Re my: > The most accurate approximate arithmetic is done by using > convergent iteration. > An iteration to divide by 3 is based on the identity > a/3=a/2-(1/2)(a/3) Matt Calder asks: > I hate to beat a dead horse, but isnt this solution the same as >the one that was offered up by an Andy?, ie. > a/3 = a/2 - a/4 + a/8 ... Basically yes. The difference between the series summation and the iteration is that the error is not accumulated but instead reduced. You can do rather well with 8 bits in iteration while you need more bits when doing the series as has been show in other postings. For comparison, you get accurate result via series summation in 16 bits using 24 words of program (see Ray Gardiner on Mar. 2., the program needs BCF STATUS,C before the first RLF) with as many cycles and 8 bytes of data space. Iteratively you can get the same in 13 words of program with max of 56 cycles and 3 bytes of data space (or less, that was just my first quick and dirty iteration code). -- Lauri --- For more info.