[time-nuts] carrier phase tracking GPS receiver

[Magnus Danielson]

From: Didier Juges <didier at cox.net>
Subject: Re: [time-nuts] carrier phase tracking GPS receiver
Date: Mon, 19 Feb 2007 17:50:04 -0600
Message-ID: <45DA37AC.2090502 at cox.net>

Bon soir Didier,

> Magnus Danielson wrote:
> > You can acheive much greater speedup by a combined frequency/phase approach.
> > You will get a very accurate frequency error estimate, so you will very
> > quickly be close enought to go into phase lock. At least if your clock isn't
> > too noisy. So, the lock-in time should not be the major concern, but rather
> > the behaviour of the full setup when running. Also, another classic trick is
> > to vary the bandwidth, so you have a much wider bandwidth in the beginning and
> > then step down towards your target bandwidth as some suitable conditions have
> > been met prior to the step.
> >
> >   
> My concern is that closing the loop faster by itself does nothing, if it 
> is not stable enough. What I meant by being able to close the loop 
> sooner was that the carrier phase data is actually better than the OCXO 
> sooner after power up. I think you or Bruce said the Allan variance of 
> the carrier phase signal could be as good as e-10 in 1 second, or 
> something like that.

Actually, it works both ways in the proposed setup. The lower (unknown)
frequency offset and lower noise of the OCXO kickstarts the GPS tracking
(which is FLL/PLL loops). The low noise of carrier phase measurement is
there, but there is an initial loss of time in order to accumulate enought of
them, so it takes time before you can actually get a more fullblown precission.
But even without those benefits you would be able to crank out much better
values even from code phase tracking by this method. Carrier tracking and
fancy corrections is just down the line.

> >> It seems that in this case, the acquisition time of the receiver will be 
> >> the most significant delay.
> >>     
> >
> > Your OCXO needs to heat up anyway. You can usually acheive a good GPS lock in
> > that time. A full-fledge correction would probably require data collected over
> > some time anyway, so don't fool yourself here.
> >   
> I was assuming the OCXO remains powered,

Yes, but you failed to mention this, so this is a whole different ballgame.

> which is fairly easy to do, but 
> in a transportable application, you cannot guaranty continued visibility 
> of the GPS satellites, or even continued connection of the GPS antenna, 
> when the system has to be transported, and in a hobby application, I am 
> not sure I can design a better set of algorithms to compensate for the 
> OCXO variations than what the Trimble or others have come up with in 
> their GPSDOs (in fact, I am sure I can't), so that holdover will 
> probably not be optimized (read: suck).

:)

Come on, where are your sense of adventure? Don't you feel like wanting to try
the wings a little bit? :)

> Also, I am not sure what happens when you move a timing receiver, I 
> guess the Thunderbolt for instance would have to do a new survey, here 
> goes an hour at best...

You better be running it with an antenna, so you can get a positioning
estimate. You probably want either a Kalman or particle filter approach here.

When you are talking about moving it around you again have shifted into a
different field and different set of solutions.

> > I was kind of expecting that question, it was just a matter of time before it
> > came up. :)
> >   
> Well, hopefully we are done with this one :-)

Indeed. Especially since I had cached up on a good answer (or so I hope).

Cheers,
Magnus