Order Tray | Contact Us | Home | SIG Lists

[time-freq] What Use Clock-Block?

John Ackermann N8UR jra at febo.com
Wed Apr 25 01:38:13 UTC 2007


Hi Keith --

Do you "need" a Clock-Block for good NTP timekeeping?

No.

But it will help quite a bit if you're looking for "ultimate" stability.
 The reason is that NTP doesn't react quickly to changes in the nominal
rate of the system clock, so if the oscillator frequency shifts due to,
e.g., a temperature change in the machine room, you'll see some wiggle
in the time as a result.  Dave Mills has commented that you can use a
normal NTP server to determine if the air conditioning has failed.

If you use a Clock-Block driven by a good external oscillator, you will
avoid that problem and improve the short-term stability of the system.

Now, we're talking about small differences here (a few hundred
microseconds perhaps) but they are noticeable if you're after the
ultimate performance.

John
----

K&J Aggett said the following on 04/24/2007 08:02 PM:
> Hi,
> 
> Having driven some country miles recently I had plenty of time to
> consider the Clock-Block and, I must confess, have some queries.  Is it
> an essential bit of kit or simply a 'must have'?  Let me explain.
> 
> The average system would probably comprise a GPS receiver, a GPS
> controlled oscillator producing, amongst other frequencies, a "standard"
> 1PPS, Clock-Block driven by the "standard" PPS and a computer and
> operating system.  Clock-Block drives the kernel clock and to set the
> actual system time in hh:mm, NTP is used to derive it from the GPS
> receiver, although I concede that this could be set from the hardware
> clock.  To make the system time accessible to other than locally running
> programs, the computer would be designated as an NTP time server on the
> network.
> 
> Now, it seems to me that there are several ways of arriving here.  One
> way is to set up the local clock as being externally disciplined and
> introduce some complications I have not yet thought through, while a
> simpler way seems to be simply set the stratum of the local clock to,
> say, 1 or 2.  Whichever way one proceeds, my understanding of NTP is
> that it 'trains' the system clock according to some well honed
> algorithms to obtain the 'best result', whatever that means.
> 
> With the 'right' GPS controlled oscillator, I can see that the system
> can ride out periods of GPS signal loss, but that is not a function
> provided by Clock-Block.  A time server with PPS implementation driven
> from GPS controlled "standard" PPS would do the same, so my question is:
> 
> "Does anyone have any idea of the in-practice 'best result' of a system
> with Clock-Block compared to that of one without?"
> 
> Having a GPS source with PPS on it, one could succumb to the temptation
> to use both Clock-Block and GPS derived PPS (or even the 'standard'
> PPS).  Big or small mistake?
> 
> Or have I got the whole thing completely wrong?
> 
> Regards,
> 
> Keith Aggett
> 
> _______________________________________________
> time-freq mailing list
> time-freq at lists.tapr.org
> https://lists.tapr.org/cgi-bin/mailman/listinfo/time-freq
> 
> 





More information about the time-freq mailing list