Hacker News Reader

Comment by Animats | Hacker News Reader

There should be an app for this.

To get direction and location, you need acoustic samples from multiple widely-spaced locations, all synchronized. Human ears are not wide enough apart to directionalize low frequencies, but with a baseline of tens or hundreds of meters from separate cell phones, it's not hard. Run a correlator, line up the samples, compute the hyperbolas of constant time offset, (like GPS and LORAN) and find the target.

This needs an app with timing accurate to a millisecond or so. Can you get that from cell phones? The built-in clock synchronization isn't that good. The GPS receiver has more accurate time, but you may not be able to get at that.

View on HN

Nit: Human sound localization is 'good' from about 200Hz to about 1600Hz. Other posters have the sound at ~440Hz, so humans should be able to localize the sound fairly well.

https://en.wikipedia.org/wiki/Sound_localization

In this case, I think they're mostly referring to "The Hum"[0] that others have been talking about, since the article is about the source of this noise. When it's lower frequency and not this easy to locate, it would be great if there were some other way to go about it.

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

I agree. For several years I have been looking on and off for a layman way to locate a source of a sound, but haven't found anything.

A solution like this could resolve many of the lingering "hum" cases. Often they are only heard by a few people, and those in charge of environmental control don't tend to take it seriously because "they don't hear it", or because they or the higher-ups are hiding something.

Acoustic consultants are extremely expensive, may require many hours if the sound is only apparent during random hours, and even then may come up with nothing.

An app that would at least be able to give a good estimate of the direction from where a sound is coming would be extremely helpful.

Sound travels, what, 30 centimeters a ms?

Bluetooth/wifi is good for at least a few meters, and should have latency of around a ms. Should be pretty easy to synchronize via that.

You need to maintain millisecond sync while the devices are widely separated, possibly for hours. Cell phone clocks are not millisecond-accurate over hours. The idea is that a group of people all download the app and walk or drive around the area of interest collecting time-stamped audio and sending it to a server.

Amusingly, data collection would be easier with analog technology. Get some VHS walkie-talkies, and, at the base station, feed all the channels into a multitrack tape recorder like musicians use. No sync problems.

For sounds transmitted through the air, ms precision is really necessary at distances of more than a few meters. Network time should be close enough at 10s or even 100 ms (roughly speaking < 10 meters of error - which is good enough to set up a local monitoring net via wifi with <1ms/<30cm of error).

I guess this is complicated by the fact that the speed of sound through the ground is ~6km/s. Rather faster than the speed of sound through the air. Multiplying those errors by 20... but 200 meters of error isn't that bad. A powerful local network with no latency would then be enough to isolate further.

Anyways, turns out that you can get gps time on androids, so this should be solvable with much better latency https://stackoverflow.com/questions/7017069/gps-time-in-andr...

Sorry, VHF.

I think you’re missing how easy it is to do these days.

It’s pretty easy to do with NTP synced devices. I’m lucky enough to have a fleet of devices sync with GPS and the PPS signal. It’s pretty amazing to/fun to do latency tests, you grab a time stamp and ship it to far side and time stamp the arrival. Voila single direction latency measurements.

Wouldn't work in this case though - the sound doesn't contain high enough frequencies to correlate well.

You'd also need microphones that pick up low frequencies.

Reply on news.ycombinator.com