[time-nuts] Time source for indoor standalone PC

Chris Albertson albertson.chris at gmail.com
Fri May 17 11:43:07 EDT 2013

If there is no way to get radio signal into the room, then buy a rubidium
oscillator.  Conect the Rb to a small notebook PC the run Linux or BSD
Unix.  Let the Rb oscillator drive NTP and get it sync'd up outside your
room and then  walk the Rb/NTP server into the room.    Because you are
isolated you will need at least three of these systems and some people
argue you need five of them.   I'd argue five is certainly better, but
three is a minimum.   Then periodically you rotate one of the systems
outside for calibration with GPS.

Inside the room you configure the three to five servers to run in "Orphan
Mode"  This wil allow them to develop a kind of consensus time based ont
the set of servers that agree.  Hence the reason for having five servers.

One real problem with a disconnected "island" is dectecting errors.  How to
know if the server is 50 or so milliseconds "off".  You can't depend on
only one.

The good news is theRubidium Oscilters are not expensive.  $100 Will get
you a working unit.  And certainly PC notbook computers are dirt cheap if
you buy older ones.

The hard part here is setting the Rb units.  They need to be GPS
disciplined when GPS is available and then flip over to "hold over" mode
when GPS goes away.  with your low-precision requiremnts that should keep
good time for over a year with GPS disconnected.   Then you take them
outside and run them for a few days with GPS.   So with six servers, one
would be outside and five inside and every two months you rotate them.
This should let you run at the millisecond level and also have the ability
to withstand two server failures.

On Thu, May 16, 2013 at 9:45 PM, Grant Waldram <grant at remobs.com.au> wrote:

> Hi folks. I wouldn't call myself a time nut, so this is really an effort to
> ask for advice from some people who know the field. My first contact with
> time synchronisation was looking at the instrumentation clocks for the
> Woomera rocket test facility when I went out there for a few (large!) hobby
> rocket launches. I can't even remember the system's name but it used a
> series of pulses of various lengths to give a unique time code. But, I
> digress...
> I've not had much need for time synchronisation over the years, but in
> recent years NTP has been able to get me by. Unfortunately I'm now faced
> with a network that needs a moderately correct clock (I'm scared of using
> the word 'acurate' around you folks!) to the order of a few seconds or so,
> but with no possibility of an external internet connection (for a number of
> reasons). At present I'm using one PC running Windows Server as an SNTP
> server to synchronise all of the devices, as it is the only device in a
> physically secure location. This is inside a security-fob protected room. I
> can't get GPS signals in there, and the Australian radio clock network was
> shut down about ten years ago. Our CDMA network was turned off in 2008.
> Right now all I can think of is GSM, and while i know it's not terribly
> accurate it seems like it might be the easiest. It also might be that I've
> got tunnel vision and there's a simpler solution out there.
> I would be quite happy with some sort of dedicated GSM/NTP-server box, and
> there are Arduino/Raspberry Pi/Linux homebuilts for that out there, but I
> have been wondering if one of the fairly common GSM USB sticks could
> somehow
> be a time source to set the server clock?
> Regards,
> Grant
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.


Chris Albertson
Redondo Beach, California

More information about the time-nuts mailing list