[time-nuts] Testing frequency using NTP
Bruce Griffiths
bruce.griffiths at xtra.co.nz
Thu Oct 2 06:58:55 EDT 2008
Ulrich Bangert wrote:
> Steve,
>
>
>> Hmmm... If I measured a 10MHz oscillator for a 1/10 second, I
>> could achieve, at best, 1ppm accuracy. Now my measuring
>> system has a non accumulating error in the ms range, say <1s,
>> so this would be totally unworkable. If I sampled for 1s,
>> best would be .1ppm accuracy...
>>
>
> this is a common misbelieve but simply not true. You need not buy the
> most advanced stuff. Even a moderate surplus counter like the RACAL DANA
> 1991/1992/1996 will get you < 1ppb (b, not m!) resolution @ 1 s. Since
> more than 30 years counters do no more simply "count" but work with lots
> clever electronic tricks inside!
>
> 73s de Ulrich, DF6JB
>
>
This belief also appears to widespread in the musical instrument making
community.
ie a counter is useless for measuring the frequency of a plucked string
because the sound doesnt last long enough for accurate measurement.
With a reciprocal counter and time interval interpolators its is
possible to make more accurate measurements than a naive analysis would
suggest.
However careful design of the signal conditioning is required to ensure
that the counter input circuit noise doesn't degrade the measurement noise.
In the particular case of a source like a plucked string it is perhaps
more effective and informative to use a sound card for the measurement.
For precision frequency comparisons NIST divide the frequencies to be
compared down to about 1Hz and time stamp the resultant signal
transitions and the leading edge of the PPS output of a GPS timing
receiver with a resolution of a few picosec. The resultant timestamps
are then processed (along with measurements taken with a similar system
that time stamps the transitions of a 1Hz signal s derived from one or
more of their frequency standards) to derive the frequencies of the
various sources.
Bruce
More information about the time-nuts
mailing list