MEMS digital microphone and Arduino compatible microcontroller: an embedded system for noise monitoring

Abstract

Noise assessment and monitoring are essential parts of an acoustician's work. They help to better understand the environment and propose better solutions for noise control and urban noise management. Traditionally, equipment used to carry out these tasks is standardized and often times is considered expensive for the early career professional. This study has developed a high-quality (and cost-effective) prototype for an embedded noise monitoring device based upon a digital I2S MEMS microphone and an Arduino compatible microcontroller named Teensy. Its small size and low power consumption are also advantages designed for the project. The system captures and processes sound in real-time and computes A and C frequency-weighted equivalent sound levels, along with time-weighted instant levels (with a logging interval of 125 ms). Part of the software handles the audio environment, while the biquadratic IIR filters present in the Cortex Microcontroller library are responsible for the frequency- and time-weightings - using floating-point for enhanced precision. The prototype performance was compared against a Class 1 Sound Level Meter (SLM), rendering very similar results to tested situations, proving a powerful and reliable tool. Improvements to the system and further testing will be conducted to refine its functioning and characterization. Ultimately, the prototype demonstrated promising performance, achieving LAeq values within a +/-0.5 dB margin over the Class 1 SLM and confirming its potential as a cost-effective solution for noise monitoring and assessment.

Full text

MEMS digital microphone and Arduino compatible microcontroller:
an embedded system for noise monitoring
Felipe Ramos de Mello1
Acoustical Engineering, Federal University of Santa Maria
Av. Roraima nº 1000, Cidade Universitária, Bairro Camobi, 97105-900, Santa Maria, RS, Brazil
William D’Andrea Fonseca2
Acoustical Engineering, Federal University of Santa Maria
Av. Roraima nº 1000, Cidade Universitária, Bairro Camobi, 97105-900, Santa Maria, RS, Brazil
Paulo Henrique Mareze3
Acoustical Engineering, Federal University of Santa Maria
Av. Roraima nº 1000, Cidade Universitária, Bairro Camobi, 97105-900, Santa Maria, RS, Brazil
ABSTRACT
Noise assessment and monitoring are essential parts of an acoustician’s work. They help to better
understand the environment and propose better solutions for noise control and urban noise management.
Traditionally, equipment used to carry out these tasks is standardized and often times is considered
expensive for the early career professional. This study has developed a high-quality (and cost-effective)
prototype for an embedded noise monitoring device based upon a digital I2S MEMS microphone and
an Arduino compatible microcontroller named Teensy. Its small size and low power consumption are
also advantages designed for the project. The system captures and processes sound in real-time and
computes A and C frequency-weighted equivalent sound levels, along with time-weighted instant levels
(with a logging interval of 125 ms). Part of the software handles the audio environment, while the
biquadratic IIR filters present in the Cortex Microcontroller library are responsible for the frequency-
and time-weightings — using floating-point for enhanced precision. The prototype performance
was compared against a Class 1 Sound Level Meter (SLM), rendering very similar results to tested
situations, proving a powerful and reliable tool. Improvements to the system and further testing will
be conducted to refine its functioning and characterization. Ultimately, the prototype demonstrated
promising performance, achieving
LAeq
values within a
±
0.5 dB margin over the Class 1 SLM and
confirming its potential as a cost-effective solution for noise monitoring and assessment.
Keywords: Arduino, CMSIS, Inter IC-Sound, I2S, Teensy, Sound Level Meter, MEMS microphone, noise
monitoring.
PACS: 43.58.Fm, 43.50.Yw, 43.60.Qv, 07.07.Df, 43.38.Kb .
1felipe.mello@eac.ufsm.br.
2will.fonseca@eac.ufsm.br.
3paulo.mareze@eac.ufsm.br.
‡This is an extended version of the original paper.
§
Cite this article: F. R. Mello; W. D’A. Fonseca; P. H. Mareze. MEMS digital microphone and Arduino compatible
microcontroller: an embedded system for noise monitoring. In 50th International Congress and Exposition on Noise
Control Engineering — Internoise 2021, pages 1–12, Washington, DC, USA, Aug. 2021. doi: 10.3397/IN-2021-2557.
doi: 10.3397/IN-2021-2557. Page 1 of 22

1. INTRODUCTION
The objective of this work is to create an embedded system focusing on sound pressure level (SPL)
and equivalent continuous sound level (
Leq
) measurements. That is, the designed software is written as
firmware into the hardware, creating a prototype dedicated for this sole purpose. In consequence, the
direct application is sound/noise monitoring, i.e. measuring, processing, and storing levels according
to sound level meter (SLM) standards. To achieve this goal, instrumentation, programming, and digital
signal processing (DSP) develop essential roles. Several branches of acoustical engineering are put
to work together as a unity. The hardware includes a microphone, a processing unit, a storage device,
and a power source. Therefore, the DSP must be meticulously programmed to fulfill time-weighting,
frequency-weighting, and fractional octave filtering.
Microelectromechanical (MEMS) microphones have inspired a new way of thinking in using sound
recording devices. Its relatively low-cost, low-power consumption, and tiny size have enabled designs
from single prototypes to massive applications in the cellphone industry, for instance. Arduino is a
microcontroller unit (MCU). As such, its development boards and kits have brought up an easy way to
program hardware and produce virtually any type of electronic device. The combination of these two
can render unlimited options when dealing with sound, from toys to professional projects. Considering
the aforementioned, this study used the Teensy board — an Arduino compatible but with much more
muscle and add-ins — to build an SLM. Applications range from home monitoring through sound
security systems to urban mapping.
2. THEORY, HARDWARE AND SOFTWARE
This section addresses the principles of the pressure-related quantities tested within this study,
as well as a brief description of the employed hardware and basic knowledge on the programming
approach used. The authors also encourage further reading, such as Pierce [
1
] and Jacobsen & Juhl [
2
],
to get a better understanding of sound.
2.1. Sound pressure and Sound Pressure Level (SPL)
Sound pressure is created by fluctuations over the static pressure of the medium, commonly air
and water, for general acoustics. As a scalar quantity, it represents a series of different potentials that
yields particle velocity. Accordingly, sound intensity is produced as the product of sound pressure and
particle velocity and is equivalent to the power per unit area [
3
]. Considering the electroacoustical
analogy (impedance mode), sound pressure is analogous to voltage. As such, its effective value (or
RMS, root mean square) is adequate for the math involved, as a time-varying quantity.
The sound pressure level (SPL) is given by
SPL =10 log10 prms
p0!2
dB (1)
with
prms =q⟨p2(t)⟩t=s1
TZT
0
p2(t) dt=v
u
t1
N
N
X
n=1
|pn|2,(2)
in which
p0
=20
µ
Pa is the reference pressure (at 1 kHz),
prms
is the root mean square (rms) pressure,
T
is the measuring period of time (or time-window in s),
p
(
t
) the instantaneous value of the sound
pressure (in Pa), and
t
is time (in s). Considering discrete values,
N
is the total number of measurements
taken and
pn
is the
n
-th pressure value. The symbol
⟨·⟩t
designates the average over time, in this case
of a squared signal. SPL is the most well-known parameter when dealing with acoustics and part of the
acoustician’s daily life. This happens because sound pressure is the easiest quantity to access within
the topic (with a microphone, for instance). The physical meaning of any SPL is related to the acoustic
energy for a given time frame. Moreover, a decibel scale compacts the range of values and a reference
(ca. threshold of hearing @ 1 kHz) sets the 0 dB (SPL). Figure 1 (a) depicts the calculus in a visual
way. The reader may access a Matlab example code by clicking on the icon or searching for the
attached files in this pdf.
doi: 10.3397/IN-2021-2557. Page 2 of 22

squaring
time
domain
for T
averaging square root
(a)
(a) RMS steps.
Time domain
Level SPL Leq for a given TLeq
Leq
Time series with fs
SPL for a given TSPL
store + processing cycle
store +
processing cycle
Leq for a given TLeq
(b)
(b) SPL vs. Leq.
Figure 1: (a) Steps to estimate the sound pressure level (SPL) out of a pressure measurement with a microphone.
(b) Time series with fs, SPL, and Leq measurement examples.
Depending on the nature of the sound being recorded, it is common to find time-averagings called
Fast (F, 125 ms), Slow (S, 1 s), and Impulse (I, 35 ms) applied to the SPL (see Section 4.1). These
kinds of averagings descend from the time in which SLMs were only analogic and used needles to
mark the measured values [4].
2.2. Equivalent Continuous Sound level (Leq)
The Equivalent Continuous Sound level (
Leq
) is a measure of total energy (of sound fluctuating
level) over a certain time, presented as a unique value (in dB) [4]. Its calculus follows
Leq =10 log10 






1
TZT
0
p2(t)
p2
0
dt





(3) or Leq =10 log10 







1
N
N
X
n=1
p2
n
p2
0







=10 log10 







1
N
N
X
n=1
wn10SPLn/10







,(4)
where
N
is the total number of discrete measurements, SPL
n
is the
n
-th discrete sound pressure
level taken, and
wn
is the fraction of total time the SPL
n
is present. Usually,
Leq
is frequency C– or
A–weighted. In such cases, it is denoted as LCeq and LAeq, respectively (see Section 4.2).
If time-averaging is not involved, SPL and
Leq
are quite similar. The main difference would then
be that SPL has a short time average, while
Leq
is evaluated for longer periods, as noticeable in
Figure 1 (b).
2.3. Sound Level Meter (SLM)
The SLM is the most well-known device used to evaluate sound
in an objective way, analogous to the voltmeter for an electrician.
Its functionalities encompass SPL and
Leq
measurements for
different time- and frequency-weighting. Moreover, its results may
be delivered as global values or in fractional octave bands (over
frequency), see Section 4.3. The main international documents
that regulate the SLM are the standards IEC 61672:2013 –
Parts 1, 2, and 3 [
5
,
6
]. Its accuracy, precision, and frequency
range of operation are defined and categorized as Class 1 and
Class 2, with the latter the least precise.
Usually, Class 1 SLM is costly due to the instrumentation and
hardware included. Well-known models (from this class) are from
the manufacturer Brüel & Kjær, for example, models Type 2240,
2270, and 2245 (from older to newer) [7], see Figure 2.
Figure 2: SLM B&K Type 2240,
Type 2270-S, and Type 2245 [7].
doi: 10.3397/IN-2021-2557. Page 3 of 22

