Hi, I've just signed up on this site. I was reading the info in the DIY section as well as the topics in this DIY forum but I can't get to understand really how this works with only reading an audio signal.
Correct me if I'm wrong, and let's begin simply with a 1x1 grid configuration (a simple point pressure sensor). The device would be receiving an audio signal with just one frequency and giving that to the carrier strip (or pad, or whatever), and when the user presses (or reduces the distance between the carrier and the antena) the device will capture that one frequency in some amplitude proportional to the pressure (the more close the antena and carrier are, the louder this signal would be).
So far so good. Let's complicate it adding one carrier so we have now 2 carriers (2 different frequencies) and one antena. The process would be the same but we will be receiving a mixed signal of 2 frequencies in amplitudes proportional to the pressure and we could calculate where is the user pressing by measuring each amplitude after we separate them from each other using FFT. And we could extent this theory to any amount of carriers and we would be able to have a sensitive strip of any longitude we want.
Again, so far so good. But here is where I stop understanding. Let's add one antena and we now have a 2x2 configuration. When I do the FFT of the mixed incoming signal I can tell which carrier the user is pressing, but can't tell which antena is he pressing.
Where am I not understanding this?
Anyway thank you for reading me and thank you for sharing all this information.
I think you get this. You had it right up until the end. The antennas are not mixed with each other. So to do an 8x8 version you need 8 audio inputs and a separate FFT on each.
Aha! I suspected that! But which device would do an FFT outside a computer and output useful info?
Ahhhh I think I got it: you need an expensive sound interface with multiple I/O's. And the FFT is performed in the computer.
Is there a suggestion on which dsp chip to use for the FFTs?
I suppose that an arduino could manage it at some level. But which chip is optimal for really fast and accurate calculation?
The DIY instructions are more aimed at research than making a finished instrument. Interfaces with a lot of I/O are not necessarily expensive—but this kind of setup with all the wires is probably not good for touring with either!
Any computer is fine for the FFTs. They all produce the same results. The best kind, is probably one you have already. My guess is a recent Arduino can probably do a 1x8 matrix at a reasonable sampling rate.