The xtal frequency is divided by 4 initially, and then TMR0 interrupts on=
=20
overflow from 255->0, meaning another divide by 256.  So minimum divide i=
n=20
this situation is 1024.  If you need 1 Hz resolution, your xtal freq shou=
ld=20
be divisible by that, assuming no pre-scaler.  If you need an interrupt e=
very=20
1/100-th of 1Hz, now you need a xtal that divides by 102400.

However, you *can* change TMR0 manually to make the overflow divisible by=
=20
something else, but that just adds another complication.  Writing to TMR0=
=20
also stops the TMR0 counter for something like 2 or 3 instruction cycles,=
=20
IIRC.  So it would be really nice if your xtal divided properly.

Clocking timer-1 and timer-2 as independent oscillators is nice cause you=
 can=20
run a low-cost, higher-speed oscillator (like RC or even internal osc) fo=
r=20
the main processing, and then use an accurate 32.768khz xtal on the secon=
dary=20
oscillator for the clock timing.  I believe your 18F452 has an internal o=
sc=20
(no external components req'd), so I would lean towards using this latter=
=20
method.  Pick your secondary oscillator xtal so you get the resolution yo=
u=20
need.

There's lots more details in here, which the datasheet and the mid-range=20
manual explains much better than I can.

Cheers,
-Neil.


On Thursday 05 June 2003 15:41, Peter Moreton scribbled:
> Hi, I'm playing about with a PIC18F452 and trying to design a very
> simple clock circuit, to show both UTC and Sidereal (astronomical) time
> on a 2x16 LCD display. Before I start writing code using the C18
> compiler, I'd like to ask for opinions for crystal frequencies that
> 'make sense' in a clock application. I was kindof expecting that a powe=
r
> of 2 crystal such as 32768 khz or better still, a 4.194304 Mhz crystal
> would be sensible, as the frequency is easily divisible into 1hz. But,
> then, reading up about the TIMER0/1/2/3 modules, I begin to doubt that
> it matters at all, since I can load a counter value appropriate to the
> xtal, and have the counter overflow, and generate an interrupt, say
> every 100'th of a second.
>
> Is my understanding correct, that using TIMER0, accurate time can be
> generated from any abitrary clock frequency?
>
> I noticed that TIMER0 (and also TIMER1) have the ability to be clocked
> independently from the CPU clock, from a 32768 khz watch xtal, is there
> any reason for this, in terms of clock applications?  Would it be the
> case that the 32768 khz xtal might be used when the main CPU clock had
> been selected for ease of generating serial (RS232?) baud rates?
>
> Sorry for all the questions.... steep learning curve for me right now..=
=2E
>
> Peter Moreton

--
http://www.piclist.com hint: To leave the PICList
mailto:piclist-unsubscribe-request@mitvma.mit.edu