Play Counter Strike 1.6, with full multiplayer, in the browser

Since everybody's headshape and mass is different, it's hard to do binaural audio that works for everyone.

Also, I'm pretty sure the brain uses small movements of the head to know where sounds come from. So you'd have to have head tracking with virtually no latency.

A bunch of speakers in a circle around you don't have any of these issues.

The latency issue would be in the timing of samples between the sound drivers, not from the game to the sound port. The sound data could take 50ms+ to make it through the DAC, but as long as it plays the signal to each driver with enough precision relative to the other (and assuming the response of the driver is sufficient), your brain would get the orientation info. The frames you're seeing are already around 50ms late for most games anyway (unless you're playing the game at hundreds of fps, which is very possible with cs1.6), and mouse latency is between 50-100ms.

Head tracking wouldn't be necessary because you're always looking straight at the screen and the camera is always aligned with your character's head. You're never going to physically turn your head to get a better angle on a sound source, you'll just turn in game.

I’m surprised to hear 50ms for audio processing. I’m a music producer and I can feel it at about 20ms of latency. (Pressing key on keyboard (musical) to hearing sound)

I would have thought video games wouldn’t have more latency than audio production software.

TBH a bunch of speakers seems like it would be much worse.

If you had an infinite matrix of speakers located at every possible point relative to your head, then you could play each sound from the exact speaker representing the correct direction, and get perfect 3D audio. Maybe it would even be sufficient to have a sphere of speakers around you, or even a circle if elevation isn't relevant in most games.

But in practice we don't have any of those. We have 4 or 5 speakers roughly arranged around the player. If one of those 4 or 5 directions happens to be exactly what you need, then great, play the sound from that speaker and you're good. But if not, then what?

The brain decides the direction of sound based primarily on the relative latency between when it is heard in each ear [0]. How do you create a precise time difference when you have 4-5 different speakers each of which can be heard by both ears?

Plus the game doesn't even usually know exactly where the speakers are located relative to the player's head. Exactly how far away are they? Are the front speakers closer than the back?

With headphones, none of this is a problem. The game can precisely control exactly what the person hears, including precisely controlling interaural time difference.

[0] https://en.wikipedia.org/wiki/Interaural_time_difference

My DDR setup -- a Windows gaming PC connected to a pretty standard stereo receiver -- has 70ms total latency between button press and audio. That's more than enough to make the game unplayable, except that of course StepMania lets me configure it to pretend button presses occurred 70ms in the past.

It wouldn't surprise me in todays market but humans are much more sensitive to audio than visual. That's why in audio production ~10ms is the max tolerable limit IIRC.

You're right, but if the head can't move, you'll aquire less spatial information. That's what I was trying to say.

Since most games don't even let you choose dedicated ASIO drivers for output, I'm not surprised at all. I think 50ms for gaming is okay.

My whole point was that the headphones move with the head, so your brain can't deduct spatial information from head movements.

If I mount your head in a fixture and play a sound, you won't be able to place it in three-dimensional space. Only on a 2D plane.

As soon as I let you move your head, you have a much better chance of guessing where the sound comes from.

Yes, I understood your argument. However, your brain does not primarily use head movement to judge location. It primarily uses interaural time difference. Head movement may have some effect (your brain is a neural network trained to use all signals available to it), but ITD is the main factor, at least for most people.

Moreover, it's not clear that 4 point sources of audio can accurately reproduce real-world effects for the purpose of head movements, either.