Hi all,

My first post in the sound section, so don't shout at me if I'm not making loads of sense! This isn't linked directly to theatre / events, but I feel that the sound section is the best area to put this post in!

As part of a group design project at uni, we're making a system to calculate positions of sound sources using a microphone array. Were having to make the audio capture hardware as part of the project, but as I'm usually hanging around the lighting forum on here, I'm not that experienced in what kind of microphones would be useful for this application.

We'll ideally need something with a fairly good response for vocal frequencies, as we might end up having to do voice recognition to be able to discriminate between different people talking. Also needs to have a fairly wide pickup range, possibly omnidirectional, dependent on microphone position in the array, but possibly some form of cardioid.

We'll be looking at getting about 6 mics to provide good positioning information, so they can't be that expensive to buy. We don't have an exact amount of how much budget we have available, as we're going to have to get a development board in, which is fairly expensive, so we would usually have to put a case in to buy in extra pieces for the project.

Any help would be appreciated, and if you need to know anything specific, or if I've missed something obvious out, then let me know!

Thanks in advance.

I would say the best thing would be a linear reference mic. if you want to workout distances and postions, this is a hard one. you will need a system like what they use for finding mobile phones. You will need to triangulate it. so three mics around you area. Im sure someone else can throw in some ideas.

The ability to position is all based upon the phase differences between signals from different microphones. A member of the group has pretty much already done this before, but using sonar. The phase bit is the key, which we obtain using the FFT from the signals, and then the triangulation (in 3D) is done from this.

The volumes at certain directions aren't very important, as the majority of the work will be in the frequency domain, so the amplitudes will all be at some form of ratio to one another anyway dependent on the way the person is facing.

I remember a episode of CSI Miami where they used mics in street fixtures to work out the shooters position. I know some bits are far fetched but there is every chance its real, often googling the weird things CSI has I find they are actaully real, like 3d holograms (well a bar that spins and displays a scan)

BUT I did a quick google and found this which might help.

Had a look , but that covers the theory and software side, which we've got sorted out so far, rather than the hardware side / what microphone is best for the situation.

On a side note, where is usually the cheapest place to purchase microphones / audio stuff online? Anywhere people have had good experiences with, with relatively short lead times?

Interesting project.

I would not think you need to spend that much on the mics as long as they're all the same. You could look at:

Very cheap & small, but well thought of CPC's little lavalier mics - omnidirectional.

Behringer C2 (matcher pair) - cardioid.

I often use both the companies in these links along with Thomann

I'd agree with Mark...

some relatively cheap electrets should be sufficient - a point backed up by this document describing an array used at Southampton.

Cheers for the replies guys. :-)

Will have a look into those with the others at our next meeting tomorrow.

I would have thought that omni mics would be a necessity here - you don't really want the mics to have nulls in which the sound would disappear. Maybe something like behringer's measurement mic - the ECM8000. Think they're around the £30 mark but I'm sure you could negotiate a discount for any of these if you were buying half a dozen.

FS43W from Maplin is so cheap that you could hae a few to play with! It needs a power supply prob 1.5 - 12 vdc at less than a milliamp

A concurrent post has been automatically merged from this point on.

Possibly the symetry of the electronics will be important, Mics the same, cables the same.

Indeed such things are in current use for both the military and civil (police) authorities for locating snipers and other gun shots

Boomerang
Gunshot location
Shotspotter

There was one system that detected where the shot had come from, then trained a gun in that direction and fired it.... It doesn't look like it went into mass production!

I was thinking along the lines of that if a microphone is placed generally on the bounds of the measurement area, any noises behind it wouldn't be wanted, but anything in front would be fairly important. I'm guessing that's what a non-omnidirectional will give me.

Ideally the mics and cabling will be identical, and the microphones are going to be sampled at exactly the same time, so that the phase information is retained as much as possible.

If you're calculating the direction that the sound is coming from via phase (time of arrival) differences, then it's vital that each microphone hears from all directions equally. The sound approaching the "front" of one mic will then approach the "back" of another one. If that second mic is deaf in that direction, you won't have two signals to compare.

Cheap omnis should do for this.

Note that if you are doing beamforming or the like, you will need to be aware of possible spatial aliasing and may need to only operate at low frequencies (20Khz response is probably not a win as it requires a painful amount of accuracy in the geometry of the mic array).

It might be worth looking into a tetrahedral array or the like (as under for ambisonic recordings), which could provide both azimuth and elevation, but the processing will be a little hairy.

Good luck.

Regards, Dan.

theoretically you should be able to use any microphone for this, but I'd guess you would want a small form factor omnidirectional mic. I wouldn't worry about frequency response too much, you can always measure the frequency response in an anechoic chamber (do you have access to ISVR? they must have one...) and just create a transfer function to compensate for any differences across frequency, so long as it's not too appalling. Are you doing the blind source segregation in reverberation? there was an interesting poster at the ASA conference in Paris this year on just such a process...