[FFmpeg-devel] [PATCH] libavformat/mxfenc: Allow more bitrates for NTSC IMX50

[Tomas Härdin]

ons 2019-08-21 klockan 12:47 -0700 skrev Baptiste Coudurier:
> Hey guys,
> 
> 
> > On Aug 19, 2019, at 9:54 AM, Thomas Mundt <tmundt75 at gmail.com> wrote:
> > 
> > Am Fr., 16. Aug. 2019 um 23:31 Uhr schrieb Tomas Härdin <tjoppen at acc.umu.se
> > > :
> > > tor 2019-08-15 klockan 13:55 +0200 skrev Thomas Mundt:
> > > > Am Do., 15. Aug. 2019 um 11:01 Uhr schrieb Tomas Härdin <
> > > tjoppen at acc.umu.se
> > > > > :
> > > > > ons 2019-08-14 klockan 22:18 +0200 skrev Thomas Mundt:
> > > > > > 
> > > > > > New patch attached.
> > > > > 
> > > > > Looks OK. I'll push in a few days if no one else has any comments
> > > > > 
> > > > 
> > > > Thanks. Would you mind porting it to branches 4.1 and 4.2?
> > > 
> > > I'm not quite sure what the process is for that. I have confirmed that
> > > the problem exists in 4.1 and 4.2 and that your patch fixes it.
> > > 
> > > I think we also might want to put a note somewhere in the documentation
> > > how to make NTSC IMX50 files.
> > > 
> 
> Yeah, it’s been an issue for quite some time, s356m mentions:
> "When used as a signal source for the type D-10 recording format, the
> bit stream is carried by SDTI-CP, as defined in SMPTE 326M, using
> recommended operating point bit rates as defined in this annex. Other
> bit rates may be used. However, users are cautioned that other system
> design parameters within the studio may not support all bit rates.
> 
> Table A.1 indicates recommended operating points to simplify studio
> operations and to provide users with a tool to be used in designing
> systems."
> 
> Then specifies the exact value of the sequence_header bit_rate_value,
> 50mit/s being “1E848h”, "To be used when compliant with EBU D84 and
> D85"

mpeg12enc.c does this, if I read it correctly. 49999840 / 400 =
124999.6 which gets rounded up to 125000.

> I don’t think it is a good idea to produce files with the wrong
> bit_rate value, and I know for a fact that many file analyzers in use
> today will simply reject the d-10 essence.

Would writing 50000000 in mxfenc for values in the range
[49999840,50000000] make it pass?

> The current code is simply a hack, and IMHO the right solution is to
> put an option to force the ratecontrol to behave as people want for
> NTSC D-10. I’ve submitted a patch for that but Michael wanted it as
> an option.

An option to set the exact desired packet size would be useful for
other things too, so sounds like a good idea.

/Tomas