[time-nuts] Building a GPSDO & trouble using Jupiter-T

Chris Albertson albertson.chris at gmail.com
Mon Jan 30 12:59:24 EST 2012


On Mon, Jan 30, 2012 at 5:16 AM, Bob Camp <lists at rtty.us> wrote:
> Hi
>
> The GPS ops signal (or 10 KHz signal)  you will be trying to track will be moving by 10's or 100's of ns per second. 10 ns per second is 10 ppb. 10 ppb at 10 GHz is 100 Hz. You need to smooth this out or your LO will be moving all over the place.

Up at the top of this thread I said you need to have a long loop
constant for exactly this reason.   It does not matter of the GPS has
a 1PPS 100PPS or 1000PPS the problem is GPS itself.   GPS simply does
not have good short term stability.   In the short term the OCXO will
be better.

The reason we build GPSDOs is to combine the best of both, the short
term characteristics of an OCXO and the long term characteristics of
GPS.     If you try and build a GPSDO using a very short loop filter
time constant then you have a GPSDO with same  characteristics of GPS
averaged over your short time constant.

In other words you select the time constant based of your frequency
stability requirements.

Look at the first blue line in the top graph.  This is very much what
a Jupiter GPS has.   If you need 10E-10 you need a 100 second constant
.   The line is straight to for 10x better accuracy you need 10x
longer time constant.

I think using an analog PLL you can get to about the 1E-10 level.  I
think a non-expert could build a good 100 second RC filter.

 Doing better is going to require a longer time constant then is
reasonable.   a uP based PLL can do much better and likely with effort
equal the performance of a Trimble Thuderbolt.     But then you can
simply buy a Thunderbolt  for $150.


I'm just starting work on a PLL that (I hope) can be built using just
an Aduino as a controller and an 4046 chip (or the 74hct9046) as the
phase detector.  The Arduino costs more than a bare uP chip but I
don't want to have to build a PCB.
-- 

Chris Albertson
Redondo Beach, California




More information about the time-nuts mailing list