[time-nuts] Testing frequency using NTP

[Bruce Griffiths]

Tom Van Baak wrote:
> Anyway, I think it would make a fun project to see how well a
> typical PC sound card actually does as a 1PPS or LF phase
> comparator and frequency counter. I wouldn't expect better
> results than a $50 surplus eBay time interval counter, but it
> would be interesting to see just how bad or even how good it
> might be. There are, after all, many timing applications where
> nanosecond and picosecond resolution isn't required.
>
> /tvb
>
>   
Tom

If one has a frequency counter with say 100uV rms equivalent input noise 
and a 1V peak 1kHz sinewave input signal the trigger jitter will be 
about 16ns rms.
Some frequency counter trigger circuits will be noisier than this.

Thus a good sound card can be expected to have somewhat lower noise than 
this when comparing 2 ~ 1kHz sinewave signals.

To improve the low frequency trigger noise of a counter a carefully 
designed zero crossing detector is required to drive the counter.
Such a zero crossing detector will be limited to a narrow range of 
frequencies unless the various filter time constants etc can be adjusted 
to optimise the zero crossing detector to suit the input frequency of 
interest.

Thus the performance of a sound card for comparison low frequency 
sinewave signals will exceed that of a $50 or even $10,000 counter 
unless suitable ZCDs are used to drive the counter inputs.

If the signal zero crossing slew rates are large enough a good counter 
may outperform a sound card even with low frequency signals.

Bruce