[time-nuts] Testing frequency using NTP

Bruce Griffiths bruce.griffiths at xtra.co.nz
Thu Oct 2 17:29:21 EDT 2008

Tom Van Baak wrote:
>> When using a sound card for frequency comparisons the zero crossing time 
>> stamp resolution is improved if the slew rate of the input signal is 
>> slow enough that several (3 or more) samples are taken in the vicinity 
>> of the zero crossing. One can then make use of  WKS (Whittaker, 
>> Kotelnikov, Shannon) interpolation to accurately calculate a high 
>> resolution zero crossing time stamp.
> Clever.
>> The instability of the sound card LO isnt completely cancelled if the 
>> zero crossings of the the 2 signals aren't coincident.
> That seems right for absolute event timing with a stereo sound card
> but I think for a frequency measurement the delay, if any, between
> channels would also cancel out (as long as the delay itself stays
> relatively fixed). We'll know for sure when someone actually tries it.
> A sound card timing experiment, vaguely related to this, is here:
> http://www.leapsecond.com/pages/sound-1pps/
> There I used a sound card to generate a 1 PPS and measured its
> ADEV. Note that this is unrelated to NTP (NTP disciplines the CPU
> or system bus clock, which is typically not the same oscillator as
> the sound card clock).
> Win32 source code: http://www.leapsecond.com/tools/1hz.c
> /tvb
> http://www.LeapSecond.com


The short term phase instabilities of the sound card LO that occur 
during the finite interval between the zero crossing times for one 
channel and the zero crossing time for the other channel can be 
significant. Also even with a zero noise sound card LO the 2 zero 
crossings occur at different times so that some correction for the 
additional phase instability of the 2 sources over this time interval is 
needed in exactly the same way as for a dual mixer time difference 
system when the 2 zero crossings aren't nearly coincident in time.
Such variations are substantially reduced when the input frequencies are 

If the input signal SNR is sufficiently high the zero crossing time 
stamp jitter noise may be as low as a few ns for a 1khz input frequency.


More information about the time-nuts mailing list