2.4. Noise monitoring
Noise and noise pollution assessments are generally made using
Leq
measurements. For such tasks,
it is common to request that a technician assess the desired area using a sound level meter. Carrying
out this way involves the disadvantages of being expensive, time-consuming, and does not granulate
1
time and space enough. To overcome these problems, innovation has trended towards the development
of autonomous wireless sensor networks for noise pollution monitoring. In general they are relatively
low-cost systems that can be deployed throughout the location of interest in order to measure sound
continuously and send data to a land server. For more information, see the references [8–10].
2.5. Microelectromechanical microphones (MEMS)
A microphone is a transducer that transforms acoustic pressure into an electric signal. This can
be accomplished either by electromagnetic induction (dynamic microphones), piezoelectric effect,
a capacitance variation (condenser microphones), or perturbations on a beam of light caused by
sound pressure waves (optical microphones). Microelectromechanical (MEMS) microphones are
manufactured using special semiconductor techniques which allow for the construction of very tiny
devices. They can be designed using any of the transduction principles aforementioned, with the vast
majority capacitive [11].
MEMS microphones can be either analog or digital. The former consist of a transducer element,
a conditioning circuit, and a pre-amplifier (see Figure 3 (a)). The latter encompass all the analog
blocks plus an analog-to-digital converter (ADC) and a digital interface (for communication). There
are three possible digital interfaces on commercially available microphones: Inter-IC Sound (I
2
S),
Figure 3 (b); Pulse Density Modulated signals (PDM), Figure 3 (c); and Time-Division Multiplex
(TDM), Figure 3 (d).
PDM microphones transmit 1-bit signals with a high sampling frequency. They must be converted to
Pulse Code Modulation (PCM) for further processing. The majority of processors are unable to do this
in real-time, making it necessary to add an audio codec chip to the signal chain [
12
]. The I
2
S format, in
contrast, has a built-in converter and transmits PCM audio through a serial port (using the I
2
S protocol
[
13
]). Both I
2
S and PDM formats allow for the transmission of two channels over a single data line.
Finally, TDM microphones send PCM audio through a serial port that allows for the transference of up
to 16 channels on a single data line.
Digital microphones dispense analog-to-digital converters, therefore they reduce both the size and
cost of the system. Furthermore, they are less susceptible to interference, making them more suitable
for applications where wifi, Bluetooth, or other wireless devices are near the microphone (e.g., on an
autonomous noise monitoring system, a cellphone, or a computer). In general, MEMS microphone
performance is stable over time (no sensitivity drifting). Its manufacturing process promotes a high
standardization, as well, resulting in closely matched devices (useful for beamforming and noise
suppression applications) [14].
2.6. Teensy 4.0 and Teensy Audio Library
Teensy is a family of Arduino compatible microcontrollers (MCU) developed by Paul J. Stoffregen
[
15
,
16
]. For this project, Teensy 4.0 was chosen due to its low cost, low power consumption, high
computing power, and small size, per Figure 5 (c). Its main features include an ARM Cortex M7 at
600 MHz; floating-point math unit (64 bits and 32 bits, FPU); USB device (480 Mbit/sec) and USB
host (480 Mbit/sec); and two I
2
S/TDM digital audio ports [
15
]. Observe Table 1 where different Teensy
and Arduino models are compared.
1Granularity is the scale or level of detail present in a set of data.
doi: 10.3397/IN-2021-2557. Page 4 of 22

ADC FILTER
HARDWARE
CONTROL
I2S
SERIAL
PORT
POWER
MANAGEMENT
Serial Data Clock
Serial Data Output
Word Clock
VDD
GND
MEMS
TRANSDUCER
Figure 3. Typical I2
S MEMS Microphone Block Diagram
A lid is then bonded over the laminate to enclose the
OUTPUT
AMPLIFIER
MEMS
TRANSDUCER
POWER
V
DD
GND
OUTPUT
Figure 1. Typical Analog MEMS Microphone Block Diagram
(a) Analog MEMS microphone block diagram. (b) I2S MEMS microphone block diagram.
(a) Analog MEMS microphone block diagram
ADC FILTER
HARDWARE
CONTROL
I2S
SERIAL
PORT
POWER
MANAGEMENT
Serial Data Clock
Serial Data Output
Word Clock
VDD
GND
MEMS
TRANSDUCER
Figure 3. Typical I2
S MEMS Microphone Block Diagram
A lid is then bonded over the laminate to enclose the
OUTPUT
AMPLIFIER
MEMS
TRANSDUCER
POWER
V
DD
GND
OUTPUT
Figure 1. Typical Analog MEMS Microphone Block Diagram
(a) Analog MEMS microphone block diagram. (b) I2S MEMS microphone block diagram.
(b) I2S MEMS microphone block diagram
MEMS
TRANSDUCER
ADC
POWER
MANAGEMENT
CLK
DATA
V
DD
GND
PDM
MODULATOR
CHANNEL
SELECT
L/R SELECT
AMPLIFIER
(c) PDM MEMS microphone block diagram.
ICS-52000
ADC
FILTER
HARDWARE
CONTROL
SERIAL
PORT
POWER
MANAGEMENT
SCK
SD
WS
TDM
WSO
CONFIGGNDVDD
(d) TDM MEMS microphone block diagram.
(c) PDM microphone block diagram
MEMS
TRANSDUCER
ADC
POWER
MANAGEMENT
CLK
DATA
V
DD
GND
PDM
MODULATOR
CHANNEL
SELECT
L/R SELECT
AMPLIFIER
(c) PDM MEMS microphone block diagram.
ICS-52000
ADC
FILTER
HARDWARE
CONTROL
SERIAL
PORT
POWER
MANAGEMENT
SCK
SD
WS
TDM
WSO
CONFIGGNDVDD
(d) TDM MEMS microphone block diagram.
(d) TDM microphone block diagram
Figure 3: Block diagrams for different types of MEMS microphones: analog, I2S, PDM, and TDM [17].
Teensy MCUs come with a built-in and open-source library for audio processing (Teensy Audio
Library). As described on its website, the Teensy Audio Library is “a toolkit for building streaming
audio projects, featuring Polyphonic Playback, Recording, Synthesis, Analysis, Effects, Filtering, Mixing,
Multiple Simultaneous Inputs &Outputs, and Flexible Internal Signal Routing” [
15
]. All audio signals
circulating on Teensy have a 16-bit resolution, 44.1 kHz of sampling frequency, and stream while
Arduino sketches are running.
Table 1: Hardware comparison between the prototype used and other Teensy and Arduino boards.
Name Processor CPU Speed Storage Memory Direct memory access (DMA)
Teensy 4.0 NXP iMXRT10621,⋄600 MHz 2 MB (Flash Memory) 1 MB 32 channels
Teensy 4.1 NXP iMXRT10622,⋄600 MHz 8 MB (Flash Memory) 1 MB 32 channels
Teensy 3.6 NXP MK66FX1M03,⋄180 MHz 1 MB (Flash Memory) 256 KB 32 channels
Arduino Due AT91SAM3X8E 84 MHz 512 KB (Flash Memory) 96 KB (SRAM) 23 channels
Arduino Mega 2560 ATmega2560 16 MHz 256 KB (Flash Memory) 8 KB (SRAM) -
1,2ARM Cortex-M7. 3ARM Cortex-M4. ⋄FPU =Floating Point Unit. Additional info can be found at PJRC store,Arduino.cc,Board db, and .
The audio library is object-oriented. Therefore, objects are used to execute audio functions, such as
input receiving, signal filtering, and output streaming. Audio programs can be created using the Audio
System Design Tool [
15
], a graphic interface that allows users to generate a signal chain and export
it as an Arduino code. The design tool also offers a brief description and documentation of all audio
objects. Additionally, it is possible to program Teensy to run as a USB audio device, allowing the user
to send and receive audio data to and from a computer, which is useful for testing and prototyping.
An example is illustrated in Figure 4, where a signal from the computer is fed into Teensy via USB,
processed by an Infinite Impulse Response (IIR) filter, and sent back to the computer (the reader may
access an Arduino example code by clicking on the icon or searching for the attached files in
this pdf) .
As an open-source library, it is possible to write and develop new audio objects that integrate
seamlessly with the others. The objects are written in C++ language. Guidelines on how to write a new
one can be found on Teensy’s website [15].
doi: 10.3397/IN-2021-2557. Page 5 of 22
PDF

