acoustic camera idea

[Dave]

I've been kicking around this idea recently, and I think it just might work. The idea is for a camera that takes pictures with sound instead of light. If you have to ask "Why?" then you don't need one, and you may be on the wrong website! ;) I just think it would be fascinating to play with.

Most light cameras use a lens, but the simplest way to focus the incoming wavefronts onto your pixel array is with a pinhole. The pinhole can be almost any material that is opaque to light at around 550 nm wavelength. The sound equivalent might be some type of sound-deadening baffle, maybe a slab of drywall, or cardboard, or other insulating board. For light, the pinhole is often literally a hole made with a pin, but for sound the wavelengths are many orders of magnitude larger, and so everything has to get correspondingly bigger. For example if our acoustic camera were tuned to pick up roughly 10 kHz audio (wavelength in air is about 1.3 inches), if we had just 32x32 pixels at 1 wavelength pitch, the pixel array would be nearly 4 feet on a side. For a focal length of 2 meters (79 inches) the pinhole would be about 20 inches in diameter, for a 30 degree field of view. Potentially you could capture these low-res images and mosaic a bunch of them together for a much larger image.

That's all well and good, but 32x32 pixels means you need 1024 total microphones! Even if they cost 50 cents a piece (less than the cheapest mic on DigiKey) and could solder each one in 10 seconds, it would cost over $1000 and take 3 hours of soldering just for the mics! Plus you might need an amp for each one, some number of ADCs, some way to pull all that data together, etc. So here's what I'm thinking: you make the imaging array out of large PCB panels that you etch yourself, and you build the mics in yourself as part of it. I'm leaning towards homebrew condenser microphones using aluminum foil as the diaphragm. You'd need to develop a good way to make an array of these mics yourself but it seems doable to me. Here's a page showing one guy's own homebrew condenser microphone, clearly not streamlined to be made hundreds at a time, but still. I think with some experimentation you could get a system down for making 100 mics on a panel in just a few minutes.

http://kt4qw.com/condenser.pdf

I think with the right mic design, you could multiplex them all together using just two inexpensive mux ICs, a single amp, and a single ADC. It would probably take several seconds to capture each image since you'd only be pulling in one pixel at a time, but the processing would be really simple. You could even do an FFT and extract out different frequencies to be displayed in different colors. Different materials would show up in different colors. Practical? No way! It would sure make some cool pictures though.

[Jack Zylkin]

This is a very cool idea! It sounds totally magical, and I would love to see what thos pocs will look like-- Too cool!!!!

I do have one concern -- like a light camera, which is black on the inside, the bellows of your camera will need to be anechoic, and that might be the most challenging part.  Doable but possibly very expensive for fancy foam panels.

--
To post to this group, send email to hive76-d...@googlegroups.com
To unsubscribe send email to hive76-discuss...@googlegroups.com
For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

[pezman]

Wow, that is a cool idea.

I should imagine that you can make an array by:

- etching "condenser plate" zones on a pcb (hexagons might be nice in practice and for symbolic purposes)

- printing a "grille" over the PCB substrate that outlines each condenser plate -- the reprap would be ideal for printing the grille

- spread adhesive on the grille and lay foil on it (gold or copper leaf might make a nice, responsive membrane)

With some clever layout, you should be able to run traces that multiplex access to the sensors.

I think that there is one major difference between traditional light imaging acoustic imaging -- you can easily generate coherent "illumination" with sound and can easily measure phase differences between acoustic sensors.  As a result, you have much more information available.  I kind of wonder it it's possible to use phased array principles to make multiple, synthetic apertures and then use something like polydioptric imaging techniques to make "sound-field" cameras that can post-process images to play with focus, "look-angle" etc.  You should be able to dispense with "lenses" and "housings" -- you just need that array (and lots' of fancy math).

[Kyle Yankanich]

I'm really curious as to the images this will produce. From what I understand, sounds scatters much more than light. It influences the air of the room, mixing and creating some chaos in the airflow. Focusing.an image of that would be.... Difficult. Akin to taking a regular photograph, but in a foggy room. Except also the fog interacts with itself. I'd also be curious as to how the "focusing" effect would change. It normal photography, as the lens is moved from the photo sensitive material, it changes the distance of focus. Sometimes, since lights different colors are different wavelengths, you get different colors focusing.differently. its called.chromatic.aberation. Since sounds wavelengths are so much larger, does that mean we'll be effectively tuning out whole parts of the audio wavelengths?

[Kyle Yankanich]

[Joshua D. Johnson]

I'd like to hear Brendans take on this. From what I understand light travels in a fairly direct line while sound doesn't.

Maybe an ultrasonic emitter and all the collectors on a concave surface?

On Oct 7, 2012 8:42 AM, "Kyle Yankanich" <kyleya...@gmail.com> wrote:

I'm really curious as to the images this will produce. From what I understand, sounds scatters much more than light. It influences the air of the room, mixing and creating some chaos in the airflow. Focusing.an image of that would be.... Difficult. Akin to taking a regular photograph, but in a foggy room. Except also the fog interacts with itself. I'd also be curious as to how the "focusing" effect would change. It normal photography, as the lens is moved from the photo sensitive material, it changes the distance of focus. Sometimes, since lights different colors are different wavelengths, you get different colors focusing.differently. its called.chromatic.aberation. Since sounds wavelengths are so much larger, does that mean we'll be effectively tuning out whole parts of the audio wavelengths?

[Sean McBeth]

Sound is direct enough to allow bats to navigate by it in the night. Sonar and ultrasound are types of acoustic imaging, and these days can make some very detailed pictures. So I have no doubt this can be done.

The implications for 3d imaging are kind of exciting. The Xbox Kinect does it's depth sensing with infrared light, presuming that brighter pixels are closer pixels. Of course, put something that is very infrared-reflective in the scene and it breaks that assumption. With sound, it's slow enough that you could clock the amount of time it takes to bounce off of things, triangulate the bounces, and come up with a 3d picture.

Let me know if you get around to working on this, would definitely like to help.

[Dave]

Jack: I think you could get away without worrying about it, but something like egg cartons would probably help a lot and be dirt cheap.

Pez: totally, phased arrays would be awesome. You'd need at least like 4x4 sensors (more would be better) so you'd need like 16 good ADCs that are simultaneous, not multiplexed, and a processor capable of pulling all that in and at least storing it in realtime. The processing is actually fairly straightforward once you have the hardware and the RAM for it. One nice thing about the pinhole camera idea is that you can switch out the mics to sense other things: thermopiles for a crude thermal camera, antennas for a microwave camera (think passive radar where the scene is illuminated by every wifi router and laptop in the building), etc. Also I believe that lots of the characteristic image artifacts and ambiguities that show up in phased array images (like weather radar, ultrasound, etc) would go away. Absolutely I'd love to work on either one though, and that's part of my motivation to look into FPGAs.

