Re: NEO series GPS modules
Eric Haskell <eric_haskell@...>
toggle quoted messageShow quoted text
SORRY TIME-NUTS is NOT a YAHOO group.
See time-nuts link below:
To subscribe or unsubscribe via the World Wide Web, visit
This is not SPAM,
Date: Tue, 26 Jul 2016 20:30:42 +0100
Subject: Re: [ukmicrowaves] NEO series GPS modules
The classic "Simple GPSDO" such as http://www.g4jnt.com/SimpleGPSDO.pdf and http://www.g4jnt.com/EvenSimplerSimplestGpsdoPossible.htm generally run with a PLL time constant of a few tens of seconds; perhaps as much as a minute. But this is not sufficient to smooth out the variations on the GPS timing signal itself, due to propagation anomalies.
See http://www.g4jnt.com/10MHz_Reference_Source_Stability.pdf for a set of short term stability measurements I made on several reference types including various GPSDOs
To smooth out GPS's own frequency shift, which can amount to a couple of parts-per-billion easily, you need to move to a time constant of tens of minutes to hours and that runs into a new problem altogether.
The simple short time constant approach allows a low cost TCXO to be used as the oscillator element since the loop is fast enough to track its drift and wobble due to thermal effects (no TCXO has perfect compensation, ~~1ppm is all they're specified to). With a time constant of minutes, it is too easy for the TCXO to wobble faster than the loop can compensate and that ends up worse than a fast loop
So you now need an ovenned source that can maintain good performance over tens of minutes to hours - and once you've got that, you have something that's probably going to be good enough on its own, with a periodic calibration / check. And then, of course, there is the lock up time. If your GPSDO has a loop time constant, say, of 20 minutes, it could well take up 2 hours to lock from a cold switch on. As a rule of thumb, 4 to 8 time constants are needed to reach fulll stability. So unless you have an intelligent computer controlled PLL that can adaptively adjust bandwidth as lock is achieved any /P event will be well underway before it is locked. So this sort of GPSDO should never be turned off, and in fact makes a very good fixed lab standard.
As really good lab standard, a GPS locked rubidium source can probably not be bettered for home use. Rb gives you parts per trillion short term stability over hours, and GPS locking with a time constant of very many hours or days corrects Rb's inherent slight error. Such a setup would probably hold its own against the older second hand Caesium source.
All of which brings in a third problem : how do you actually do a PLL with a time constant that long? Analogue R/C is out of the question, and if you try to do it with D/A converters and / or DSP, there is the quantisation issue. (One failing of the VE2ZAZ design) The best D/A converters are only 16 bit, and you'll need a finer resolution than that or a very narrow adjustment range. Pulse width mod couldy be made to work, but only with a PWM of very much higher resolution than the hardware ones provided on-board chips - which means doing it in software / firmware. And that high a resolution means a fast clock.
It ain't simple, doing it properly! But for practical purposes, you can get away with a lot less.
On 26 July 2016 at 19:51, Neil neil@... [ukmicrowaves] <ukmicrowaves@...> wrote: