[time-nuts] Ublox F9P multi-band GPS receiver

[Bob kb8tq]

Hi

Been there / done that as well. The “gain slope” as you approach the magic trip point is 
a big gotcha. Since the effective gain depends a lot on the time offset, it’s not something 
you really want in the middle of a (should be linear) control loop ….

Having a really rotten oscillator with a whole lot of jitter would make it a little easier as 
far as gain slope issues … :) Yes, actually it’s the amount of jitter vs the PPS quantization 
interval that matters. A (much) higher frequency clock driving the PPS generator would 
also make the gain slope less of an issue.

Bob

> On Jan 22, 2019, at 12:11 PM, Poul-Henning Kamp <phk at phk.freebsd.dk> wrote:
> 
> --------
> In message <73DB44DE-F81E-4A3C-B27A-C9875DA91120 at n1k.org>, Bob kb8tq writes:
> 
>> Once you do replace the TCXO, you then are very dependent on a sawtooth correction 
>> to run your GPSDO. The PPS becomes one big long hanging bridge and thus is not useful. 
> 
> Well...
> 
> If you do not have the sawtooth, what you can do instead is try to steer your
> external clock to give you *maximum* jitter on the PPS.
> 
> That situation arises when you get the GPS to "waffle" between which
> two clock-cycles of the external clock it should put the PPS.
> 
> It is incredibly dependent on the GPS receiver internals, and it is
> very hard to keep stable, but you _can_ do it.
> 
> I know, because I did :-)
> 
> -- 
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> phk at FreeBSD.ORG         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe    
> Never attribute to malice what can adequately be explained by incompetence.