Kyle/Josh: waves are waves, they all act the same. Yes, the images would be strange compared to visible light. Visible light's tiny wavelengths mean you can resolve much smaller features. Chromatic aberration happens with sound exactly as it does with light (thus typically choosing an average wavelength of interest when making a light camera, whether using a refracting glass lens, a diffracting zone plate or pinhole), so you would only operate in a portion of the spectrum of sound, same as we only see the tiniest little sliver of the EM spectrum (no IR, UV, heat, X-rays, etc etc, even though they are otherwise identical). If your wavelength is on the order of 2 inches then any surface whose rough features are smaller than that will appear smooth / shiny / specular, as opposed to diffuse. But almost all of the calcs for light cameras should map directly to all other wave phenomena including sound.

If anybody has any interest in lending a hand or kicking around ideas, let me know. For now this is more a thought experiment while I get some other projects off my plate, but I'd love to work on it at some point and if anyone wants to help, it'll be more likely to be successful. I have basically all the calcs taken care of already for the pinhole/zone plate case as well as the phased array.

[Joshua D. Johnson]

That's fair. Dave was talking about homebuilt condenser microphones though.. Sonar and ultrasound work at higher frequencies.

Why not start with a single emitter and collector from a known distance and see if you can get accurate readings then go on from there?

[Dave]

I bet the homebrew condenser mics could be made to handle ultrasonic without much trouble, if that's what you wanted to do. Probably wouldn't even have to do anything different unless you wanted it to be optimal.

That's my plan to start with: set up a sound source somewhere (little piezo buzzer or something), put up a pinhole of the appropriate size and material, at the right distance from the mic. Move the mic around manually to different positions on a grid, and assemble a picture. I have no doubt whether or not it would work, it's just a matter of how well. Try holding up a helium balloon near your ear, the ambient room sound deadens dramatically, it's pretty eerie.

-Dave

On Sunday, October 7, 2012 9:39:47 AM UTC-4, J.Johnson wrote:

That's fair. Dave was talking about homebuilt condenser microphones though.. Sonar and ultrasound work at higher frequencies.

Why not start with a single emitter and collector from a known distance and see if you can get accurate readings then go on from there?

On Oct 7, 2012 9:27 AM, "Sean McBeth" <sean....@gmail.com> wrote:

Sound is direct enough to allow bats to navigate by it in the night. Sonar and ultrasound are types of acoustic imaging, and these days can make some very detailed pictures. So I have no doubt this can be done.

The implications for 3d imaging are kind of exciting. The Xbox Kinect does it's depth sensing with infrared light, presuming that brighter pixels are closer pixels. Of course, put something that is very infrared-reflective in the scene and it breaks that assumption. With sound, it's slow enough that you could clock the amount of time it takes to bounce off of things, triangulate the bounces, and come up with a 3d picture.

Let me know if you get around to working on this, would definitely like to help.

On Sunday, October 7, 2012, Joshua D. Johnson wrote:

I'd like to hear Brendans take on this. From what I understand light travels in a fairly direct line while sound doesn't.

Maybe an ultrasonic emitter and all the collectors on a concave surface?

On Oct 7, 2012 8:42 AM, "Kyle Yankanich" <> wrote:

I'm really curious as to the images this will produce. From what I understand, sounds scatters much more than light. It influences the air of the room, mixing and creating some chaos in the airflow. Focusing.an image of that would be.... Difficult. Akin to taking a regular photograph, but in a foggy room. Except also the fog interacts with itself. I'd also be curious as to how the "focusing" effect would change. It normal photography, as the lens is moved from the photo sensitive material, it changes the distance of focus. Sometimes, since lights different colors are different wavelengths, you get different colors focusing.differently. its called.chromatic.aberation. Since sounds wavelengths are so much larger, does that mean we'll be effectively tuning out whole parts of the audio wavelengths?

--

To post to this group, send email to hive76-discussion@googlegroups.com
To unsubscribe send email to hive76-discussion+unsubscribe@googlegroups.com

For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

--
To post to this group, send email to hive76-discussion@googlegroups.com
To unsubscribe send email to hive76-discussion+unsubscribe@googlegroups.com

For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

[Joshua D. Johnson]

I don't know anything about the electronics in this instance but would like to help build the mechanical bits, certainly the pinhole mic. I have 1/2" square by 4' foam rods left over from a quadratic diffuser build if you want them.

[Kyle Yankanich]

I think this is an awesome project, and I'm really excited to see how it works out, and of course available to help in any way I can. Dave, I think the trouble I'm having with this vs the bat/whale thing is that they measure distance to an object, using a relatively known frequency that they emit. In my mind, a camera is passive. It measure existing light levels, so ideally a sound-camera would do the same.  It wouldn't so much as show the physical shape/properties of the room, but rather be a more .... abstract image. Almost a heat map of a 3 dimensional space, showing where sound is coming from.

The other one is just a depth-map of a room, which is neat as hell, but already kind of done in various ways.

[Dave]

Kyle, yes, this is a passive sensor. As with light cameras you can always provide your own illumination if there's not enough signal out there already for you to pick up. This is not about depth, but measuring the energy coming from different regions. Seeing the shape and structure of the environment. Abstract. I'm not expecting this to replace light cameras...

[Dan Shookowsky]

I haven't seen every post in this discussion, so my apologies if this has already been discussed, but this is currently being done to image babies in the womb - http://en.wikipedia.org/wiki/3D_ultrasound.  Might be some literature that can be leveraged in choosing frequencies / imaging the result.

[Jack Zylkin]

Err this seems a bit different from regular ultrasound -- although pez's idea for a coherent phased array is getting there.  I have never heard of an ultrasound machine that used an actual aperture, for one thing.  Or audible frequencies.  Or one that is the size of a large room.  I think this project is hot

[Dan Shookowsky]

See, that's why I should've RTFT.  I missed the part about audible frequencies and a room-sized device.

[Doctor]

This whole thing sounds pretty cool. It would be cool to see a real-time image of, say, someone talking in a room, turning about... It would let you see how the sound reflects about the room.

[Dave]

One disadvantage we may have is related to the type of sensor. In a digital light camera, each pixel is basically a small solar cell, and whenever a photon of sufficient energy is absored, it knocks an electron loose and traps it in an electron well. In a real light camera the integration time is how long you allow the electron wells to collect more electrons before you measure the contents. If that time is too low then you don't get much signal, and if it's too high then the well fills up and saturates, but the great thing is it allows you to basically count photons, with extremely low noise. In my proposed approach we would be sampling the mic channel many times per period of the acoustic frequency of interest and doing a little signal processing, and each sample has a certain small ADC noise. That noise floor basically sets the threshold below which we just can't reliably pick up a signal, unlike the ideal light imager that can potentially just keep capturing more photons and accumulating a stronger signal before measuring it if you just leave the 'shutter' open long enough.

This is not a difference between light and sound, just a difference in the electronics we have available to us. The idea works, this just means the tech is not as advanced as today's imaging chips. We'd be in exactly the same boat if we were building our own light camera by hand (I, for one, would have no idea how to build an electron well from scratch). If anybody has any ideas of how to approximate something like this with sound I'd be curious to hear though. As it is, this is just the acoustic equivalent of having a camera with poor low-light sensitivity, so we can only shoot brightly lit scenes.

[Kyle Yankanich]

Always remember to board up an abandoned electron well.
Also, did some googling, found some neat examples:

http://www.acoustic-cartography.com/documentation.html

http://www.lmsintl.com/testing/testlab/acoustics/high-definition-acoustic-camera
Same camera, but amazing youtube video to watch. They basically use 36
microphones on a polar array, and use beamforming techniques to
localize the sound, and can even choose which frequency to highlight.
http://www.lmsintl.com/acoustic-beamforming

-- Kyle Yankanich

On Sun, Oct 7, 2012 at 10:29 PM, Dave <dgs...@gmail.com> wrote:
> One disadvantage we may have is related to the type of sensor. In a digital light camera, each pixel is basically a small solar cell, and whenever a photon of sufficient energy is absored, it knocks an electron loose and traps it in an electron well. In a real light camera the integration time is how long you allow the electron wells to collect more electrons before you measure the contents. If that time is too low then you don't get much signal, and if it's too high then the well fills up and saturates, but the great thing is it allows you to basically count photons, with extremely low noise. In my proposed approach we would be sampling the mic channel many times per period of the acoustic frequency of interest and doing a little signal processing, and each sample has a certain small ADC noise. That noise floor basically sets the threshold below which we just can't reliably pick up a signal, unlike the ideal light imager that can potentially just keep capturing more photons and accumulating a stronger signal before measuring it if you just leave the 'shutter' open long enough.
>
> This is not a difference between light and sound, just a difference in the electronics we have available to us. The idea works, this just means the tech is not as advanced as today's imaging chips. We'd be in exactly the same boat if we were building our own light camera by hand (I, for one, would have no idea how to build an electron well from scratch). If anybody has any ideas of how to approximate something like this with sound I'd be curious to hear though. As it is, this is just the acoustic equivalent of having a camera with poor low-light sensitivity, so we can only shoot brightly lit scenes.
>

[Dave]

Yeah there are some cool commercial products out there (all of them beamforming-oriented). But there's virtually nothing that's been done by hobbyists in this field. My cheesy transmitting acoustic phased array I brought in a few months ago was probably the only amateur phased array I've actually come across anywhere, which is kind of sad. I've been learning about FPGAs largely so I can do a better one some day. But there's just something attractive to me about the pinhole sound camera for some reason.

[Jack Zylkin]

your pinhole will be way better than a phased array.  Acoustic phased array is more analogous to a radar than a camera. I mean, you don't see many cameras out there without lenses or apertures.

On Sun, Oct 7, 2012 at 10:56 PM, Dave <dgs...@gmail.com> wrote:

Yeah there are some cool commercial products out there (all of them beamforming-oriented). But there's virtually nothing that's been done by hobbyists in this field. My cheesy transmitting acoustic phased array I brought in a few months ago was probably the only amateur phased array I've actually come across anywhere, which is kind of sad. I've been learning about FPGAs largely so I can do a better one some day. But there's just something attractive to me about the pinhole sound camera for some reason.

--
To post to this group, send email to hive76-d...@googlegroups.com
To unsubscribe send email to hive76-discuss...@googlegroups.com
For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

--
Jack Zylkin
usbtypewriter.com
Philadelphia, PA

[Kyle Yankanich]

If you're looking for something more sci-fi, and actually more
functional than a pinhole:
http://en.wikipedia.org/wiki/Zone_plate

-- Kyle Yankanich

[Dave]

:) My calcs are actually all done for zone plates, and a pinhole is just the first zone of a zone plate, so it's smaller and easier to make. Zone plates add more gain and can improve the resolution some but are larger and a bit harder to make. One nice thing about a pinhole is that if you make it as an iris mechanism you can use it to change your focus and field of view (zoom) by adjusting its size and moving the imaging plane back and forth. Harder to do with a zone plate.

[pezman]

http://www.mrpinhole.com/zp.php

[Dave]

Yeah, those calculators are a good start. Once you start making unusual assumptions (nonstandard optical formats, huge wavelengths, etc) or needing more information (focus distance, feature resolution, field of view, etc) it's easier to just roll your own. I think for audio, starting out with a pinhole is probably better though -- easier, and more likely to work will. The material will likely have to be fairly thick, which I'm guessing would introduce more artifacts with a zone plate than a pinhole.

-Dave

[Joshua D. Johnson]

I've never done the acoustic version but pinhole cams and camera obscuration are fun. Most people just make an aperature plate with a bunch of holes or just poke a hole and enlarge till it works.

Now I want to build a digital pinhole cam....

Josh

[Dave]

I had heard of but never really seen camera obscura until a couple of weeks ago, actually. Thinking about all this stuff prompted me to try it out. I was in my basement and blacked out all the windows with cardboard, with a 'pinhole' cut in one of the pieces. It wasn't the best example I'm sure but it was kind of cool to be basically standing inside a camera looking at the world projected upside-down and backwards on my floor.

When you say making a digital pinhole camera, do you mean using a commercial imaging sensor (webcam etc) with a pinhole instead of a lens? Coincidentally, a few weeks ago one of the Sparkfun videos showed a SF employee that made a digital camera with a single light sensor in a cardboard box, and I think some type of lens (I believe he mentioned pinholes though). He built a little 2-axis positioner that would scan the sensor back and forth, and a Processing program to build up the image on his computer. Sort of neat, although very difficult to make out any features other than the bright sky out the window and dark walls. If I knew him I'd probably ask to borrow his setup and attach a mic to it instead of a light sensor. Really curious to see what an audible sound camera can do, even crudely.

[pezman]

I think that you could make something quick that is basically a round table and a mic on a string.  The table rotates slowly and winds a string around a post in the center.  The string pulls the mic inward at a more-or less fixed amount per revolution.  If you are able to resolve the angular position reasonably well, then you should be able to build up an image of the steady-state sound-field.

Old turntables are pretty quiet -- they might make a nice platform if you can figure out how to drive the platter slowly (and quietly).

[Joshua D. Johnson]

I meant a small camera obscuration/large pinhole with a digital camera taking a picture of the image.
We used to make them with old scanners as the back, you get some neat effects. You can also change the shape of the hole to squares or your initials etc and it does cool things (there is a name for it, I forget) you can also do several holes with filters or colored glass over. I would just want to capture that digitally..

Josh

On Oct 8, 2012 10:46 PM, "Dave" <dgs...@gmail.com> wrote:

I had heard of but never really seen camera obscura until a couple of weeks ago, actually. Thinking about all this stuff prompted me to try it out. I was in my basement and blacked out all the windows with cardboard, with a 'pinhole' cut in one of the pieces. It wasn't the best example I'm sure but it was kind of cool to be basically standing inside a camera looking at the world projected upside-down and backwards on my floor.

When you say making a digital pinhole camera, do you mean using a commercial imaging sensor (webcam etc) with a pinhole instead of a lens? Coincidentally, a few weeks ago one of the Sparkfun videos showed a SF employee that made a digital camera with a single light sensor in a cardboard box, and I think some type of lens (I believe he mentioned pinholes though). He built a little 2-axis positioner that would scan the sensor back and forth, and a Processing program to build up the image on his computer. Sort of neat, although very difficult to make out any features other than the bright sky out the window and dark walls. If I knew him I'd probably ask to borrow his setup and attach a mic to it instead of a light sensor. Really curious to see what an audible sound camera can do, even crudely.

