[time-nuts] Divide by five ->Ensemble

Magnus Danielson magnus at rubidium.dyndns.org
Mon Nov 10 18:46:32 EST 2014

On 11/09/2014 10:54 PM, Chris Albertson wrote:
> On Sun, Nov 9, 2014 at 12:45 PM, Bob Camp <kb8tq at n1k.org> wrote:
>> Hi
>> The main point is that NTP picks *one* source from among it’s batch of
>> inputs and uses that. The ADEV of the output can be no better than the ADEV
>> of the output.
> The statement above is not correct.  NTP does not select just one clock.
> The statement below is correct and is  what NTP actually does.   If you
> look at the output from "ntpstat" you might think NTP selects one clock but
> internally it's not going that.  The display is misleading.
> In the case of an ensemble of clocks combined with a better approach the
>> ADEV of the output can be better than the ADEV of the best clock in the
>> group.
> It's best to read this http://www.eecis.udel.edu/~mills/ntp/html/warp.html
> In simple terms it searches for consensus range of time where all the error
> bars of the various clock overlap and then eliminate clocks who d't agree
> with the consensus.  Of those still "in" it figures out and kind of
> weighted average.
> I think you could do the same thing.  First find the set of 10MHz
> oscillators who are in phase with each other to within some statistical
> limit and then you compute the weighted average phase
> Maybe start with "Clock Select Algorithm".

Free-running oscillators won't be agreeing on phase. As they have 
different independent frequency that we observe without steering, they 
won't agree on phase. NTP on the other hand assumes that the sources is 
sources of UTC, and thus have one way or another been coordinated with 
that unifying phase, then and only then it makes sense to compare the 
phase of the individual sources, but it's not really independent "clocks".

What you can do is to have a set of free-running oscillators, use them 
to build a ensemble paper-clock average, which when weighing them 
against the stability of the ensemble clock the individual stability 
will expose itself, and that ensemble clock will have some weighted 
frequency, and the individual clocks offset from that average frequency 
would be known, and momentary phases and prediction of those phases into 
the future can be done, and well deviations from that gives new info.

However, the average frequency won't be a true frequency, and if the 
clocks is way to similar, there will be considerable "common mode" 
effect from the environment, but it will be some improved stability in 

If you compare it to some outside clock, long-term frequency and phase 
adjustments can be done to align it up to some time-scale, and that 
source would also give a measure of the ensemble stability and precision.

This is similar to how EAL is a ensemble-clock out of about 450 atomic 
clocks, and that a few of them is considered "phase-accurate" is used to 
correct into the TAI clock.


More information about the time-nuts mailing list