HardwareX 10 (2021) e00236

Contents lists available at ScienceDirect

HardwareX
journal homepage: www.elsevier.com/locate/ohx

Modular system for UV–vis-NIR radiation measurement with

wireless communication
J.S. Botero-Valencia ⇑, M. Mejia-Herrera
Grupo de Sistemas de Control y Robótica, Instituto Tecnológico Metropolitano, Medellín, Colombia

a r t i c l e i n f o a b s t r a c t

Keywords: Light wavelengths like Ultraviolet (UV) and Near-Infrared Radiation (NIR) are some topics
Spectrum measurement of great interest for research in renewable energy, agriculture, architecture, interior design,
Internet of things and psychology. Due to the light spectrum influence in all these fields, it is necessary to
Ultraviolet-UV
develop instruments that facilitate their remote measurement and storage using Internet
Near Infrared-NIR
of Things technology. In this work, a modular system of sensors for UV–vis-NIR radiation
Visible-VIS
Modular
measurement is presented. The system includes six multi-spectral sensors that allow 54
different measures. The acquired data can be sent to the cloud in real-time or stored in a
micro SD memory in separate files (per sensor) to facilitate its reading, each data is
time-stamped using Unix format, synchronized with a Network Time Protocol (NTP) server.
The sensor enclosure was standardized in installation and all of them except the AS7265
have the same size, and were covered with a 1/8” PTFE sheet to take advantage of its dif-
fuser characteristic. Finally, the sensors were mounted on a standard 2020 extruded alu-
minum guide rail, this rail allows to fix sensors in different distances and arrangements,
bringing adaptability to the system.
Ó 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).

1. Hardware in context

Lighting is variable of interest in different fields of study, such as renewable energy, indoor cultivation, chemistry, interior
design, and even psychology [1–3]. Because light can be composed of radiation at different wavelengths, its quality can affect
a system in various ways. For example, different UV radiation proportions in a lighting system could trigger health problems
[4,5], alter the energy production of solar panels [6,7], cause vision problems [8], or cause early flowering or other modifi-
cations in some plant species [9,10,2]. That is why different commercial systems have been developed to study the compo-
sition of lighting systems or their behavior [11]. However, these systems are expensive or do not have real-time
communication. These factors make it difficult development of research and integration into control systems to acquire
the data automatically. Additionally, open hardware projects, in general, are a new way to increase knowledge democrati-
zation and expand the knowledge frontier faster. For this reason, it is necessary to support emerging initiatives where inde-
pendent studies and research can be developed without the need for large and expensive laboratories. Additionally, one of
the advantages of generating open hardware, such as the one developed in this work for the growth of research or data anal-
ysis, is the possibility of adding new measures or adjusting parameters that in most cases are complex tasks in instruments.

⇑ Corresponding author.
E-mail address: [email protected] (J.S. Botero-Valencia).
@juansezoh (J.S. Botero-Valencia)

https://doi.org/10.1016/j.ohx.2021.e00236
2468-0672/Ó 2021 The Author(s). Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

Specifications table:
Hardware name Portable low-cost modular UV–vis-NIR acquisition device
Subject area  Engineering
 Instrumentation
 Radiation measures
 Internet of things
Hardware type  Measuring physical properties and in-lab sensors
 Field measurements and sensors
 Electrical engineering and computer science
Open source license Creative Commons Attribution-ShareAlike license
Cost of hardware $229.25 USD
Source file repository https://doi.org/10.17605/OSF.IO/J362K

