from A-B-C without missing B, but even if I would miss B at higher speeds then I can assume, that the direction has not changed since the last reading. The trouble is only if I miss 2 states then.
It is not totally clear to me what you mean by A-B-C, we are talking about an encoder with A and B outputs only ?.

If you are talking about A-B-C as various states of a and b, then you are right except where you say "not cumulative".

On a 1000 pulses/rotation system I measured up to 7 step error in ONE rotation when making frequent starts and stops. Most of these errors were due to the original design mindlessly believeing that b = direction and a = clock would work and. I had errors after that was fixed, do not remember how gross, those were removed by moving the logic to a PLD without change (except, of course, the faster response).

Erik

Erik