[time-nuts] Sawtooth Error Correction, WAS: Capturing NMEA and TICC timestamp data in time-correlated way?

[Forrest Christian (List Account)]

Let me see if I can give this a go.   Some of this will likely be
oversimplified and not exactly accurate.  Consider this a simplified 101
level class taught by an assistant and not the professor.

Most GPS receivers with 1PPS outputs are only able to align the 1PPS to the
nearest edge of some internal clock.   In addition, in most GPS receivers,
that internal clock is not synchronized to the 1PPS signal in any way.

For instance, if you had a crappy GPS receiver which only had a 1MHz clock,
the GPS receiver could only align the 1PPS output to the nearest uS (1 /
MHz), and since the timing signal from the Satellites are much better than
this, the 1PPS won't actually be aligned as closely as the GPS receiver
knows it could be.   For instance the GPS might pick a particular edge of
the 1Mhz clock, but it's also going to know that it really wanted to output
a certain amount of time before or after that specific pulse.   This data
is what is being output as correction data.

So, the GPS receiver in effect is saying 'I did the best I could outputting
the pulse, but if I could have, I would have outputted the pulse this
amount of time earlier or later'.   Not all GPS receivers support providing
this information, but at least some that do send it in NMEA between the
pulses.  Note that at least some of the GPS receivers also tell you in
advance of the errant pulse so one could correct it in real time.    (I'm
hedging a bit about how many receivers do this, as I don't have enough
experience with enough different receivers to know how many do or don't).

The term 'sawtooth' comes from how the uncorrected error looks when you
measure the 1PPS signal.   Generally the error will start out really low
(or negative) then grow until it has reached a point where one clock worth
of phase drift has accumulated, then the GPS will say 'oh, it's better to
use the next pulse', at which point the error will immediately return to
the low point since all of the accumulated phase error has been removed by
outputting on the next clock.  Thus a sawtooth...  slope from low to high,
then an immediate drop.   Note that the phase drift can also go the other
way, so it starts high and then goes low.   And, sometimes you'll see the
direction of the phase drift change as the GPS clock changes frequency
slightly.  That's what causes the hanging or suspension bridges.   If you
find a graph of this online, you'll see the sawtooths are going one way on
one side of the bridge, and the other way on the other side.

On Sun, Sep 15, 2019 at 5:21 AM Andrew Kohlsmith (mailing lists account) <
aklists at mixdown.ca> wrote:

> I’m interested in learning more about how this sawtooth comes about, how
> it’s predicted (so you can correct for it) and so on. Are there
> approachable documents/sites/source with this information? I’m still a
> time-nuts newb.
>


-- 
- Forrest