Lighting studies usually focus on a specific wavelength, usually related to some research problem. Most cases use accu-
rate and expensive equipment that sometimes tends to be complex for operation or some applications.Some articles present
low-cost devices that facilitate the acquisition of lighting information for some wavelength [12–16]. Nevertheless, they are
specific for some operation range(s). The following article presents the construction of a modular system for the acquisition
of light information in UV-NIS and NIR bands with wireless communication for the transmission of the collected data, local
storage on a micro SD memory, and a Li-Po battery for emergency operation. The designed system owns nine sensors that
can be distributed to satisfy different requirements of light quality studies. However, due to their modular nature, it is pos-
sible to add more sensors, modify the position and orientation, as well as the wavelengths or range of interest. System sensor
distribution consists of six multi-spectral sensors, two UV and two with NIR measurements. Therefore the system processes
a total of 54 readings when all nine sensors operate together.Another essential factor in the development of light acquisition
systems is the use of multi-spectral sensors. Such sensors permit the reconstruction of SPD from the available channels using
artificial intelligence techniques [17] or measures to characterize light sources such as TM30-18 [18].
Lighting studies usually focus on a specific wavelength, usually related to some research problem. Most cases use accu-
rate and expensive equipment that sometimes tends to be complex for operation or some applications. Some articles present
low-cost devices that facilitate the acquisition of lighting information for some wavelengths [12–16]. Nevertheless, they are
specific for some operation range(s). The following article presents the construction of a modular system for the acquisition
of light information in UV-NIS and NIR bands with wireless communication for the transmission of the collected data, local
storage on a micro SD memory, and a Li-Po battery for emergency operation. The designed system is suitable for indoor oper-
ation, and owns nine sensors that can be distributed to satisfy different requirements of light quality studies, reconstruct
spectrum wavelength behavior, or estimate electric efficiency in a specific location by gathering 32 different bands. However,
due to their modular nature, it is possible to add more sensors, modify the position and orientation, as well as the wave-
lengths or range of interest. Which brings adaptability to any specific study case without major modifications, reducing
the number of needed devices when deploying diverse lighting research due to the IC bus. The system sensor distribution
consists of six multi-spectral sensors, two UV and two with NIR measurements. Therefore the system processes a total of
54 readings when all nine sensors operate together. Another essential factor in the development of light acquisition systems
is the use of multi-spectral sensors. Such sensors permit the reconstruction of SPD from the available channels using artificial
intelligence techniques [17], measures to characterize light sources such as TM30-18 [18], evaluation of visual performance
at some facilities [19,20], the effect of some power distributions over ocular diseases [21], analysis in situ for some illumi-
nation systems [22], as well as other applications that might require the use of a spectrometer which is more expensive than
the proposed system.

2. Hardware description

The system presented in this work is composed of different sensors to measure UV–vis-NIR radiation to facilitate the
acquisition of such measurements or integrate it to tasks of automatic control of artificial sources. In total, the system owns
up to nine sensors: four of them are multispectral, two principally UV, and three mainly in VIS. The multispectral sensors are
the AS7262 with 6 channels in the VIS, the AS7263 with 6 channels in the NIR, the AS7265 with 3 photodetectors each with 6
channels, and the AS7341 with 11 channels in the VIS–NIR region. In the UV is the LTR390 and the VEML6075, these also
provide compensation values in other regions of the spectrum. In the VIS, the VEML6030 is used, with an approximate
response between 450–650 nm, and two TSL2591 are used, these have two channels, one between 500 and 950 nm and
the other between 650 and 950 nm, the purpose of using two TSL2591 is to verify changes in distance, that is, although
the system is modular, it is always proposed to use these two sensors and place them as far apart as possible to estimate
a change in the region measured by the total array of sensors.
2
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

Table 1
Accuracy

Sensor Type Accuracy [%] Unit Measures

LTR390 Experimental 24 counts 2
TSL2591 Experimental 13 counts 2x2
VEML6075 Experimental 15 counts 6
VEML6030 Experimental 15 counts 1
AS7265 Theoretical 12 counts 18
AS7263 Theoretical 12 counts 6
AS7262 Theoretical 12 counts 6
AS7341 Theoretical 15 counts 11

The use MCU is an ESP32 WROOM, selected due to its versatility, ease of acquisition in the market, and because it sup-
ports BT and WiFi communication. Considering that the I2C address of all multi-spectral sensors is the same and cannot be
modified, it was necessary to use the TCA9548A multiplexer, as shown in 1, the multiplexer was also used to connect the two
TSL2591 sensors, the sensors VEML6030, VEML6075, while LTR390 were connected directly to the I2C bus of the ESP32,
using I2C OpenLog for data storage. It is important to mention that the design of system assembly does not need soldering,
and all the sensors used have a Qwiic connector which facilitates their connection. The connector and the LiPo battery charg-
ing system is integrated into the development system selected in this project. Which makes it easier to use, increases usabil-
ity and portability. Additionally, due to the SD card implementation, whenever the device lose internet connection, the data
will still be gathered and saved into local storage for later study and comparison with the information sent via wifi to the
cloud services provider (Ubidots in this case). This backup improves measurement storage security and facilitates research
by integrating different data analysis tools provided by the cloud services. For the mechanical assembly, an enclosure was
designed for each sensors, taking into account the field of view (FOV), and facilitating the access of the connectors (Qwiic).
The enclosures of the sensors were covered with a 1/8 ”sheet of PTFE. Finally, using wedges, and M3 screws, the sensors can
be assembled anywhere in a proposed structure of 2020 extruded aluminum guide rail. It is wanted to highlight, the idea
that the sensors were in separate enclosures allows these sensors to be used more versatilely, the mounting style is just
one example, for ease of installation, nevertheless, mounting can be adapted to different experimental conditions. One of
the main advantages of this project is the diversity of measurements in the same system. Including 31 bands in specific
wavelengths used to know characteristics of a particular source, reconstruct SPD, or determine measures that need or com-
bine UV–vis-NIR, bringing the possibility of updating calculation methods to estimate novel illumination metrics proposed
by the academic community. In the acquisition model, the possibility of storing in SD for a greater volume of data, or send it
to the cloud, can support and facilitate research in multiple areas such as architecture, light engineering, and psychology,
among others. The proposed architecture for data acquisition uses a blocking network which implies reading each sensor
independently, the delay between measurements is about three seconds. And each measure is stored along with its respec-
tive timestamp to keep the variables chronology, facilitate the posterior analysis or use and ensure data collection. The sys-
tem publishing frequency is a variable parameter in the code and can be adjusted for any desired application. Although the
used sensors gather data down to the 0.25 Hz range in the default configuration, data is gathered and published with less
than four seconds of difference, which is an advantage with respect to most of the commercial systems that do not store
information or have an internet connection [24]. Additionally, as a security factor, when one of the sensors fails, an alert mes-
sage indicates which one has to be replaced to keep tracking sensor correlation and therefore system accuracy. Table 1 shows
the accuracy for each of the sensors used in this hardware, the experimental or theoretical type, refers to whether the accu-
racy was taken from the data-sheet (theoretical) or obtained by comparison (experimental), the number of measurements
delivered by each sensor is also displayed. The proposed hardware provides several advantages:

 Supports the acquisition of UV–vis-NIR radiation for renewable energy, agriculture, architecture, interior design, and psy-
chology purposes.
 Easy to integrate with smart systems to control artificial light sources, and thus improve the comfort conditions of the
occupants.
 The system is modular and sensors can be removed to simplify acquisition and adapt them to a specific application.
 It can save the sensor data in an SD memory for saving energy for longer time data acquisition or when there is no Wi-Fi
connection.

3. Design files

Fig. 1 shows the wiring between the electronic elements of the device. The main components are the ESP32 and the mul-
tiplexer TCA9548A. The sensors used are, AS7262, AS7263, AS7265 and AS7341 (multispectral), mainly UV, LTR390 and
VEML6075, mainly VIS TSL2591 (2) and VEML6030. Additionally, the OpenLog is necessary to store the data in an SD memory
and a LiPo battery to back up the power in case of failures, the charger is integrated into the ESP32 board.
3
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

Fig. 1. Schematic.

3.1. Design files summary

Design file name File type Open source license Location file
0_Case_external .stl GNU GPL 3.0 https://osf.io/h9zwy/
0_Case_internal_Adafruit .stl GNU GPL 3.0 https://osf.io/by5ap/
0_Case_internal_Sparkfun .stl GNU GPL 3.0 https://osf.io/8p3he/
0_Case_external_AS7265 .stl GNU GPL 3.0 https://osf.io/pd76k/
0_Case_internal_AS7265 .stl GNU GPL 3.0 https://osf.io/r8keh/
0_Acrylic_Base_Generic .stl GNU GPL 3.0 https://osf.io/2d6tb/
1_Assemble .png GNU GPL 3.0 https://osf.io/n5h73/
1_Schematic .eps GNU GPL 3.0 https://osf.io/u8fet/
3_Code .zip GNU GPL 3.0 https://osf.io/qjf82/

 The 0_Acrylic_BaseGeneric consists of a generic 3D printed grid plate for additional hardware if needed. Although the file is
mainly designed for a 3D printer, this piece can be laser cut Table 2.
 The 0_Case_external correspond to the external part of the enclosure for all the used sensors except the AS7265, and also
serves as a frame designed for positioning a sheet of Teflon Table 2.

Table 2
Listed references and respective STL file for replication of the system.

Reference STL Amount

Acrylic base 0_Acrylic_BaseGeneric 1
Sensor Enclosure 0_Case_external 6
Adafruit Sensor holder 0_Case_internal_Adafruit 3
Sparkfun Sensor holder 0_Case_internal_Sparkfun 3
AS7265 Sensor Enclosure 0_Case_external_AS7265 1
AS7265 Sensor holder 0_Case_internal_AS7265 1

4
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

 The 0_Case_internal_Adafruit is the internal part of the enclosure for the used Adafruit sensors (i.e. TSL2591, LTR390,
AS7341, AS7262, AS7263), and complement the 0_Case_external Table 2.
 The 0_Case_internal_Sparkfun is the internal part of the enclosure for the used Sparkfun sensors (i.e. VEML6030,
VEML6030), and complement the 0_Case_external Table 2.
 The 0_Case_external_AS7265 correspond to the external part of the enclosure for the AS7265, and also serves as a frame
designed for positioning a sheet of PTFE. Table 2
 The 0_Case_internal_AS7265 consists on the internal part of the enclosure for the AS7265, and complement the 0_Case_ex-
ternal Table 2.
 The 1_Assemble consists in graphical aid to the system construction that contains the 3d assembled system.
 The 1_Schematic for the connection of the electronic components.
 In the 3_Code.ino file, is a C code to save into an SD memory, and post gathered information into the cloud.