Figure 4: Example of an audio signal chain generated on Audio System Design Tool. The I
2
S object is not used
but must be present to allow for audio updates and streaming.
3. PROTOTYPE
This section describes the prototype developed for sound monitoring. It is comprised of a Teensy
microcontroller, a digital I
2
S MEMS microphone, an SD card module, a portable battery, and a simple
on/offswitch. First, the technical specifications of the components and their interconnections are
characterized. Ultimately, the functionalities of the audio objects programmed for sound pressure level
calculations are explained.
3.1. Components
The prototype uses a digital I
2
S MEMS microphone to capture sound. A Sipeed MSM261S4030H0
evaluation board was chosen due to its availability, low cost, and ease of connectivity with Teensy 4.0.
Its frequency response and picture can be found in Figures 5 (a) and (b), respectively, while its main
specifications are detailed in Table 2. Readers may listen to a recording example via the following
file (voice speech, acoustic guitar, and environmental noise).
Table 2: MEMS microphone Sipeed MSM261S4030H0 technical specifications [18].
Parameter Limits /Data
Min. Nom. Max. Unit Condition
Directivity Omnidirectional — —
Sensitivity -27 -26 -25 dB dBFS @ 1 kHz 1 Pa
Operation voltage 1.6 – 3.6 V —
Frequency range 100 – 10k Hz —
Signal-to-noise ratio — 57 — dB 20 kHz bandwidth, A-weighted
Total Harmonic Distortion — — 1 % 100 dB SPL @1 kHz, S =nom., Rload >2k
Acoustic Overload Point — 124 — dB SPL 10% THD @1 kHz, S =nom., Rload >2k
Maximum SPL 140 dB SPL —
A simple SD card Arduino module is used to store the measurements taken, per Figure 5 (d). The
module has seven pins, two of which are for power (GND, +3.3 V, and +5 V), two for control (CS and
SCK), and two for data transfer (MOSI and MISO). Teensy 4.0 controls both the microphone and SD
card module via serial connections. A 4400 mAh capacity USB mobile battery is used to power the
system, see Figure 5 (e).
3.2. Audio objects for sound pressure level calculations
Sound pressure level calculations are held by two objects designed to integrate the Teensy
Audio Library. They are
AudioAnalyzeSoundLevelMeter
and
AudioAnalyzeOctaveBands
. The first
evaluates time-weighted SPL and equivalent continuous SPL (
Leq
), with frequency-weightings Z, A,
or C. The second evaluates
Leq
in one (
1
/
1
) or one-third (
1
/
3
) octave bands, with frequency-weightings Z,
A, or C, as well. They are configured using the following methods (some are common to both objects):
–setLogInterval(float)
: receives a value (in seconds) to configure the integration time for
Leq
calculations and the rate at which the object returns data (e.g.,
setLogInterval(1.0f)
set the
object to calculate and return a value every second) — common to both objects;
doi: 10.3397/IN-2021-2557. Page 6 of 22

MSM261S4030H0 frequency response @ 94 dB SPL
Sensitivity (dB FS/PA)
20
10
0
-10
-20
100 Frequency (Hz)
200 500 1k 2k 5k 10k
2
(a) MSM261S4030H0 I S digital MEMS microphone
frequency response.
2
(b) MSM261S4030H0 I S digital MEMS microphone
next to a 25 mm diameter coin.
(a) MSM261S4030H0 I2S digital MEMS microphone
frequency response.
MSM261S4030H0 frequency response @ 94 dB SPL
Sensitivity (dB FS/PA)
20
10
0
-10
-20
100 Frequency (Hz)
200 500 1k 2k 5k 10k
2
(a) MSM261S4030H0 I S digital MEMS microphone
frequency response.
2
(b) MSM261S4030H0 I S digital MEMS microphone
next to a 25 mm diameter coin.
(b) MSM261S4030H0 I2S digital MEMS
microphone next to a 25 mm diameter coin.
(d) SD card module dimensions.
5.1 cm
3.1 cm
(c) Teensy 4.0 dimensions.
3.6 cm
1.8 cm
(c) Teensy 4.0 dimensions.
(d) SD card module dimensions.
5.1 cm
3.1 cm
(c) Teensy 4.0 dimensions.
3.6 cm
1.8 cm
(d) SD card module dimensions.
(e) H'Maston pro 10000 mAh and
Advansat 4400 mAh power banks.
(f) Components assembled inside the prototype case.
(e) H’Maston pro 10000 mAh and Advansat
4400 mAh power banks.
(e) H'Maston pro 10000 mAh and
Advansat 4400 mAh power banks.
(f) Components assembled inside the prototype case.
(f) Components assembled inside the prototype
case.
(g) Prototype fully assembled.
(g) Prototype fully assembled.
Figure 5: Frequency response of the Sipeed microphone, pictures of the prototype components, and the fully
assembled prototype.
doi: 10.3397/IN-2021-2557. Page 7 of 22

–setFreqWeighting(int)
: configures the frequency-weighting used for calculations. The
options are Z_WEIGHTING,A_WEIGHTING, and C_WEIGHTING — common to both objects;
–setSensitivity(float)
: receives a calibration factor that adjusts the sensitivity of the system;
–setTimeWeighting(float)
: configures the time-weighting used for SPL calculations. The
options are FAST and SLOW (specific to AudioAnalyzeSoundLevelMeter object); and
–setFractionalOctaveBands(int)
: sets the fractional octave bands used for calculations. The
options are ONE_OCTAVE and THIRD_OCTAVE (specific to AudioAnalyzeOctaveBands object).
All data inside the objects are evaluated using floating-point precision (32 bits), ensuring adequate
math steps. Filtering is accomplished by Infinite Impulse Response filters (IIR) implemented using built-
in functions from the Cortex Microcontroller Software Interface Standard DSP library for
Cortex-M
processors (CMSIS [
19
]) and coefficients previously calculated in Matlab. The objects are placed
inside an Arduino sketch and can be used in parallel, which makes the system extremely versatile and
easy to use.
4. ALGORITHM VALIDATION
This section describes the tests conducted to validate the sound level meter algorithms. In all tests,
an appropriate digital signal was applied to Teensy through a USB connection, and sound pressure
levels were retrieved via serial communication (
Serial.println()
) for further plotting and analysis.
The process is the gray-box method, i.e. activation and response, as observed in the signal chain in
Figure 6 (a).
The procedure and acceptance limits described in IEC 61672:2013 (Parts 1 and 2) were used to
evaluate the time- and frequency-weighting filters [
5
,
6
]. Moreover, ANSI/ASA S1.11-2014 (Part 1),
IEC 61260:1-2014, and IEC 61260:2-2016 standardized methods were used to evaluate one- and
one-third octave band-pass filters [20, 21], as seen in Figure 6 (b).
Teensy 4.0
PCM
audio
USB output
Computer
USB input
SPL values
(serial communication)
USB output
USB input
(a) Signal flow for the gray-box method.
Inside Teensy
Sound
MEMS
microphone
Teensy 4.0
PCM/I2S
audio
SD card
RMS, Max or Peak
Z, A or C
freq.-weighting
time-weighting
octave filters
1/1 or 1/3 oct
Fast or Slow
ADC
Leq
Level
(b) Designed SLM workflow.
Figure 6: (a) Signal chain used for all electrical-procedural tests. (b) SLM processing steps.
4.1. Time-weighting filters (Fast &Slow)
According to IEC 61672-1, a steady 4 kHz sinusoidal signal followed by sudden silence must be
used to evaluate the decay rates of time-weighting filters. The design goals for decay rates of Fast and
Slow time-weightings are 34.7 dB/s and 4.3 dB/s, respectively, with acceptance limits of +3.8 dB/s
doi: 10.3397/IN-2021-2557. Page 8 of 22

and -3.5 dB/s for Fast, and +0.8 dB/s and -0.7 dB/s for Slow. For this test, a 30-second digital signal
(10 seconds for the sine wave and 20 seconds of silence) was generated on Matlab and applied to
Teensy via USB. The microcontroller was configured to assess time-weighted sound pressure levels
and return a value every 2.9 milliseconds, which corresponds to approximately 128 samples per log.
The curves and decay rates for each filter are presented in Figure 7. It is possible to notice that the
filters implemented are working properly, with the subsequent decays observed exactly as the standard.
0 2 4 6 8 10 12 14 16 18 20
Time [s]
60
70
80
90
100
SPL [dB ref. 20 µPa]
Decay curves for Fast and Slow time-weightings
4 kHz sine wave
SPL Fast
SPL Slow
Fast decay rate: 34.7 dB/s
Slow decay rate: 4.3 dB/s
Figure 7: Decay curves for time-weighing filters and rates Slow and Fast).
4.2. Frequency-weighting filters (A &C)
Section 9.5 of the standard IEC 61672-2 describes two methods for the evaluation of frequency-
weighting filters using electrical signals. In this work, the second method was used. The test consists of
inserting a series of steady sinusoidal signals with the same amplitude and within the 10 Hz to 20 kHz
frequency range into the device. Frequency increments of one-twelfth octave were used, resulting in a
total of 133 test frequencies.
The microcontroller was configured to return time-averaged equivalent sound pressure level (SPL)
every 0.1 seconds. Subsequently, the SPL levels for each frequency were averaged and normalized by
the 1 kHz SPL value. Finally, the values were plotted along with the Class 1 design goals, defined in
Table 3 of the IEC 61672-2. The tests were conducted for both A and C frequency-weightings. Figure 8
shows that the filters are working as expected, within the Class 1 mask. For the record, the Z-weighting
means zero (or not any) weighting, which would render a flat line on zero along the full frequency rage.
101102103104
Frequency [Hz]
-80
-60
-40
-20
0
Normalized level
[dB ref. 1 @ 1 kHz]
A-weighting filter validation according to
IEC 61672-1 Class 1 acceptance limits
Teensy
Class 1 Mask
(a) A frequency-weighting validation.
101102103104
Frequency [Hz]
-40
-30
-20
-10
0
Normalized level
[dB ref. 1 @ 1 kHz]
C-weighting filter validation according to
IEC 61672-1 Class 1 acceptance limits
Teensy
Class 1 Mask
(b) C frequency-weighting validation.
Figure 8: Frequency response of the Frequency-weighting filters and Class 1 acceptance limits.
4.3. Octave and fractional octave band filters
Considering that the algorithms implemented in Teensy only return SPL values, the method described
in IEC 61260:2-2016
2
was used to validate the one- and one-third octaves band-pass filters. The
procedure consists of applying a series of constant sinusoidal signals to each filter and then measuring
2Section 7.2.1.4, analogous to the one in Annex D of previous ANSI S1.11:2004 (R2009) standard.
doi: 10.3397/IN-2021-2557. Page 9 of 22

