What if your ears could blink?
Chameleon is a variable hearing protection device that changes its attenuation according to the environment. Chameleon is targeted towards users in periodically loud environments—environments with significant noise exposure, but with periods of quiet where communication is key, such as construction sites, factories and machine shops.
Chameleon has won several prizes, including SYDE People's Choice, and the Norman Esch Entrepreneurship Award.
I was the product lead in this project, and played a pivotal role in all aspects of product design including defining requirements & benchmarks, designing the attenuation system, measurement circuit, and control logic, and implementing standard test procedures and analyzing the results.
The bulk of the first term (4 months) of this project was spent refining our problem, researching user requirements, as well as learning about acoustics physics. Though we started the Final-Year Design Project (FYDP) in September 2016, we didn't arrive at a reasonable problem area until early October. This time was spent researching music, audio and health spaces, and meeting audiologists, audio electronics engineers, and acoustics professors to come up with a reasonable and realistic design problem statement:
Workers in variably loud environments, where communication is key, often do not wear hearing protection in order to more easily communicate, and to avoid the tedium of constant removal and re-application.
We will design a hearing protection device to dynamically and effectively attenuate relevant moderate-to-loud noise as it occurs, and to not attenuate significantly at normal ambient sound levels.
User Requirements & Benchmarks
To validate that we were solving a real problem, we interviewed and surveyed over a dozen potential users working in the target environments. Many users admitted openly that they didn't regularly use hearing protection—even when they knew they should—because it was inconvenient and uncomfortable. If they were wearing hearing protection and needed to talk to a coworker, they would need to take the hearing protector off. Because of this tedium, many people wouldn't put on hearing protection during the loud periods between conversation.
Other reasons cited were because of comfort—standard hearing protectors can put a lot of pressure on the head and feel isolating, while earplugs can be hard to insert, especially when wearing work gloves. Findings from these interviews were corroborated by a number of academic sources, which study the comfort, and social aspects of hearing protection 1 2 3 4 5 6.
From all our research we were able to define six major areas that the product should perform in, and user requirements in each. We took these categories and set benchmarks based on government standards, the behaviour of other devices and other research.
|Product Attribute||User Requirement||Metric||Unit||Min||Target||Max|
|Attenuation||Shall protect against excessively loud noise||Reaction Time||ms||0||100||1000|
|Communication||Shall allow communication without removal in periods when communication is possible||Open Attenuation||NRR||-||0||6|
|Threshold to Close||dB (SPL)||77||85||90|
|Threshold to Open||dB (SPL)||50||55||65|
|Comfort||Should be comfortable to wear for a full work day||Weight||Grams||-||245||330|
|Ear should not touch inner cup||Inner height||mm||63||75||-|
|Cost||Should be competitively priced relative to similar products||Cost of Device||$||-||50||300|
|Should function for an entire workday||Operational time||Hours||8||12||-|
|Measurement Accuracy||Should accurately measure noise level||Measurement Error at 4kHz||dBSPL||0||-||3|
Durability was also an area of concern throughout the design process, given the use environment, but it was an oversight that we never explicitly defined durability benchmarks.
Our first prototypes were of the measurement circuit—a system to calculate the Sound Pressure Level (SPL) in A-weighted decibels (dBA) from the mic's signal. The design of this piece was based on that of a standard noise meter. In order to have a useful dB value that represents SPL as a human ear might hear it, the incoming sound signal is put through a special band-pass filter, known as an A-weight filter.
I took a circuit I found online for this filter, double checked it had the right poles/zeroes, and simulated it in CircuitLab.com to verify its behaviour. The filter worked as expected, though it did have a constant negative gain of about -6dB. This is not a problem since the signal would need to be amplified before passing through the filter in the first place.
Next, in order to find the amplitude of the incoming signal, I added a peak detector to the circuit. This removed the phase information, and left us with a DC signal representing the peak volume in the environment.
This DC signal was then input into an
analogRead pin of an Arduino UNO. Since all the gains in the circuit are known, we can easily calculate the voltage at the output of the microphone. Since the sensitivity of the microphone is given, the incoming noise level in dB(A) can be calculated like so:
This took a little tweaking since the component values and mic input voltage weren't precise. In the end we were able to get a relatively accurate measurement of the noise level reaching the microphone which we verified the SPL this using an app the CDC recommends, NoiSee.
I often got questions when demo-ing the prototype about why the filtering was implemented in analog circuitry as opposed to digitally. The answer for the first prototype is that the Arduino Uno was the only microcontroller we had access to at this point, and that I was more familiar with analog filters vs. digital filters. Once we had decided to implement the final prototype using a Teensy 3.2, which has enough processing power to do this kind of filtering (and I had become more familiar with digital filtering) I decided to keep this section of the design the same so I could spend more time working on parts of the prototype that didn't work yet. So once the measurement circuit was designed, not much changed (other than the specific op-amps to deal with new power requirements).
First Full Prototype
Our first prototype used a standard issue ear-muff, retro-fitted with two aluminum disks. Our goal early on was to make a device which could attenuate at all values between maximum an minimum attenuation. This could hypothetically be achieved by rotating the disks, and aligning them to allow enough sound to pass through. Using a mic placed inside the ear-cup, we measured the SPL reaching the ear, and used the Arduino to control a motor and rotate the top disk.
