com/peter/PDF/Mathcad - t1p1 pid mrplc.pdf
The 'noise' does not excite any un-modeled poles because my software simulator doesn't have any but a real system has higher frequency poles that can get excited by the 'noise'. The drive will not like the 'noisy' control signal either.
This 'noise' affects how high you can crank up the gains, especially the derivative gain.
Most derivative 'noise' is actually feedback non-linearity or sample jitter
So what do you want on your motor?
Speed control systems like moving conveyors don't need to use a derivative gain. A PI controller will do. In this case you can get by with a lower PPR encoder but when gearing the master should always have a high PPR encoder. This way the slave can compute accurate feed forwards from the master's encoder.
PLC Kid, if you are gearing to a master the master should have a high PPR encoder but if the drives themselves are just doing speed control then buy the higher PPR where you don't need to spend extra money.
I would need more data to give a more exact answer.