[time-nuts] GPSDO using 100Hz

WarrenS warrensjmail-one at yahoo.com
Tue Nov 25 04:18:45 UTC 2008


Didier

Nicely put.
I'd guess there is a good chance you are mosly correct, In any case correct about the end effect.
(The 100 Hz Phase Jitter is not what would normally be called 'pure' jitter or noise) 
Some day I'll look closer to see what is happening in more detail, but for now at least, 
what I'm doing works for me, and I have taken advantages of what  IT  is doing 
for whatever reason it may have for doing it.

The Jitter is NOT Random Jitter. View a plot of the 1 Hz correction phase sawtooth error signal,  
Mine, as do all I think,  has about a 4 second nominal sawtooth waveform period that slows down to zero Hz, 
then reverses direction and then starts increasing again giving a waveform from near DC  back to a 1 second period.
(note that a sawtooth of DC and one with a period of 1 second is in fact the same thing when sampled at 1 Hz) 
Now the interesting thing is that a plot of the the 100 Hz Phase error jitter signal looks just like the 1 Hz plot, 
except it has a nominal 4 to 10 Hz sawtooth waveform on it, instead of the 1/4 Hz sawtooth.
It also goes thru the same FM cycle, the only difference is that the nominal freq is about 50 times greater.
I also  found interesting that the cause of the 1 Hz sawtooth wave form appears to be totally due to aliasing.
That is there is a nominal say 4 to 5 Hz sawtooth on the 100 Hz that is non synchronous (most of the time) to the 1 Hz 
and when sampled at 1 Hz, only a single sample of the 100 Hz waveform value is output. 
In effecting this causes the 1Hz output to be the product of the beating between the 1 Hz signal and the 4 to 10 Hz 
nominal sawtooth that is on the 100 Hz waveform. 

The net effect to a Phase  filter is that with either the 1 Hz or the 100 Hz output rate, the sawtooth wave form, 
its shape, its peak to peak values are about the same in either mode, the only difference is that one sawtooth is at a 
nominal 1/4 Hz and one is at a nominal 5 Hz.   If you understand and accept that,  It becomes real clear why the 
100 Hz signal is so much easer to filter.
What starts out as about a +- 50 ns peak to peak phase noise, can be averaged down to around 1 ns phase jitter by averaging
 (filtering) from 100 to 200 samples. This works most of the time, i.e as long as there are 10 + cycles of the sawtooth in the sampling period.

Note: this is the results from my old 8 channel Oncore receiver running in fixed mode at 1Hz  and 100Hz. I have little or no experience using other receivers.
  
Warren

*******************
Warren,

This is what I believe, feel free to set me straight :-)

The 100Hz is done by synthesis from the GPS receiver's crystal oscillator,
which is usually not on a harmonic of 100Hz, at least not precisely (cheap
crystal) so the receiver will usually generate each pulse to the best of
it's ability to line up to 100Hz (so these will have somewhat deterministic
jitter due to the difference between 100Hz and the divided crystal), and
only once per second will try to align the average pulse train to GPS. There
is pure jitter 99 times out of a hundred, and actual correction the 1/100
time.  The advantage is that the once-per-second correction is burried under
(spread by would be a better term) a fair amount of 100Hz noise, so it's
probably easier to filter, allowing the use of a faster filter than a 1 PPS
output for the same level of 1 PPS attenuation. You can't use the scope to
determine if the jitter is pure jitter or a GPS correction, but I bet a TIC
feeding a PC would.

The point is that older receivers in particular simply don't have the
horseower to update the timing solutions 100 times/second, or 10,000 times
per second. 

Didier KO4BB

-----Original Message-----
From: time-nuts-bounces at febo.com [mailto:time-nuts-bounces at febo.com] On
Behalf Of WarrenS
Sent: Monday, November 24, 2008 6:48 PM
To: Discussion of precise time and frequency measurement
Subject: [time-nuts] GPSDO using 100Hz

Bruce

My Oncore's phase is definitely different on each 100Hz cycle. A digital
storage scope shows it very well.

Some things that come to mine:

1) Not all Oncores are the same, mine is an old 8 channel one. (Don't
remember the number, it is the one with the sawtooth capability)
2) The actual error amount is updated each 1 sec and the processors still
dithers that to get the 100 Hz to be close to the correct value
3) Other TBD 

Warren
*******************



More information about the Time-nuts_lists.febo.com mailing list