[time-nuts] Thunderbolt vs Z3801A

Magnus Danielson magnus at rubidium.dyndns.org
Sun Jul 20 21:25:17 EDT 2008

Dear Didier,

Didier Juges wrote:
> In my unqualified opinion, Trimble did something very smart with the
> Thunderbolt. 
> Most previous GPSDOs use a stand alone GPS receiver, with its own CPU clock,
> to generate the 1PPS signal, to which a separate OCXO is servoed via PLL and
> occasionaly smart software.
> The basic problem I see with this is that because the GPS receiver is
> clocked with a (20, 25, 40 MHz, or anything in between, pick your choice)
> crystal of generally dubious quality, the PPS signal is aligned with one
> edge of that CPU clock, and therefore the PPS signal error will fluctuate
> over that 1/f period. The GPS receiver may know by how much the PPS is off
> (and some tell you so), but it cannot align the PPS between two successive
> CPU clock edges, it has to choose one, so the PPS is adjusted with quantum
> leaps of 1/f, typically 20 to 50 nS. 
> Because Trimble integrated the GPS receiver and the OCXO, they use the OCXO
> as a clock to run the processor, therefore that source of error is not
> there.

First of all, so called sawtooth correction overcome part of this 
problem, but only to the degree by which the sawtooth correction reports 
corrections (often in steps of 1 ns) where as the Thunderbolt seems to 
report in 10 ps steps during locked conditions.

There is another aspect to the use of a good OCXO rather than standard 
TCXOs and that is that the lower noise allows for a better "position" in 
space and time as there is less noise in the input signal. Especially at 
lower frequencies/longer times.

Locking the LO to the GPS allows for a different kind of clock error 
handling than normal OEM GPS recievers typically use.

> The ultimate performance is not better than a perfectly optimized
> "conventional" GPSDO + OCXO, but excellent performance can be achieved at
> lower cost, and without having to deal with certain tradeoffs.


> Now, I am sure there are other, more subtle details that explain why the
> Z3801A is generally superior to the Thunderbolt, but that I cannot
> explain...

It is strange considering the 100 ps resolution TI counter in the Z3801A 
with the TCXO-GPS beating and high sawtooth error of the Oncore VP.

Locking the OCXO up and using the GPS receiver as a phase comparator to 
GPS time is a more direct approach than the acquward PPS method. You 
always gain alittle by doing it through the LO.


More information about the time-nuts mailing list