On 01/06/2011 08:00, William "Chops" Westfield wrote:
>> >  why waste my time with 'C' when assembler is just as good.
> Well, the current example wants
>      M =3D (f0 * 232)/fc
> Where f0 and fc are 32 bit numbers (some constant, some not.)  That's
> not particularly a line that I'd like to translate into assembler.  (I
> suppose it depends on how much access you have to a good set of
> libraries...)  (And then C will easily go low-level and let you
> separate the bytes efficiently (one way or another), supporting my
> first statement.)
>
Olin probably has such an assembler library and can write the entire DDS=20
app inc "GUI" with keypad, rotary encoder and graphic panel in a coffee=20
break. The rest of us would be better using C or JAL on 16F

232/fc  is presumably a constant.

f0 is the part that changes.

fc is 50,000,000
fo is  likely less than 150,000,000

232  =3D 4,294,967,296


You don't want to convert to floating or lose any resolution.

Pre-calculating a constant

4,294,967,296/50,000,000 =3D 85.89934592, approx 86 as integer, loses too m=
uch

so it does need 64 bit int maths if you do (f0 * 232)/fc  properly. This is=
 possible in C or JAL

However if fo max is 150MHz then  that is 1/28.63311530667 of 232

Or if fo =3D 30MHz max 1/ 143.16557653333333333333333333333

There isn't a way to scale to 32bit maths unless the frequency step is=20
integer multiple of scaling..


Minikits in Australia have a PIC kit with DDS chip premounted for this=20
kind of application.  It's not very complicated.
I've written a similar application in JAL that outputs in packed BCD via=20
serial to a radio rather than a DDS.

http://www.minikits.com.au/kits4.html

http://www.minikits.com.au/kits5.html#iqdds

I'll probably do a PIC based DDS myself for the softrock RX/TX I have.


--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.