[ntp:questions] Re: A question abouth the Allan Intercept

David L. Mills mills at udel.edu
Fri Nov 25 19:33:34 UTC 2005


See: Mills, D.L. Adaptive hybrid clock discipline algorithm for the 
Network Time Protocol. IEEE/ACM Trans. Networking 6, 5 (October 1998), 
505-514. Further details are on the NTP project page at 


Mauro Fiacco wrote:
> Hi to All,
> I have a question about the Allan Intercept. I understand how it is used
> in the NTP code (i.e. tuning the NHPFL) and I also understand it's use
> on characterise oscillators (wander-jitter threshold... basically, its
> stability). 
> However I do not understand, if the value of CLOCK_ALLAN, used in the
> NTP, represents an approximation of the intercept of the oscillator (the
> component under the hood) or more generally of the time discipline
> system (which includes processor interrupt latencies, network cards
> jitter, clock jitter... temperature variations... platform resolution
> and precision... more?)
> In the first case, to measure it, I would need a (e.g.) frequency counter,
> in the second case, I would need a pretty accurate NTP server and
> analysing the NTP rawstats to derive the Allan Intercept.
> I believe that the Allan Intercept tries to model the time discipline
> system not of the oscillator... am I correct?
> Thank you,
> Regards.
> Mauro

More information about the questions mailing list