Ambisonics support?

[vectorwarrior]

What I don't understand from the webpage is if it is always encoded in 4 channels, or more? And what is the point of all this anyways? I must say it's all a bit confusing - e.g. why isn't this stored as normal channels instead of the "encoded format"?

The key is to realizing that you are used to thinking in terms of channels (e.g. stereo, 5.1), but binaural audio is a representation of a whole 360 degree sphere soundstage, which is then decoded into the required channel output. It is channel agnostic, a single ambisonic output can be decoded into ANY channel output, it's universal, which is what makes it so appealing.

The minimum number of channels for ambisonics is 4. Increasing the number of channels creates more accurate positioning in the sphere, that is all. For instance, 4 channel (called 1st order) ambisonic audio is not accurately positioned at all, sounds sort of smear into surrounding areas, so you don't get pinpoint positioning. To get more accurate positioning you need to use higher order ambisonics:

1st order - 4 channel
2nd order - 9 channel
3rd order - 16 channel
etc
here is an image showing how the numbering works: https://www.audiokinetic.com/images/201 ... dering.png


So, to answer your question: "why isn't this stored as normal channels instead of the encoded format"
It's because an ambisonic output can be used to translate to ANY output signal that the user has. No need to make separate 2.0, 5.1, 7.1, 7.1.4 mixes of things, one ambisonic mix can work for everything AND be future proof for any future channel output. Also, because ambisonic represents a sphere around the listener, it's also super useful for VR, as you can just 'point' the listener towards a specific direction of sound in the sphere... all with only 4 channels (or more if you want more positional accuracy).

Hope that helps make a bit more sense. Again, the reason this is interesting to me personally is that I work in game audio and 3d audio (which includes, but isn't limited to ambisonics) is going to be a MAJOR push for the soon to be released next generation of consoles from MS and Sony.

[mgw38]

What I don't understand from the webpage is if it is always encoded in 4 channels, or more? And what is the point of all this anyways? I must say it's all a bit confusing - e.g. why isn't this stored as normal channels instead of the "encoded format"?

Ambisonics (Ambisonic B-format to be correct) is a speaker layout agnostic audio format. It encodes the entire soundfield with respect to the head location and independent from the location of the speakers. You need that for audio work in immersive media formats (games, cinematic VR, etc). You can also think of it as an exchange format. Once you have it in Ambisonics you can accurately reproduce it with any speaker layout.

Edit: btw, the “encoding” step is essentially a spherical Fourier transform. In a regular Fourier transform you decompose a soundwave into its fundamental wave components. For Ambisonics you decompose a spherical soundfield into its fundamental spherical wave components.

[MeldaProduction]

Thanks for the info. Makes more sense now. BUT technically something is doing some decoding/encoding, which makes me think that it's not exactly true. If it would be already decoded, then it would still be usable for all formats, just after encoding. So it seems to me, that the whole point is to "render" encoded format, so that the target systems are having less workload, which is reasonable. BUT it would seem more logical to work with not-encoded format until the last step. Thing of a limiter at the end of the chain, that one would be the encoder.

Also mgw38 - I quite don't follow your reasoning about FFT. It seems to me that the encoding is just a linear projection matrix, I don't see a reason for any hardcore spectral stuff , or am I missing something?