replicating FB architecture?

Hal Murray hmurray at
Tue Mar 31 08:33:27 UTC 2020

> For less than 50 ms drift over 3 months, it looks like an OCXO will likely
> hit the max drift wall:

If you have an OCXO with a wire coming out of it, I don't know a simple way to 
get than into ntpd.

I see 3 approaches.  There may be more.

1) You can connect up the OCXO to the clock on the PC.  Then you can use 
ntpd's local clock or orphan mode.  You either have to get an OCXO that 
matches the frequency your PC wants or add a PLL box to convert it to what you 
need.  Good OCXOs tend to be available in things like 5 MHz or 10 Mhz.  Maybe 
a few Telcom frequencies, but they are much less common.  PLL boards are 

2) You can turn your OCXO into a GPSDO external of ntpd.  Pick your favorite 
GPS to emulate.

3) You can do something like divide the OCXO output down to 1 Hz and feed that 
into a PPS input on your PC.  ntpd expects PPS inputs to be on the second 
boundary.  So you either have to steer your PPS so that it ticks at the second 
boundary, perhaps by adjusting the divider, or by just letting the PPS go 
where it goes and fixing things up in software.


Of course, plan 2a is to start with a real GPSDO rather than a raw OCXO.  I 
don't know of any of them that are specified at holdover times in months.  
That would be an interesting experiment.

If this is a one-off and you are willing to use recycled gear, look on eBay 
for Trimble Thunderbolt or HP Z3801A.  Both are well known but I haven't seen 
any holdover data on a month scale.

eBay also has a bunch of home brew GPSDOs from China.  I don't know how good 
they are.


For 3 months, you may have to worry about wall power going away.  A UPS will 
solve that.

These are my opinions.  I hate spam.

More information about the users mailing list