CODEC WG session IETF 98 - Chicago March 30, 2017 9:00-10:00 Chairs: Tim Terriberry, Mo Zanaty Area Director: Ben Campbell Note Taker: Jamieson Brettle Jabber Scribe: Ben Campbell Agenda: Adminstrative, Ambisonics Draft Ambisonics in Ogg Opus (Andrew Allen) https://tools.ietf.org/html/draft-ietf-codec-ambisonics-02 https://www.ietf.org/proceedings/98/slides/slides-98-codec-ambisonics-02.pdf Reviewed basics of ambisonics Ogg Opus channel mapping for ambisonics Channel Mapping 2 - More bits are allocted for the omnidirectional channel Channel Mapping 3 - Offers a mixing/demixing transform matrix for encode/decode Mapping 2 uses Mapping 1's mapping table Question: Q: When we get to HOA with a more channels, can we make it more efficient? A: Yes by defining the 'A' matrix to put your channel count into a much more efficient usage. Can use few channels after encoding then demux to the original number of channels. Q: Changes to the draft to account for channel mapping 255. 255 use the same channel mapping table that channel mapping 1 has. This would work for mapping 2, but problematic for 3. A: There was discussion how to address this, and currently the mapping matrix is removed from the current draft. The proposal could have the mapping table related to the streams that were encoded. Q: Can we update RFC 7845 (Ogg Opus spec) to clarify handling of unknown channel mappings. A: Mark Harris says he will send text to update the draft to the list. Q: The matrix should be in big/little endian. Need to specify. Need to specify column or row first A: We have specified it is column wise. We will specify endianness Q: I think the formula in fig 2 in the draft is incorrect. The order and degree based on K & N. A: This should be from the ambic spec, it will be updated. Will double check, but believe the formula listed is correct Q: Explain how the presence or absence of Non-diegetic audio is signaled A: By the channel count - N+1^2, or N+1^2 + 2 Q: In channel mapping 2 what happens with non-diegetic A: The final 2 channels. Q: Could the non-diegetic be 2 mono? A: Yes you could on the decoder. But on the encoder they're paired. Q: Can we do mixed order ambisonics with mapping 3 and non-diegetic A: It would be part of the mixing matrix Q: In Mapping 3 can you just say I want 8 channels? Which would make it difficult to use non-diegetic and partial order? A: You would need to submit channels zeroed out. Need to explicitly call this out in the document