I've had good success with an LP PIC and a 32khz Xtal as an RTC !

I feed the OscOut signal thru an op-amp buffer to the RTCC input with the
divider set to 128.  ( I use an TL27L1 .. can just barely buffer 32Khz, but
draws only 10-15 ua while doing so.  This MAY be overkill, but some LP Xtals
are sensitive to load, and run at multiples of their resonant if loaded too
heavily. )

I then poll the RTCC register, and when bit 7 is set, 128 x 256 32Khz clocks
have occurred.

DO NOT test bit7 of the RTCC.  Move it into a temporary register, and test
that.  That way, the lesser significant bits of the RTCC are unaffected.
( Testing the RTCC clears it as well. ) If you're a little late polling the
RTCC, little is lost, as the RTCC just wraps around and maintains those
bits for the next count.  Yes an individual "seconds" rollover may occur a
little late, but the error is NOT cumulative, merely a small offset.


( Here follows a mix of Microchip and Parallax code )


                mov     OPTION,#100111b         ; assign to pin
                                                ; trigger on 0->1
                                                ; assign prescaler to RTCC
                                                ; set prescalar to 1:128
check_seconds
               mov     w,RTCC
               mov     temp,w
               jb      temp.7,:rollover
               ret
:rollover


The above option and example was used for a 2 PIC board, in which the 32Khz
PIC was not doing duty as the RTC. ( He was a UART ! )  If a single
"programmable RTC" is desired, then the opamp is not necessary, and the
option assignments change, but idea remains the same.


Alan

--

Alan Rothenbush             |   There must be an ideal world, a sort of
Academic Computing Services |   mathematicians's paradise, where everything
Simon Fraser University     |   happens as it does in textbooks.
Burnaby, B.C., Canada       |                            Bertrand Russell