the relative attenuation of each one. The frequency of the
i
-th test signal (
fi
) is related to the center
frequency of the analyzed filter by
fi/fm=hG1/(bS )ii(5)
where
fm
is the filter midband frequency,
G
is the octave ratio (
G10
=10
3/10
was used),
b
is the
bandwidth designator (1 for one octave and 3 for one-third octave), and
S
is the number of test
frequencies per filter.
A Matlab code was written to automate the measurements. The test signals were 2 seconds long,
and a total of 49 test frequencies were used for each band-pass filter. The results for all the bands
are illustrated in Figures 9 and 10. They show that all filters comply with the standard for Class 1
performance (Class 2 and old Class 0 are also depicted for the sake of comparison).
To test the stability of the filters, the 16 Hz to 63 Hz and 1 kHz frequency
1
/3
-bands were evaluated
30 times each. Figure 11 depicts the mean together with the respective confidence interval for 99.73%
(shaded areas). It is possible to notice that approaches used in the project rendered high precision (i.e.
the dispersion is almost null) while maintaining its values within the Class 1 mask requirements.
102103104
Frequency [Hz]
-20
-15
-10
-5
0
Normalized Level [dB ref. 1]
Frequency response for 1/1 octave band-pass filters
(a) All 1/1 octave filters plotted together.
562 750 1000 1334 1778
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 1 kHz
(b) Class 1 validation of 1/1 octave filter
with mid-band frequency of 1 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(a) All 1/1 octave filters plotted together.
102103104
Frequency [Hz]
-20
-15
-10
-5
0
Normalized Level [dB ref. 1]
Frequency response for 1/1 octave band-pass filters
(a) All 1/1 octave filters plotted together.
562 750 1000 1334 1778
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 1 kHz
(b) Class 1 validation of 1/1 octave filter
with mid-band frequency of 1 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(b) Class 1 validation of 1/1 octave filter with
mid-band frequency of 1 kHz.
9 12 16 21 28
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 15.5 Hz
(c) Class 1 validation of 1/1 octave filter
with mid-band frequency of 15.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
18 24 32 42 56
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 31.5 Hz
(d) Class 1 validation of 1/1 octave filter
with mid-band frequency of 31.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(c) Class 1 validation of 1/1 octave filter with
mid-band frequency of 15.5 Hz.
9 12 16 21 28
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 15.5 Hz
(c) Class 1 validation of 1/1 octave filter
with mid-band frequency of 15.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
18 24 32 42 56
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 31.5 Hz
(d) Class 1 validation of 1/1 octave filter
with mid-band frequency of 31.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(d) Class 1 validation of 1/1 octave filter with
mid-band frequency of 31.5 Hz.
35 47 63 83 111
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 62.5 Hz
(e) Class 1 validation of 1/1 octave filter
with mid-band frequency of 62.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
70 94 125 167 222
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 125 Hz
(f) Class 1 validation of 1/1 octave filter
with mid-band frequency of 125 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(e) Class 1 validation of 1/1 octave filter with
mid-band frequency of 62.5 Hz.
35 47 63 83 111
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 62.5 Hz
(e) Class 1 validation of 1/1 octave filter
with mid-band frequency of 62.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
70 94 125 167 222
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 125 Hz
(f) Class 1 validation of 1/1 octave filter
with mid-band frequency of 125 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(f) Class 1 validation of 1/1 octave filter with
mid-band frequency of 125 Hz.
Figure 9: Frequency response of the Sipeed microphone, pictures of the prototype components, and the fully
assembled prototype [1/2].
doi: 10.3397/IN-2021-2557. Page 10 of 22

141 187 250 333 445
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 250 Hz
(g) Class 1 validation of 1/1 octave filter
with mid-band frequency of 250 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
281 375 500 667 889
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 500 Hz
(h) Class 1 validation of 1/1 octave filter
with mid-band frequency of 500 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(g) Class 1 validation of 1/1 octave filter with
mid-band frequency of 250 Hz.
141 187 250 333 445
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 250 Hz
(g) Class 1 validation of 1/1 octave filter
with mid-band frequency of 250 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
281 375 500 667 889
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 500 Hz
(h) Class 1 validation of 1/1 octave filter
with mid-band frequency of 500 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(h) Class 1 validation of 1/1 octave filter with
mid-band frequency of 500 Hz.
1125 1500 2000 2667 3557
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 2 kHz
(i) Class 1 validation of 1/1 octave filter
with mid-band frequency of 2 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
2249 3000 4000 5334 7113
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 4 kHz
(j) Class 1 validation of 1/1 octave filter
with mid-band frequency of 4 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(i) Class 1 validation of 1/1 octave filter with
mid-band frequency of 2 kHz.
1125 1500 2000 2667 3557
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 2 kHz
(i) Class 1 validation of 1/1 octave filter
with mid-band frequency of 2 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
2249 3000 4000 5334 7113
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 4 kHz
(j) Class 1 validation of 1/1 octave filter
with mid-band frequency of 4 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(j) Class 1 validation of 1/1 octave filter with
mid-band frequency of 4 kHz.
4499 5999 8000 10668 14226
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 8 kHz
(k) Class 1 validation of 1/1 octave filter
with mid-band frequency of 8 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
8997 11998 16000 21336 28452
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 16 kHz
(l) Class 1 validation of 1/1 octave filter
with mid-band frequency of 16 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(k) Class 1 validation of 1/1 octave filter with
mid-band frequency of 8 kHz.
4499 5999 8000 10668 14226
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 8 kHz
(k) Class 1 validation of 1/1 octave filter
with mid-band frequency of 8 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
8997 11998 16000 21336 28452
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/1 octave filter validation: fm= 16 kHz
(l) Class 1 validation of 1/1 octave filter
with mid-band frequency of 16 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(l) Class 1 validation of 1/1 octave filter with
mid-band frequency of 16 kHz.
Figure 9: Frequency response of the Sipeed microphone, pictures of the prototype components, and the fully
assembled prototype [2/2].
102103104
Frequency [Hz]
-20
-15
-10
-5
0
Normalized Level [dB ref. 1]
Frequency response for 1/3 octave band-pass filters
(a) All 1/3 octave filters plotted together.
794 1000 1259
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1 kHz
(b) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(a) All 1/3 octave filters plotted together.
102103104
Frequency [Hz]
-20
-15
-10
-5
0
Normalized Level [dB ref. 1]
Frequency response for 1/3 octave band-pass filters
(a) All 1/3 octave filters plotted together.
794 1000 1259
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1 kHz
(b) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(b) Class 1 validation of 1/3 octave filter with
mid-band frequency of 1 kHz.
Figure 10: Frequency responses and validations according to ANSI/IEC masks for 1/3-octave band-pass filters
[1/5].
doi: 10.3397/IN-2021-2557. Page 11 of 22

12 16 20
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 15.5 Hz
(c) Class 1 validation of 1/3 octave filter
with mid-band frequency of 15.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
16 20 25
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 20 Hz
(d) Class 1 validation of 1/3 octave filter
with mid-band frequency of 20 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(c) Class 1 validation of 1/3 octave filter with
mid-band frequency of 15.5 Hz.
12 16 20
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 15.5 Hz
(c) Class 1 validation of 1/3 octave filter
with mid-band frequency of 15.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
16 20 25
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 20 Hz
(d) Class 1 validation of 1/3 octave filter
with mid-band frequency of 20 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(d) Class 1 validation of 1/3 octave filter with
mid-band frequency of 20 Hz.
20 25 31
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 25 Hz
(e) Class 1 validation of 1/3 octave filter
with mid-band frequency of 25 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
25 32 40
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 31.5 Hz
(f) Class 1 validation of 1/3 octave filter
with mid-band frequency of 31.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(e) Class 1 validation of 1/3 octave filter with
mid-band frequency of 25 Hz.
20 25 31
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 25 Hz
(e) Class 1 validation of 1/3 octave filter
with mid-band frequency of 25 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
25 32 40
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 31.5 Hz
(f) Class 1 validation of 1/3 octave filter
with mid-band frequency of 31.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(f) Class 1 validation of 1/3 octave filter with
mid-band frequency of 31.5 Hz.
32 40 50
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 40 Hz
(g) Class 1 validation of 1/3 octave filter
with mid-band frequency of 40 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
40 50 63
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 50 Hz
(h) Class 1 validation of 1/3 octave filter
with mid-band frequency of 50 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(g) Class 1 validation of 1/3 octave filter with
mid-band frequency of 40 Hz.
32 40 50
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 40 Hz
(g) Class 1 validation of 1/3 octave filter
with mid-band frequency of 40 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
40 50 63
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 50 Hz
(h) Class 1 validation of 1/3 octave filter
with mid-band frequency of 50 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(h) Class 1 validation of 1/3 octave filter with
mid-band frequency of 50 Hz.
50 63 79
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 62.5 Hz
(i) Class 1 validation of 1/3 octave filter
with mid-band frequency of 62.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
64 80 101
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 80 Hz
(j) Class 1 validation of 1/3 octave filter
with mid-band frequency of 80 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(i) Class 1 validation of 1/3 octave filter with
mid-band frequency of 62.5 Hz.
50 63 79
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 62.5 Hz
(i) Class 1 validation of 1/3 octave filter
with mid-band frequency of 62.5 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
64 80 101
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 80 Hz
(j) Class 1 validation of 1/3 octave filter
with mid-band frequency of 80 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(j) Class 1 validation of 1/3 octave filter with
mid-band frequency of 80 Hz.
Figure 10: Frequency responses and validations according to ANSI/IEC masks for 1/3-octave band-pass filters
[2/5].
doi: 10.3397/IN-2021-2557. Page 12 of 22

