[time-nuts] Advantages of GNSS ???

[Poul-Henning Kamp]

--------
In message <8e74832c-58e6-c2ef-f68c-48e116718e51 at LeapSecond.com>, Tom Van Baak 
writes:

>The plot is beautiful. The reason this delay line technique isn't used 
>much anymore is that AFAIK the Dallas chips are no longer produced. So 
>almost every uses s/w sawtooth correction now.

So... this is where it gets complicated isn't it ?

If you saw-tooth compensate the PPS in hardware, and then use it
to latch some digital counter, you may actually have made your
performance worse, because you lost the "dithering" of the counter
provided by the sawtooth, and all else being equal, just created a
new "hanging bridge" problem.

If instead you latch your counter on the uncompensated PPS and apply
the saw-tooth compensation in software, you most likely dither the
+/-1 count noise almost out of existence.

Compensating the saw-tooth in hardware only[1] makes sense (IMO) if you
feed the compensate PPS into an analog PLL and want to reduce the
"hanging bridge" effect to the resolution of your delay-line.

And if you implement an analog PLL in time-nuts territory
where the timeconstants are measured in minutes and hours, you
have far bigger challenges than the hanging bridge from the GPS...


[1] Or, I guess, if you want to feed the PPS to your HP5370 without
having to postprocess the TI measurements for the sawtooth.

-- 
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.