Harvesting Sound Energy From Passing Cars





Introduction: Harvesting Sound Energy From Passing Cars

About: I am a University of South Carolina Mechanical Engineering Graduate Student doing ultrasonic wave propagation research in the field of structural heath monitoring. I happen to enjoy making things in my spare...

There is energy everywhere around us and in many different forms. Many devices have been developed to harvest light, wind, waves, and more. One unusual place of energy harvesting is from passing cars. As cars pass by some of their energy is released in form of sound. Even though the overall energy maybe small it can be harvested. In this Instructable I will show how to apply the solution of Euler–Bernoulli beam theory to design a cantilever beam to oscillate at such a frequency to adsorb sound waves as well as converting its mechanical motion into electricity.

Step 1: Materials and Prices

The materials to make one or a similar device can vary based on what someone wants. The most important part is that the oscillator (described in next step) is cut to proper dimensions and firmly clamped to an object so that its vibrations will have little dampening (Less dampening = more power). A few of my components were salvaged from used metal plates and old circuit boards. I encourage testing out your own designs with varying parts. Anyway, this is what I used totaling $39.37:

  1. 20 pcs piezoelectric wafers..............................7.07 (need for number of oscillators made)
  2. 20 pcs MB6S full wave rectifiers......................6.43
  3. 10 pcs Printed circuit board (PCB)...................5.99
  4. Tripod................................................................Already owned (suggested but not needed)
  5. 4 pcs 1/4 x 2 inch bolt.......................................4 x 0.25 (estimated)
  6. 4 pcs 1/4 inch washer.......................................4 x 0.10 (estimated)
  7. 3 x 0.75 x 0.75 inch square rod.......................Already owned (Not needed)
  8. 2 pcs 3 x 2 x 0.25 inch plates..........................Already owned (Not necessarily needed)
  9. Hookup wires...................................................Already owned
  10. 2 pcs 10K ohm resistors..................................Already owned (Could use others)
  11. 2 pcs 4.7 micro F capacitors...........................Already owned (Could use others)
  12. 24 x 12 x 0.062 inch Nylon 6/6 sheet...............18.48
  13. 6061-T6 1mm thick Aluminum sheet................Already owned

One will also need a device to cut, drill, solder, glue and more. I suggest one watch/read this and recognize what tools are needed for what project that it may be applied to. For example, changing the 0.25 inch metal plates to 0.5 inch wood could potential work just as fine if metal working tools are not available.

Step 2: Creating the Oscillator

The oscillator is the fundamental piece on this device. However, it is nothing more than a cantilever beam (a protruding beam with one fixed end) with designed dimensions. This step explains the calculation of the dimensions.

To begin, sound not only can be seen within a time domain but also a frequency domain. When one records sound it is in a time domain, meaning it can be plotted on a time vs amplitude graph. To transform the audio sound from its time domain to the frequency domain a Fast Fourier Transform (FFT) was preformed. This was done in MATLAB software, but other free software is available. After obtaining the sound of passing cars using an audio recorder the signal (sound waveform) was transformed into the frequency domain (seen in FFT from passing vehicles). From that image we see that the the highest amplitudes lay within 0 to 75 Hz and 100 to 250 Hz. From this a beam can be designed where it's natural frequency would correspond with the peak power frequencies from passing cars.

For experimentation I choose to make 3 oscillators: an aluminum 6061-T6 dual 8 inch that operates at 20.55 Hz, a Nylon 6/6 dual 8 inch at 8.12 Hz, and a Nylon 6/6 2 inch and 3 inch at operates at 129.99 and 57.77 Hz respectively. Note: these are just the first mode resonance peaks. This was calculated using the equation in the "1mm thick natural vibrations of various materials" image. The equation is the derived expression for a rectangular cross-sectional beam from Euler–Bernoulli beam theory. C is the vibration mode constant (multiple modes exist) for the first vibration mode. Also in the image are length vs frequency plots of 1 mm thick Nylon 6/6, 6061-T6 aluminum, and 304 stainless steel. The two metals (aluminum and stainless steel) behave similarly. Nylon, because of its lower elastic modulus has a much lower natural frequency trend.

