hi,
As I understand it, the 1second accuracy obtained by using software will not correct for any drift in the frequency of the crystal itself.
Programming tweaking only sets the program timing to precisely 1 second at a particular crystal frequency.
If the crystal frequency drifts due to say ambient temperature changes so will the 1 second period.???
Actually that's a *strength* of my 1-second system because it uses a long constant which allows the clock speed to be fine tuned very exactly. The problems of xtal temp drift and xtal ageing exist in all simple clock designs, but at least with my software it's very easy to trim the clock speed where you want it - unlike other systems like 32kHz xtals etc.
Of course you could make that long constant into a long variable, and get the PIC to auto adjust it over months to compensate xtal ageing, (or temperature compensate) which would be quite easy.
Roman's method is an excellent "software trimmer" method.
I've been able to obtain accuracy of about 1/2 second per month while experimenting with various "software trimmer" methods these last few years using 'standard' 50-ppm crystals. When I get to the point after several months where I'm adjusting the "correction count" up or down 1 count each month without improving the accuracy I figure I'm chasing temperature drift at that point.