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

[Bob kb8tq]

Hi

Without a deep dive into the physics of GPS, there is not a good way to demonstrate 
just *why* the solution is as bad as it is on a second to second basis. There is no magic
“silver bullet” that will suddenly turn it from an ADEV of 1x10^-9 into an ADEV of 1x10^-13
at one second. The system ( Ionosphere, onboard clocks, orbits, transmitters, coding algorithm,
signal to noise ….) just isn’t good enough. Given that 1 ns is *way* better than the spec
it was designed against, you really can’t complain.

====

Basics of GPS:

1) You lock on to at least 4 sats ( = find the carrier / work out the doppler)
2) You demodulate the spectrum spreading to get the code clock
3) You use the code clock to get the data
4) From the data you work out:
	a) What sat is it
	b) Where is it
	c) What time does it think it is
5) Based on all that you can figure out where you are and what time it is. 

When spread, the signals are *way* below the noise floor. Even after de-spreading
they are not so great S/N in a pretty narrow bandwidth. What’s described above is 
not an easy task.

Code phase simply means you are workin out time against the code clock. That’s what 
all the magic little brains do. The TBolt simply takes out the need for a precision time
measurement between the PPS and the OCXO. They can do that because they built
the receiver from scratch. 

Bob


> On May 30, 2019, at 1:58 PM, life speed via time-nuts <time-nuts at lists.febo.com> wrote:
> 
>   2. Re: GPS 1PPS, phase lock vs frequency lock, design (Bob kb8tq) 
> 
> Hi
> 
> The TBolt is a very unique design. It directly uses code phase information 
> against the OCXO. The net result is really no different than the “correction
> message” approach, but it is a different implementation. Since you can’t 
> *buy* the guts of a TBolt to strap into a DIY GPSDO, it’s not generally part
> of a “I want to build a GPSDO from scratch” conversation.
> 
> Bob
> 
>> On May 29, 2019, at 9:50 AM, Alberto di Bene <dibene at usa.net> wrote:
>> 
>> On 2019-05-29 14:53, Attila Kinali wrote:
>>> The saw-tooth correction is the error of the PPS signal, as generated by
>>> the hardware, and where it really should be. The clocks of most GPS receivers
>>> are in the order of 20-60MHz and are usually unsteered TCXOs (or even XO for
>>> the cheap ones). Hence the granularity at which the PPS can be generated
>>> is fixed. The saw-tooth correction gives you a higher accuracy (or removes
>>> noise) from what you would get without.
>> 
>> Am I correct if I suppose that the Trimble Thunderbolt, which uses the 10MHz OCXO as clock for the processor, does not need any saw-tooth correction ?
>> 
>> TNX
>> 
>> 73  Alberto  I2PHD
>> /<<< http://www.weaksignals.com >>>/
> Just to be clear, I am an electrical engineer working on a commercial new product design, which already has a high-performance 10MHz OCXO as part of the product.  Although I realize much of this list is composed of DIY and hobbyists, it has always been clear to me there are some smart people participating in these discussions.  I am in the investigation phase of a feature enhancement to the product that is unfamiliar to me, hence my questions.
> Bob,
> Would you care to elaborate on "directly uses code phase information against the OCXO"?
> I have noticed that a product my company is currently manufacturing uses a "GPS receiver" containing a TCXO, which we then use the 1PPS to discipline a low-end 100MHz OCXO.  When I investigated the performance of our current design it did not meet my requirement of nominal phase coherence of two separate receivers.
> It would seem that a GPS receiver containing an oscillator may not make sense for a design that already contains a high-end 10MHz OCXO, rather the implementation should use the oscillator that is already part of the design.
> Lifespeed
> 
> | 
> | 
> |  | 
> time-nuts Info Page
> 
> 
> |
> 
> |
> 
> |
> 
> 
> _______________________________________________
> 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.