These 3 oscillators will be tested to see how each will perform in laboratory and field testing.

Step 3: Device Hardware and Assembly

Each hardware component with it's dimensions are shown if the first picture. Based on your desired operating frequencies will determine the length and material. For my oscillators I choose 6061-T6 aluminum and Nylon 6/6. For one of my oscillators (a second nylon) the dimensions of the lengths are 3.5 inches and 2.5 inches instead of the 8.5 inches in the dimensions figure. The oscillator in the pictures was not used for testing. Regardless, it is assembled in the same way.

Assembly is straight forward and easy. The series of pictures shows how the components are assembled. Before the piezoelectric sensors are bonded on the oscillator, a test fit should first done. Then a pencil will be used to mark the point where the top of the base plate meets the oscillator. Next they device will be dissembled. Bond the piezoelectric sensors just above this line to avoid them from being crushed. Once they are glued reassemble the device. Wires now can be soldered to the wafers. Insure a tight clamp on the oscillator by securely tightening the nuts.

Step 4: Electric Components (Power Generating)

The electrical system consists of the oscillator, resistor, and capacitor (in image). The oscillator will act as the voltage supply, however the voltage produced is alternating. To convert the voltage to DC a MB6S full wave rectifier will be used. The effects of the rectifier can be seen in the second image where the signal was measured after a flick was applied to the oscillator.

The resistor and capacitor were added to measure the system. The system follows the equation above (where τ = R*C) assuming the voltage in (Vin) is a constant power supply. However the voltage in will be sporadic but there can be a somewhat close estimation of the actual voltage produced. Using the equation above and inputting Vin = 1 v, C = 4.7 μF, R = 10 kΩ results in a near full charge within a half of a second. Realistically, since the voltage in varies, the time it takes to reach that value will take longer, however the voltage gain will eventually level out as it reaches the input voltage.

Step 5: Electric Components (Voltage Recording)

Early testing using a voltmeter seemed to drain the capacitors' voltage. To decrease the drain and to better record the data an Arduino nano along with a SD card will be used. Before each test the capacitors are drained using the two jumper wires seen in the image above (white wires left of SD modular). The code is attached as well as a simple wiring schematic. This device was only used in field testing.

Step 6: Testing and Results

Before I took the device to the road for experimentation I decided to preform laboratory testing first to see where the resonance frequencies were and the normalized voltage amplitudes between the oscillators. To record the data, I borrowed an oscilloscope after a few failed attempts with the Arduino device (too low of voltage to be detectable at a too low sampling rate).

Design parameters for the test follows; Each oscillator was exposed to a sub woofer as it played a frequency sweep from 1 to 200 Hz over a duration of 20 seconds. The voltage from the piezoelectric wafers was measured at a sampling rate of 40 kHz from the wafers themselves and not the capacitors. This was to see not the stored voltage but the voltage being actively produced.

The website for generating the audio frequency sweep: http://onlinetonegenerator.com/frequency-sweep-gen...

Reviewing the data the true resonance peaks can be measured (from frequency plots), seen in the table. The aluminum provided the most accurate results, however all the nylon resonance peaks varied in error from 10.78 to 20.32%. This is likely due to a wrong value of Young's modulus since the raw material did not come with a data sheet and the value was guessed from the lower range of Nylon property values.

The aluminum oscillator's second mode, with a C=4.694^2 and a resulting peak of 126 Hz, had significantly higher amplitude.

Since the output voltage to the sub woofer was unknown but remained constant throughout the testing, the signals can be compared to see which oscillator produced the most power. The aluminum oscillator, even though it was it's second resonance mode, produced significantly more power varying from 60% to 70% more from its nylon counterparts. All the nylon oscillators produced nearly the same power, regardless of their resonance frequencies. The nylon producing lower voltage could be due to their higher damping.

After the data was gathered and analyzed the device was then prepared for field testing.

The device was taken to the same street and location where the audio waveform FFT was taken from. Before each experiment the capacitors were shorted out to discharge their voltage. Voltage was recorded every second for 10 minutes for each oscillator. The time of testing occurred at 10:32 am till 11:11 am under sunny conditions with sporadic wind gusts of 0 - 5 mph. After the data was collected the voltage was then plotted against time.

