>The correct method of dividing by 255 is this situation is as follows:
>
>        To calculate:   x * y /255
>
>1. Multiply 8-bit x by 8-bit y to get 16-bit resultant.
>2. Add low 8-bits to 16-bit resultant.
>3. Round (i.e., increment top 8-bits if msb of low 8-bits is set).
>3. Take result as most significant 8-bits.
>
>You will note this is essentially the same as incrementing the top 8-bits
>if the msb of the low 8-bits is set, except for the effect of rounding.

... I believe you mean if the msb of the high 8 bits.
        This thread seems to be getting unnecessarily complicated, with
Michael's comment being missed in the shuffle:
        If you want to divide by 255, this is the same as multiplying by
1/255. Now 1/255 = 1/256 + 1/65280. Your max value is 65535, ergo Michaels
comment that simply dividing by 256 (using the msbyte) would be accurate to
1 bit. If you want better accuracy, you can round the result up by adding
one if the original 16 bit value was >32767 (msb is set).

        Jim Nagy
        Elm Electronics
  ICs for Experimenters
http://www.elmelectronics.com/