Does anybody know if any of the Sparkfun GPS-modules is usable as a precise clock for time-stamp purposes (sub-second accuracy when locked to >4 satellites)?
Yes and no.
The PPS output will be once per second and indicate the time, but to get sub-second you will need to synchronize an external timer.
The GPS sentences are accurate, but not synchronized (i.e. the lat/long/time will be precise and match, but will not come out at any exact time).
Depending on how accurate you need it you could use a micro to discipline a VXCO from the PPS output for a cheap and simple setup. Or you can use something like a OXCO (ovenized crystal oscilator) or a RbXO (rubidium oscilator) to get crazy accurate. I am undertaking a project to start with a VXCO then move to a OXCO when I get a higher quality GPS module.
The dallas semi part can do 2ppm (and I think can be calibrated closer).
They have a web demo - it drifted just over 4 minutes in 4 years:
http://www.maxim-ic.com/products/timers/DS3231_demo/
Macetech has it on a breadboardable module (out of stock at the moment):
http://hackaday.com/2009/10/27/parts-ch … le-ds3231/
http://docs.macetech.com/doku.php/chronodot
Or there would be the SMD route. There are a few more variants (3232, etc.) but they have square wave output which would be accurate to sub-seconds if synced with the pps.
A Venus GPS plus one of these (sharing a rechargeable battery backup) might be interesting.