Analysis of the Energy harvesting field experiment results seem to reveal the mild gusts of wind, that accelerated the voltage increase, especially on the 8 inch nylon oscillator (which was the most prone to such effects). These voltage increases are indicated with the highlighted circles in the images. From the charging capacitor equation, the slope will approach 0 when maximum voltage is obtained. Therefor a power of two polynomial tread lines are added and extrapolated past the recorded data to have a somewhat rough estimate of the input voltage. These numbers and tread lines are shown in another image. The 2 inch and 3 inch nylon oscillators did not have a conclusive tread line. Where the slope turns to 0 are indicated and corresponding voltages recorded. The 8 inch nylon left oscillator preformed the best with an estimated input voltage of 0.275 volts, most likely due to energy gathered from wind. Oddly enough the right 8 inch nylon oscillator performed much worse with an estimated input voltage of 0.16 volts. The aluminum oscillators performed consistently with an estimated input voltage of 0.225 volts.

Step 7: Conclusions

In the laboratory testing the various oscillators were examined to see the accuracy of the calculations and to compare the voltages. The aluminum oscillators produced significant more power on their second vibration mode than any of the nylon oscillators. Also the aluminum oscillators suffered little error while the nylon oscillators error ranged from 10-21% possibly due to a wrong property value for Young's modulus (E).

Once in the field, the 8 inch Nylon oscillator had an maximum input voltage of all of roughly 0.275 volts of power harvested from the sound of passing cars. However the energy produced was certainly effected by a 0 - 5 mph breeze which produced significant jumps in input voltage of the 8 inch Nylon oscillators. The aluminum and 2 inch and 3 inch Nylon oscillators seemed to be less resistant to the effects of wind. The most consistent results was from the aluminum oscillators which both reached an estimated input voltage of roughly 0.225 volts. The 2 and 3 inch Nylon oscillators were inconclusive on the precise input voltage as their tread lines failed to reach an extrapolated peak.

Improvements could be made to optimize the vibration modes to better match the energy available since higher frequencies contain more energy. Also more metals should be explored since they have less damping.

From the experiments it is clear that harvesting wind with this device was more effective. If the oscillators were lengthened, significantly more power could be harvested from a mild breeze. However, the point of this project is to not only create a nifty device but to also open the mind to abstract ideas. Even though sound harvesting of cars might not be a useful way to produce power doesn't mean that the concept is completely useless. The fact that the device is actively removing sound energy means that for high traffic areas such as highways, prone to noise pollution, could benefit from a similar device. If sound is being turned into mechanical motion the volume could potentially be decreased in such an environment.

Step 8: Update: Raw Voltage Produced

Due to questions and concerns regarding the voltage in, I went out and collected the raw signals from the wafers themselves using an oscilloscope. Sure enough the wind produced the highest amplitudes seen in the long nylon figure (8 inch oscillator). However, no passing vehicles could be seen in that oscillator's signal. Passing cars can be seen in the other figures. It appears that the aluminum preformed the best.

Hopefully this resolves a few questions in the comment section.

Explore Science Contest 2017

Second Prize in the
Explore Science Contest 2017

Invention Challenge 2017

Runner Up in the
Invention Challenge 2017



  • Clocks Contest

    Clocks Contest
  • Creative Misuse Contest

    Creative Misuse Contest
  • Oil Contest

    Oil Contest

35 Discussions

how to generate more electricity from this at least 50 volts .if possible give me full information and circuit on my email

We're going to innovate this experiment. With our background in electronics, we're going to produce more voltage than this experiment. Any suggestions is appreciated. Thanks.


11 months ago

Could be a less expensive means of powering remote equipment than photovoaics

As an engineering exercise, this is very interesting. I also liked the detailed presentation. But probably The sum total of energy spent in the manufacture of all the components used in the device and that spent in assembly will be greater than that scavenged by this device through it's entire lifetime! The big problem we face today is primarily with economical energy storage, not as much with generation.

1 reply

Great Instuctable and looks fun to try out!

Thanks and please make more instructables

How about putting the collection device ON A CAR...lots of wind noise can be heard from a moving car.

