[FFmpeg-user] XDCAMHD422 MXF files - MXF IndexTableSegments have the same InstanceUIDs (not unique)

Marton Balint cus at passwd.hu
Mon Mar 14 20:49:19 EET 2022



On Mon, 14 Mar 2022, Christoph Gerstbauer wrote:

> Hello everyone!
>
> I am currently testing with XDCAMHD422 mxfs generated by ffmpeg 4.4 and 5.0.
> I am testing these MXFs against the SMPTE ST377 Standard.
> My testtool is the MXF Analyser from the IRT which can test the MXF against 
> Standard Compliance SMPTE ST377 MXF.
>
> The MXF files generated by FFMPEG are almost "green" to ST377 but one single 
> issue comes up all 10 seconds in the file:
> "This Index Table Segment has the same Instance ID value of
> ad.ab.44.24.2f.25.4d.c7.92.ff.29.bd.00.0f.00.00 as the Index Table Segment
> located at byte offset 78605312. The current Index Table
> Segment is not a repetition of the earlier Index Table Segment"

This is a known issue with the MXF muxer of ffmpeg.

> My questions:
> 1.) Is it possible to control via syntax to write only ONE IndexTableSegment 
> (Essence Container) into a MXF file? And does this make sence for XDCAMHD422 
> mxfs (GOP based video, GOP=12)?

It is generally preferred to write an index segement to every partition 
and duplicate all the index segments in the footer, so a smart application 
does not have to seek through all the paritions to get an index of the 
whole file. Raw2Bmx can do that for example, so you should consider 
encoding to mpeg2 with ffmpeg, and muxing with raw2bmx, you will have more 
compliant MXF output.

> 2.) OR is it possible to fix the issue in a way that IF more than ONE Index 
> Table Segments are written, then every Index Table Segment has a unique 
> Instance UID to be SMPTE ST377 standard conform?

This would be desirable, I will submit a patch to ffmpeg-devel, does not 
seem too complicated. Maybe can you test it?

Thanks,
Marton


More information about the ffmpeg-user mailing list