79 100 126
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 100 Hz
(k) Class 1 validation of 1/3 octave filter
with mid-band frequency of 100 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
99 125 157
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 125 Hz
(l) Class 1 validation of 1/3 octave filter
with mid-band frequency of 125 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(k) Class 1 validation of 1/3 octave filter with
mid-band frequency of 100 Hz.
79 100 126
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 100 Hz
(k) Class 1 validation of 1/3 octave filter
with mid-band frequency of 100 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
99 125 157
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 125 Hz
(l) Class 1 validation of 1/3 octave filter
with mid-band frequency of 125 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(l) Class 1 validation of 1/3 octave filter with
mid-band frequency of 125 Hz.
123 155 195
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 155 Hz
(m) Class 1 validation of 1/3 octave filter
with mid-band frequency of 155 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
159 200 252
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 200 Hz
(n) Class 1 validation of 1/3 octave filter
with mid-band frequency of 200 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(m) Class 1 validation of 1/3 octave filter with
mid-band frequency of 155 Hz.
123 155 195
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 155 Hz
(m) Class 1 validation of 1/3 octave filter
with mid-band frequency of 155 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
159 200 252
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 200 Hz
(n) Class 1 validation of 1/3 octave filter
with mid-band frequency of 200 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(n) Class 1 validation of 1/3 octave filter with
mid-band frequency of 200 Hz.
199 250 315
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 250 Hz
(o) Class 1 validation of 1/3 octave filter
with mid-band frequency of 250 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
250 315 397
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 315 Hz
(p) Class 1 validation of 1/3 octave filter
with mid-band frequency of 315 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(o) Class 1 validation of 1/3 octave filter with
mid-band frequency of 250 Hz.
199 250 315
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 250 Hz
(o) Class 1 validation of 1/3 octave filter
with mid-band frequency of 250 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
250 315 397
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 315 Hz
(p) Class 1 validation of 1/3 octave filter
with mid-band frequency of 315 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(p) Class 1 validation of 1/3 octave filter with
mid-band frequency of 315 Hz.
318 400 504
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 400 Hz
(q) Class 1 validation of 1/3 octave filter
with mid-band frequency of 400 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
397 500 629
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 500 Hz
(r) Class 1 validation of 1/3 octave filter
with mid-band frequency of 500 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(q) Class 1 validation of 1/3 octave filter with
mid-band frequency of 400 Hz.
318 400 504
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 400 Hz
(q) Class 1 validation of 1/3 octave filter
with mid-band frequency of 400 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
397 500 629
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 500 Hz
(r) Class 1 validation of 1/3 octave filter
with mid-band frequency of 500 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(r) Class 1 validation of 1/3 octave filter with
mid-band frequency of 500 Hz.
Figure 10: Frequency responses and validations according to ANSI/IEC masks for 1/3-octave band-pass filters
[3/5].
doi: 10.3397/IN-2021-2557. Page 13 of 22

500 630 793
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 630 Hz
(s) Class 1 validation of 1/3 octave filter
with mid-band frequency of 630 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
635 800 1007
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 800 Hz
(t) Class 1 validation of 1/3 octave filter
with mid-band frequency of 800 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(s) Class 1 validation of 1/3 octave filter with
mid-band frequency of 630 Hz.
500 630 793
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 630 Hz
(s) Class 1 validation of 1/3 octave filter
with mid-band frequency of 630 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
635 800 1007
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 800 Hz
(t) Class 1 validation of 1/3 octave filter
with mid-band frequency of 800 Hz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(t) Class 1 validation of 1/3 octave filter with
mid-band frequency of 800 Hz.
993 1250 1574
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1.25 kHz
(u) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1.25 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
1271 1600 2014
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1.6 kHz
(v) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1.6 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(u) Class 1 validation of 1/3 octave filter with
mid-band frequency of 1.25 kHz.
993 1250 1574
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1.25 kHz
(u) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1.25 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
1271 1600 2014
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 1.6 kHz
(v) Class 1 validation of 1/3 octave filter
with mid-band frequency of 1.6 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(v) Class 1 validation of 1/3 octave filter with
mid-band frequency of 1.6 kHz.
1589 2000 2518
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 2 kHz
(w) Class 1 validation of 1/3 octave filter
with mid-band frequency of 2 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
1986 2500 3147
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 2.5 kHz
(x) Class 1 validation of 1/3 octave filter
with mid-band frequency of 2.5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(w) Class 1 validation of 1/3 octave filter with
mid-band frequency of 2 kHz.
1589 2000 2518
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 2 kHz
(w) Class 1 validation of 1/3 octave filter
with mid-band frequency of 2 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
1986 2500 3147
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 2.5 kHz
(x) Class 1 validation of 1/3 octave filter
with mid-band frequency of 2.5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(x) Class 1 validation of 1/3 octave filter with
mid-band frequency of 2.5 kHz.
2502 3150 3966
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 3.15 kHz
(y) Class 1 validation of 1/3 octave filter
with mid-band frequency of 3.15 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
3177 4000 5036
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 4 kHz
(z) Class 1 validation of 1/3 octave filter
with mid-band frequency of 4 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(y) Class 1 validation of 1/3 octave filter with
mid-band frequency of 3.15 kHz.
2502 3150 3966
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 3.15 kHz
(y) Class 1 validation of 1/3 octave filter
with mid-band frequency of 3.15 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
3177 4000 5036
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 4 kHz
(z) Class 1 validation of 1/3 octave filter
with mid-band frequency of 4 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(z) Class 1 validation of 1/3 octave filter with
mid-band frequency of 4 kHz.
Figure 10: Frequency responses and validations according to ANSI/IEC masks for 1/3-octave band-pass filters
[4/5].
doi: 10.3397/IN-2021-2557. Page 14 of 22

3972 5000 6295
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 5 kHz
(aa) Class 1 validation of 1/3 octave filter
with mid-band frequency of 5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
5044 6350 7994
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 6.35 kHz
(bb) Class 1 validation of 1/3 octave filter
with mid-band frequency of 6.35 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(aa) Class 1 validation of 1/3 octave filter with
mid-band frequency of 5 kHz.
3972 5000 6295
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 5 kHz
(aa) Class 1 validation of 1/3 octave filter
with mid-band frequency of 5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
5044 6350 7994
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 6.35 kHz
(bb) Class 1 validation of 1/3 octave filter
with mid-band frequency of 6.35 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(ab) Class 1 validation of 1/3 octave filter with
mid-band frequency of 6.35 kHz.
6355 8000 10071
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 8 kHz
(cc) Class 1 validation of 1/3 octave filter
with mid-band frequency of 8 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
7943 10000 12589
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 10 kHz
(dd) Class 1 validation of 1/3 octave filter
with mid-band frequency of 10 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(ac) Class 1 validation of 1/3 octave filter with
mid-band frequency of 8 kHz.
6355 8000 10071
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 8 kHz
(cc) Class 1 validation of 1/3 octave filter
with mid-band frequency of 8 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
7943 10000 12589
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 10 kHz
(dd) Class 1 validation of 1/3 octave filter
with mid-band frequency of 10 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(ad) Class 1 validation of 1/3 octave filter with
mid-band frequency of 10 kHz.
9929 12500 15737
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 12.5 kHz
(ee) Class 1 validation of 1/3 octave filter
with mid-band frequency of 12.5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
12709 16000 20143
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 16 kHz
(ff) Class 1 validation of 1/3 octave filter
with mid-band frequency of 16 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(ae) Class 1 validation of 1/3 octave filter with
mid-band frequency of 12.5 kHz.
9929 12500 15737
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 12.5 kHz
(ee) Class 1 validation of 1/3 octave filter
with mid-band frequency of 12.5 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
12709 16000 20143
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 16 kHz
(ff) Class 1 validation of 1/3 octave filter
with mid-band frequency of 16 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(af) Class 1 validation of 1/3 octave filter with
mid-band frequency of 16 kHz.
15887 20000 25179
Frequency [Hz]
-20
-15
-10
-5
0
Normalized level [dB ref. 1]
1/3 octave filter validation: fm= 20 kHz
(gg) Class 1 validation of 1/3 octave filter
with mid-band frequency of 20 kHz.
Filter response
Class 0 mask
Class 1 mask
Class 2 mask
(ag) Class 1 validation of 1/3 octave filter with mid-band frequency of 20 kHz.
Figure 10: Frequency responses and validations according to ANSI/IEC masks for 1/3-octave band-pass filters
[5/5].
doi: 10.3397/IN-2021-2557. Page 15 of 22

16 20 25 32 40 50 63 80
Frequency [Hz]
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
Normalized level [dB ref. 1]
1/3 Octave-band tests (30 runs): mean and confidence interval for 99.73%
(ah) 1
/3-octave bands: {15 – 63} Hz.
850 875 900 925 950 975 1000 1025 1050 1075 1100 1125 1150
Frequency [Hz]
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
Normalized level [dB ref. 1]
1/3 Octave-band tests (30 runs): mean and confidence interval for 99.73%
1 kHz
Class 1 mask
Class 2 mask
(ai) 1
/3-octave band: 1 kHz.
Figure 11: 1/3 Octave bands test with 30 runs: mean and confidence intervals (shaded area) for 99.73%.
5. PERFORMANCE TESTS
This section describes the tests conducted to evaluate prototype performance. Initially, the system’s
power consumption was measured and used to calculate the battery capacity necessary for one week of
continuous operation. As a second step, the CPU consumption by the program was estimated using a
built-in function from Teensy Audio Library. To follow, a comparison was carried out against the B&K
Class 1 SLM Type 2240, evaluating SPL and
Leq
, considering acoustic testing. Finally, the prototype
was assembled and used for noise monitoring over long periods (30 minutes, 1 hour, and 8 hours). Four
distinct soundscapes were considered. From the field tested cases the device proved to be stable and
reliable.
5.1. Power and CPU consumption
A USB digital power meter with a precision of 0.01 V and 0.01 A (USB digital tester J7-C) was
used to verify the power consumption of the system. All measurements were held with the prototype
connected to a computer with five minutes duration and five different configurations. Teensy was set to
run at 600 MHz and 150 MHz, and an estimation of the CPU consumption was calculated using the
Audio Library built-in function
AudioProcessorUsageMax()
. The results in Table 3 show that Teensy
maintains around the same consumption for all configurations.
With a clock speed of 150 MHz, it is possible to reduce the current consumption to 40 mA while
still running the firmware. This setup was considered in order to estimate the battery capacity required
to power the system during an entire week. The formula
C
=
xT
was used, where
C
is the battery
capacity (mAh),
x
the current drawn (mA), and
T
the intended period in hours. Considering that a
week has around 168 hours and including a 20% margin, the device draws 40 mA and the battery
capacity should be at least 8000 mAh.
Table 3: Power consumption and CPU load of the system for five configurations.
Configuration @ 600 MHz Voltage (V) Current (mA) Load resistance (Ω) Power (W) CPU (%)
Idle w/o microphone 5.09 100.00 50.90 0.50 0.50
Idle with microphone 5.09 100.00 50.90 0.50 0.50
Processing w/o microphone 5.10 90.00 56.66 0.45 24.00
Processing with microphone 5.10 90.00 56.66 0.45 24.00
Processing with microphone and SD card 5.10 100.00 51.00 0.51 24.00
Configuration @ 150 MHz Voltage (V) Current (mA) Load resistance (Ω) Power (W) CPU (%)
Idle w/o microphone 5.12 40.00 128.00 0.20 0.50
Idle with microphone 5.12 40.00 128.00 0.20 0.50
Processing w/o microphone 5.12 40.00 128.00 0.20 96.00
Processing with microphone 5.12 40.00 128.00 0.20 96.00
Processing with microphone and SD card 5.12 40.00 128.00 0.20 96.00
doi: 10.3397/IN-2021-2557. Page 16 of 22

