Notices

I'm not sure about the processing power and time resolution, but in terms of geometry i know it should be possible to use the stereo speakers and the microphone on the N900 to acquire a 2 dimensional image from the surroundings by sonar.


It could work kinda like this: you emit a pulse of sound on the stereo speakers with a different frequency on each speaker, then you monitor the mic for the arival of the sound from the speaker and any subsequent echoes on the respective frequencies, with the time difference between the original sound and the echoes and the speed of sound in air you should be able to calculate the distance between whatever bounced the sound and each of the speakers, with those 2 distances its just a matter of simple trigonometry to calculate the 2d position of what the sound hit.



To make the image more interesting to see, you could also take in consideration how loud the echoes were, perhaps displaying the resulting image with a colored background and having any areas corresponding to echoes be on a grayscale representing the loudness, or perhaps have a black background and color the echoes mapping a color gradient to the ratio of the loudness between the left and right echoes (at one end it would be an echo pretty much only from the left speaker and on the other end from the right, with the middle for when both are equally loud).



If the framerate is fast enough, you don't even need to just do single pulses, use a pseudo-random number generator to pick the frequencies and then look for the spikes in the recorded sound for the frequencies in the order they came out; if t is really fast it might even sound close to white noise.


Btw, to help with environment interferences, it might be a good idea to analyze the environment noise for a little bit before starting and look for more silent frequency bands and avoid the noiser ones.

This is quite difficult...

You will probably want to send a coded pulse anyway as you can then cross correlate to try to pick out the reflection arrival time more easily (unless you can generate very large amplitude sounds, but that might be intrusive ). The problem (apart from how loud you can emit, how well you can receive, and ambient noise, is then reflections, and whether you can tell them apart.

That is almost imposible, you need frequency in ultrasound to eliminate the ambient noise and all false echoes... i don't know if the N900 is capability to handled ultrasound...

You wouldn't need to eliminate all the problems in order to get something useful, or at least interesting. I think I'll download soundmeter and see if I can't measure some ultrasonics just the same.

Okay, I just looked, and iTunes App Store has a Sonar Ruler app. They say it sends out a click, so I assume that means normal human hearing range.

Okay last post door a while!
http://labs.laan.com/wp/2009/08/sona...re-with-sound/

Is the site that describes an iPhone app.it mentions that the limitations are object size, accuracy, etc. It specifically mentions that it can't measure a person. I imagine that an app for our device would have similar limitations, even with ultrasonics.

Hi i remember that there was such program for Windows Mobile
http://nerdipedia.com/tiki-index.php?page=Sonar
and as i remember it realy worked for some phones

I'm not 100% sure, but that Windows one seems to do 1d plus time instead of trully 2d

All of them do 1D (from reading the links posted above), where the 1D is the round-trip time. 2D or 3D is theoretically possible by performing multiple 1D "pings" with different send/receive pair locations, and then performing triangulation on the return signals. You'd have to look quite carefully at the required distance between the different transmit/receive points to determine whether you could resolve the differences in round-trip time.

The iPhone one also specifically says that it won't do well against small targets (like a human) and works best against large planar (and reflective) targets such as walls.

I suggested a coded pulse in the generic sense and a chirp is the best example of what you could use, make sure your frequency range is as large as possible (start and end freqs for the chirp) to get as narrow as possible a cross correlation peak (and therefore be able to resolve the returned signal in time accurately).

Some links that might be vaguely relevant:
http://www.ecse.monash.edu.au/centre...99/sensrev.htm

There is two emitters, stereo speakers, you know their offset in relation to the receiver, the mic, so you first wait for the first pulse in the respective frequencies of both speakers (you can do it in parallel as long as each speaker use a different frequency or set of frequencies) that is comming directly from the speakers and any pulse after the first one for each speaker will be echoes, the time difference between the first and the echoes is the distance from the respective emitter: 2 distances from two points you know the coordinate with a bit of trig gives you the 2d coordinates.