[DonTrackbiker]

  One thing about pinholes:  They diffract.  The pinhole will have to be several wavelengths in diameter.  The blurring by diffraction in radians will be roughly the pinhole diameter in wavelengths.  This is in addition to blurring caused by the size of the pinhole and closeness of the pinhole to the sensor array.  I recommend an ultrasonic frequency, perhaps 40-60 KHz, possibly even higher, so that the pinhole can be small enough to focus an image without a diffraction problem.

 

  Condenser microphones will probably woek at these frequencies, preferably with a modification.  Remove the cloth-like or felt-like material from the front surface.  This exposes a hole in the front surface.  Either enlarge the hole or remove the front surface.  The hole and air volume between the front surface and the diaphragm for a Helmholz resonance, and the microphone rols off at 12 dB/octave above that.

  Caution - the diaphram may be close to the front surface and is probably fragile.

 

  Loudspeakers tend to get more directional as frequency increases.  Maybe get a few piezoelectric ceramic discs from piezo tweeters.  Maybe make a small curved line array of those, and rotate the line array to scan the room with ultrasound.

 

  If this project needs highpass filters, my favorite is the "equal component value" Sallen Key.  My favorite op-amp for this is the TLO84.  The formula is

F=1/(2pi*SQR(*R1*R2*C1*C2)), which is 1/(2pi*R*C) when the frequency determining resistors are equal and the frequency determining capacitors are equal. With the two frequency determining resistors being 10K and the two frequency determining capacitors being 470 pF, the nominal cutoff frequency is 34 KHz.

 

 - Don Klipstein

[DonTrackbiker]

On Monday, October 8, 2012 7:18:19 PM UTC-5, pezman wrote:

http://www.mrpinhole.com/zp.php

 

  That calculator appears to me to recommend a pinhole size in wavelengths around .57 times the square root of wavelengths from the pinhole to the sensor array.

 

 - Don Klipstein

[Jack Zylkin]

Don has a good point but the ratio of the pinhole size to the sensor will still need to be small relative to the sensor or else the image will be blurry -- the best case resolution on the sensor you can hope for on a pinhole camera is equal to the pinhole size, but in this case thats at least 20 in. So, that will be the limiting factor on resolution  -- not the lamda/2 microphone spacing.  

--

[pezman]

Ooh -- even simpler -- use a foucault pendulum for scanning.  It would take about 16 hours to build up the image at this latitude, but it sure would be simple.

[Dave]

Good stuff! I'll probably need some time to wrap my head around all of this. Great discussion though!

Don: Making the pinhole "several" wavelengths in diameter shouldn't be a problem. In my 10kHz example the wavelength is about 1.3" and the pinhole is just over 20", or around 15 wavelengths across. If I understand your math right from what you observed of that calculator, the pinhole diameter according to that rule of thumb using a 2m (~79") focal length would be about 0.57*sqrt(focal length/wavelength)*wavelength = about 5.8" across. Quite a bit smaller than 20".

I'm having trouble understanding how to visualize blur in terms of radians. Any insight there? "The blurring by diffraction in radians will be roughly the pinhole diameter in wavelengths." So if the pinhole is 15 wavelengths across then we're looking at around 15 radians. What does that look like? Is this different from what Jack is describing, which sounds to me like the diffraction-limited resolution? My calculator is putting that at 1.22 times the pinhole diameter (or 1.22 times the width of the narrowest zone, for a zone plate), around 25". My understanding was that that means the smallest feature this thing could hope to discern is about 25", but that that wasn't directly related to the pixel pitch. This is clearly not my area of expertise. ;)

Another thing that makes it tricky is that if you are imaging very close objects you will need to adjust the distance between the pinhole and the sensor array to something other than the pinhole's focal length. For this example, imaging something about 16' away using a 6.5'-focal length pinhole, in order to focus on it you would actually need to move the pinhole about 11' away from the pinhole. Shallow depth of field. Higher frequencies will help as usual, but on the other hand higher frequencies attenuate much faster too (not to mention just not being as cool as audible frequencies).

Perhaps the "illumination" source could be more wideband, and then some signal filtering could cut out the frequencies of interest. Maybe even something mechanical.

-Dave

[Jack Zylkin]

The blurring I was talking about is due to straight up trigonometry, not diffraction. Just imagine  a tiny point of light at infinity (ie the sun) -- it will blur out to the size of the aperture.

[DonTrackbiker]

On Monday, October 8, 2012 10:25:14 PM UTC-5, DonTrackbiker wrote:

  One thing about pinholes:  They diffract.  The pinhole will have to be several wavelengths in diameter.  The blurring by diffraction in radians will be roughly the pinhole diameter in wavelengths.

 

  I meant roughly the reciprocal of the pinhole diameter in radians.

 

 - Don Klipstein (d...@donklipstein.com)

[DonTrackbiker]

On Sunday, October 7, 2012 12:55:56 PM UTC-5, Dave wrote:

Kyle, yes, this is a passive sensor. As with light cameras you can always provide your own illumination if there's not enough signal out there already for you to pick up. This is not about depth, but measuring the energy coming from different regions. Seeing the shape and structure of the environment. Abstract. I'm not expecting this to replace light cameras...

 

  In this sort of application of imaging location of high frequency sound sources, I have a liking more for a pinhole than a zone plate.  This is because a zone plate appears to me frequency-specific while a simple pinhole generally works well at

frequencies higher than the frequency for which its size is optimized for.  Also,

I seem to think a zone plate is more directional than a simple pinhole.

 

  One thought I have :  At 1st provide a source of "ultrasonic illumination", and afterwards improve this to remotely locate sources of ultrasonic emissions,

such as leaks of pressurized gases or "ultrasonically buzzing" electronic power supplies or CFLs, or to "acoustically videorecord" travels of mice or large roaches.

 

  In development for actual industrial application, I like to think of finding leaks

of pressurized gases, or acoustic signs of vermin skittering behind objects that they can hide behind.  Maybe mechanical things producing a "tick" or "crack sound" when they malfunction.  Especially if this gets translated to some sort of

"video recording" with time notation.

 - Don Klipstein

 

[DonTrackbiker]

On Sunday, October 7, 2012 9:29:30 PM UTC-5, Dave wrote in part:

 We'd be in exactly the same boat if we were building our own light camera by hand (I, for one, would have no idea how to build an electron well from scratch). If anybody has any ideas of how to approximate something like this with sound I'd be curious to hear though. As it is, this is just the acoustic equivalent of having a camera with poor low-light sensitivity, so we can only shoot brightly lit scenes.

 

  We don't need to build microphones from scratch for 10 KHz, and probably don't

for even ~50 KHz.  It appears to me that there is a fair chance to modify some

cheap condenser microphones to make them work at least reasonably well (or better than before) at ~50 KHz.