5.2. Measurement comparison against the Class 1 SLM B&K Type 2240
Aiming to evaluate the developed system’s performance, a set of parallel measurements was
conducted using both the prototype and a Class 1 Brüel & Kjær Type 2240
3
sound level meter,
per Figure 12 (a). Due to the global pandemic situation (from March, 2020, to the time of publication),
the authors have had no physical access to the acoustic laboratory. Therefore, the measurements took
place at home using a test bench. It is comprised of a small chamber with a loudspeaker on top (model
JBL 4TR10A), with partial foam covering (to control for cavity resonances), and with two symmetrical
openings for the microphones, ensuring a very similar sound field at each side. Moreover, before each
measurement, the prototype’s sensitivity was adjusted to ensure that both SLMs returned the same SPL
value at 1 kHz.
First, prototype performance across the spectrum was verified. For this task, a noise with a flat
frequency response was generated in Matlab and filtered into 1/6-octave band signals ranging from
100 Hz to 16 kHz. All signals were adjusted to generate the same SPL levels in Type 2240 when played
through the speaker.
LAeq
values from both the developed system and Type 2240 were assessed and
compared. From the results, a correction curve was implemented to adjust the prototype’s frequency
response. Values below 100 Hz were not assessed due to loudspeaker limitations.
Following, several measurements were held to compare
LAeq
,
LAF-max
, and
LC-peak
values
4
returned
by the SLMs, see Figures 12 (b) to (h). Four types of signals were used. The first was a flat noise
filtered into 1/3-octave band signals ranging from 100 Hz to 12.5 kHz (adjusted to generate the same
SPLs). This test was repeated three times and basic statistical analysis was performed. The second
was a flat noise filtered into 1/1-octave band signals ranging from 100 Hz to 8 kHz. Next, to verify the
response for a broadband signal, a flat noise filtered between 100 Hz and 16 kHz was used. Finally,
to ensure the system’s performance for environmental noise detection, a city soundscape recording
was reproduced several times (it can be found on freesound.org), varying its level from 6 dB above the
background noise (54 dB A-weighted) to 94 dB, in 6 dB increments.
The results presented in Figure 12 show that the prototype achieved an outstanding performance for
the frequency and amplitude range considered, as observed in Figures 12 (b) and (c). For the broadband
and city soundscape measurements, the results for
LAeq
are within
±
0.2 dB, and
LAFmax
are within
±
0.3 dB in comparison to Type 2240 (Figures 12 (e) and (f)).
LCpeak
showed a maximum error of 1 dB.
In addition, it should be noted that further tests must be performed in the laboratory to characterize
prototype responses. Ultimately, the proposed system has the potential to be in accordance with Class 1
SLMs, when concerning the 1/3-octave bands acceptance limits, while complying with Class 2 in the
frequency range.
100 200 400 1000 2500 6350 12500
Frequency [Hz]
-6
-4
-2
0
2
4
6
Relative level [dB ref. 1]
1/3 octave flat noise: mean and confidence interval for 95%
(b) Mean error between the prototype and Type 2240
(a) Illustration of the test bench used for comparison tests. for flat noise filtered in 1/3 octave bands.
Error Class 1 Limits Class 2 Limits ±0.5 dB
SD card Display
Prototype Type 2240
Loudspeaker
+ cavity
(a) Illustration of the test bench used for comparison
tests.
100 200 400 1000 2500 6350 12500
Frequency [Hz]
-6
-4
-2
0
2
4
6
Relative level [dB ref. 1]
1/3 octave flat noise: mean and confidence interval for 95%
(b) Mean error between the prototype and Type 2240
(a) Illustration of the test bench used for comparison tests. for flat noise filtered in 1/3 octave bands.
Error Class 1 Limits Class 2 Limits ±0.5 dB
SD card Display
Prototype Type 2240
Loudspeaker
+ cavity
(b) Measurement error between the prototype and
Type 2240 for flat noise filtered in 1/3 octave bands.
Figure 12: Test bench and comparison measurements results [1/2].
3
Since the Type 2240 has a frequency range from 8 Hz to 16 kHz, a low-pass filter with a cut-offfrequency of 16 kHz
was applied to the prototype algorithm prior to the SPL calculations in order to guarantee a fair comparison. The free-field
microphone correction was also removed from both systems.
4
Peak is different from the maximum sound level. For peak, neither time-weighting nor RMS averaging occurred. It is
the true peak of the sound pressure.
doi: 10.3397/IN-2021-2557. Page 17 of 22

100 200 400 1000 2500 6350 12500
Frequency [Hz]
84
86
88
90
92
94
96
LAeq [dB ref. 20 µPa]
Mean LAeq for flat noise filtered in 1/3 octave bands
(c) Mean LAeq values measured by the prototype and Type 2240
for flat noise filtered in 1/3 octave bands (3 runs).
LAeqprototype
LAeq2240 Class 1 limits Class 2 limits
125 250 500 1000 2000 4000 8000
Frequency [Hz]
84
86
88
90
92
94
96
LAeq [dB ref. 20 µPa]
LAeq for flat noise filtered in 1/1 octave bands
(d) LAeq values measured by the prototype and Type 2240
for flat noise filtered in 1/1 octave bands.
LAeqprototype
LAeq2240 Class 1 limits Class 2 limits
(c) Mean LAeq values measured by the prototype and
Type 2250 for flat noise filtered in 1/3 octave bands (3
runs).
100 200 400 1000 2500 6350 12500
Frequency [Hz]
84
86
88
90
92
94
96
LAeq [dB ref. 20 µPa]
Mean LAeq for flat noise filtered in 1/3 octave bands
(c) Mean LAeq values measured by the prototype and Type 2240
for flat noise filtered in 1/3 octave bands (3 runs).
LAeqprototype
LAeq2240 Class 1 limits Class 2 limits
125 250 500 1000 2000 4000 8000
Frequency [Hz]
84
86
88
90
92
94
96
LAeq [dB ref. 20 µPa]
LAeq for flat noise filtered in 1/1 octave bands
(d) LAeq values measured by the prototype and Type 2240
for flat noise filtered in 1/1 octave bands.
LAeqprototype
LAeq2240 Class 1 limits Class 2 limits
(d) LAeq values measured by the prototype and Type
2250 for flat noise filtered in 1/1 octave bands.
SPL values for broadband signal
88.3
90.4
109.9
88.4
90.5
110.7
(e) SPL values measured by the prototype and Type 2240
for flat noise filtered between 100 Hz and 16 kHz.
LAeq LAFmax LCpeak
40
60
80
100
120
SPL [dB ref. 20 µPa]
Prototype Type 2240
LAeq values for a city soundscape recording
59.5
64.6
70.4
76.5
82.5
88.5
94.5
59.7
64.7
70.6
76.6
82.6
88.6
94.6
(f) LAeq values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
40
50
60
70
80
90
100
110
SPL [dB ref. 20 µPa]
Prototype Type 2240
(e) SPL values measured by the prototype and Type
2240 for flat noise filtered between 100 Hz and 16 kHz.
SPL values for broadband signal
88.3
90.4
109.9
88.4
90.5
110.7
(e) SPL values measured by the prototype and Type 2240
for flat noise filtered between 100 Hz and 16 kHz.
LAeq LAFmax LCpeak
40
60
80
100
120
SPL [dB ref. 20 µPa]
Prototype Type 2240
LAeq values for a city soundscape recording
59.5
64.6
70.4
76.5
82.5
88.5
94.5
59.7
64.7
70.6
76.6
82.6
88.6
94.6
(f) LAeq values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
40
50
60
70
80
90
100
110
SPL [dB ref. 20 µPa]
Prototype Type 2240
(f) LAeq values measured by the prototype and Type
2240 for a city soundscape recording (+6 dB every
run).
LAFmax values for a city soundscape recording
72.3 72.7
75.0
81.0
86.9
92.6
98.4
72.6 73.0
75.0
81.0
86.9
92.6
98.3
(g) LAFmax values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
40
50
60
70
80
90
100
110
SPL [dB ref. 20 µPa]
Prototype Type 2240
LCpeak values for a city soundscape recording
102.1 102.3 102.4
104.9
110.7
116.5
123.1
101.9 102.1 102.2
105.8
111.6
117.5
123.4
(h) LCpeak values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
90
100
110
120
130
SPL [dB ref. 20 µPa]
Prototype Type 2240
(g)
LAFmax
values measured by the prototype and Type
2240 for a city soundscape recording (+6 dB every
run).
LAFmax values for a city soundscape recording
72.3 72.7
75.0
81.0
86.9
92.6
98.4
72.6 73.0
75.0
81.0
86.9
92.6
98.3
(g) LAFmax values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
40
50
60
70
80
90
100
110
SPL [dB ref. 20 µPa]
Prototype Type 2240
LCpeak values for a city soundscape recording
102.1 102.3 102.4
104.9
110.7
116.5
123.1
101.9 102.1 102.2
105.8
111.6
117.5
123.4
(h) LCpeak values measured by the prototype and Type 2240
for a city soundscape recording (+6 dB every run).
1234567
Measurement number
90
100
110
120
130
SPL [dB ref. 20 µPa]
Prototype Type 2240
(h) LCpeak values measured by the prototype and Type
2240 for a city soundscape recording (+6 dB every
run).
Figure 12: Test bench and comparison measurements results [2/2].
5.3. Noise monitoring for long periods
Since the intended purpose of the prototype is to be an autonomous noise monitoring system, the
device was assembled and placed to measure in four different scenarios: in a shopping mall food court,
on the facade of a residence during morning and evening periods, and in a bedroom throughout the
night. The goal was to evaluate the system’s reliability and capability for measurement over long
periods.
The device was set to return
Leq
values and fast-weighted SPL values with A, C, and Z frequency
weightings for all measurements. The integration period was set to 125 ms. Thus, eight values were
saved for each parameter every second. Later, in post-processing, the global Leq was calculated.
Measurement in the food court was taken during lunchtime at the Park Lagos mall (Cabo Frio,
doi: 10.3397/IN-2021-2557. Page 18 of 22

