[ntp:questions] best gps receiver for time synchronization
unruh-spam at physics.ubc.ca
Wed May 13 17:25:42 UTC 2009
"David J Taylor" <david-taylor at blueyonder.not-this-part.nor-this.co.uk.invalid> writes:
>David Woolley wrote:
>> David J Taylor wrote:
>>> - with operating system calls, the main clock can be read with a
>>> precision of 1 millisecond (although the ticks may be only 15
>>> milliseconds). There are higher resolution counters which can also
>>> be read.
>> The architectural resolution is much higher than this. I think you
>> are seeing actual 1ms clock ticks, as the result of ntpd forcing the
>> use of multi-media timer mode.
Since the same machine can run Linux or BSD whose resolution is usec or
nsec, yes, the hardware can do better. The question is how good is the
software in the kernel. If I do a timestamp on an event, how accurate is
Is it msec? Is it 15msec?
>David, yes the hardware is better, but I was asked what the system clock
>call was and the structure it returns only goes down to milliseconds.
What the structure returns need have nothing to do with the accuracy.
Linux system calls can return a structure with nsec but the accuracy of
the linux clock is no better than usec.
>I'm not "seeing" anything.
More information about the questions