jtaylor said:
Just got a little programmer kit with pic for $20 CDN. Whoopee.
Anyway, I don't know much about these chips, have been googling a lot.
What I need to know now is how accurate and stable the internal oscillator
is?
The pic will be part of a complicated clock. One of the complications will
be running at some percentages of true time, ranging from about 50% to 800%.
It would be nice if I could use the internal clock, and once each unit is
built, run a calibration on it and set a value somewhere that that unit
would use as the 100% value from then on.
Is this needed - would the internal oscillator be out by more than say 1
part in 3600 (1 sec/hr)?
Would the internal oscillator drift very much? Temperature will be pretty
much constant, if that matters...
Its quite easy to build a little program/circuit to calibrate the internal
4MHz oscillator for many pic chips, assuming there is an OSCCAL word. I
wrote a little program that supports two buttons, and puts out a square
wave. One button tweaks up osccal, and the other tweaks it down. The result
is stored in eeprom. You read it out aftewards and use that as your OSCCAL
value.
One problem with this approach, however, is that the internal oscillator is
quite temperature dependent, so your clock will drift around if the weather
changes.
You can get temperature compensated crystal oscillators; then, you only need
to use one pin as the oscillator input, and the clock will be far more
accurate.
Another possibility is to use the powerline AC as a clock. It drifts some
small percentage, but is usually very accurate overall. There are tech notes
on doing this; basically, you connect a pin up using a 4.7Meg resistor, and
use the pin change interrupt to increment your clock. It'll change 120 times
a second in the US, or 100 times a second in Europe.
Regards,
Bob Monsen
