[time-nuts] 1PPS for the beginner
eb at telight.com
Tue Aug 14 16:56:40 EDT 2018
I have often wondered about sawtooth correction and other means to deal
with it besides reading the offset data (I know about the implementation
with a digital programmable delay line - pretty slick). Then I wondered
about what would happen if the GPS receiver's internal clock, instead of
being from a drifty TCXO, was synchronized and "perfect," synthesized
from the output 10 MHz. It seems to me that wouldn't work though,
because eventually the accumulated errors would just flip the offset up
and down around the true mean, at some very slow rate, rather than more
rapid, but unpredictable sawtooths and hanging bridges that can more
readily be filtered out.
Perhaps what's needed is a clock that is reasonably stable, like a TCXO,
that can be dithered over some range and rate so that the sawtooths
never get stretched out far enough to become bridges. If the dithering
is in a predictable or controlled form, it seems maybe the
filtering/processing could be improved.
Does this make any sense?
More information about the time-nuts