[hwillis]

> Whoever cracks this particular nut without headphones and manages to create a bubble of 2 cubic meters or so that is silent from outside interference will make a lot of money.

You can tell roughly how far away a sound source is by measuring the curvature of the wavefront, since sound radiates out in a sphere. The farther away you are, the lower the parallax between two sensors.

If you have sensors with 20 cm separation (each ear) then you would need a minimum sample rate of 34.3 kHz to easily correlate sounds within 2 m. With 5 cm separation (to discriminate sounds coming from your sides), you'd need 550 kHz.

Since sound waves are ~10-10,000x lower frequency than that, you need to be able to measure pressure very precisely to distinguish two waves. Rough rule of thumb is 10x, so you'd need up to 1/50,000 precision microphones, or 16 bit. Thats an absolute minimum and you'd probably want ~20 bits.

You might need a DSP and this trick works much better with some high range microphones (electrets are maybe sufficient), but all of that is well within the range of possibility... so I'm kind of surprised nobody has done this yet. I might take a shot at it.

The tricky part is the noise floor, which is only 10-1000x (in terms of voltage from the microphone) below the things youre trying to measure. That means youve got to gather dozens to thousands of samples to determine what should get through, so you need to hear something for up to tenths of a second to decide to let it be audible.

> My 'ideal' soundsucker (sonic black hole) is a ceiling mounted device that projects a cone of silence. One possible way in which it could work is by using a phased array of speakers to 'fake' a larger one. But that ideal will likely never be reached due to limitations in physics, imagine the problem as applied to a wavy surface of water: create a wave pattern that cancels out the wavy surface in one circular area without touching the water directly.

You'd probably want a full network of microphones around your room for this to work, as standing waves will be set up at audio frequencies. It's nontrivial to handle reflections with data from a single spot.

You could do this if you could handle the reflections, but you would need sensors spread around the room (to detect sound before it was already in the cone) and it would only work at a single head height. You'd do it by intercepting the sound, sending opposite waves from above that reached the sound as it propagates across the silenced volume. If you're more than 10s of cm away from the band of silence, the desynchronization will start letting sound through.

It would also be less effective at higher frequencies unless you had head tracking, because the wavelength of sound is not far off from the size of your head.

[jacquesm]

Are you planning an experimental setup? I'd be happy to collaborate on something like this. Agreed about the sensors inside the room, you'd need them in more than one place and quite possibly also at multiple levels to be able to figure out the height of the transmitter. Ideally the sensors can be stuck on flat surfaces and go through a calibration routine where they 'chirp' at each other to figure out their relative positions.

Wireless is hard for this kind of stuff due to latency, wires are ugly but practical and for a first run I would definitely prefer a wired solution over none at all assuming it is even possible. An alternative configuration would be a cylindrical shape that is 'noise free', with speakers radiating outward, microphones would be set in a secondary circle around the first one.

I imagine this sort of setup would use a very large amount of computing power to make it work.

[hwillis]

Unfortunately I'm underwater on work and with my own stuff for the next many months... Electrical engineering is a time consuming side hustle.

Wireless only has latency if you let it! Plenty of bandwidth outside wifi and bluetooth stacks, and communication over infrared is always an option. You are correct that wifi and bluetooth have way too much latency, though- 50 ms is 15 meters of distance at the speed of sound.

Computing power is surprisingly relaxed. The problem of figuring out what waveforms to output from a line of speakers to create a given distribution of sounds is actually just a fourier transform, funnily enough. Creating a volume of sounds is just a 3d fourier transform. Both are unreasonably fast to compute, given how powerful they are.

If you want to know more about how the math works and what it applies to, check out fourier optics. IMO its one of the coolest natural phenomena.

[jacquesm]

Thank you for the pointers. This is all super interesting to me, both because I really am bothered by sound pollution on a daily basis and because I think we have all the parts in place to really deal with this, even if it is anything but a trivial problem. And in a way those are the best problems to be working on.

[Anechoic]

You can tell roughly how far away a sound source is by measuring the curvature of the wavefront, since sound radiates out in a sphere. The farther away you are, the lower the parallax between two sensors.

That only works in the free field. In a small room (defined as dimensions being larger than ~1/6th to 100% of a wavelength) you no longer have waves, the entire room is pressurized. And as you mentioned, modal behavior is also an issue.

If you're trying to measure the parallax of outdoor (free field) waves, that works in a truly free field, but once you have reflections (from, say, the ground) you have to account for that as well.

[hwillis]

You are referring to the reactive near field region, lambda/2pi, which is used for antenna characterization. At the lower end of human hearing (20-40 Hz, using 40) that figure is 1.36 meters. Any room larger than that -literally all of them, unless your ceiling is very low- will propagate a traveling pressure wave instead of just pressurizing.

Further, that figure isn't very helpful for fluid acoustics. Within the equivalent distance you instead would get tons of nonlinear effects, because fluid flow dominates acoustic transmission. In addition to heavily changing transducer loading, things like vorticity also start to dominate. The net effect is that near field issues arise much earlier, at more like a third of a wavelength. Still, only the very lowest audible waves and quite small rooms create non-acoustic behavior.

> once you have reflections (from, say, the ground) you have to account for that as well.

Ish. Only for quite high frequency sounds which change very quickly. Otherwise reflections tend to mostly just overlap with the primary source. For low frequency waves the distance between the microphone and your ears is much smaller than the wavelength, so you don't need to worry about multiple waves very much.

[danielheath]

> Only for quite high frequency sounds which change very quickly.

I have young kids at home. Those are the exact sounds I’d want to cancel!

[jacquesm]

That's still < 4 KHz.