Ok, I have a possible solution for you.  As long as 11 cycles per increment
is OK for you, and you don't need the extra precision, I can see a way of
doing this.

1: Forget division.  If you have time critical code that requires division,
a RISC processor is the wrong solution.

2: Remember that you can implement a 0-255 counter.

Here's what you can do:  take the value from the ADC, and throw it in some
working register for the sake of argument,  call it counter.  Then use a
nested loop structure where the internal loop is an 11 cycle delay (make it
a bit less so that the cycles in the outer loop make it 11 cycles (this
counter can be done in W)  then loop through that delay 0-255 times (dec
counter each time until you hit 0), giving you your 0-2805 cycle delay, then
you add an external 400 cycle delay.

I think that should do it.

I had a problem a lot like that on one of my midterms or finals once...

--Brendan

----- Original Message -----
From: "Jonathan Starr" <JonathanS@ELECTRONIX.CO.UK>
To: <PICLIST@mitvma.mit.edu>
Sent: Wednesday, June 12, 2002 8:41 AM
Subject: [PIC]: maths for parsing value to delay


> This damn project seems to be more trouble than it's worth, if anyone
mathematically minded can help me with this I would be very grateful.
>
> Okay, I have a ADC input that has full range (ie. 0 to 255) and basically
it needs to cause a delay like this:
>
> ADC_VAL                Delay in cycles
>
> 0                              400
> 255                          3200
>
> So this really means that each decimal '1' of ADC_VAL equates to 11 cycles
extra delay, ok so its not exactly 11 but it'll do, it'll only be 5 cycles
off at the end.
>
> Now here comes the part that is giving me grief, I can do 3205 cycles with
a two tier nested delay, but there are two values that affect the delay
time, the inner and outer loop variables.
> I think the way to do it would be too divide the required delay into 255
to get a carry (whole numbers) and the rest would go into the inner loop ie.
the stuff after the decimal point to 3 sig figs.
>
> This would be all well and good except for the fact that Nikolai
Golovchenko's division code uses a whopping great 400 cycles.
> I can't think of anything I can do, as I require two values I can't even
do a computed jump into a 255 long retlw table.
>
> Is anyone able to give me a poke towards a solution?
>
> Jonathan Starr
>

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics