AmbiXchange Discussion Minutes

Remarks on the proposed transfer format for Ambisonics during the Ambisonic X-change Meeting by Alois Sontacchi

Tuesday, 20th of May 2008, IEM, CUBE from 13 to 16 o’clock

Discussion Panel: Andrea Capra, Fons Adriaensen, Jan Schacher, Martin Neukom Oliver Frei, Martin Rumori, Klaus Hollinetz, Thomas Musil, Franz Zotter, Winfried Ritsch, Johannes Zmölnig, Alois Sontacchi

Minutes: Alois Sontacchi

Within the first step remarks form each discussion panel member concerning the presented exchange format has been selected.

Hereafter the given remarks and recommendations are summarized:

  • Jan Schacher: Human readable character should be used for the additional information within the file header. Separation between meta data and audio-content is requested to provide the possibility to use standard audio editors to modify exchanged sound files.
  • Fons Adriaensen: An acceptable file format related to a larger community is primary required. Therefore usage of a standard audio format which is loadable by standard software and a separate file for metadata is advisable. There should be only one format in the file. This should be done by the author. Normalization etc. where you have free choice should be left open. - Franz: I remember the opposite: We shall define and declare the normalization used. i.e.: it shall be well defined instead of left open Proposed format with the weighting matrix might be easily abused (not only commercial misusage)
  • Andrea Capra: Agreed on the addressed remarks by F.A.
  • Martin Neukom: No further remarks.
  • Winfried Ritsch: The proposed format with the matrix provides the option of storing performances consisting of statically and dynamically rendered sound sources. The number of channels for exchange/transmission can be reduced largely and losslessly within this format. Any modification concerning the proposed format should provide the independence of any exchange container-format.
  • Johannes Zmölnig: The repeated headers within the proposed format assist the streaming aspect.
  • Franz Zotter: If there is no further information within the audio file the numbering of the transmission channels have to be considered and agreed on. Furthermore the choice of normalization has to be discussed, too i.e. full normalization (overall sum is equal one), big normalization (the max. value in one direction is ceiled by one) Normalization related to Furse-Malham-Set, Condon-Shortly phase
  • Thomas Musil: The orientation of the coordinate system and the counting direction of azimuth and elevation have to be defined.
  • Franz Zotter + Fons Adriaensen: A reference radius (proposed by Hannes Pomberger) can be considered to allow for distance coding with minimized numerical problems. This information only refers to the audio material, not to playback.

Within the discussion:

  • Compatibility to older material shall can be neglected in the first step.
  • For dissemination of the exchange format, conversion tools and standard playback tools will be accessible.
  • Fons Adriaensen + Hannes Zmölnig: Redundant information should be avoided – make it simple as possible.
  • Franz Zotter + Fons Adriaensen + Hannes Zmölnig: Concerning the numbering of channels a fixed order of channels will be helpful. The row-wise listing based on the triangular arrangement of spherical harmonics (i.e. W=1,X=2,Z=3,Y=4, etc) might be intuitive.
  • Based on the numbering: n_m =nn+n+m+1 each channel number and placement is ideal defined (whereby n_m is a positive integer; could alternatively start from 0).
  • Alois Sontacchi + Franz Zotter + Fons Adriaensen: Based on the proposed numbering of the channels the order of the transmitted channel could be easily recalculated: e.g. channel 13: ceil(sqrt(13)-1)=3. The cut-off order N is determined from the maximum channel index in the same way.
  • Fons Adriaensen + Hannes Zmölnig: Within the format: each channel is labeled with a string identifier: "bmft". For additional channels, other identifiers have to be used (e.g. "sub"). Generally the channels must appear in ascending order. Channels may be left out (e.g. for mixed-order formats).
  • Fons Adriaensen: Concerning mixed orders: there is a convenient example given within the Furse Malham-Set.
  • Franz Zotter: Inconsistencies with the definition of mixed-order horizontal and vertical resolutions have to be well-defined. (or fixed).
  • Fons Adriaensen + Hannes Zmoelnig: The term B-format does not restrict the order to one. Therefore it could be used as well-accepted label for the Ambisonics channels.
  • Fons Adriaensen: Allowing onls for a finite set of layouts (2D, hemisphere, 3D, mixed-order), the layout could be reconstructed from the pattern of given "bfmt" indices.
  • Thomas Musil demonstrates 24ch half-spherical harmonics synthesis with symmetry reduced 5th order set.

First discussion results:

  • The human readability of any additional metadata is of great importance.
  • The orientation of the coordinate system should be defined as follows.

The x-axis faces forward, the y-axis to the left and the z-axis to the top. The azimuth angle is counted positively to the left (anti clock-wise). The elevation is zero for the horizontal plane and counted positively facing upwards.