[time-nuts] GPS 1PPS, phase lock vs frequency lock, design (Bob kb8tq) (life speed)

[Bob kb8tq]

Hi

Ok

> On May 31, 2019, at 4:48 PM, life speed via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> On Friday, May 31, 2019, 5:18:32 AM PDT, Bob kb8tq <kb8tq at n1k.org> wrote:  
> 
> Hi
> 
> One term you keep tossing up is “nominal phase coherence”. Typical GPS will do “phase coherence” 
> at the 10 ns level to a fairly high degree of confidence (90 something percent) with a number of 
> footnotes. You have never mentioned what your requirement is so it’s a bit tough to know what
> you are up against. If you are trying for “a couple of degrees at L band” from 0.1 second on out, 
> GPS simply isn’t going to do that, no matter what you do. 
> 
> Bob
> 
> Hi Bob,
> As you know, phase coherence has bandwidth and noise floor.  If you looked at two VCOs phase-locked to a common 10MHz reference, they would be coherent within the bandwidth of the PLL down to the in-band noise floor pedestal.  Beyond the bandwidth of the PLL they would not be coherent, and the in-band phase coherence would have a noise floor as defined by the PLL architecture and multiplied up reference noise.
> So I guess by nominal phase coherence I am trying to say that two GPSDO modules locked to the same GPS satellites (same physical location, probably even the same external antenna - and by antenna I mean that in the classic sense of the term, not GPS receiver) would exhibit phase coherence within a very small bandwidth.  From what we've discussed this would be tiny fractions of a Hertz.
> I understand that a GPS timing signal does not have the greatest jitter/phase noise, which I think is what you meant by "10ns level to a fairly high degree of confidence".  But we've also discussed that GPS would be best used to phase-lock a high-performance OCXO, where the PLL bandwidth is so low that the crossover point is where the OCXO ADEV far-out in time (1,000 sec?) is worse then the GPS.  My goal would be, with such a correctly-implemented system, to accomplish phase coherence within the BW of the PLL, which admittedly might only be milliHertz.  But if you displayed the 10MHz outputs of 2 modules on an oscilloscope (or ran them into a mixer or any other method of phase detection) they would be coherent.  The phase noise at offsets greater than a few milliHertz would of course not be coherent (correlated), but that is not the requirement.  I have other methods to implement wide-bandwidth phase coherence when that is needed.

You have two factors, one is the bandwidth of the loop. The other is the degree to which two identical GPS receivers track each other. In term of classic coherence, the units would be micro hertz rather than mili hertz. 


> One other term I've seen referenced is "navigation" vs. "timing" GPS receivers, is this just the error/jitter in the PPS?  TCXO vs. OCXO clocking the receiver?
> 
> You said:
> "The TBolt simply takes out the need for a precision time measurement between the PPS and the OCXO. They can do that because they builtthe receiver from scratch."
> It would seem this is somewhat along the lines of what I want to do.  Not necessarily build the RF receiver, but implement the code correction and digital PLL to arrive at results more in line with a "timing" than "navigation" receiver.

A DIY receiver that “beats” a conventional off the shelf device likely involves custom ASIC’s. That’s not outside the range for a commercial project. It is a 
“couple of million” dollar sort of hit to the budget. Since timing and navigation are very much linked at the root of the system, there isn’t much you would do
for a custom timing receiver that they don’t already do on the “timing” versions of the standard modules (and chip sets).

More or less:

If you are only making a couple thousand a year, modules are likely the most cost effective approach. Past that up into the couple thousand a month range, stock chip sets are probably your most effective solution. If you get into the tens of thousands per month, custom fiddling with stock chip sets comes in. At around a million a year, chip development is probably your best approach.

Bob


> Lifespeed
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.