I have several sound sensor modules and I assume they output in analog the relative SPL. I'd like to have Arduionos around house to pick up the noises such as doors slamming and compare them - shared to a master (Mega or ESP12). If the difference is sudden in the temporal sense, it interprets it as a sound event.
The nodes will NOT be placed at regular intervals in space; the calc would have to take that into account. Difference between nodes is then calculated X, Y, and Z. For X scale it means planes of the sensors along a north or south wall. Y, the east and west. Z is comparing floors. The one with greatest value is considered against the triangulation then, if the loudest is in the cross-section calculated from the 3 planes, it interprets it as a positive located sound. If not, it takes the triangulated sector, and interprets that area as the possible source.
The more diffuse the sound is (low delta) the system would concur the sound must be distant or at least outside the home. I may use LC networks to create crossovers to differentiate between low and high frequency sounds.
The analog ones won't output in response to sound pressure level? I left out a question on purpose. The real question was if it was feasible. Responses tend to discourage me, but that's only one of few irons in my fire, so I will work on my thermostat...
Typically, you'd do it with timing rather than amplitude/volume. Sound travels at approximately 1 foot per millisecond but you can look-up the actual speed.
The analog ones won't output in response to sound pressure level?
Yes, there are 3 popular kinds of sound modules -
There are boards that amplify the (millivolt) signal from the microphone and add a bias so it can be read by the Arduino. (The Arduino can't read negative voltages so the AC audio signal is biased to 2.5V.)
There are boards that put-out a DC voltage that's proportional to the volume (the actual audio is "filtered-out"). This is sometimes called an "envelope follower" or a "peak detector" circuit. These are easy to use, but it may be a little slow so it might not work for timing.
There are boards that put-out a digital pulse when the loudness exceeds a threshold. These are also very easy to use if the loudness is predictable and consistent enough.
If I do it by timing I'd have a timecode attached to each packet sent to the Nexus (calculator) for processing. The hard part then is to synchronize perfectly the time down to the millisecond each node.