[ntp:questions] Dispaying nanoseconds

David L. Mills mills at udel.edu
Fri Feb 16 17:54:53 UTC 2007


Eugen,

Remember, the filegen data may appear to the nanosecond, but the data 
have already been spiked by the random bitstring below the precision factor.

The ./scripts directory in the distribution has some scripts to massage 
the filegen files, but they are 20 years old. I have a set I use myself 
to produce Matlab friendly files, but these are rather crude. When I 
whip them to public shape, I will park in the ./scripts collection.

Dave

Eugen COCA wrote:

> On Feb 15, 2:35 am, "David L. Mills" <m... at udel.edu> wrote:
> 
>>David,
>>
>>Note that the nonsignificant bits at the low end of the fraction word
>>are purposly filled with a random bitstring. My machine has a precisioni
>>-19, so there are 19 meaningful bits in the fraction along with a
>>carefully manufactured 13 bits of fuzz. This is done both to make it
>>harder to predict timestamps and smooth out tiny wiggles due to roundoff
>>and bias. It doesn't make sense to split the microseconds in the
>>readouts. For what it's worth, the statistics files display offsets to
>>the nanosecond.
>>
>>Dave
>>
> 
> 
> Dave,
> 
> clear for me too. I'll use the information in loopstats file to
> display the offsets with nanoseconds precision.
> 
> Are there any .pl or .php scripts already tested to display the
> offsets from the loppstats file ?
> 
> Thanks !
> 




More information about the questions mailing list