[time-nuts] Jupiter GPS Settings For Best Timing Performance
df6jb at ulrich-bangert.de
Thu Apr 10 08:04:49 EDT 2008
I do not know whether your design enables you to measure this with the
necessary precision or not, but in general questions for the sources of
such effects are more easily answered if NOT the observed frequency
fluctuation is specified but the sudden phase offset value that the
regulation loop is reacting to by the frequency fluctuation.
I have dealt a lot with such sudden phase offsets in my DIY GPSDO
because I wanted to be able to recognize outliers which should not have
an influence on the loop response. The experience out of it is:
1) Sometimes there will be sudden phase offsets that are more or less
due to errors in the phase comparator circuits, like a sudden glitch
that makes the phase comparator measure a wrong value. Errors like that
are produced for example by spikes on the power lines. It is
characteristic for that kind of error that it has a certain statistical
probability to appear but when it appears it has a negative influence on
a SINGLE phase measurement only and does not produce a "history". Errors
like this may come from the receiver circuitry as well. A analogue pll @
10 kHz with a time constant of several seconds should not react heavily
on effects like that.
2) Then there is a second source of sudden phase offsets that is more
serious: If there is a divider chain after the LO which most GPSDOs will
have down to 1 pps or 10 kHz in the G3RUH design then there are chances
that one or more flip-flops in the divider chain may change its state
due to an unwanted glitch. Because the divider chain has a "memory" in
the sense that its next state always depens on the previous state, this
will result in a permanent phase offset that a "normal" pll will react
on (try to remove it) ALTHOUGH the LO's FREQUENCY may be perfect all
over the time. It is characteristic for this kind of effect that the
measured phase offset is an integer multiple of the lo's period length.
3) Depending on fundamental changes in the geometry of the observed
sats, i.e. a sat dissappears below the horizont or a new one appears
above the horizont, the receiver must "re-think" its strategy on which
signals to rely more and on which less, which gives rise to sudden
changes in its output timing in the order of typically 20-30 ns. I have
observed a lot of events when the phase offset suddenly changes by 30 ns
or so stays so for some minutes and then falls back by the same amount
as if the receiver had now "learned" the new situation.
The things that you apply to like Satellite Elevation Mask Control
a.s.o. have only an impact if 3) is the issue that you observe. So you
should try to check whether 2) or 3) is the case.
> -----Ursprüngliche Nachricht-----
> Von: time-nuts-bounces at febo.com
> [mailto:time-nuts-bounces at febo.com] Im Auftrag von David Smith
> Gesendet: Donnerstag, 10. April 2008 13:05
> An: time-nuts at febo.com
> Betreff: [time-nuts] Jupiter GPS Settings For Best Timing Performance
> I've been working on the performance of my portable GPSDO
> (very similar
> to G3RUH design) and have got down to a reasonable level. Testing
> against an HP3815A, it will remain within better than 1x10^-10 for
> periods of 10-15 mins.
> However, there are times when it seems to get a "hit" and it
> will then
> move up to ± 5x10^-10 before settling down again. The
> movement seems to
> follow the time constant of the PLL loop filter, possibly indicating
> that it's coming from the Jupiter GPS?
> Reading the Jupiter doco (Zodiac GPS Receiver Family
> Designer's Guide),
> there seem to be several Zodiac Binary messages that can be used to
> alter the GPS's performance (e.g. Satellite Elevation Mask
> Control). I
> already have a processor talking to the GPS card and displaying
> position, time, gridsquare, status etc. It would be a simple
> matter to
> have it tweak the parameters as required.
> Does anyone have any experience with this?
> David Smith
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-> bin/mailman/listinfo/time-nuts
> follow the instructions there.
More information about the time-nuts