[time-nuts] Testing frequency using NTP Bruce GPS ps
bruce.griffiths at xtra.co.nz
Sat Oct 4 06:34:41 UTC 2008
Mike Monett wrote:
> Bruce Griffiths <bruce.griffiths at xtra.co.nz> wrote:
> > Mike
> > They actually use an augmented form of GPS common view for which
> > the GPS PPS signal and its timing variations are largely common to
> > both locations and thus largely cancel when comparing the
> > frequencies at the customer site with the standards at NIST.
> > With a good timing receiver (and antenna location) the sawtooth
> > corrected PPS signal timing noise can be as low as a few nanosec.
> Is that like the CNSC02-O1 High Performance PPS system? From their
> web site, they state:
> "Provides for dynamic hardware correction of the 1PPS quantization
> ("sawtooth") error. This reduces the noise on the 1PPS pulses from
> (typically) +/-27 nsec (15 nsec 1-sigma) to +/- 11 nsec (3 nsec
> 1-sigma). More importantly, eliminates periods of bias error caused
> by the quanization error going through a "zero beat" period that can
> last one to two minutes about every 10 minutes or so."
> That paragraph is very interesting. The "zero beat" error is
> particularly nasty. I tried to get more information, but it wasn't
> really clear how their correction method fixes this problem. Could
> you explain it a bit more?
The GPS receiver can only position the leading edge of the PPS pilse to
the nearest transition of the receiver crystal oscillator or harmonic
However it can estimate the resultant timing error with a resolution of
1ns (for the M12 GPS timing receiver and its derivatives).
The resultant sawtooth error can either be corrected in hardware as in
the CNSClockII or it can be corrected in software (if the PPS transition
is time stamped with sufficient precision).
The NIST system has a time stamp resolution of 30ps or so software
correction for sawtooth error is possible.
After sawtooth correction their is still some residual, largely random,
noise of a few ns rms with a well sited antenna.
> I am working on a new technique that might give one or two orders of
> magnitude improvement locking to the 1PPS signal. It might give
> sub-nanosecond locking in a very low cost system. But the zero beat
> is a serious problem, so I am interested to learn as much as
> possible how other solutions work. Maybe the simplest is to just
> apply some jitter to the GPS crystal oscillator to keep it off zero
> > The "all in view" technique will reduce the noise contribution to
> > the comparison somewhat.
> That term is a bit confusing. Does it mean using all satellites that
> are in view, or does it mean the NIST receivers see the same ones
> that the user sees?
Common view means that the 2 GPS timing receivers use the same GPS SV
for timing purposes.
Observing schedules have to be agree upon before hand.
This technique originated with early receivers that could only track one
SV at a time.
All in view means that each GPS receiver use all SVs that it can see to
derive its timing.
With modern timing receivers sufficient information can be logged so
that the timing info from each individual SV can be extracted from the
PPS time stamp sequence if desired.
> > The timestamp resolution of better than 30ps or so ensures that
> > time stamp quantisation noise is negligible.
> > It also allows, in principle at least, standalone 3 cornered hat
> > comparisons of the frequency instabilities of the 3 sources
> > connected to the customer instrument.
> Another term I must study and learn how to do: "3 cornered hat". It
> sounds like a very powerful technique, ideally suited for a fairly
> simple program and some data logging.
It is if and only if the phase variations of 3 (or more ) oscillators
are statistically independent.
It is then possible to derive the individual phase variances of each
oscillator from the the 3 (or 0.5*N*(N-1) for N oscillators) sets of
phase difference measurements between pairs of oscillators.
In practice all oscillators may share the same ambient thermal
environment so that their finite oven gains will produce small cross
Even humidity and atmospheric pressure variations may affect the
frequency of an OCXO.
> > For more detail see:
> > http://tf.nist.gov/timefreq/service/fms.htm
> Thanks very much for the link. It is curious they don't seem to
> spend much effort on correcting the user's frequency errors. They
> just want to report how much they are off.
> Why is that? You'd think they would perform a more valuable service
> by applying advanced techniques to adjust the user's equipment to
> minimize the error, then report and certify the actual result. Is
> there some reason they want to leave the user's equipment
> > Only carrier phase GPS techniques are potentially capable of
> > picosecond noise levels.
> Now we are getting very interesting. How do you do that?
Most GPS receivers track the carrier phase variations internally.
Some make it available to the user (eg. Novatel Superstar GPS receivers,
Rockwell/Navman Jupiter GPS receivers ).
However since these refer to the phase variations between the GPS
carrier and the receiver LO, they are usually only useful if the
receiver LO is phase locked to the OCXO output frequency.
> > However there are a large number of effects that have to be taken
> > into account and data reduction and correction is very complex.
> What if we just want to stabilize an oscillator frequency, without
> caring much what the exact phase offset is from the USNO? The high
> carrier frequency should make it much easier to lock a rubidium or
> crystal oscillator, and it should give much lower phase noise.
There are commercial carrier phase disciplined oscillators available (eg
from Quartzlock) if you can afford them.
One of these purports to have better short to medium term stability than
a passive hydrogen maser.
> Of course, auroras and other disturbances would be more significant,
> but here I'm mainly interested in getting a good lab reference to
> measure and compare the performance of other commercial oscillators.
> This is a separate topic, but as long as we are talking about
> precision signals, do you happen to know what kind of distribution
> amplifiers are used at the USNO to distribute the signals from their
> cesium and hydrogen standards?
No idea, as they don't say but the Spectradynamics amplifiers are used
by national standards labs in several countries.
> > Bruce
> Thanks very much for your help. It might seem like a lot of
> questions, but I'm pretty sure you will be very pleased with the
> results. I think I can beat just about all the other methods except
> carrier phase techniques.
> Best Regards,
> Mike Monett
More information about the time-nuts