[time-nuts] GPS Locked and Unlocked Performance Comparison

[Magnus Danielson]

From: SAIDJACK at aol.com
Subject: Re: [time-nuts] GPS Locked and Unlocked Performance Comparison
Date: Wed, 13 Feb 2008 14:03:04 EST
Message-ID: <c1c.2b8a250b.34e498e8 at aol.com>

Said,

> I wonder how the Miller unit would perform if the GPS is kept alive during  
> holdover.
>  
> If the Jupiter continues to generate the 10KHz and 1PPS from it's  internal 
> TCXO even without GPS reception then I would expect the holdover  performance 
> to degrade to that of the TCXO if the unit goes into antenna-related  holdover?

This is indeed important for any GPS diciplining, how you handle loss of
signal. Regardless if it is on 10 kHz or PPS level. The TCXO can be the clock
you actually track...

> Unless the Miller design removes the 10KHz reference from the PLL if there  
> are less than a certain number of sat's received etc. Should be possible to add 
>  that feature by monitoring the RS232 output, gating the 10KHz, and clamping 
> the  EFC control voltage -  if that feature is not already  there.

Hold of the EFC is the important thing. Regardless the type of error
indication, it needs to act fairly quickly not to infect the state of the
loop too much. It will always do that one way or another.

There are many mechanisms that can reduce signal quality or remove the
possibility to track signals.

Relying on the AC-aspects of 10 kHz XORed with a '0' or '1' isn't very good.
The averaging of the aspect ratio will dominate over time for the achieved
frequency. Over time will power voltage also steer things.

> That's a nice behavior trait of the Motorola receiver: it shuts off it's  
> 1PPS when there are less than 4 Sats etc (in TRAIM mode).

Badly needed.

> It's surprising what kind of performance can be achieved with the 10KHz  
> locking the OCXO through an Exor gate. Then again that design get's to compare  
> phase 10.000 times more often per second than all of the other 1PPS based PLL's  
> :)

Indeed. While the time error is not updates as often but 10 kHz is a nicer
frequency to work with. The information rate is not different, but the flip
energy is dispersed and more easilly filtered.

Cheers,
Magnus