[time-nuts] Z3801A performance plots

Tom Van Baak tvb at LeapSecond.com
Fri Jan 25 19:18:24 EST 2008


> That looks like a classical underdamped PLL response.  You usually see this
> sort of hump in a phase-noise plot when your PLL is about to break into
> oscillation. :)
> 
> Either someone at HP was careless with the loop-filter constants and came
> close to running out of phase margin, the Z3801A is not configured properly,
> or (who knows?) something fundamental about the disciplining process changed
> when selective availability was turned off.
> 
> It looks like the disciplining loop bandwidth should be reduced, and/or the
> phase margin should be increased.  It's at least an order of magnitude too
> wide.
> 
> -- john, KE5FX

S/A is one factor; remember that all the HP SmartClocks were
manufactured back when S/A was on. So the optimal loop constant
today could safely be a bit longer from the optimal loop constant
back then. I assume S/A won't ever be turned on again?

For time-nuts newcomers, see:
http://www.leapsecond.com/pages/saoff/
or
http://www.google.com/search?hl=en&safe=active&q=gps+selective+availability

Another factor is the OCXO itself. I assume HP was conservative
and picked a firmware loop constant to cover the spec for the
10811 OCXO inside the GPSDO. If you happen to get a 10811
that is better than usual, as #4 was, then the filter appears today
to be less  than optimally tuned.

One of the reasons I did that Z3801a oscillator experiment
http://www.leapsecond.com/pages/z3801a-osc/
was to show that not all OCXO are the same, but all meet spec.

A final factor might be the environment that the GPSDO is
designed for. The SmartClocks were designed for cell sites
which perhaps have their own set of extreme weather conditions,
To keep look you want to err on the side of a short time constant
rather than assuming scummy days.

This brings up an issue. When amateurs make GPSDO they almost
always can do better than a manufacturer because they can tweak
the loop-filter to exactly match their particular GPS receiver, the type
of antenna, it's location, the level of environmental extremes, and
the measured quality of their LO.

A GPSDO manufacturer, on the other hand, must create a product
that meets their published specs, regardless of actual temperature,
location, or variation in the OCXO that they purchase from a third
party.

The Trimble Thunderbolt and Jackson Fury allow one to tweak the
filter parameters. In this way, an amateur user can tweak the loop
parameters. To do this reliably, though, usually requires a good
local frequency reference that is more stable than the GPSDO itself
for tau up to about 1 hour; an Rb or Cs, perhaps.

/tvb




More information about the time-nuts mailing list