[time-nuts] Simple holdover ideas

Murray Greenman murray at rakon.co.nz
Wed Feb 13 20:21:10 EST 2008


With many GPS modules with 1pps the pulses keep coming even if the fix
is lost, and so your precious GPSDO attempts to track these erroneous
pulses, and you lose your nice holdover performance. The pulses can be
wildly off.

If you know when to hold, the micro driving the D-A and OCXO EFC input
can easily hold the last good value, or (as in the commercial GPSDOs)
use a Kalman filter arrangement to continue moving the D-A in an open
loop manner.

In one of the units I built, I had a HOLD input to the GPSDO micro, and
used another micro to monitor the GPS module (Motorola Binary Protocol).
If the fix went below 3D, the number of satellites in the fix below 5,
or the DOP was poor, this micro would simply raise a HOLD output, and
not lower it until proper conditions had been restored for at least a
minute.

It worked well. Of course the error on 24-hour holdover was larger than
desired, for the much more frequent dropouts a few minutes long, the
performance was all you could desire. I achieved about 1us for one hour.

The trouble is, this isn't really practical unless you write all your
own code.

Murray ZL1BPU




More information about the time-nuts mailing list