 - Don Klipstein (d...@donklipstein.com)

[Dave]

Interesting thought, thanks! As for making my own condenser mics, the idea was that if we were making an imaging array with a large number of pixels, like 32x32=1024 for instance, buying even the dirt-cheapest commercial mic would get very expensive. If we were making them ourselves, we may be able to come up with a method for making an array all at once. For example let's say we lay out a large PCB with 100 of our custom mics' condenser plates. We etch all 100 chemically, in parallel, in just a couple of minutes, and etching 100 takes the same time as 1. Then we apply a laser-cut spacer sheet, again all 100 at once since it's one piece. Then we apply a single membrane of aluminum foil over the whole board and press it between rubber sheets to stretch it taught until it cures, and finally peel off the foil between elements. We didn't have to buy or install 100 of anything (much less a thousand). Just a thought.

[Chris Thompson]

Would Hive76 and Dave be interested in doing this project with some money from a Knight Arts grant?

http://www.knightarts.org/knight-arts-challenge/philadelphia

I'll fill out the form and all. This is just one step of many. We would need to find matching funds if accepted. The practical side is easy. We have a source for fiscal sponsorship if needed.

The pitch would be that we would make an acoustic camera, and take pictures of the city. We could overlay them on real photos captured at the same time, or just make abstract looking things. They would be visible on a gallery on Hive76.org and hopefully printed out for display and an opening. I think I ave the chops as a digital artist to take any of the data we collect and make some beautiful images.

What say you? How feasible is some kind of acoustic camera? I can leave technical details off the application.

Chris Thompson, eagleapex.com
GPG key available
"Obscurity is a far greater threat to authors and creative artists than piracy" -Tim O'Reilly

On Thu, Oct 11, 2012 at 6:13 AM, Dave <dgs...@gmail.com> wrote:

Interesting thought, thanks! As for making my own condenser mics, the idea was that if we were making an imaging array with a large number of pixels, like 32x32=1024 for instance, buying even the dirt-cheapest commercial mic would get very expensive. If we were making them ourselves, we may be able to come up with a method for making an array all at once. For example let's say we lay out a large PCB with 100 of our custom mics' condenser plates. We etch all 100 chemically, in parallel, in just a couple of minutes, and etching 100 takes the same time as 1. Then we apply a laser-cut spacer sheet, again all 100 at once since it's one piece. Then we apply a single membrane of aluminum foil over the whole board and press it between rubber sheets to stretch it taught until it cures, and finally peel off the foil between elements. We didn't have to buy or install 100 of anything (much less a thousand). Just a thought.

[Jack Zylkin]

Hey check out this mic for $0.29

http://www.surplus-electronics-sales.com/index.php?main_page=product_info&products_id=906

Still not sure how you would read 1000 of them at the same time.  Maybe you should make a crude CCD circuit?

[Dave]

Jack,

The inventor in me loves the idea of coming up with a novel way of building up a thousand mics in a simple process, with no need for individual soldering or components, and using only raw materials that I already own. The realist in me acknowledges the fact that I may never get around to it, and the simpler the task, the more likely it is to succeed. Nice find!

You're absolutely right, the real tricky part here seems to be getting all this data captured without breaking the bank. There are a couple of nice ADCs I've used that have an SPI interface with like 8 separate channels (not multiplexed) that can be daisy-chained, so you could get maybe 32 channels sampled simultaneously for let's call it $100 in ADC chips. If you had 32 actual ADC inputs you might be able to rig up some kind of multiplexing system so you could scan through a column at a time. Might get pretty pricey though. Ideas are definitely welcome. On a CCD or CMOS imager each pixel has its own electron well which accumulates a signal corresponding to the energy it has received -- I wonder if there's some reasonable equivalent for sound that uses minimal components? That way you could go through row by row and read out the charge in each "bucket". Sounds like a crap-ton of components though. Getting it built by robot would probably be worth it. Minimizing the number of amps and ADCs and such would really help keep the cost and complexity (and odds of failure) down.

-Dave

[Dave]

Chris,

I think this is an awesome idea. I had been starting to see this largely as something you might do a live installation at something like the PAFA event, since it's probably impractical for any real functional use. The two things that strike me immediately as being possibly good subjects to acoustically image using ambient sound are traffic (cars zipping between buildings, by statues, etc) and crowds (like at a sports arena). Maybe a concert. Ooh, an orchestra pit, with different frequencies coming from different areas. It would be nice to have the scene actually illuminated by the city life itself.

No idea where matching funds would come from.

There would be a lot of technical decisions that would have to be made, obviously. I strongly believe this is feasible in some form, but it's hard to predict how well it will work or what kind of practical sacrifices we will have to make for our implementation. I think this would be cool as all get out though. Count me in for sure.

-Dave

[Sean McBeth]

Perhaps a completely analog signal, ala NTSC or PAL. Eeh, I dont really have a suggestion on how that would work.

[DonTrackbiker]

  I saw worse:  http://www.whizkidtech.redprince.net/zoneplate/

 

  If I understand that correctly, the central pinhole should have radius of the square root of the product of focal length and wavelength.  This appears to me that radius in wavelengths should be square root of focal length in wavelengths.  Diameter is twice the radius.

 

  Using Dave's example that he showed me for focal length of 2 meters and wavelength of about .034 meter, the whizkidtech site appears to me to recommend a pinhole radius of .26 meter, or diameter of .52 meter.  This is regardless of number of zones to use in a zone plate.

 

  If this pinhole is a simple pinhole, I would make the size half that, or diameter of .26 meter.

 

  This sounds to me that resolution at 2 meters behind the pinhole would be about or a little under half a meter.  At 2.5 meters behind, I seem to think around .55-.56 meter with a pinhole diameter of .55-.56 meter.  (The Mr. Pinhole site appears to me to suggest pinhole size about 40% smaller still.)

 

  .55 meter at 2.5 meters behind the pinhole is resolution of about 12.5 degrees,

or 1.1 meters at 5 meters in front.

 

  A zone plate sounds to me likely to improve resolution, but I am concerned about field of view narrowing down almost to resolution size.  I am still researching that - I'm not certain that zone plates are that bad for field of view.

 

 - Don Klipstein (d...@donklipstein.com)

[Chris Thompson]

Awesome.

I'm submitting the two question form tonight. Here's the text:

http://apps02.knightfoundation.org/arts/

Project Title

An acoustic camera that will covert soundscapes into images.

Describe your idea below 

We at Hive76 propose to use our members' technical expertise to create a focused array of microphones for the collection of soundscapes. This audio data will then be coupled with photographs, or displayed alone to visually illustrate the sound of the city of Philadelphia, quiet spaces, social gatherings, and any other audibly interesting space. This project is currently in the feasibility stage but would require funds for fabrication and development. We would obtain additional funding from our members and fundraising events.

Sound good?

Chris Thompson, eagleapex.com
GPG key available
"Obscurity is a far greater threat to authors and creative artists than piracy" -Tim O'Reilly

[DonTrackbiker]

On Thursday, October 11, 2012 8:45:21 PM UTC-5, DonTrackbiker wrote:

On Monday, October 8, 2012 10:45:40 PM UTC-5, DonTrackbiker wrote:

On Monday, October 8, 2012 7:18:19 PM UTC-5, pezman wrote:

http://www.mrpinhole.com/zp.php

 

  That calculator appears to me to recommend a pinhole size in wavelengths around .57 times the square root of wavelengths from the pinhole to the sensor array.

