Mark Jenkins wrote:

> I was hoping to use a hall effect sensor because of the the ease of
> setting up this type of system with 4 magnets set 90 degs apart on
> each wheel, i imagine this would make the software harder to write and
> may require additional hardware so may not be practical or possible,

  No, that sounds fine, particularly as I presume the digital hall
sensors have hysteresis and therefore debounce the signal nicely.

  If the PIC has to debounce the sensor, that would complicate things as
you may have to develop a PLL-style system to determine the actual speed
and ignore "bounce" suggesting a higher speed.  This could otherwise be
a problem if the "bounce" at a lower speed was comparable to the actual
signal at a higher speed.

  I must confess, no solution has came to my mind last night.

  The general approach I would be looking at however, is to count
transitions on each wheel sensor (if the sensors are such as to have
equal "mark" and "space" sectors, transitions in both senses could be
counted to double the resolution) in a common "sum" counter, and also
apply the transitions to a "difference" counter which counts up for the
left wheel and down for the right.

  Let's say you want to detect a 10% speed difference.  Every time the
"sum" counter reaches or exceeds 40 (each wheel has generated 20 counts,
or perhaps one has generated 19 and the other 21), you subtract 40 from
the sum counter and look at the difference counter.

  If the difference counter reads 6 or more, you reset it to 4 and
indicate a speed difference.  If it reads 3 or 4, you subtract 1, and if
it reads 5, subtract two.  If the difference counter is negative, the
same algorithm applies to the negative numbers.

  The effect of this algorithm is that small variations in speed will
cause the difference to "slop" around in the range -2 .. +2, variations
consistently in one direction will cause the count to oscillate between
-3 and -2 or +2 and +3, a difference greater that 5% will cause it to
run up and oscillate between 4 and 5 (or negative thereof) while at
exactly 10% difference, it will consistently sit at +/-5.

  Any more, and it will sit at +/-6 or greater and indicate accordingly.
--
  Cheers,
        Paul B.