The first iteration of the control logic attempted to create a pseudo-PID controller using the target SPL as the reference signal, and would rotate the disks until the incoming SPL matched the reference. Due to the placement of the motor though, the noise of the motor could be heard inside the ear-cup, and made it so the device would never find an equilibrium point.
The second controller iteration did away with continuously variable attenuation, and focused on trying to either attenuate, or not attenuate. This worked better, but the noise of the motor still influenced the sound inside the cup. Because of this, we had to write a line in the controller to ignore inputs immediately after opening for the device to work properly. Also, since reaction time was a priority, this design had too much inertia to close in a reasonable time frame.
Although this prototype was noisy and slow, it did afford some attenuation in the closed position—albeit with some slit leaks—validating our hypothesis of variable attenuation by opening and closing a hole. Our second prototype focused on designing a better attenuator/actuation method, and refining the circuit and controller code. We also spent some time validating different attenuation methods, and testing the device.
Our second concept was an evolution of the first, with a new actuation method. Since the first prototype was slow to react, I decided to replace the motor with a solenoid. I found out that using a solenoid to get a disk to rotate is harder than it sounds.
After a bunch of math and searching for a solenoid with a long enough stroke length, I finally designed a device that did what we wanted. When the solenoid turned on, the disk would rotate to open. When released, a counterweight would close it—satisfying the need to fail safely. The idea was that using PWM control of the solenoid, we could have a partly open device.
The problem with this design though was that it could not achieve both requirements of response time, and fail safely simultaneously. By setting the default position of the disk, we could achieve fast response, but would sacrifice the fail safe. The solenoid here was also very heavy, and would very quickly put us over our weight restriction. We had to come up with a better design.
A Fork in the Road
Around the same time as we were evaluating the solenoid-piston actuator, we prepared an experiment to determine what attenuator was actually the best. We had taken all our design ideas, and evaluated them based on our criteria using a decision matrix. We took the top three, and ran them through a rough a qualitative acoustic test. We retrofitted these designs into an ear-muff, and played loud (but not harmful) noise. We then and asked participants to evaluate the attenuation heard for each design. While this method was not entirely scientific, it was a quick process that provided interesting data.
|Only Aperture (control)||Not much noticeable attenuation, some slight muffling|
|Solid Plug||Good Attenuation. Comparable to real HPD|
|Shutter (closed)||Good Attenuation. Similar to plug|
|Shutter (open)||No noticeable attenuation. Similar to control|
|Shutter (partly closed)||No Noticeable attenuation until aperture size was less than a millimetre. Some muffling as aperture got smaller. Reflects research by N. Trompette|
|Pie-Slices (closed)||Some Attenuation. Seal wasn’t perfectly formed (sizes were misaligned)|
|Pie-Slices (open)||No Noticeable attenuation. Similar to control|
|Pie-slices (partly closed)||No Noticeable attenuation until aperture size was less than a millimetre. Similar to partly-closed shutter.|
From only a few of these tests, it became clear that the "Pie slices" design (used in the solenoid prototype) was not the greatest performer, nor did the "Shutter" design succeed in achieving variable attenuation.
Backed by real science
While looking at the results of these tests, I came across research that was able to quantify the trends we were seeing. It suggested that the majority of attenuation variability essentially comes from closing small slit leaks, and variability decays exponentially as the aperture size increases 7.
That makes sense intuitively—take the analogy of closing a window to block noise from a party outside. There is no noticeable difference in noise level until the window is nearly completely closed. In order to achieve useful continuously variable attenuation, we would need incredibly accurate control of an actuator in the first few degrees of rotation.
Since continuously variable attenuation was not absolutely necessary to the success of the device (and deadlines were approaching), we decided to abandon continuously variable attenuation for a simpler "binary attenuation" design—the best of which was a plug-type design.
More details coming soon
Hsu, Yeh-Liang et al. "Comfort Evaluation Of Hearing Protection", International Journal of Industrial Ergonomics, vol.33, pp. 543-551 (2004) ↩
Park, Min-Yong et. al. "An Empirical Study of Comfort Afforded by Various Hearing Protection Devices: Laboratory versus Field Results", Applied Acoustics, vol. 34, pp. 151-179 (1991) ↩
C. Stephenson and M. Stephenson, "Hearing loss prevention for carpenters: Part 1 - Using health communication and health promotion models to develop training that works", Noise and Health, vol. 13, no. 51, p. 113, 2011. ↩
D. Gower and J. Casalvi, "Speech Intelligibility and Protective Effectiveness of Selected Active Noise Reduction and Conventional Communications Headsets", Human Factors: The Journal of the Human Factors and Ergonomics Society, vol. 36, no. 2, 2016. ↩
Acton, W. J., "Effects of Ear Protection on Communication", The Annals Occupational Hygeine, vol. 10, pp. 423-429 (1967) ↩
E. H. Berger, "The Effects of Hearing Protectors on Auditory Communications", Aearo Company (1979) ↩
Trompette, N., Barbry, J.L., Sgard, F., Nelisse, H., "Sound transmission loss of rectangular and slit-shaped apertures: Experimental results and correlation with a modal model". The Journal of the Acoustical Society of America, vol. 125, no. 31 (2009) Available: http://asa.scitation.org/doi/full/10.1121/1.3003084 ↩