 

  I saw worse:  http://www.whizkidtech.redprince.net/zoneplate/

 

  I just tried http://www.mrpinhole.com/zp.php  for 2 meters from zone plate to

"film", and wavelength 3.4 centimeters, with 3 zones.  (one ring hole around a

central pinhole.)

 

  I interpret the results as having the ring around the central spot having outer

dimeter of around 11 inches.

 

 

[Dave]

Sounds good to me.

[DonTrackbiker]

On Thursday, October 11, 2012 8:45:21 PM UTC-5, DonTrackbiker wrote:

On Monday, October 8, 2012 10:45:40 PM UTC-5, DonTrackbiker wrote:

On Monday, October 8, 2012 7:18:19 PM UTC-5, pezman wrote:

http://www.mrpinhole.com/zp.php

 

  That calculator appears to me to recommend a pinhole size in wavelengths around .57 times the square root of wavelengths from the pinhole to the sensor array.

 

  I saw worse:  http://www.whizkidtech.redprince.net/zoneplate/

 

  I just tried http://www.mrpinhole.com/zp.php  for 2 meters from zone plate to

"film", and wavelength 34 millimeters, with 3 zones.  (one ring hole around a

central pinhole.)

 

  I interpret the results as having the ring around the central spot having outer

diameter of around 11 inches.  I'm not sure correctly.

 

  I just now tried for single pinhole from http://www.mrpinhole.com/holesize.php

 

  With wavelength of 34 mm as opposed to .00056 mm and "focal length" of 2000 mm, based on http://www.mrpinhole.com/holesize.php and extrapolating by square root for wavelength being 34 mm rather than assumed .00056 mm, pinhole diameter comes up as 466 mm.  That's about 1.88 times square root of product focal length and wavelength.  When I said .57 before, I may have been low by a factor of square root of 10.

 

 - Don Klipstein (d...@donklipstein.com)

[Jack Zylkin]

Dave,

The audio equivalent of a ccd is called a bucket brigade-- you can get bucket brigade chips but they are usually single input single output -- to implement an audio delay.

From what i understand though the circuit is just a single fet and cap per "bucket" : 
http://www.synthdiy.com/files/2003/MN3005.pdf

Maybe you could make one with discrete parts.

[Dave]

Those seem a lot smaller than the numbers I am getting from my spreadsheet, odd.

I haven't seen any evidence to suggest that a zone plate has a more narrow field of view, my understanding was that it acts just like a convex lens (albeit through a different mechanism), and the FoV of a lens or pinhole is based purely on the focal length and the size of the 'film'. I'm not much of an expert on this stuff unfortunately, so I'll be very curious to see what you turn up. It certainly may be possible that it exhibits worse vignetting or something, though I have no idea.

[Dave]

Jack,

It would require a lot of parts and I'd need help, experimentation, or both, but it seems feasible. My concern about the CCD-like approach is that it's a crap ton of data to gather at audio frequencies. Also there's the issue that each pixel keeps recording even while you are reading out. Maybe there's a good way around that, like disconnecting the condenser array's biasing supply or something (brain is not operating fully right now, sleepy).

Potentially maybe you could give each 'pixel' (mic) a simple envelope follower (maybe just a diode and a resistor, or maybe need a filter of some sort first...). The idea being that if each pixel has an envelope follower then you just sample the envelope of each pixel, which doesn't have to happen at audio frequencies. You'd lose the ability to do certain types of post processing (FFTs and such), and it would use lots of components, but they'd be relatively cheap and you could relax on the sampling rate and processing horsepower requirements.

Maybe next open house we can kick around ideas in person. Lots of these things are just judgment calls that have to be made.

[DonTrackbiker]

On Thursday, October 11, 2012 9:53:04 PM UTC-5, Dave wrote:

Those seem a lot smaller than the numbers I am getting from my spreadsheet, odd.

I haven't seen any evidence to suggest that a zone plate has a more narrow field of view, my understanding was that it acts just like a convex lens (albeit through a different mechanism), and the FoV of a lens or pinhole is based purely on the focal length and the size of the 'film'. I'm not much of an expert on this stuff unfortunately, so I'll be very curious to see what you turn up. It certainly may be possible that it exhibits worse vignetting or something, though I have no idea.

 

  Followup by me:  I searched for photos taken with zone plates.  Some show some signs of directivity, with brightness and/or contrast and/or resolution deteriorating as angle from axis increases.  However, this does not appear as bad as I feared.  Many images even show little or no degradation off-axis.

 

  Now I feel in favor of a zone plate with an open zone or 2 around the central hole.

 

 - Don Klipstein (d...@donklipstein.com)

[Dave]

Sounds good. It seems like adding that first or second exposed zone improves the resolution quite a bit without making it too much more bulky. Gotta try some mic experiments soon. Hopefully it's as you predicted at open house and it will be easy to make a good strong acoustic barrier at these frequencies.

[Dave]

Has anybody got a favorite op-amp to use for mic preamps? I'm thinking of designing a small board with an 8-channel ADC (16-bit), and several of these boards could be ganged together. So I'll need 8 mic preamp circuits that are as small and cheap as possible (low component count etc) while having acceptable quality. Bonus points if it comes in, say, a 4-channel IC to save on parts. I'd prefer SMD to thru-hole. Any recommendations?

In planning this project out I'm starting to talk myself into the phased array approach after all. Less cool in some ways, but just a lot more practical in too many other ways to ignore (cost, size, speed of capture, bandwidth...). Current plan is to design a simple 8-channel ADC board around the ADS1178 ADC which I have some experience with. SPI interface, cheap, 16-bit, up to 52kHz sample rate, and I could fairly easily daisy-chain up to 4 boards together (32 channels) on the same bus. A microcontroller could then pull the data off and save it to an SD card for later processing on a computer. You could create stills or even animations at several Hz playback rate, with false color based on the frequencies present at each pixel location. I think it should be fairly doable, definitely more advanced than any amateur phased array I've ever heard of, and if nothing else the daisy-chainable 8-channel digital mic preamp board could probably be reused by all kinds of other projects.

Thanks in advance for any opamp suggestions!

[Brendan Schrader]

I like the INA217, BurrBrown/TI makes them. They're cheap and come in SMD packages.

http://www.digikey.com/product-search/en?x=0&y=0&lang=en&site=us&KeyWords=Ina217

[pezman]

I used to be partial to the TLO84C quad op-amp -- cheap, low noise,
super-high impedance etc. There are probably better amps out there
nowadays.

For a better quad op-amp, there is the OP470 -- but they are about $5
per package.

[pezman]

BTW, if we order, let's order a bunch -- always good to have amps around.  It's about the only sort of old-school chip that I'm apt to use.  I'd be happy to kick in.

Also also -- there are a bunch of precision amps already kicking around the space.  A lot of these are special instrument amps -- low noise. low offset, but not very fast.

Also (unrelated to the acoustic camera), we have lots of traditional op-amps.  These are "new" (i.e. untouched, but probably twenty or thirty years old).  They are interesting because they are models that are often used in old guitar effects and prized for their special, vintage quirks.

