-
Notifications
You must be signed in to change notification settings - Fork 101
Description
Problem to Solve
Currently, a Left-Right symmetric loudspeaker geometry may end up in an AllRAD designed decoder without Left-Right symmetry. As far as I understand, this may happen during the VBAP phase, during the triangulation, when 4 loudspeakers, e.g. {L1, L2, R2, R1}
happens to be coplanar. In such case, the actual triangulation ends up choosing one of the two triangles configuration, e.g. {L1, L2, R2}
and {R1, R2, L1}
, and eventually produces a matrix that have lost the Left-Right symmetry.
Current Workaround
The best work-around, so far, is to add a virtual loudspeaker, Z
, in such a way that the triangulation of {L1, L2, R2, R1,Z}
produces 4 triangles and eventually, to downmix Z
onto L1, L2, R2, R1
.
Proposed Solution
One possible way forward is to adapt VBAP in a way similar to 6.1.2.3 QuadRegion
from ITU-R BS.2127-1, i.e. by allowing quadrilateral to be handled and to create its panning accordingly.
Additional Information
In order to reproduce explicitly the issue, consider for instance an AllRAD design given those 4 side loudspeakers {L1, L2, R2, R1}
: {110, 0}
, {90, 60}
, {-90, 60}
, {-110, 0}
. Add for instance 3 front loudspeaker to complete the sphere coverage: {30, 0}
, {-30,0}
and {0,0}
. Note also that in that case, adding a 8th loudspeaker in e.g. {180, 30}
re-establishes the symmetry of the decoder.