[time-nuts] TIC resolution impact on GPSDO's performance

[Magnus Danielson]

From: Dr Bruce Griffiths <bruce.griffiths at xtra.co.nz>
Subject: Re: [time-nuts] TIC resolution impact on GPSDO's performance
Date: Wed, 27 Dec 2006 01:15:48 +1300
Message-ID: <45911274.5030605 at xtra.co.nz>

> Magnus Danielson wrote:
> > From: "Poul-Henning Kamp" <phk at phk.freebsd.dk>
> > Subject: Re: [time-nuts] TIC resolution impact on GPSDO's performance
> > Date: Tue, 26 Dec 2006 08:14:12 +0000
> > Message-ID: <36958.1167120852 at critter.freebsd.dk>
> >
> >   
> >> In message <4590BB56.5070809 at xtra.co.nz>, Dr Bruce Griffiths writes:
> >>     
> >>> Poul-Henning Kamp wrote:
> >>>
> >>>       
> >>>> I think what you are trying to express is that the frequency from
> >>>> the internal Xtal (at times) is in an overtone of the 1Hz PPS, which
> >>>> gives rise to hanging bridges.
> >>>>
> >>>>   
> >>>>         
> >>> Poul-Henning
> >>>
> >>> Surely you mean harmonic? An overtone is not necessarily a harmonic.
> >>>       
> >> I meant overtone, becuase they are not in any harmonic relationship,
> >> the XO is wandering around whereas the (ideal) PPS is, supposedly,
> >> rock stable.
> >>     
> >
> > The only real way to describe it is to say that the XO and the PPS is
> > asynchronous to each other. Those who has freshen up on their greek for
> > technicians will know that this means "not the same clock" which is quite
> > accuratly what we have. ITU-T Rec. G.700 has a few interesting sections in it.
> >
> > The use of words such as harmonic or overtone should not be used since they
> > is normally used to describe the various frequencies within one signal where
> > as words as synchronous and asynchronous is to be used mainly for pair of
> > signals and describes their relative timing. Naturally, for a pair of signals
> > to be synchronous does not require them to have the same frequency, just a
> > fixed ratio between those frequencies.
> >
> > It does happends that a single signal is said to be synchronous, but the
> > wording which should have been used is isochronous (same clock). Then we have
> > the lovely words of plesiochronous and mesochronous. :)
> >
> > Cheers,
> > Magnus
> >
> > _______________________________________________
> > time-nuts mailing list
> > time-nuts at febo.com
> > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> >
> >   
> Magnus

Bruce,

> So the correct description of the cause of hanging bridges is something 
> like?
> 
> As the frequency of the clock which controls the timing of the receiver 
> PPS pulse output drifts and approaches pleisiosynchronism with the 
> "true" frequency of the PPS signal, the rate of change of the PPS timing 
> error due to quantisation decreases reaching zero should the clock 
> achieve synchronism with the "true" PPS frequency.
> Thus hanging bridges occur when the clock frequency approaches 
> pleisiosynchronism with the "true" PPS signal frequency and then drifts 
> away from pleisiosynchronism with the PPS signal..

This is much better even if it is not more readable, but that is in the
beholders eye.

However, there is one thing which may not be all that clear. For each PPS pulse
to be generated, the firmware needs to decide which clock-transition will best
approximate the internal time solution. This means that it continously makes a
frequency justification. PDH guys would call this bit justification and SDH
guys would call this pointer justification. Justifications of this kind is
known for one thing, being bad of relasing detailed phase information. The lack
of transition causes the worstcase quantization error to manifest itself.
The high frequency error causes many justifications and when they are smoothed
out they give a higher degree of phase information. Having to rely on such a
justification mechanism isn't very nice. The negative sawtooth error output
really removes much of the problem and especially relating to the hanging
bridge problem becomes less of an issue by using the information. However, the
negative sawtooth error signal is limited in resolution so on modern receivers
the full potential of the receiver frontend.

> When the frequency of the clock controlling the timing of the receiver 
> PPS pulse output is synchronous with the PPS frequency, then neither 
> sawtooth corrections or hanging bridges occur, however there may be a 
> fixed offset of upto 1/2 a clock cycle between the "correct" and the 
> actual leading edge of the receiver PPS output pulse.

The hanging bridge of the sawtooth correction is much smaller than that of the
PPS only. A hanging bridge of the PPS regulation can still output phase errors
and thus reduce the effective quantization error. The effect is there thought,
and for a shorter time.

> Please correct and simplify these statements as required.

A certain NZ product has my main attention right now. They have some problem
with a ringmod or some ringing... they use old tools and old magic and takes a
hell of a time to acheive. Back to the TV.

But yes, it could probably be better explained.

> The only problem that I can see is we still need to define how close to 
> synchronism the timing clock frequency and the "true" PPS signal 
> frequency system must be to be considered pleisiosynchronous.

As always, what degree of synchronousity is meaningfull for us?

Also, do we really *have* to care? Is this *really* how this problem should be
solved? Just because this is how receivers traditionally have indicated time,
is this really the best solution to compare a clock, via a for our purposes a
really poor clock? I think not.

Cheers,
Magnus