[Dave]

Just an update. I've talked myself into going the phased array route, just due to cost and practicality issues. I started putting some code together: it will run far slower than realtime and in fact will likely be excruciatingly slow, but that's fine with me. I've got a good start on my phased array processing system, code-wise. I'm shooting for simplicity, not efficiency or speed.

Here's basically what I can do so far:

- Specify one or more pre-recorded sound files and their desired locations in space.

- Specify a mic array geometry: currently just linear or grid arrays, number of mics is unlimited but I'm starting with 8 since I could conceivably test it by borrowing existing audio gear. Each mic has an arbitrary position in 3D space. Fancier array geometries can easily be added, the math doesn't change a bit.

- Simulated mic output is generated from the specified sound sources and their locations. Attenuation with distance isn't modeled, and I don't add any noise or mic gain patterns in simulation, but otherwise this should be quite close to reality. Multipath isn't modeled, this is just pure speed of sound.

- Specify a 3D location in space to focus the array, and automatically generate an audio file corresponding to it.

- Specify a virtual camera field of view and resolution, and produce the appropriate time delays for each mic corresponding to each pixel.

I need to implement a couple of things in order to actually start producing output images from the simulated audio inputs, but it's really close. If you focus the array on one of the sources and listen to the output compared to the unfocused output you can tell that the sound that's in focus is a bit more full, but the out-of-focus sound is only a little quieter to the ear. Using simple methods I can compare the response of the focused vs unfocused output, and in my test case the focused output is either 40% stronger or 100% stronger than the unfocused audio depending on how I evaluate it. So that seems promising as far as imaging goes: that corresponds to a very useful range of pixel intensities. If I added basic matched filters and could actually search for a known signal like a chirp I'm pretty confident it would be many times stronger than that. Might be fun to play with eventually.

So, very soon I should be able to generate images from simulated sound. Hopefully then I will be happy enough with the results to try borrowing an 8-channel audio input device and capture the real thing. After that I can decide whether to go ahead with the 32-channel audio input board design. The great thing about being able to simulate everything is that I can work out a lot of the details like array spacing beforehand, and get some idea of what will work the best. Should be fun!

-Dave

On Saturday, October 6, 2012 10:38:32 PM UTC-4, Dave wrote:

I've been kicking around this idea recently, and I think it just might work. The idea is for a camera that takes pictures with sound instead of light. If you have to ask "Why?" then you don't need one, and you may be on the wrong website! ;) I just think it would be fascinating to play with.

Most light cameras use a lens, but the simplest way to focus the incoming wavefronts onto your pixel array is with a pinhole. The pinhole can be almost any material that is opaque to light at around 550 nm wavelength. The sound equivalent might be some type of sound-deadening baffle, maybe a slab of drywall, or cardboard, or other insulating board. For light, the pinhole is often literally a hole made with a pin, but for sound the wavelengths are many orders of magnitude larger, and so everything has to get correspondingly bigger. For example if our acoustic camera were tuned to pick up roughly 10 kHz audio (wavelength in air is about 1.3 inches), if we had just 32x32 pixels at 1 wavelength pitch, the pixel array would be nearly 4 feet on a side. For a focal length of 2 meters (79 inches) the pinhole would be about 20 inches in diameter, for a 30 degree field of view. Potentially you could capture these low-res images and mosaic a bunch of them together for a much larger image.

That's all well and good, but 32x32 pixels means you need 1024 total microphones! Even if they cost 50 cents a piece (less than the cheapest mic on DigiKey) and could solder each one in 10 seconds, it would cost over $1000 and take 3 hours of soldering just for the mics! Plus you might need an amp for each one, some number of ADCs, some way to pull all that data together, etc. So here's what I'm thinking: you make the imaging array out of large PCB panels that you etch yourself, and you build the mics in yourself as part of it. I'm leaning towards homebrew condenser microphones using aluminum foil as the diaphragm. You'd need to develop a good way to make an array of these mics yourself but it seems doable to me. Here's a page showing one guy's own homebrew condenser microphone, clearly not streamlined to be made hundreds at a time, but still. I think with some experimentation you could get a system down for making 100 mics on a panel in just a few minutes.

http://kt4qw.com/condenser.pdf

I think with the right mic design, you could multiplex them all together using just two inexpensive mux ICs, a single amp, and a single ADC. It would probably take several seconds to capture each image since you'd only be pulling in one pixel at a time, but the processing would be really simple. You could even do an FFT and extract out different frequencies to be displayed in different colors. Different materials would show up in different colors. Practical? No way! It would sure make some cool pictures though.

[Chris Thompson]

From what I understand, this looks great! 

via mobile. 

--

[pezman]

Hey, you may have just solved my problem of finding an application for the Stellaris Launch pad. It ought to be able to do phased array calculations in real time.

[Dave]

What sort of phased array calculations? This was my Core 2 Duo laptop I was talking about: old, but I doubt the Stellaris has an edge over it. It might be doable with FPGAs or GPUs but probably only if you stripped it way back. You could do something with it for sure, I was doing basic transmit mode with an Arduino Pro Mini and there are lots of different applications. Receive-mode imaging is another animal though. It pretty much requires ridiculous brute force. Lends itself to parallel processing though.

[Dave]

Check it, yo:

http://creamyrobotgoodness.com/media/random/acousticCamera/animatedgif-test-8x8.gif

This is a half-second animation of the output of my phased array code. The input data is two mono recordings of my voice ('testing' and 'experiment'), placed at simulated locations 10 meters away from the array and a few meters apart. The 'camera' has 80x80 degrees field of view (totally arbitrary without changing the array). In this case the array has 8x8 elements (mics), but the results of a much smaller array (like 16 channels, maybe even less) may be good enough. The output is currently at 10 frames per second but that's arbitrary as well, you could probably crank it up to 1kHz if you wanted to watch individual echoes bounce back off of things. Still tweaking and refining, but I think this is a great start. Hopefully in not too long I can do color (for different frequency bands).

[Joshua D. Johnson]

I have no way of contributing to this project but find it very useful. I'm sure audiophiles of all flavors would be very interested in a higher res version with some sort of color registration/calibration.

Could this be used with IR as well? IR cans are very expensive.

Josh
Objects-unlimited.com

[Dave]

IR as in infrared? IR headphones? I'm not familiar.

[Dave]

[Dave]

[Dave]

[Chris Thompson]

That's super cool



via mobile.

[Joshua D. Johnson]

IR sensors., Not audio.
Seems like if you are going to develop a system capable of capturing sensor data in this way it could be applied to other realms. Just sayin

Josh

On Dec 14, 2012 9:06 AM, "Dave" <dgs...@gmail.com> wrote:

IR as in infrared? IR headphones? I'm not familiar.

[Kyle Yankanich]

Couldn't you just use Infrared film and a regular camera for that instance? The movie "Soy Cuba" was shot in B&W and Infrared film: http://youtu.be/eOLVm_9UcRw  Different films can reproduce different ranges of IR, so I guess depending on an IR sensor that can capture a broader spectrum of IR this would be interesting.

