[ntp:questions] Re: Interpretation of frequency error / drift_comp

Maarten Wiltink maarten at kittensandcats.net
Thu Feb 23 17:28:13 UTC 2006

"Daniel Kabs" <daniel.kabs at gmx.de> wrote in message
news:43fdd2d6$0$496$9b4e6d93 at newsread4.arcor-online.net...
> Maarten Wiltink wrote:

>> Colour me unimpressed. Get yourself a newsreader (NNTP client) and shake
>> off both your ignorance and your dependence on Google.
> You don't seem to like google! What's wrong with it?

Google Grooups should have remained a Usenet archive and search engine.
I appreciate those two functions, but they made a mess of the posts
submitted through them.

> I am not ignorant. Actually, I used a newsreader (Mozilla News) to post
> my article. And I searched the newsserver but it holds only three month
> worth of posting and there was nothing relevant found.

Okay, I missed that. A million other people think Google Groups invented
its functionality and have never even heard of Usenet. I was too quick to
class you as one of them.

> Are you now willing to share some of your knowledge about the
> "frequency" variable, please?

Sure. It's (the current guess of) the difference between clock speed and
real time speed, in PPM relative to the internal clock.

For the sake of example, consider a timer that ticks 1000 times per
second. On every tick, 1 ms would be added to the internal time,
expressed as a (large) number of quanta. The size of a quantum is
determined by the resolution of the clock. NTP timestamps have a
resolution of 2**32 quanta per second, an OS's resolution is generally
worse, but still better than the precision of the clock.

The frequency offset is scaled from PPM to clock precision and added
to the increment. So a value of +40 would result in 1.040 ms being
added to the internal time on every tick, compensating for a timer that
is slightly slow - or for lost ticks. I expect there is some quantisation
noise if 1.040 is not an even number of quanta, or actually I would
expect that not all frequency offsets are reachable.

The above is my understanding of it, some of it based on hearsay (here)
and some of it conjecture. I eagerly await corrections to my own
frequency. No disciplining, please.

Maarten Wiltink

More information about the questions mailing list