Car ultrasonic distance transducers as parametric speakers

[Recyclojunk64]

I recently obtained about a hundred ultrasonic distance transducers that were designed to be used in a car reversing system. I was hoping to use some of them to build a parametric speaker (modulates audio frequencies with an ultrasonic frequency (40kHz for these transducers) and projects it as a narrow beam of sound that's only heard by people in the path of it). Somehow it's distorted when it hits an object and becomes audible.

I got the idea here: http://hackaday.com/2012/03/05/speechjammer-puts-an-end-to-annoying-speakers/ , although I'm not necessarily going to limit it's applications to just speech-jamming 

I plan to use only about 30 or so. However, the transducers I have are in somewhat waterproof housings (as opposed to the bare pizeos), and I'm guessing that they're meant to have a wide projection angle since they’re meant to detect objects/children not directly in front of them. Do you think this could be a problem to the projection of the ultrasonic sound? Hopefully there's someone here who has some experience in sound waves...

EDIT: here are some images.

Transducer:


And here are the packages they came from.

Full size: http://www.recyclojunk64.comxa.com/Images/Boxes.jpg
I came across these boxes out on someone's verge collection (same house as the one pictured on the packaging). BTW, these are meant to be sold with the main unit (that goes in the car), to attach to a trailer/caravan. They are useless on their own. I guess the guy who ordered/designed them over-estimated the number of people who would want to attach one to the back of their 10 metre trailer. I have other plans for all those pizeo alarms as well 

[David_AVD]

I have a project in mind that could use some of these sensors.  Not sure if you want to get rid of some or if you're anywhere near me (Brisbane) ?

[ve7xen]

The principle should be similar to that of a phased array antenna, as often used in radar. Wikipedia tells me this is called 'beamforming' for sonar, basically what you want to do here (with audio modulation). It works based on destructive/constructive interference, and typically is used with a fairly omnidirectional radiator, so as many patterns as possible can be generated. Seems to make sense that you'd actually want it to be as isotropic as possible, so you are in complete control of the output pattern. I'm not totally sure on what effects would 'demodulate' it back to audible sound, but I imagine there's plenty of nonlinearity laying around reality to make it happen. You should be able to experiment with the modulation part of it with a single transducer / driver.

Technically, this shouldn't be all that hard to imagine with the parts that you've got, but I think actually realizing it would be a pretty large project. Audio frequency is relatively low, so the DSP performance requirements and analog design wouldn't be that stringent if you do it that way, though still fairly substantial and you've got the many channels of D to A problem to deal with. And plenty of highly parallel code. A fast PWM running straight into the transducer drivers might work fairly easily, if you have a way to do the signal processing quickly enough.

Doing it in the analog domain with delays could be 'simpler', but then you've got to deal with large tunable delay networks and tuning that would not be very fun, and it's practically unadjustable (lengthy tuning required to change parameters).

Fun experiment if you've got some signal processing background I think (I don't), otherwise a serious challenge.

[Marco]

Just in case you weren't aware there was also a kickstarter for this kind of speaker with open design files, Soundlazer (site is down for me at the moment though).

[mark03]

It would help to know the center frequency and bandwidth of the transducers.  From that, and their diameter, you can determine what the radiation pattern of a single transducer will be.  In particular, if the wavelength in air is much larger than the diameter (c/fo >> D where c is ~ 350 m/s) then they will radiate a single broad lobe without nulls in the far field (near field is a different story).

The previous poster is on the right track with the phased-array idea.  Note, however, that to localize the energy in three-dimensional space you require a *2-D* array, so the task is even more complex than already described.  Practical 1-D ultrasound arrays, e.g. in medical imaging, have rectangular elements which are much longer in one direction than the other, and a cylindrical lens on the front to provide a weak (high F number) fixed focus in one plane.  Then the phased-array part is used to steer in the other plane.

If these are in the right frequency range you could experiment with meteorology applications.  Google "ultrasonic rain gauge" for some interesting work using rayleigh backscatter to estimate droplet size and rainfall rate.

[Experimentonomen]

You are basically trying to replicate the "Audio Spotlight" thing from what i can tell, Correct ?

[DenzilPenberthy]

A mate who I was at uni with did pretty much this exact thing for his final year Masters project. If you were interested I could ask him if he has an electronic copy of the write-up...

[Recyclojunk64]

Thanks for all the replies guys.
I haven't done much work on it since my first post, other than put some transducers into an array.



This is really just a mock-up, the final version will probably have 37 transducers (an extra ring of them) and also a better substrate than MDF.



The original idea was to try to drive them in parallel with a sound-card and amp, but the frequency response of the sound-card would barely even get past 18kHz. A test on my function-generator with amplitude modulation proved fruitless as well, as far as audible sound production goes.

Technically, this shouldn't be all that hard to imagine with the parts that you've got, but I think actually realizing it would be a pretty large project. Audio frequency is relatively low, so the DSP performance requirements and analog design wouldn't be that stringent if you do it that way, though still fairly substantial and you've got the many channels of D to A problem to deal with. And plenty of highly parallel code. A fast PWM running straight into the transducer drivers might work fairly easily, if you have a way to do the signal processing quickly enough.