..and GREAT Instructable!

1 reply

On a car - that would increase car's air resistance. So, yes, you would get electrical energy (with very poor efficiency) at a cost of burning more gasoline.

instead of the pizos, how about using speakers, they will generate power, using them as microphones.

My first thoughts on this? A fence made up of the metal plates running alongside a highway or motorway for us in the UK. These plates then finally being hooked up to a battery system every so many yards to then feed a high efficiency but low power requirement LED lighting system for night use of the road. Possibly to power emergency road signage too. It could easily complement the solar powered systems we have in the UK. It would also have the added benefit of cutting down on some of the traffic noise that would otherwise reach residential areas.

This device won't "suck up" sound but rather absorbs (some of) the sound that hits it. Think of sound as a ripple in a pond and this harvester is a 'stick' you are using to collect the 'ripple energy' (said not for your benefit MA, but some other commenters :) )

Acoustic energy for everyday noises is typically quite small. For example, assume you measure 85 dB (dB re 20 microPa) at the location of your energy harvester. Assuming spherical spreading of an idealized acoustic source That would correspond to about 0.3mW of power for every square meter of surface. Your largest beam looks to be about 10 square inches (.0065m^2) which means that in the best case, the harvester has about 20 microWatts of potential acoustic power impinging on it. 20microWatts at your 0.2 volts is 0.1mA . Although there will undoubtably be losses, it's interesting to see 'best case'

Cool project!

1 reply

Sorry, that's 2 microWatts of acoustic power and 9 microAmps at .2Volts

The rectifying bridge drops your voltage by two diode drops; you might do better with Schottky diodes (lower Vfwd).

However, the big question is how much power is developed. You could try inserting a current meter + a resistor across the cap, but I suspect it will be very hard to measure because both the voltage and current are quite low.

A better approach would be to put various size resistors across the capacitor, while measuring the voltage on the capacitor - these will simulate voltagae and power when carrying a load. Suppose you measure 0.2v sustained with a 100Kohm resistor; that would indicate that it generates 2uA of current at that load, or 0.4 uW. Or suppose that it shows 0.1v with a 10K resistor - that would be 10uA and 1uW. Suppose you could get 0.15v with a 1K resistor (unlikely), that would be 150uA and 23uW.

The forward drop of most diodes are not specified at such low currents, so I don't know if a Schottky bridge would help or not, but you might try it.

I wonder if it's helpful to have a tuned mechanical resonator, when your sound looks more like white or pink noise, so you are extracting only a thin frequency slice of the broad spectrum sound energy. The small peaks in the FFT can be meaningless noise (noise as in signal to noise ratio, not as in sound) that disapear if you average many samples, or due to one set of tires passing at a given time, not to be repeated often.

And finally, piezo elements can produce much higher voltages; I wonder if you have managed a very efficient coupling between your mechanical oscillators and the piezo element. You want as much as possible of the mechanical vibrations to be damped/absorbed by the piezo. Try another mounting approach.

Thanks for the thought provoking instructable.

1 reply

Thanks! and that's exactly what I plan to do once I get my hands on an oscilloscope so that I can test the raw voltage from the wafers. (Which means I can cut out my weak circuits and electronics background haha)

Yes, designing a device to operate on a wide freqeuncy band would be way more powerful. My first design consisted of 5 oscillators of various lengths to collect a few more slices of the FFT range... but decided to keep it simple and went with dual oscillators.

And I have a hard time seeing much more than a half a volt from the wafers in an experiment like this. There just isn't enough strain produced from pressure waves.

Thanks again and now I know what a Schottky bridge is! haha

If it looks like a speed detector then it might slow drivers to the speed limit. Thereby increasing its value! ?

1 reply

Great work, turning sound into energy that is just going to waste.


11 months ago

How I long to see this developed to the point of practicality! Highways and subway lines produce sound in limitless volume.

1 reply

And if you live by the train tracks at least there's one up side!


11 months ago

Because wind provides a big boost in output I can forsee many of these devices along heavily used highways where the wind generated by the vehicles adds most of the energy. Wind from vehicles changes direction and velocity rapidly. It could be that resonant metal strips are more efficent in capture than typical spinning blades.