[time-nuts] GPSDO control system

Magnus Danielson magnus at rubidium.dyndns.org
Sun Mar 23 19:45:01 EDT 2014

On 24/03/14 00:08, Charles Steinmetz wrote:
> Chris wrote:
>> 4) why use a PPS as the communications link from GPS to GPSDO?
> Generally, because that is the only precision timing output you get from
> a GPS unit.  The models with a 10kHz output were prized by the "simple
> GPSDO" crowd precisely for the fact that the phase-locked 10kHz signal
> made the control loop much easier to design (or, made an analog loop
> *possible* to design).
>> Can I find a way to move the OXCO into the guts of a GPS receiver.
> I don't know.  Can you?  That is essentially what the Thunderbolt does,
> and it brings with it the tremendous advantage that it removes the
> sawtooth error -- which is why some of us think so highly of that
> design.  But now you are really down and dirty, designing your own GPS
> receiver.  Are you competent (and, if so, willing) to do that?  (You are
> very, very far from a "simple" GPSDO if you are considering this.)
> Honestly, I never understood why all timing receivers aren't designed
> that way.  It seems to be such a superior way to do things that I'd have
> thought the other approaches would have died out once everybody knew
> about it.  Even if Trimble has a patent, I can't imagine it would be
> that expensive to license.  But even Trimble doesn't use it on all of
> their timing receivers.
> Anybody know why this scheme hasn't been used more widely?  Is there
> something I'm missing?

For the more expensive GPS receivers, they have been using OCXO, some 
even suspended OCXOs. Also, as an option you can connect an external 
clock for additional stability, and for some you can steer this clock.

Many of the OEM GPS boards is just... el cheapo.
So, you need to look beyond those to find other solutions.

For a GPS receiver itself, it doesn't need a clock to be sharp on time, 
as most of them is designed to handle a clock at the wrong rate. What 
they however do is learn and compensate for phase and frequency errors.
Also, the performance of the clock will affect the produced results, and 
ẃhen doing precision measures of code-phase for L1 and L2, the benefit 
is certainly there.


More information about the time-nuts mailing list