Rio de Janeiro, Brazil). The evaluation started at 12:20 pm and ended at 12:50 pm, rendering a total
of 30 minutes. This period was chosen since it is the most typical lunch break for workers in Brazil,
when the daily food court activity is at its peak. Seat occupancy and prominent sound sources during
the measurement session were assessed qualitatively and recorded on a notepad. In the beginning,
about 30% to 40% of the seats were occupied, over the second half of the time period, 40% to
60% occupancy was observed. Noise sources included conversations, background music, and kitchen
operations. Weather conditions such as temperature and relative air humidity were verified using the
weather.com database (20°C and 76%, respectively).
The residence where the facade noise was evaluated is located near the main access highway to
the city of Cabo Frio, Brazil. Accordingly, measurements were performed at the times in which most
workers transit to and from work; specifically, from 9:00 am to 10:00 am (for the morning period) and
from 6:00 pm to 7:00 pm (for the afternoon/evening period). Noticeable sound sources in the morning
were vehicles (cars, motorcycles, and trucks), construction sites, animals (birds and dogs), as well as
neighborhood activity, while in the evening, vehicles. The weather conditions for the morning were
20°C and 90% relative air humidity, and for the evening, 21°C and 82% relative air humidity.
The last measurement was held with the device placed on a wooden desk inside a bedroom. The
evaluation occurred from 10:25 pm to 6:18 am (approximately 8 hours). During this period, a wind
storm befell. Ultimately, all tests were performed successfully, and the prototype operated flawlessly.
SPL and Leq plots throughout the time for all evaluations are presented in Figure 13.
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court A-weighted noise levels
76.1 dB 74.7 dB 74.7 dB
(a) Food court A-weighted noise levels
between 12:20 pm and 12:50 pm.
LAF LAeq,10min LAFmax = 93.7 dB LAeq = 75.3 dB
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court Z-weighted noise levels
78.7 dB 77.5 dB 77.5 dB
(b) Food court Z-weighted noise levels
between 12:20 pm and 12:50 pm.
LZF LZeq,10min LZFmax = 93.8 dB LZeq = 78.1 dB
(a) Food court A-weighted noise levels between
12:20 pm and 12:50 pm.
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court A-weighted noise levels
76.1 dB 74.7 dB 74.7 dB
(a) Food court A-weighted noise levels
between 12:20 pm and 12:50 pm.
LAF LAeq,10min LAFmax = 93.7 dB LAeq = 75.3 dB
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court Z-weighted noise levels
78.7 dB 77.5 dB 77.5 dB
(b) Food court Z-weighted noise levels
between 12:20 pm and 12:50 pm.
LZF LZeq,10min LZFmax = 93.8 dB LZeq = 78.1 dB
(b) Food court Z-weighted noise levels between
12:20 pm and 12:50 pm.
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court C-weighted noise levels
78.5 dB 77.2 dB 77.2 dB
(c) Food court C-weighted noise levels
between 12:20 pm and 12:50 pm.
LCF LCeq,10min LCFmax = 93.7 dB LCeq = 77.9 dB
75.3 dB 78.1 dB 77.9 dB
LAeq LZeq LCeq
0
20
40
60
80
SPL [dB ref. 20 µPa]
93.7 dB 93.8 dB 95.7 dB
Food court noise levels comparison
(d) Food court frequecy-weighted noise levels comparison.
LAFmax LZFmax LCFmax
50
60
70
80
90
100
(c) Food court C-weighted noise levels between
12:20 pm and 12:50 pm.
0 5 10 15 20 25 29.7
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Food court C-weighted noise levels
78.5 dB 77.2 dB 77.2 dB
(c) Food court C-weighted noise levels
between 12:20 pm and 12:50 pm.
LCF LCeq,10min LCFmax = 93.7 dB LCeq = 77.9 dB
75.3 dB 78.1 dB 77.9 dB
LAeq LZeq LCeq
0
20
40
60
80
SPL [dB ref. 20 µPa]
93.7 dB 93.8 dB 95.7 dB
Food court noise levels comparison
(d) Food court frequecy-weighted noise levels comparison.
LAFmax LZFmax LCFmax
50
60
70
80
90
100
(d) Food court frequency-weighted levels comparison.
Figure 13: Sound pressure levels and Leq for four different noise monitoring scenarios [1/3].
doi: 10.3397/IN-2021-2557. Page 19 of 22

0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade A-weighted noise levels (morning)
55.9 dB 55.2 dB 54.9 dB 56.6 dB 56.0 dB 58.8 dB
(e) Facade A-weighted noise levels
between 9:00 am and 10:00 am.
LAF LAeq,10min LAFmax = 79.3 dB LAeq = 55.8 dB
0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade Z-weighted noise levels (morning)
66.6 dB 64.4 dB 64.3 dB 69.0 dB 64.4 dB 65.7 dB
(f) Facade Z-weighted noise levels
between 9:00 am and 10:00 am.
LZF LZeq,10min LZFmax = 89.9 dB LZeq = 66.0 dB
(e) Facade A-weighted noise levels between 9:00 am
and 10:00 am.
0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade A-weighted noise levels (morning)
55.9 dB 55.2 dB 54.9 dB 56.6 dB 56.0 dB 58.8 dB
(e) Facade A-weighted noise levels
between 9:00 am and 10:00 am.
LAF LAeq,10min LAFmax = 79.3 dB LAeq = 55.8 dB
0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade Z-weighted noise levels (morning)
66.6 dB 64.4 dB 64.3 dB 69.0 dB 64.4 dB 65.7 dB
(f) Facade Z-weighted noise levels
between 9:00 am and 10:00 am.
LZF LZeq,10min LZFmax = 89.9 dB LZeq = 66.0 dB
(f) Facade Z-weighted noise levels between 9:00 am
and 10:00 am.
0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade C-weighted noise levels (morning)
65.6 dB 62.7 dB 63.0 dB 68.1 dB 63.0 dB 64.9 dB
(g) Facade C-weighted noise levels
between 9:00 am and 10:00 am.
LCF LCeq,10min LCFmax = 89.2 dB LCeq = 64.9 dB
55.8 dB
66.0 dB 64.9 dB
LAeq LZeq LCeq
0
20
40
60
SPL [dB ref. 20 µPa]
79.3 dB
89.9 dB 89.3 dB
Facade noise levels comparison (morning)
(h) Facade frequecy-weighted noise levels
comparison for the morning period.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(g) Facade C-weighted noise levels between 9:00 am
and 10:00 am.
0 10 20 30 40 50 60
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Facade C-weighted noise levels (morning)
65.6 dB 62.7 dB 63.0 dB 68.1 dB 63.0 dB 64.9 dB
(g) Facade C-weighted noise levels
between 9:00 am and 10:00 am.
LCF LCeq,10min LCFmax = 89.2 dB LCeq = 64.9 dB
55.8 dB
66.0 dB 64.9 dB
LAeq LZeq LCeq
0
20
40
60
SPL [dB ref. 20 µPa]
79.3 dB
89.9 dB 89.3 dB
Facade noise levels comparison (morning)
(h) Facade frequecy-weighted noise levels
comparison for the morning period.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(h) Facade frequency-weighted noise levels
comparison for the morning period.
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade A-weighted noise levels (evening)
57.3 dB 58.4 dB 55.7 dB 55.3 dB 55.5 dB 55.7 dB
(i) Facade A-weighted noise levels
between 6:00 pm and 7:00 pm.
LAF LAeq,10min LAFmax = 73.7 dB LAeq = 56.5 dB
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade Z-weighted noise levels (evening)
62.4 dB
(j) Facade Z-weighted noise levels
between 6:00 pm and 7:00 pm.
LZF LZeq,10min LZFmax = 81.3 dB LZeq = 64.5 dB
(i) Facade A-weighted noise levels between 6:00 pm
and 7:00 pm.
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade A-weighted noise levels (evening)
57.3 dB 58.4 dB 55.7 dB 55.3 dB 55.5 dB 55.7 dB
(i) Facade A-weighted noise levels
between 6:00 pm and 7:00 pm.
LAF LAeq,10min LAFmax = 73.7 dB LAeq = 56.5 dB
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade Z-weighted noise levels (evening)
62.4 dB
(j) Facade Z-weighted noise levels
between 6:00 pm and 7:00 pm.
LZF LZeq,10min LZFmax = 81.3 dB LZeq = 64.5 dB
(j) Facade Z-weighted noise levels between 6:00 pm
and 7:00 pm.
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade C-weighted noise levels (evening)
61.5 dB
(k) Facade C-weighted noise levels
between 6:00 pm and 7:00 pm.
LCF LCeq,10min LCFmax = 79.0 dB LCeq = 63.7 dB
56.5 dB
64.5 dB 63.7 dB
LAeq LZeq LCeq
0
20
40
60
SPL [dB ref. 20 µPa]
73.7 dB
81.3 dB 79.5 dB
Facade noise levels comparison (evening)
(l) Facade frequecy-weighted noise levels
comparison for the evening period.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(k) Facade C-weighted noise levels between 6:00 pm
and 7:00 pm.
0 10 20 30 40 50 60
Time [min]
30
40
50
60
70
80
SPL [dB ref. 20 µPa]
Facade C-weighted noise levels (evening)
61.5 dB
(k) Facade C-weighted noise levels
between 6:00 pm and 7:00 pm.
LCF LCeq,10min LCFmax = 79.0 dB LCeq = 63.7 dB
56.5 dB
64.5 dB 63.7 dB
LAeq LZeq LCeq
0
20
40
60
SPL [dB ref. 20 µPa]
73.7 dB
81.3 dB 79.5 dB
Facade noise levels comparison (evening)
(l) Facade frequecy-weighted noise levels
comparison for the evening period.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(l) Facade frequency-weighted noise levels comparison
for the evening period.
Figure 13: Sound pressure levels and Leq for four different noise monitoring scenarios [2/3].
doi: 10.3397/IN-2021-2557. Page 20 of 22

