[time-freq] 1pps/10MHz discipline

[richard allen]

Chris,

Thanks for your comments.  I agree about the long filter times. I 
have built a similar system at work for a 16.384 MHz OCXO but I only 
needed about 10e-9 to 10e-10 stability and got away with a 120 second 
software filter.  I want this home brew one to be as good as my z3801.
   

I think I'll probably use the front end phase stuff that Shera used 
but I'm planning on switching to one of the more modern processors 
like the 18f452 or 18f8722 since I use them at work and have a C 
compiler for them.  

The oscillator that I'm using is a Datum 10 MHz pulled from a Lucent 
'lockbox'.  It's EFC is 0-5 volts for a +- 10 Hz swing. They were 
available in the Dallas area several years back and also have a good 
Motorola timing receiver.  I bought several at $25 each and wish I 
had bought more.  No one has figured out how to make them fly so I 
just ripped out everything but the osc and gps rx.

Richard

"Christopher Hoover" <ch at murgatroid.com> wrote:
(08/24/2006 01:12)

>Richard,
> 
>I was using Reflock II with r2_flex_01 to lock a VCO to my house
>standard (10 MHz).
> 
>But recently I got interested in the performance of the r2_enhpps_30_1,
>and I began taking some measurements.
> 
>I'm using an 5 MHz MTI 260 OXCO, with a 15 Hz tuning range from -10 V ..
>+ 10 V, negative coefficient.    I tweaked R22 to get some gain out the
>EFC opamp to get a swing close to 0 V to 10 V.
> 
>I'm not sure which model Isotemp you have, but the MTI 260 is an
>interesting case -- it is a very good OXCO. Almost anything you do to
>the EFC on this oscillator short of giving it a clean signal from a
>clean voltage source (e.g. a 10-turn pot) hurts its short term
>stability, which is in the neighborhood of 5E-12.
> 
>My initial measurements show that the Reflock II - MTI 260 system has a
>stability of ~ 5E-9 (Allan deviation) for tau < 10s, when locked to the
>uncorrected (i.e. the uncorrected for sawtooth error) 1 PPS signal from
>an M12+, .  This seems reasonable considering that I'm using a 8-second
>averaging and that the stability of a typical GPS receiver is ~
>10^(-8)/tau.   Given that I'm not bothering with sawtooth correction on
>the 1 PPS, this actually seems pretty good.
> 
>When locked to the very clean 1 PPS from my Z3801A (it comes from
>dividing the OXCO not from the GPS rcvr),  the short term stability of
>the Reflock II - MTI 260 system falls to ~ 1E-7 for tau < 10 s.
> 
>I'm not sure why the system is performing so poorly in this second case.
>I may be doing something wrong, but I have to wonder if the dithering in
>the 1 PPS signal from the M12+ doesn't improve the performance of the
>TIC block or PWM block  (or both) in the Reflock II design by way of
>averaging.    That's only a guess; I need to do more investigation.
> 
>The bottom line is that you need a fairly long integration time in the
>loop to take full advantage of GPS in a GPS-disciplined OXCO.  However,
>the maximum integration time you can use is limited by drift in your
>OXCO.  Somewhere around 100 s is a good place to start for a decent
>OXCO.   
> 
>The Shera design has an advantage in that it handles filtering in a
>microcontroller and can do longer time constants than the Reflock II can
>alone.   It also discards phase jumps in the 1 PPS signal.  (They do
>happen.)
> 
>If you married Reflock II with a microcontroller and added your own HDL
>to do TIC and PWM (or convince Luis to release his source so you can
>borrow those blocks), you can match the Shera design closely.  The
>difference in DAC implementations in the two designs should be
>considered, too.
> 
>-ch
> 
>