Eriks comment: done in SW, this does not cause problems, even if you miss part of this oscillation stuff, all it does is inc or dec the result by 1.

In my case it resulted in disaster. I had to move it out of the uC and into a CPLD just to correctly get the position. The uC simply could not respond fast enough to these oscillations. It was not easy to see (most stops were without oscillation across the border, some with just a bit and - once an hour or so - the really bad ones) your eyes get strange when you look at a scope for hours, but I am certain that I saw higher than 500 kHz. To see the problem, you need an oscillation at the one case of 16 where "a is clock and B is direction" fails and EVERY oscillation will increase or decrease the count by one.

BTW this was documented in a previous discussion on this subject abt a year ago.

My documentation is the property and reside at a former client and it would take hours to recreate it - time I do not have. Draw out all possible transitions from forward to reverese and you will see where the problem is.

Have fun,

Erik