[time-nuts] LPRO-101 with Brooks Shera's GPS locking circuit

Tom Van Baak tvb at leapsecond.com
Wed Dec 20 14:35:32 UTC 2006


Hal writes:
> It might be possible to avoid hanging bridges by dithering
> the sawtooth. I'm thinking of something like a heater under
> the xtal for the GPS unit that gets driven by a medium
> frequency - slow relative to the normal sawtooth but fast
> relative to the PLL time constant.  This somehow feels
> not-good, but I can't see the obvious screwup.

Yes on the heater; No on the medium frequency.

Bruce writes:
> Avoidance of hanging bridges requires that the GPS receiver
> oscillator frequency not be a Hamonic of 1Hz at any time.
> Using a heater as you suggest will not solve this problem.
> Ideally it would be best if it were not a rational multiple of
> 1Hz either. Its generally simpler and easier to use the PPS
> sawtooth correction, if available. Resultant residual jitter is
> closer to Gaussian.


OK, here's the real story. If you've ever spent hours or
days on the bench watching the sawtooth error, noting
that the pitch of the teeth changes over time, and even
holding your breath waiting for one of those cute little
hanging bridges to show up -- it will be obvious to you
that any sudden change in the ambient temperature,
or even supply voltage, or perhaps shock or vibration,
will stop the "hang" in its tracks. Then the receiver
goes back into its classical rapid sawtooth mode. And
I second PHK's comments that in most cases, like
older GPSDO, sawtooth is your friend.

I suggested in an older thread that one place a resistor
(as a heater) above the xtal on the PCB -- and when
a bridge is detected -- you pulse some power into the
resistor to disrupt the temporary harmonic between
GPS and the xtal that causes the bridges in the first
place.

To be honest it was an obvious suggestion but one
that I hadn't proved. And like Bruce says, if you have
a receiver with sawtooth correction, and can use it,
the need to worry about the character of the sawtooth
is reduced.

But since I hate to make suggestions that I haven't
actually tried, I fired up my old SHOWTIME Oncore
VP to put my money where my mouth is.

The trick is to monitor the 1PPS phase and calculate
the delta phase each second (which is a kind of
measure of the size and pitch of the teeth). When
the sawtooth is coarse and there is cycle wrapping
every few seconds there is no problem.

But when phase moves slowly and the teeth get large
you tend to get a long ramp that is detrimental to
1PPS averaging. And if there is sign reversal during
one of these long ramps you get a hanging bridge,
even more detrimental to 1PPS averaging.

I used a 53131 TIC and wrote code to monitor this:
when the phase steps get down to just a few ns and
stay that way for a while you have the makings of
slow ramp or a bridge. It is at this point that you
blow some heat on the xtal. And, like magic, the
ramp or bridge goes away, and you go back to a
healthy sawtooth mode.

For a view of the heater experiment see:
http://www.LeapSecond.com/pages/vp/heater.htm

For a basic sawtooth example see:
http://www.LeapSecond.com/pages/m12/sawtooth.htm

In this case, I used my eye to detect the beginnings
of a ramp or a bridge and used a hair dyer to make
a quick pulse of hot air over the PCB.

If you were to make this part of a sawtooth solution
you can easily see how replacing my eye with some
code and the hair dryer from a meter away with a 50R
resistor on top of the xtal would do the trick quite
nicely.

/tvb






More information about the Time-nuts_lists.febo.com mailing list