Re: [AVTCORE] I-D Action: draft-ietf-avtcore-clksrc-06.txt

[Magnus Westerlund <[e-mail]>]

Hi,

Thanks for writing up this text. I think this are making clear the
calculations that are needed and the difference between NTP and TAI
usage. So I am happy.

Cheers

Magnus

On 2013-09-27 01:25, Kevin Gross wrote:
> Sorry again. I missed this part of the thread. The implementations we've
> done so far all use TAI timestamp reference clocks. I've now worked
> through the details for an NTP reference.
> 
> The era of accurate timekeeping really only started in 1972 (and is
> well-defined retroactively to 1970). Not everyone agrees on
> what happened before then so I think we do need to give the RFC 868
> figure to ensure interoperability when using NTP.
> 
> Here's proposed replacement text with all the gory details:
> 
>     The signalling optionally indicates a media clock offset value.
>      The offset indicates the RTP timestamp value at the epoch (time of
>     origin) of the reference clock. To use the offset, implementations
>     need to compute RTP timestamps from reference clocks. To simplify
>     these calculations, streams utilizing offset signaling SHOULD use a
>     TAI timestamp reference clock to avoid complications introduced by
>     leap seconds. See [draft-ietf-avtcore-leap-second] for further
>     discussion of leap-second issues in timestamp reference clocks.
> 
>     To compute the RTP timestamp against an IEEE 1588 (TAI-based)
>     reference, the time elapsed between the 00:00:00 1 January IEEE 1588
>     epoch and the current time must be computed. Between the epoch and 1
>     January 2013, there were 15,706 days (including extra days during
>     leap years). Since there are no leap seconds in a TAI reference,
>     there are exactly 86,400 seconds during each of these days or a
>     total of 1,356,998,400 seconds from the epoch to 00:00:00 1 January
>     2013. A 90 kHz RTP clock for a video stream would have advanced
>     122,129,856,000,000 units over this period. With a signaled offset
>     of 0, the RTP clock value modulo the 32-bit unsigned representation
>     in the RTP header would have been 2,460,938,240 at 00:00:00 1
>     January 2013. If an offset of 23,465 had been signaled, the clock
>     value would have been 2,460,961,705.
> 
>     In order to use an NTP reference, the actual time elapsed between
>     the 00:00:00, 1 January 1900 NTP epoch to the current time must be
>     computed. 2,208,988,800 seconds elapsed between the NTP epoch and
>     00:00:00 1 January 1970 [RFC 868]. Between the beginning of 1970 and
>     2013, there were 15,706 days elapsed (including extra days during
>     leap years) and 25 leap seconds inserted. There is therefore a total
>     of 3,565,987,225 seconds from the NTP epoch to 00:00:00 1 January
>     2013. A 90 kHz RTP clock for a video stream would have advanced
>     320,938,850,250,000 units over this period. With a signaled offset
>     of 0, the RTP clock value modulo the 32-bit unsigned representation
>     would have been 1,714,023,696 at 00:00:00 1 January 2013.
> 
>     If no offset is signalled, the offset can be inferred at the
>     receiver by examining RTCP sender reports which contain NTP and RTP
>     timestamps which combined define a mapping. The NTP/RTP
>     timestamp mapping provided by RTCP SRs takes precedence over that
>     provided by the SDP, however the media clock rate implied by the SRs
>     MUST be consistent with the rate announced in the SDP.
> 
> 
>




-- 

Magnus Westerlund

----------------------------------------------------------------------
Multimedia Technologies, Ericsson Research EAB/TVM
----------------------------------------------------------------------
Ericsson AB                | Phone  [phone]
Färögatan 6                | Mobile [phone]
SE-164 80 Stockholm, Sweden| mailto: [e-mail]
----------------------------------------------------------------------