[time-nuts] GPS DO Alternatives

Charles P. Steinmetz charles_steinmetz at lavabit.com
Sat Dec 8 22:12:13 UTC 2012

Chris wrote:

>The goal, well my goal is to build a GPSDO to this set of requirements
>1) well under 1/2 the cost of the t-bolt.
>2) can be made with common parts and skills most people have
>3) is completely modifiable (open source software)
>    *   *   *
>if you could use the GPSDO a a local frequency standard for a 
>counter that has 8 digits you are doing "well enough".  If you can 
>get to 10 digital it is pretty good

OK, if we assume that a Thunderbolt is < $200 these days, the GPSDO 
module (no PS, no antenna) needs to be "well under" $100.  So, 
$75-80?  And to be clearly good enough for an 8-digit counter with 
one-digit overrange using gate times of 1 and 10 seconds, it should 
have an ADEV ~ 1e-10 at tau = 1 and 10 seconds.  Go down one power of 
ten for each digit over 8, and up one if the counter does not have overrange.

The performance at tau = 1 and 10 seconds will be largely dependent 
on the quartz oscillator used, assuming that its performance over the 
next two or three decades allows one to set the time constant of the 
discipline loop out in the 500 second range.

1e-10 is a bit more than one order of magnitude worse than an average 
Thunderbolt.  [1e-12 (what you would need for a 10-digit counter with 
one digit overrange) is somewhat better than you can expect from an 
average Thunderbolt.  1e-11 (what you would need for a 10-digit 
counter with no overrange, or a 9-digit counter with one digit 
overrange) is just about what an average Thunderbolt delivers.]

The holdover performance of the DIY box would likely not be as good 
as the Thunderbolt, because any sub-$100 implementation that can be 
made with common parts and the skills most people have is unlikely to 
include Kalman filtering or other predictive steering during holdover 
(but as I said in a previous message, holdover performance may not be 
required by many amateur time nuts, who can simply restrict their use 
of the standard to periods when it is performing nominally).

As we have discussed many times here, the control loop is much easier 
to design if it is comparing the 10 MHz output to a frequency much 
higher than 1 Hz.  This is why the original Miller DIY GPSDO 
(http://www.jrmiller.demon.co.uk/projects/ministd/manual.pdf) is able 
to achieve such good results (comparable to an average Thunderbolt) 
with a very simple design -- it uses Jupiter GPS engines with a 
disciplined 10 kHz output.  I believe the PC card may still be 
available, for those who can find a Jupiter to use it with.  (Note 
that Mr. Miller also supplies a ready-made version of the GPSDO with 
Axtal OXCO, but that one costs well more than a Thunderbolt.)

The design goals stated above are ambitious, and I suspect it will be 
a challenge to come in on budget even with the relaxed stability 
requirement compared to a Thunderbolt.  But I do hope you give it a 
go, and will be most interested to see what you come up with.

Best regards,


More information about the time-nuts mailing list