[ntpwg] Timing Definitions

David L. Mills mills at udel.edu
Sun Feb 18 21:32:44 PST 2007


Kurt,

You raise an important point. There is no such thing as accuracy in the 
product, only a statistical estimation of the resolution, precision and 
other statistics. There is no specific accuracy statement possible in 
this thing. It's like saying, how pretty is your girlfriend.

Dave

Kurt Roeckx wrote:

> On Sun, Feb 18, 2007 at 10:33:38PM +0000, David L. Mills wrote:
>
>> todd,
>>
>> That's exactly what I had in mind, and that's why I phrased the values
>> as I did in the spec. Note also jitter and wander are formally defined
>> and so is delay, dispersion and distance.
>
>
> I've looking for which terms are defined in the draft, and this is what
> I came up with:
> - resolution: Let p be the number of significant bits in the second
> fraction. The clock resolution is defined 2^(-p), in seconds.
> - precision: The clock precision is defined as the running time to 
> read the
> system clock, in seconds. Note that the precision defined in this way can
> be larger or smaller than the resolution. The term rho, representing the
> precision used in the protocol, is the larger of the two.
> - offset: The offset theta represents the maximum-likelihood time 
> offset of
> the server clock relative to the system clock.
> - delay: The delay delta represents the roundtrip delay between the client
> and server.
> - dispersion: The dispersion epsilon represents the maximum error inherent
> in the measurement. It increases at a rate equal to the maximum 
> disciplined
> system clock frequency tolerance PHI, typically 15 PPM.
> - jitter: The jitter psi, defined as the root-mean-square (RMS) average of
> the most recent time offset differences, represents the nominal error in
> estimating theta.
> - wander: The RMS average of past frequency offset differences
> represents the oscillator frequency stability or frequency wander
> OMEGA
> - distance: The synchronization distance LAMBDA = EPSILON + DELTA / 2
> represents the maximum error due all causes.
>
>
> So, what I couldn't find was one for accuracy. The distance is
> probably closest to that.
>
> But I want to give an example, so I know I understand things right. If
> my kernel returns the time in microseconds, but only updates it say at
> 100 Hz. I think my resolution is 10 ms, but would set p to 6, so I get
> .0156 seconds. If I can read the clock say 1000 times a second, I think
> my precision would be 1 ms. But since I'd read the same value 10 times,
> I'd have to set rho to the resolution.
>
>
> Kurt
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://support.ntp.org/pipermail/ntpwg/attachments/20070219/dc803404/attachment.html 


More information about the ntpwg mailing list