For anyone interested, also check out UV photography. Here's a difference of UV vs Regular vs IR photography:

-Kyle Yankanich
http://hive76.org
Publicist

[Dave]

Sorry for my post fail!

This relies on being able to sense the phase of the signal. It can be done but the sensors I know of that can do it are super expensive. Interestingly, microwaves are basically the same sorts of frequencies as audible sound. 802.11 (2.4GHz) is a couple of inches long for example, and this is basically how radar works, with antennas instead of mics.

[Joshua D. Johnson]

I would want to use it for IR Thermal imaging.. so film is too slow. Cams start at $1200 for a crappy one and quickly break $10K.

[Dave]

For thermal, the wavelength is way bigger than visible light, but not quite big enough to just treat it as RF and use antennas and such. Building a zone plate to focus thermal radiation would be practical, even easy, but you'd need an array of tons of temperature sensors, probably not practical for DIY. If you could find some cheap sensor that could give you the phase of the signal, you could go the phased array route. They do have drawbacks compared to traditional pixel-array-based cameras though.

[Joshua D. Johnson]

Hmm. I did see some arduino based projects that would do it but they relied on a servo to move the sensor for each pixel capture.

Keep up the interesting work!

Josh

On Dec 14, 2012 10:27 AM, "Dave" <dgs...@gmail.com> wrote:

For thermal, the wavelength is way bigger than visible light, but not quite big enough to just treat it as RF and use antennas and such. Building a zone plate to focus thermal radiation would be practical, even easy, but you'd need an array of tons of temperature sensors, probably not practical for DIY. If you could find some cheap sensor that could give you the phase of the signal, you could go the phased array route. They do have drawbacks compared to traditional pixel-array-based cameras though.

[Kyle Yankanich]

Joshua:

http://www.kickstarter.com/projects/andyrawson/ir-blue-thermal-imaging-smartphone-accessory?ref=card

-Kyle Yankanich
http://hive76.org
Publicist

[Dave]

Looks like they're using the 16x4 Melexis chip for thermal. I think it has an SPI and/or I2C interface, so if you could get ahold of that there'd be no need to wait for a Kickstarter or anything, super simple to set up.

The acoustic camera code is coming along well. I just implemented the bandpass filtering last night and have almost got it cranking out RGB images.

[Joshua D. Johnson]

Yeah, I saw it. I want to look further. It would need to actually have temp ref/range on side, be relatively accurate and work with my phone. They make some nice promises there..

[Dave]

The work is progressing. A few weeks ago I set up a linear 4-mic array with Brendan's help to verify some of this stuff. My program works with it, although the results aren't as good as in simulation. There are some things I can fix next time, like better matching of the mics, placing the sound sources farther away in the far-field, adding more mics, etc. Some things may be tough to deal with though, like background noise and multipath. In theory if I knew there was a noise source in a specific location it should be pretty easy to effectively create a null there and suppress it, and I look forward to trying that out some time, but that won't help with general background noise.

I've added a matched filter pixel evaluator, which should come in handy for things like active sonar or other times when I know what sound I'm looking for. That will be fun to play with eventually, building a 3D map of the room using sound.

Scroll about half-way down and see an example of the image from simulated linear mic arrays with 2-, 4-, 8-, and 16 mics. It gives you an idea of how much improvement you get in spatial resolution by adding more mics.

http://wiki.hive76.org/Acoustic_Camera

I'm having some trouble with a perceived catch-22 with phased arrays. It seems like higher signal bandwidth should result in better sound source isolation... but a grid array is really designed for a specific target frequency, with lower frequencies causing defocussing and higher frequencies resulting in grating lobes. So it kind of seems like I both need high bandwidth, and can't use high bandwidth. What am I missing? Do I just have to use a less regular mic array geometry? I've seen that adding random jitter to the array can help break up the patterns in the artifacts some, but I wonder just how far that can go.

Still cranking...

-Dave

[Chris Thompson]

So we didn't get the chance to apply for the knight arts grant: http://www.knightarts.org/community/philadelphia/finalists-named-in-knight-arts-challenge-philadelphia?utm_medium=twitter&utm_source=twitterfeed

All those other ideas are lameballs by the same old groups in philly. It's a shame they are afraid to do something new. 

But we're making it anyway, right? Lets figure out a design and budget and all that stuff. Maybe a way to have it pay for itself. 

When shall we blog the idea?

--

To post to this group, send email to hive76-d...@googlegroups.com
To unsubscribe send email to hive76-discuss...@googlegroups.com
For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

--

Chris Thompson, eagleapex.com
GPG key available
"Obscurity is a far greater threat to authors and creative artists than piracy" -Tim O'Reilly

[pezman]

Most of the semi-finalists seemed to be "fostering a sense of community by [fill in the blank] ..." rather than building a frappus.

If you were using the camera to "take pictures" of the voices local choirs or what-not and using the images to "foster a sense of community by posting on Facebook the musical sounds that emanate from choral faces", I think the camera would be in the mix.

Agreed that there the awards seem to do little to "foster" innovation of any real sort.

To post to this group, send email to hive76-discussion@googlegroups.com
To unsubscribe send email to hive76-discussion+unsubscribe@googlegroups.com

For more awesome goto http://groups.google.com/group/hive76-discussion?hl=en

[Chris Thompson]

I'm totally going to make some sweet art with it though

Chris Thompson, eagleapex.com
GPG key available
"Obscurity is a far greater threat to authors and creative artists
than piracy" -Tim O'Reilly

>>> To post to this group, send email to hive76-d...@googlegroups.com

>>> To unsubscribe send email to

>>> hive76-discuss...@googlegroups.com

>>> For more awesome goto
>>> http://groups.google.com/group/hive76-discussion?hl=en
>>
>>
>>
>> --
>>
>> Chris Thompson, eagleapex.com
>> GPG key available
>> "Obscurity is a far greater threat to authors and creative artists than
>> piracy" -Tim O'Reilly
>

> --
> To post to this group, send email to hive76-d...@googlegroups.com
> To unsubscribe send email to hive76-discuss...@googlegroups.com

[Sean McBeth]

Well you see, Mike, if said person attempting the fostering isn't a part of the community they are attempting to foster, then it clearly doesn't exist at all, so by creating a project that fosters a sense of it, you're really inventing a new thing entirely.

Or something.

On Mon, Jan 14, 2013 at 1:24 PM, pezman <mikeh...@gmail.com> wrote:

--
To post to this group, send email to hive76-d...@googlegroups.com
To unsubscribe send email to hive76-discuss...@googlegroups.com

[Kyle Yankanich]

McBeth is confusing me.

But I think he's making sense. We're not fostering a community. We're
building it. Does anyone have experience with Grant writing? Might be
worth going after more of these types of things.

-Kyle Yankanich
http://hive76.org

[Dave]

Bummer! That would have been cool to do it on a larger scale. But yes, we should still move forward and do it anyway.

I'd like to blog it soon but I'm trying to work out a couple of things first. There seem to be one or two things I haven't properly wrapped my head around yet, so I'm trying to read up and play with simulations to fill in the gaps. I feel like I'm relatively close, but not there yet. I will try to make open house this week if I can, maybe we can kick around some ideas for how to proceed once I've got it working better.