4 ms·

That makes sense, since Atmos is a discrete set of channels designed for speakers in specific locations around the listener. The alternatives are: - record a

That makes sense, since Atmos is a discrete set of channels designed for speakers in specific locations around the listener. The alternatives are: - record a real sound space using a binaural mic head, then simply use 2-channel output. Head tracking not supported, HRTF is physical not modeled/virtual. - build a 3D virtual sound space. Apply the HRTF in post-production with output to 2 headphone channels. Head tracking can't be supported, but there's a lot more flexibility. - encode a potentially large number of discrete channels and their positions in 3D space (including reverb and room dynamics) and let the listener's system apply the HRTF. Would support head tracking, but the audio format would be super complex, as would be the rendering software. Reverb calculations in particular could eat up a lot of CPU. - render in post-production to a discrete number of channels that will be placed in known positions around the listener. This supports theaters and home surround setups, can be implemented in headphones using an HRTF, and would support head tracking - all at the same time. The only drawback is that in headphones, it sounds like speakers around you. Because that's exactly what the HRTF is doing with the input signal. It's easy to see why the last option is what we've landed on. It's the most flexible for the widest number of scenarios, it's a straightforward standard, and it's not too hard to emulate using an HRTF.

Shish2k 6y ago
> encode a potentially large number of discrete channels and their positions in 3D space This does seem like a lot - but at the same time, it doesn’t seem like much compared to the advances in graphics over the past decades? I feel like if audio had kept up with video, we should have an equivalent to Physically Based Rendering / raytracing by now...
- jasonwatkinspdx 6y ago
  I straight up blame Creative. They stagnated any innovators/challengers back in the sound card era, set the outline of how game audio apis work, then became irrelevant. Aureal was shipping technology superior to what we have today 25 years ago. It'd be trivial to do what they were doing in software today. I really wish someone like MS would step up here and show some leadership by just shipping a spatial audio API in directx that's closer to what's actually possible.
Cthulhu_ 6y ago
> That makes sense, since Atmos is a discrete set of channels designed for speakers in specific locations around the listener. I think that's the issue as well; in a room with speakers all around, the sound has space to move and bounce about and merge so the individual speakers are less obvious, but in headphones it's pretty much a straight line into your ear. Might just need a bit of fine tuning and partial merging of multiple channels / directions.
jasonwatkinspdx 6y ago
So, there's another possibility, the road we didn't go down: use a sound field representation based on spherical harmonics. You'd get something roughly the quality of Atmos with just 4 channels. It's a shame cinema didn't go this route, as trivial circuits can fold it down to quad or stereo, but more elaborate setups can render it specific to any arbitrary speaker geometry that's desired.
angus-prune 6y ago
I'm afraid that you're wrong about how Atmos works. Previous Dolby standards worked as you describe, but Atmos is most similar to your alernative 2. Atmos has up to 128 audio tracks which are each positioned and moved within a virtual 3D space. The Atmos reciever then processes this 3D space and outputs a number of channels tailored to your specific speaker set up, wherever they are positioned (and however many* speakers you have). I don't know if Atmos for headphones currently includes HRTF processing, but there is no technical reason it couldn't. *(I believe state of the art limit for home systems is 64, but this is a hardware issue, rather an a limit of the format)

Shish2k 6y ago

> encode a potentially large number of discrete channels and their positions in 3D space This does seem like a lot - but at the same time, it doesn’t seem like much compared to the advances in graphics over the past decades? I feel like if audio had kept up with video, we should have an equivalent to Physically Based Rendering / raytracing by now...

jasonwatkinspdx 6y ago
I straight up blame Creative. They stagnated any innovators/challengers back in the sound card era, set the outline of how game audio apis work, then became irrelevant. Aureal was shipping technology superior to what we have today 25 years ago. It'd be trivial to do what they were doing in software today. I really wish someone like MS would step up here and show some leadership by just shipping a spatial audio API in directx that's closer to what's actually possible.

Cthulhu_ 6y ago

> That makes sense, since Atmos is a discrete set of channels designed for speakers in specific locations around the listener. I think that's the issue as well; in a room with speakers all around, the sound has space to move and bounce about and merge so the individual speakers are less obvious, but in headphones it's pretty much a straight line into your ear. Might just need a bit of fine tuning and partial merging of multiple channels / directions.

jasonwatkinspdx 6y ago

So, there's another possibility, the road we didn't go down: use a sound field representation based on spherical harmonics. You'd get something roughly the quality of Atmos with just 4 channels. It's a shame cinema didn't go this route, as trivial circuits can fold it down to quad or stereo, but more elaborate setups can render it specific to any arbitrary speaker geometry that's desired.

angus-prune 6y ago

I'm afraid that you're wrong about how Atmos works. Previous Dolby standards worked as you describe, but Atmos is most similar to your alernative 2. Atmos has up to 128 audio tracks which are each positioned and moved within a virtual 3D space. The Atmos reciever then processes this 3D space and outputs a number of channels tailored to your specific speaker set up, wherever they are positioned (and however many* speakers you have). I don't know if Atmos for headphones currently includes HRTF processing, but there is no technical reason it couldn't. *(I believe state of the art limit for home systems is 64, but this is a hardware issue, rather an a limit of the format)