[ntp:questions] Accuracy of audio tones via VOIP

Doug Calvert dfc-list at douglasfcalvert.net
Wed Jul 10 06:26:33 UTC 2013

On Tue, Jul 9, 2013 at 10:35 AM, Robert Scott <no-one at notreal.invalid> wrote:
> The NIST has a telephone service that carries the audio of WWV.  When
> this service is accessed by a landline or a cellphone the playback of
> the audio is tightly locked to the phone network system clock.
> Therefore the user can count on fairly accurate audio frequency
> rendering of the NIST standard audio frequency tones.
> But what about accessing this NIST phone service via Skype and other
> VOIP technologies?  In these technologies it is impossible to tightly
> lock the playback to the system clock because of the indeterminacy of
> internet delays.  So how does Skype on my desktop computer work?  It
> can use the local oscillators (system or soundcard) to manage
> playback, but that is not guaranteed to be so good.  Over the long
> haul it is possible for Skype to lock its playback speed to the source
> clock, but that is a slow process, as SNTP applications show.  Over a
> short 3-minute phone call there is very little trimming of the
> playback rate that is possible using only the arrival timing of the
> internet packets.  So I have to wonder if the playback frequency of
> the NIST tones over the telephone via VOIP are potentially in error by
> as much as the local quartz oscillator frequencies?  Does anyone have
> any insight on how Skype and other VOIP systems manage record/playback
> rate synchronization?
> _______________________________________________
> questions mailing list
> questions at lists.ntp.org
> http://lists.ntp.org/listinfo/questions

"Digital modems, such as Digital Subscriber Line (DSL), cable and
wireless modems, cannot synchronize using ACTS. For computers with
Internet access, the Internet Time Service should be used to
synchronize to NIST."[1]

I always interpreted the first sentence as an exclusion of all
non-POTS connection methods (VOIP/DSL/etc). Ignoring that the second
sentence seems like a clear answer.

[1] http://www.nist.gov/pml/div688/grp40/acts.cfm

More information about the questions mailing list