[time-nuts] GPSD-Rb

[Chris Albertson]

On Wed, Aug 22, 2012 at 7:47 AM, Michael Tharp <gxti at partiallystapled.com>wrote:

>
> The design I'm thinking of is to have a separate microcontroller clocked
> by the Rb that will generate a third pulse-per-second


It is very hard to write a divers in software.  You have to use assembly
language and you have to mmake sure that EVERY path in every branch is has
exactly the same number of clock cycles.   And then you have to be lucky
that you can work out an exact integer division.

The above has been done but always the uP can't do anything else but the
divide down loop.  If you need some other background task use a second uP
chip.

I think a better way to do holdover is to use a stable OCXO that is phas
locked to some timing source and then, in real time you might switch to the
"best" source.   If one of your source is a 10MHz Rb then why divide that
down to 1Hz?  Use the 10Mhz output to drive the OCXO.

I can't draw a diagram so I'l use even more words... The "final" OCXO has
10Mhz output and is controlled by a PLL.   The PPL has two 10Mhz inputs
(references) it can choose from (1) the Rb and (2) a 10Mhz GPSDO.  I think
you need a uP to control the switch and also likely to manage the loop
constants inside the PLL.

I've been thinging about the because I have a few GPSes and an Rb also.
 That said this is not a high priority because in all the time I've been
doing this I've not once seen a need for holdover.  GPS seems to be 100%
reliable.

What I really want is a portable standard.  So rather then the above
I'mthinking about a GPS disiplined Rb.  Then I can unplug the GPS and take
the Rb where it is needed.



Chris Albertson
Redondo Beach, California