0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom A-weighted noise levels
51.4 dB 49.9 dB 49.5 dB 45.4 dB 45.4 dB 45.8 dB 46.3 dB 46.3 dB
(m) Bedroom A-weighted noise levels
between 10:25 pm and 6:18 am.
LAF LAeq,60min LAFmax = 83.0 dB LAeq = 50.2 dB
0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom Z-weighted noise levels
59.1 dB 57.8 dB 57.6 dB 55.6 dB 55.5 dB 55.8 dB 55.8 dB 55.8 dB
(n) Bedroom Z-weighted noise levels
between 10:25 pm and 6:18 am.
LZF LZeq,60min LZFmax = 86.5 dB LZeq = 58.2 dB
(m) Bedroom A-weighted noise levels between
10:25 pm and 6:18 am.
0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom A-weighted noise levels
51.4 dB 49.9 dB 49.5 dB 45.4 dB 45.4 dB 45.8 dB 46.3 dB 46.3 dB
(m) Bedroom A-weighted noise levels
between 10:25 pm and 6:18 am.
LAF LAeq,60min LAFmax = 83.0 dB LAeq = 50.2 dB
0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom Z-weighted noise levels
59.1 dB 57.8 dB 57.6 dB 55.6 dB 55.5 dB 55.8 dB 55.8 dB 55.8 dB
(n) Bedroom Z-weighted noise levels
between 10:25 pm and 6:18 am.
LZF LZeq,60min LZFmax = 86.5 dB LZeq = 58.2 dB
(n) Bedroom Z-weighted noise levels between
10:25 pm and 6:18 am.
0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom C-weighted noise levels
54.5 dB 54.0 dB 53.5 dB 51.9 dB 51.8 dB 52.1 dB 52.4 dB 52.4 dB
(o) Bedroom C-weighted noise levels
between 10:25 pm and 6:18 am.
LCF LCeq,60min LCFmax = 86.5 dB LCeq = 53.7 dB
50.2 dB
58.2 dB
53.7 dB
LAeq LZeq LCeq
0
10
20
30
40
50
60
SPL [dB ref. 20 µPa]
83.0 dB
86.5 dB 87.4 dB
Bedroom noise levels comparison
(p) Bedroom frequecy-weighted noise levels comparison.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(o) Bedroom C-weighted noise levels between
10:25 pm and 6:18 am.
0 120 240 360 474
Time [min]
40
50
60
70
80
90
SPL [dB ref. 20 µPa]
Bedroom C-weighted noise levels
54.5 dB 54.0 dB 53.5 dB 51.9 dB 51.8 dB 52.1 dB 52.4 dB 52.4 dB
(o) Bedroom C-weighted noise levels
between 10:25 pm and 6:18 am.
LCF LCeq,60min LCFmax = 86.5 dB LCeq = 53.7 dB
50.2 dB
58.2 dB
53.7 dB
LAeq LZeq LCeq
0
10
20
30
40
50
60
SPL [dB ref. 20 µPa]
83.0 dB
86.5 dB 87.4 dB
Bedroom noise levels comparison
(p) Bedroom frequecy-weighted noise levels comparison.
LAFmax LZFmax LCFmax
40
50
60
70
80
90
(p) Bedroom frequency-weighted noise levels
comparison.
Figure 13: Sound pressure levels and Leq for four different noise monitoring scenarios [3/3].
6. CONCLUSION
This study presented a small embedded system based on a MEMS digital microphone and
Teensy microcontroller for applications in noise monitoring (or general SPL measurement).
Procedural/electrical testing revealed that the system fulfills Class 1 SLM weighting and filtering
IEC standards.
Acoustic test bench measurements with a calibrated Class 1 SLM led to spectral and overall
correction filter coefficients which achieved approximately the same SPL and
Leq
values as the Class 1
(although limited in frequency up to 12.5 kHz, in this case within Class 2 frequency range). After
the filtering, relative levels ranged
±
0.5 dB, in general, which are exceptional results, given the
relatively small budget of the prototype. Finally, field measurements revealed adequate performance
with low power and CPU consumption, enabling long-term measurements (depending on the mobile
supply device). Further testing will be carried out in the laboratory when it is possible to return to
in-person activities. Moreover, support files for this project can be accessed via the GitHub repository
https://github.com/eac-ufsm/internoise2021-MEMS .
ACKNOWLEDGMENT
The authors want to thank all the support from the Acoustical Engineering Program at the Federal
University of Santa Maria (UFSM, Brazil), as well as its scholarship programs FIPE and FIEX, which
assisted this project. In special, the first author wants to acknowledge the Internoise 2021 for the grants
for students of Latin America and for electing this project as a winner. Thanks to Denison Oliveira and
Fernando Diaz from HBK/Brüel & Kjær for the support concerning the Type 2240. Finally, a special
thanks also to Joseph Lacey for the text proofreading.
doi: 10.3397/IN-2021-2557. Page 21 of 22

REFERENCES
[1] A. D. Pierce. Acoustics – An Introduction to Its Physical Principles and Applications. Springer
International Publishing, 2019. ISBN 978-3030112134. doi: 10.1007/978-3-030-11214-1.
[2] F. Jacobsen and P. M. Juhl. Fundamentals of General Linear Acoustics. Wiley, 2013.
[3] G. Müller and M. Möser. Handbook of Engineering Acoustics. Springer-Verlag Berlin Heidelberg,
Berlin, 2013. ISBN 978-3540240525. doi: 10.1007/978-3-540-69460-1.
[4] D. A. Bies, C. Hansen, and C. Howard. Engineering noise control. CRC press, 2017.
[5] IEC. Electroacoustics – Sound level meters – Part 1: Specifications. Standard IEC 61672-1:2013,
International Electrotechnical Commission, 2013. International standard.
[6] IEC. Electroacoustics – Sound level meters – Part 2: Pattern evaluation tests. Standard IEC 61672-
2:2013, International Electrotechnical Commission, 2013. International standard.
[7] HBK – Brüel & Kjær. Instrumentation: Sound Level Meter, 2021. URL https://www.bksv.com.
[8] S. Santini and A. Vitaletti. Wireless Sensor Networks for Environmental Noise Monitoring. In 6th
GI/ITG KuVS Fachgespraech Drahtlose Sensornetze, pages 98–101, Aachen, Germany, Jul. 2007.
[9] J. Farrés. Barcelona noise monitoring network. In EuroNoise 2015, Maastricht, NL, May 2015.
[10] E. Vidaña-Vila, J. Navarro, C. Borda-Fortuny, D. Stowell, and R. M. Alsina-Pagès. Low-cost distributed
acoustic sensor network for real-time urban sound monitoring. Electronics, 9(12):1–25, 2020. ISSN
2079-9292. doi: 10.3390/electronics9122119.
[11] P. Malcovati and A. Baschirotto. The evolution of integrated interfaces for MEMS microphones.
Micromachines, 9(7):1–20, n. 323, 2018. ISSN 2072-666X. doi: 10.3390/mi9070323.
[12] P. von Pflug and D. Krischker. Aspects of the use of MEMS microphones in phased array systems. In
Proceedings of Internoise 2017, Hong Kong, China, Aug. 2017. URL http://bit.ly/int2017-mems-array.
[13] Philips Semiconductors. I2S Bus Specification (Revised), 1996. Application note.
[14] Knowles. Sisonic design guide, 2017. Application Note (AN24).
[15] PJRC – P. J. Stoffregen. Teensy 4.0 Development Board,Teensy Audio Library,Audio System Design
Tool,&Creating New Audio Objects, 2021. URL https://www.pjrc.com.
[16] A. Z. Leão and W. D’A. Fonseca. Study of the system identification technique implemented in
Arduino Due and Teensy 3.6 microcontrollers (original: Estudo da técnica de identificação de sistemas
implementada em microcontroladores Arduino Due e Teensy 3.6). Acústica &Vibrações, 32(49):5–14,
2017. ISSN 1983-442X. URL http://bit.ly/teensy-av-2017.
[17] J. Lewis. Analog and Digital MEMS Microphone Design Considerations, 2013. Analog Devices,
Technical Article (MS-2472).
[18] MEMSensing Microsystems. MSM261S4030H0 – I2S digital output mems mic. with multi modes
(datasheet), v 1.7, 2018.
[19] Cortex Microcontroller Software Interface Standard. CMSIS DSP Software Library, 2021.
[20] IEC. Electroacoustics – Octave-band and fractional-octave-band filters – Part 1: Specifications.
Standard IEC 61260-1:2014, International Electrotechnical Commission, 2014. International standard.
[21] IEC. Electroacoustics – Octave-band and fractional-octave-band filters – Part 2: Pattern-evaluation tests.
Standard IEC 61260-1:2016, International Electrotechnical Commission, 2016. International standard.
Cite this article
○
F. R. Mello; W. D’A. Fonseca; P. H. Mareze. MEMS digital microphone and Arduino compatible
microcontroller: an embedded system for noise monitoring. In 50th International Congress and
Exposition on Noise Control Engineering — Internoise 2021, pages 1–12, Washington, DC,
USA, Aug. 2021. doi: 10.3397/IN-2021-2557.
doi: 10.3397/IN-2021-2557. Page 22 of 22
BibTeX file .