Understanding VBAP & VBIP
Those algorithms perform a pairwise (2D) or triplet-wise panning (3D). They select the two or three speakers surrounding the source, and feeds them the source signal with gain differences.
If a source is perfectly aligned with one loudspeaker, then only this speaker is fed with the source signal. In 3D, when a source is aligned with the arc between two speakers, only those two will be used.
With VBAP/VBIP, the speakers surrounding the source are considered as a vector base, and the source direction is used to calculate the gain values. These algorithms only consider the azimuth and elevation of the sources and the speakers.
The main difference between VBAP and VBIP resides in the way the gains evolve with the source position. VBAP was designed to offer an accurate perception of the source position for frequencies below 700 Hz, whereas VBIP optimizes this criterion for frequencies over 700 Hz.
The figure below shows the way the gain associated with two speakers (respectively at azimuth -30° and +30°) evolves in VBAP or VBIP when moving a source between those two speakers.
This algorithm performs an amplitude panning using the two layers surrounding the source. On each of these two layers, it applies a pairwise amplitude panning (VABP 2D) on the two speakers closest to the source. It then applies a level weighting (or crossfade) between the two selected layers, based on the elevation of the source.
If a source is perfectly aligned with one loudspeaker, then only this speaker is fed with the source signal. When a source is positioned between two speakers, only those two will be used. Otherwise, the algorithm will always use the four loudspeakers surrounding the source.
Choosing the right algorithm for Amplitude Panning
Amplitude Panning Optimization
This section offers several ways to deal with the usual optimizations needed for amplitude panning algorithms. These advices applies to VBAP, VBIP, LBAP and Angular 2D, as they all rely on amplitude panning.
Choosing an amplitude panning algorithm over another is mainly driven by the speaker layout.
Apparent Source Width
Amplitude algorithms always use the smallest number of loudspeakers, the source signal can be fed from one, to up to two (2D) or three / four loudspeakers (in 3D), depending on the algorithm.
Therefore, when moving a source accross the setup, its apparent width will vary, going from one speaker to multiple speakers.
On inhomogeneous layouts (i.e. that feature ereas with more space between speakers than others), the source width can also vary from a denser area to a sparser one.
Both of these issues can be solved by using the spread parameter to obtain a constant apparent source width across the setup. However, spreading the source over too many speakers can cause a critical signal coloration.
For very inhomogeneous setups, an other solution is to tend towards a more regular repartition by using fewer speakers in the denser area. This avoids having to spread the source signal over too many speakers.
Try finding the right balance between speakers density, the spread and the delay correction.
Gain and Delay Compensation
Those algorithms were designed to only takes into account the azimuth and elevation of the speakers. Whatever a speaker's distance to the center is, it will be fed with the same source signal level, at a given direction.
In most setups, the loudspeakers distances to the center are not constant, thus altering the setup homogeneity and signal coherence at the center of the venue. Using the ‘GAIN COMPENSATION’ and ‘DELAY COMPENSATION’ can be useful to balance this effect.
When using the spread parameter, adding delay correction can also help deal with the comb filtering coloration (due to sending the same signal to multiple speakers), especially when applying a high spread value.