Doing it in the analog domain with delays could be 'simpler', but then you've got to deal with large tunable delay networks and tuning that would not be very fun, and it's practically unadjustable (lengthy tuning required to change parameters).

Fun experiment if you've got some signal processing background I think (I don't), otherwise a serious challenge.

So do I have to drive them all individually? I could probably could manage that with all the hardware that came with the transducers (and a lot of time and effort), but the actual signal processing may pose a problem, especially since I wanted this to be portable. If I did it this way I'd probably do it with DACs and multiple amps rather than muck around with adjusting analogue delay lines.

Just in case you weren't aware there was also a kickstarter for this kind of speaker with open design files, Soundlazer (site is down for me at the moment though).

Had heard of the soundlazer' but it seems that their actual design files (such as the schematic and parts list) aren't available to the public yet (so much for opensource).

It would help to know the center frequency and bandwidth of the transducers.  From that, and their diameter, you can determine what the radiation pattern of a single transducer will be.  In particular, if the wavelength in air is much larger than the diameter (c/fo >> D where c is ~ 350 m/s) then they will radiate a single broad lobe without nulls in the far field (near field is a different story).

The output wavelength would probably be maybe 9mm or so. The diameter of the transducers is 15mm, almost twice the size. I guess this means that the projection will have more throw than flood, but I'm pretty sure they were designed to project it in wide angle to cover as large an area as possible

I will probably have to use something like the cones the SoundLazer uses then:

You are basically trying to replicate the "Audio Spotlight" thing from what i can tell, Correct ?

Correct



I will probably try modifying a soundcard for a higher frequency response and drive the speakers through a modified amp for now and see if that does anything. I have heard that I need to use SSB modulation for it, but can't seem to find information on how SSB is different from normal amplitude modulation. Otherwise I might have to try something else more complicated, such as driving them individually, maybe hacking the original controllers to drive them.

[mark03]

It would help to know the center frequency and bandwidth of the transducers.  From that, and their diameter, you can determine what the radiation pattern of a single transducer will be.  In particular, if the wavelength in air is much larger than the diameter (c/fo >> D where c is ~ 350 m/s) then they will radiate a single broad lobe without nulls in the far field (near field is a different story).

The output wavelength would probably be maybe 9mm or so. The diameter of the transducers is 15mm, almost twice the size. I guess this means that the projection will have more throw than flood, but I'm pretty sure they were designed to project it in wide angle to cover as large an area as possible

This is good for your purposes; it means each element of your array will have a fair bit of directionality all by itself.  You won't be able to steer the beam much side-to-side by phasing the elements, but you didn't need to do that anyway.

The piezo transducers are pretty high impedance; to get decent power you usually drive with high voltage (tens of volts at least).

[ve7xen]

The original idea was to try to drive them in parallel with a sound-card and amp, but the frequency response of the sound-card would barely even get past 18kHz. A test on my function-generator with amplitude modulation proved fruitless as well, as far as audible sound production goes.

Do you have any microphones that work at these frequencies? It would probably be worthwhile to figure out what frequency you get maximum output at (if you don't have full specs). I think these transducers are pretty narrowband. As mentioned you'll probably also need a fair amount of power if you're counting on nonlinearities to produce audible output.

Fun experiment if you've got some signal processing background I think (I don't), otherwise a serious challenge.

So do I have to drive them all individually?[/quote]
If you want it to be parametric, then yes. Otherwise you've just got an in-phase array. You'll get a different pattern, and it might be the pattern that you want or could be dimensioned to be, much like how you build a multi-element antenna vs. use a phased array. Loudspeaker modeling software might actually be able to handle that if it's what you're trying to do (ie. just increased directionality). This is generally something speaker designers try to minimize.

Not very knowledgeable on this, just throwing ideas out there.

[wemme]

Hello,
FYI These are available for about 1-3 USD from aliexpress some suppliers have some data available also.

Most transducers are designed to run at 40Khz (or other set frequency) I'm not sure if that is a resonant frequency or not but running them off their designed frequency will usually reduce their output level. also note manufactures quoted dB spl may be quoted at different distances keep in mind inverse square law (6dB drop when you double the distance)

You can use another transducer as a microphone (connect it to a scope) and you can then tune the frequency to a peak.

One version i have seen of the sound spot light used two arrays one with a 40Khz carrier and another running the same carrier Frequency modulated with audio 300Hz - 3Khz at the convergence point the audio signal should be audible. this may be easy to try with a decent signal generator.

Piezos usually also run at higher voltages but keep in mind power dissipation as piezos are electrically similar to capacitor and so Xc will reduce as the frequency increases.

I am interested in building a high power ultrasonic dog/cat repelling "fence" to keep the them off my lawn.
has any one had any experience with this subject?

So far my experiments have either ended up with cats doing somersaults and Piezos speakers venting with flame   Perhaps too much power 
Cheers