In the agilent chip, they have described a filter that bars very fast transients from disturbing the count. I want to implement that idea

A very dangerous idea. If you have 47 up transients and 48 down transients in a few microseconds (as I have seen - numbers approximate) how do you know that you are going to come out -1 with such a filter.

Through trying many approaches I have concluded that the only safe way to read an ab encoder is with a very fast (C)PLD.

Erik

PS if you are relying on the a is clock, b is direction scheme you WILL fail in one of 16 (or is it 8) reversal cases.

Every time we discuss encoders, I as if someone please will work out all the reversal cases and publish the graph. Not being involved with encoders at this time myself, I can not afford the time; however if someone will, I'll gladly proofread it.

Erik