[time-nuts] M12 100 Hz vs. 1PPS measurements

WarrenS warrensjmail-one at yahoo.com
Fri Dec 5 12:00:57 UTC 2008



 
In message <02c801c956c9$4aa089a0$6401a8c0 at WSOffice>, "WarrenS" writes:

>The short of it is that every once in a while the 100 Hz syncs up
>for a short time to the 1 Hz

As far as I have ever been able to tell, that is not how it
works.

The 100Hz and 1Hz are generated with the same hardware and same
behaviour relative to software.

Back when I got my first 100Hz capable Oncore, I measured the
100Hz output against 100Hz generated by the TVB PPSDIV driven by
my Cs using my HP5370B and GPIB to collect all samples.

The conclusion was that the 100Hz was kicked not once per second
but much more often.

If the 100Hz had been synched once per second, both FFT and
ADEV/MDEV would flag the 1Hz signal resulting from that, and
they didn't, and neither does Toms data.

I also found, as Tom did, that using the 100Hz was as good as doing
the saw-tooth correction, and without getting into a long-winded
treaties about the finer points of modulus arithmetic, this also
indicates that the 100Hz tickled more often than 1/s and quite
likely at every period.

I tried to determine if the 100Hz indeed was tickled every period,
I expected it to show up in a histogram of the 100Hz signals absolute
or relative periods, but there were too much noise to confirm that.

Maybe Tom can try to repeat that experiment ?

Simply measure a sequence of zero-deadtime period widths of the
100Hz and make a histogram of the absolute widths and relative
widths, (the difference in width between period N and N-1.)

If we can confirm that the 100Hz signal is updated every cycle,
then that explains why Toms and my measurement found the averaged
100Hz to be as good as the negative sawtooth: sqrt(100 - ) ~= 10

Poul-Henning

-- 
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.

Poul
BTW 
>"The The short of it is that every once in a while the 100 Hz syncs up
>for a short time to the 1 Hz

BTW the 3 times a day or so sync up the above is referring to is the reason 
that every once in a while the 100 Hz is only as good as the 1 Hz signal 
for a short while. The 100 Hz is defiantly update or at least different on every cycle, 
and has about the same peak to peak noise, which is why if nothing but just 
average the 100 Hz before using it for the 1Hz update, things would get a better.
WarrenS


More information about the time-nuts mailing list