Attempting to isolate the main cause may be useful.

Consider the following trials
- Running several PICS together at relatively precisely controlled
temperature for several days and see if they track.
- Swapping crystals between a "good" and "bad" one and see if the
error transfers.
- Trying a single much more highly specced crystal to see if this
improves things.
- Making a simple crystal oscillator similar to PICs using the same
crystal and an inverter and run the PIC from this.
Similar results?

A standard wrist watch (sitting on an oven on your wrist) can achieve
better than 1 second per day over extended periods.
.


-----Original Message-----
From: Colan Tredrea <tredrea@AGENG.UNP.AC.ZA>
>I need to add the following points:
>
>1. I am unable to make substantial changes to the design.  I need to
>improve the accuracy of what I already have, ie the simple crystal
>oscillator using PIC16C63 Timer1.
>
>2.  The errors I have mentioned have occured over a small ambient
>temperature range of approximately 25 deg C +- 10 deg C, at the very
>most.  I would like to maintain an accuracy of inside 1 second per
day
>over the range 10 to 40 deg C.
>
>Thank you.
>
>Colan Tredrea
>University of Natal
>South Africa


......>
>My original message follows:
>
>I am trying to improve the accuracy of a real time clock operating
via
>Timer 1 on a PIC 16C63.  I would like to keep the drift inside of +-
>1s per day.   So far, I have had little success.  I am not sure if
the
>problem is one of instability or drift.

......