[ntpwg] Timing Definitions

David L. Mills mills at udel.edu
Sun Feb 18 10:22:44 PST 2007


todd,

We are on a tangent here.

The definitions I use are deliberately measurable and unambiguous. They 
are also consistent with the intricately crafted statistics budget in 
the spec, the UDel report and the book. There is nothing specific to 
present or future architecture or product. The standard will not change 
with the introduction of new architecture or product. That's the 
rigorous intent of the specification in the first place.

If you find something in the budget in error, very please honk. If you 
find something inconsistent in the name or interpretation of the error 
statistics, very please honk. Whatever honk you find must be rock hard, 
measurable and consistent with analysis.

Dave

todd glassey wrote:

> Dave -
> ----- Original Message -----
> From: David L. Mills
> Cc: ntpwg at ntp.isc.org
> Sent: Friday, February 16, 2007 8:59 AM
> Subject: Re: [ntpwg] Timing Definitions
>
>
> Greg,
>
> The terms resolution and precisions are defined very narrowly in the 
> NTP documents. Resolution is the reciprocal of the system clock 
> frequency, usually one nanosecond with modern hardware and operating 
> systems (timespec).
>
> With the new architectures however the resolution will change. The 
> question becomes one of how to notate that and where it should be 
> disclosed.
>
>
> Precision is defined as the time to read the system clock, currently 
> around 500 ns to 1000 ns with modern hardware and opeating systems. It 
> is important to make that distinction when calibrating the residual 
> errors for accuaracy assessment.
>
> These also are specific to current and previously available systems 
> but the standards will change and since there is no process for 
> characterizing a NTP Server yet this is a reasonable commentary of Greg's.
>
> Symmetricom and all of the HW based Time Systems providers need these 
> data points for creating characterizations of their systems and the 
> performance and so these additions are reasonable and warranted for 
> supporting the commercial users of NTP.
>
> The specification document is very clear about this.
>
> Dave
>
> Greg Dowd wrote:
>
> Some standard uses for reference with ntp:
>
> precision: This defines the smallest measurement which can actually be
> made. It helps to define the number of significant bits in a
> measurement.
> resolution: This defines the total number of bits available to express
> a measurement. It tells you nothing about the value of any of those
> bits.
> accuracy: This defines how well a measurement agrees with a timescale
> (e.g., UTC).
>
>
>
> Greg Dowd
> gdowd at symmetricom dot com (antispam format)
> Symmetricom, Inc.
> www.symmetricom.com
> "The current implementation is non-obvious and may need to be improved."
>
>
>
>
>
>
>
> ----------------------------------------------------------------------------
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at support.ntp.org
> https://support.ntp.org/mailman/listinfo/ntpwg
>
>
>
>
> ------------------------------------------------------------------------------
>
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at support.ntp.org
> https://support.ntp.org/mailman/listinfo/ntpwg




More information about the ntpwg mailing list