> Better designed internal oscillator, less sensitive to 
> temperature and
> voltage change (perhaps temperature compensated?).
>
> Don't count on it for RS-232 and other timing applications 
> if you plan
> to run it to its extremes in temperature or voltage - it 
> can vary as
> much as +/-3.2% or so ( see figures 18-39 to 18-42 for 
> graphs) which
> is greater than the 2.5% I usually restrict myself to when 
> talking
> about RS-232 acceptable clock error.
>
> Still, if you drive it with a reasonable voltage, and stay 
> within a
> restricted temperature range, then it should work 
> reasonably well (ie,
> +/- ~1%).

Almost just saying "what he said". Almost.

RS232 with N bits will just mis-sample at ~~= 100/(2N+1) % 
total error or about 5% total error.
That's error between both ends so if you got opposing 3%+ 
per end you would be dead, as Adam says.

Also, your received needs to be a bit smart or lucky to work 
right out to bit edges OK.
Really clever code can try to resynch on bit data 
transitions, but much isn't that clever, and that may fail 
for some data. If the drifts are slow, as you'd expect, then 
bit timed clock speed recovery can allow much better 
results. Even cleverer code needed.

Also as he says, if you can get both ends to drift the same 
way (eg common temperature change) or limit drift then all 
may be well.




        Russell


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