Location and duplication of inner codec AudioSampleEntries about iamf

[cconcolato]

If you look at the AudioSampleEntry class, it defines:

• channel count (mandatory)
• sample rate (mandatory)
• channel layout (optional)
• DRC (optional)
• Downmix (optional)

As discussed during the call, we could specify that the channel count is the maximum channel count. For sample rate, I would imagine that all substreams have the same sample rate, as otherwise time alignment would be a problem (except in some cases). For all optional information, we could say that they should not be present or as it was suggested that they would be overridden by similar IAC information?

[sunghee-hwang]

We don't need the optional fields like OPUS-ISOBMFF.

[tdaede]

We should investigate what downmixing options (if any) are representable with existing mappings and ISOBMFF boxes.

[sunghee-hwang]

• RFC 7845 (Ogg Opus, CMF = 1) and RFC 8486 (Ambisonics, CMF = 2 or 3) provides stereo downmix matrices as spec text (not delivered)
• ISOBMFF defines 'dmix' box for downmix matricies as a child of Audio sample entry. Any M x N downmix matries and any coefficients can be delivered via this box, where, M = targetChannelCount and N = baseChannelCount. But N is not defined in this box but is defined in ChannelLayout box ('chnl'). As summarized, any M x N downmix matrices and any coefficients can be delivered by using two optional fields ChannelLayout('chnl') and DownMixInstructions('dmix') of Audio sample entry. However, my understanding is that M is limited to 8 because 'targetlayout' of 'dmix' is based on ISO/IEC 23091-3 which defines only up to 7.1ch. (not define height channels).

[cconcolato]

As discussed, we have an agreement to specify that only codec specific info is stored (no loop, no override). We might reconsider if new use cases are brought forward.

[tdaede]

The no loop approach is in the spec.