Andreas Pockberger wrote:

> The trick is that there are two sensors for each wheel so you will get
> 2 impulsstreams out of one wheel the phases are 90 degrees shifted

                                 |  Reading point
                                 V
             ___     ___     ___     ___     ___     ___     ___     _
Stream A|___|   |___|   |___|   |___|   |___|   |___|   |___|   |___|
           ___     ___     ___     ___     ___     ___     ___     ___
Stream B__|   |___|   |___|   |___|   |___|   |___|   |___|   |___|

               Anti-clockwise <--|--> Clockwise

> to make an encoding you have to look which edge comes first and then
> you read the second egde depending on the direction one of the egdes
> comes before the other one

  The precise logic expression eludes me at present, but a truth table
relating current to previous state goes something like:

 (A;B)            Current:
Previous ||(0;0) (0;1) (1;0) (1;1)
         |+----------------------
 (0;0)   |   NC    CW   ACW   Err        NC=No change
 (0;1)   |  ACW    NC   Err    CW        CW=Clockwise
 (1;0)   |   CW   Err    NC   ACW       ACW=Anti-clockwise
 (1;1)   |  Err   ACW    CW    NC       Err=Error (ignored)

  The error condition where both appear to change at once is a vexed
problem.

  If you presume that it will occur due to slow processing, then you
must allow that you have missed the actual direction of that move.  In
this case it is best to update the "previous" state with the current
one, and wait for the next move.  This would most likely happen when the
encoder is moving rapidly and the next move will at least be registered
as the correct direction.  For a mouse, this is perfectly reasonable as
absolute accuracy is not required in comparison to smooth movement.

  If OTOH, you presume that a simultaneous change in state on both is a
noise (debounce) phenomenon, than it is handled by *not* updating the
previous state; when a reasonable "new" state is eventually detected,
the movement direction is determined accordingly.  This means however
that if the encoder really *did* move two steps between polling, the
next step will be indicated in the *wrong* direction.  Oh well.
--
  Cheers,
        Paul B.