On Tue, 11 May 1999, Seyler Jean-Yves wrote:

> I'm looking for ideas for a specific project :
>
> I have to check for serial inputs while out-puting one or more PWM depending
> on the input(s).
>
> In fact, my first application should be an attempt to read 1 RC pulsed
> command, ie :
>
> 1) Measure the "n"st positive pulse among 7 overall pulses ; synchro being
> made by a 1st larger pulse and length of the usefull pulse varying between 1
> and 2 ms ; overall time = 10 to 20 ms
>
> 2) While outputing a PWM at 10 to 100 Hz, from #0 to #100% duty cycle
>
> 3) That's all !
>
> 4) I've thought of an absolute timing that each individual task could check
> and determine if their allowed time have passed. That shouldn't be very
> accurate but enough for me ?

You may wish to check out:

http://www.interstice.com/~sdattalo/technical/software/pic/pwm256.txt

This an all-software, 256-level pwm routine capable of 0 to 100% duty
cycles. Since it doesn't use interrupts, it's suitable for the 12 bit
core. Roughly half of the 256 cycles are available for other (isochronous)
tasks.

The only problem I can see with it is that it may be too fast for your
application. For example, with a 20Mhz clock this routine can produce a
nearly 20khz pwm'd square wave. I suppose you could slow the pic down a
factor of 200 or so, but I don'tsee this as being too practical.

Another possibility is to use the 16-bit delay loop that we were
discussing just last week. There was one solution that a) created a delay
with 3 instruction cycle resolution and 16-bits of dynamic range and b)
provided additonal cycles for simultaneously performing other tasks. The
number of 'additional cycles' can be varied between 1 to ~240 out of every
256 cycles. In fact, this number could vary depending on the time required
for each task. I unfortunately don't have time to expand upon the details
(gotta go to the Mchip seminar) - perhaps someone else may?

Scott