For something like that, I am quite sure you could open hand of some
cash and install the microcontroller very close to the sensor, probably
in the same enclosure, then, you transmit the information to the outside
world...

90,000 counts per revolution of what? I hope you don't mean 90,000
counts in a single revolution of the encoder shaft, are you?  If the
answer is yes, without using gear reduction box (the slack would be
terrible), with a minimum slot size+separation = 0.01" you are talking
about a rotor disk with aprox 286 inches in diameter... (90,000 x 0.01 =
900 inches / PI = 286"), 7 meters...  If the answer still yes, man, your
bigger problem is not only noisy environments.. but mechanical
instability... a 0.001 degree is something very, very small, probably it
would need one of those suspended 50 tons concrete block as a stable
base to operate such device... :)  Just to have an idea, imagine an
object the size of a car, aprox 4 meters in length, rotating it, the
bumper would move 34 mm per degree, and only 0.034mm (34 um) per 0.001
of degree... It means to have aprox 30 steps per millimeter... very
small, I believe just a breeze can bend a heavy concrete distribution
power post this much on the top... now imagine it reduced to a toy car
with only 4 inches in length.

"Montaigne, Mike" wrote:
>
> What do you recommend if you get an illegal condition?
> On a mouse, you probably could just throw away the illegal condition,
> but on an absolute position (like we are doing,) my first guess is
> try and determine if the following condition is legal.  If it is, you had
> a noise spike, if it isn't you probably had an legitimate count and the
> program can increment/decrement accordingly.
>
> We are (also) looking at using PIC to process our encoder counts -
> especially since we can use it to keep track of our count also.
> Our biggest problem is to try and get a encoder to give us 90,000
> counts per revolution (we want to use/display 360.000 degrees)
> The only encoder we have found is by Cannon, uses a laser diode
> and costs $3-5K