>2. MCHIP has some good divide/multiply routines in the app notes.  Check
>AN544.
>
>A GOTCHA:  You may find that 8 bit unsigned integers leave a lot to be
>desired when doing this calculation.  I use the formula
>
>C = (f-32)*9/5
>
>If you divide first, then multiply you'll truncate your result,  giving an
>error.  But if you multiply first then divide, you'll exceed the
>capabilities of 8 bit math.  I always use 16 bit numbers when doing
>temperature, and may need to use 16 bit signed integers when doing
>temperatures.  Math routines for signed 16 bit multiplication in Assembler
>are out there, but they are hairy.  This quickly leads one to the conclusion
>that C is more appropriate than Assembler for doing any kind of math.

I try to avoid signed arithmetics as much as possible - I offset all
variables so that the most negative value corresponds to zero and then
grows up. It is also advantageous for comparison to various limits and so.
I remove the offset just before the numbers are displayed, transmitted or
output other way.

Say you need -40F..+100F, e.g. -40C..+38C. The offset variables Foff and
Coff will stay in range 0..140 and 0..78 respectively. Now the formulas are:

Foff = (Coff * 9) / 5

Coff = (Foff * 5) / 9

Of course the result in parenthesis is 16 bits, you need 8x8 unsigned
multiply and 16/7 (or 16/8) unsigned divide routines. Just after all the
arithmetics, subtract 40 from the result, if negative display minus sign
and make the value positive.

Josef

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements