Way back in January, in message
<ubenfwmpp9.fsf@helium.physik.uni-kl.de> Aschwin Gopalan
<gopalan@PHYSIK.UNI-KL.DE> wrote:

> Channel A
>           ++-------++      ++-------++      ++-------++
>           ||       ||      ||       ||      ||       ||
>           ||       ||      ||       ||      ||       ||
>           ||       ||      ||       ||      ||       ||
> ----------++       ++------++       ++------++       ++-------
>
>
> Channel B
>      ++-------++      ++-------++       ++-------++
>      ||       ||      ||       ||       ||       ||
>      ||       ||      ||       ||       ||       ||
>      ||       ||      ||       ||       ||       ||
> -----++       ++------++       ++-------++       ++-------
>
> This is for clock-wise rotation for example.
>
> As you can see, if you trigger on the rising edge of A, B is always
> high. For CCW rotation, at the rising edge of A, B is always low. So
> you can use one of the channels as a trigger and the other as
> direction output. All what is left is the debouncing.

Looking at edges in this way is asking for trouble, hence you have to
write things like debouncing routines etc. Really you should always look
at examining the levels.

A nice (fast!) compromise is documented in one of the Parallax data
sheets, which is:

Is the Current reading different from the previous?
If so:
  Take the previous reading
  shift it to the right
  xor it with the current reading.
  Check bit 0 (or 2, 4, 6 etc...)
  If it is 0, then decrease the counter, else increase it.

This works wonderfully. Consider the first two transitions in Aschwin's
example (assume that count starts at 5, and the switch bounces a lot!)

     A: LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLHLLLLHHHHLLHHHHHHHHH

     B: LLLLLLLLLLHLLLLHHHLLHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Change: NNNNNNNNNNYYNNNYNNYNYNNNNNNNNNNNNNNNNNNNNNNNYYNNNYNNNYNYNNNNNNNN

Count:  5555555555655556665566666666666666666666666676666777766777777777

There's absolutely no way that the bounce can induce slippage, which it
would with an edge-detecting algorithm.

Using such a mechanism I've been keeping a serial-controlled resistor
up to date and reading encoders at up to 12,000 pulses/sec on a 4MHz
'84! (that's assuming one channel out of the 4 it's monitoring is
changing, and guarantees that any pulse is seen)

Fun!




Stuart.
--
Stuart Tyrrell Developments        Stuart@stdevel.demon.co.uk
PO Box 183, OLDHAM. OL2 8FB        stdevel@locust.co.uk (Mobile <> Email)
Orange: 0976 255 256 (9am-9pm)     http://www.stdevel.demon.co.uk
*Analogue joystick i/face 39.95*   *Using an IR keyboard via PS2Mouse 24.95*