4. Bill of materials

Table 3

Table 3
Bill of electronic components and suppliers web pages.

Designator Component Qty Unit cost Total cost Source of material

TSL2591 Light sensor 2 $ 6,95 $ 13,90 https://t.ly/UKcD
LTR390 UV sensor 1 $ 4,95 $ 4,95 https://t.ly/ICpA
AS7341 10-channel 1 $ 15,95 $ 15,95 https://t.ly/qrm1
AS7262 VIS sensor 1 $ 25,95 $ 25,95 https://t.ly/CHRk
AS7263 NIR sensor 1 $ 25,95 $ 25,95 https://t.ly/6MhM
AS7265X 18-channel 1 $ 64,95 $ 64,95 https://t.ly/qUxr
VEML6030 Light sensor 1 $ 5,25 $ 5,25 https://t.ly/Un1V
VEML6075 UV sensor 1 $ 7,50 $ 7,50 https://t.ly/TgUy
TCA9548A Mux 1 $ 11,95 $ 11,95 https://t.ly/DF4l
Qwiic cable Cable 10 $ 1,50 $ 15,00 https://t.ly/wck7
Qwiic OpenLog SD log 1 $ 16,95 $ 16,95 https://t.ly/fvaa
ESP32 MCU 1 $ 20,95 $ 20,95 https://t.ly/1lhR
Total $ 229,25

5. Build instructions

The system was designed as a modular assembly that allows different sensor configurations and gives adaptability to
many scenarios making it easier to develop lighting research. The structure is composed of three rectangular rail beams
of size 20 mm  20 mm  200 mm organized as in ‘H’ form. The beam unions consist of a couple of corner connection pieces
and a trapezoidal nut juncture that allows fixing the beams and sensors placed on them as shown in Fig. 4b.
Beneath the ‘H’ structure, there is an acrylic sheet with a grid of perforations for M3 screws designed for the location of
electronic components like batteries and micro-controllers. Nevertheless, due to the hole distribution, it is possible to add
more elements if needed.
After structure construction sensor enclosure should be 3D printed, in this case, the task was done by using FDM (Fused
deposition Material) technology. The printing process should be done using a proper set up to fit and align the models and
reduce post processing job. The sensor cases consist of three components: (1) an internal part for fixing a sensor, (2) an exter-
nal piece that protects the sensor, and allows positioning of a (3) sheet of Teflon that aids a homogeneous light distribution
over the sensor. Each sensor case was designed to be placed whether on top, or at the laterals of the beams as shown in
Fig. 4a, and each case STL file is available for 3D printing. To fix all the pieces it is necessary to use M3 bolts and nuts, some
of them should be carefully inserted into the internal part of the cases. Although the external component is the same for
every used sensor (except the AS7265), the inner piece varies according to each sensor geometry. It is highly recommendable
to plug the sensors before closing the enclosure, to avoid cable damages or non desired stretches or bendings.
After the support structure assembly, and sensor enclosure constructions, the sensors should be placed in the ‘H’ structure
by using trapezoidal nuts as mentioned before. Although different configurations can be done, we recommend symmetrical
arrays that allows comparable illumination information, as shown in Fig. 4a. Finally, all the additional electronic components
should be placed on the acrylic sheet and carefully plug together by following the Fig. 1, and the Fig. 2 shows the mechanical
assembly of the proposed system.

5
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

Fig. 2. 3D assembly and position of the components.

6. Operation instructions

Operation of the developed system does not require any additional task than turn on the system and charge batteries
occasionally. Nevertheless, before first time turning on, it does require adjusting some specific snippets of the available code
to configure parameters such as Wifi ID, Wifi password, and the names for each sensor output log. This process can be done
quickly by following the subsequent step-by-step information. Although the main code uses other files such as libraries and
headers, these files should not be modified unless you need some specific hardware modification. Otherwise, and for the
intended application, we will focus only on the Code.ino for the before-mentioned parameter configuration.

1. Download and install Arduino IDE, and set up LinKit One by following the instructions of the manufacturer.
2. Open Code.ino using Arduino IDE or any desired text editor
3. Modify Wifi network setup at lines 39 and 40 replacing, ‘‘Name” for your wifi ID and ‘‘Pass” for your private network pass-
word, see listing 1.
4. At line 41, 42, and 93 respectively. Replace ‘‘Token” with your web cloud provider token (Ubidots in this case), ‘‘Client”
with your MQTT client name, and ‘‘industrial.api.ubidots.com” with your MQTT broker, see listing 1.
5. In the line 44, the user can modify the sampling period, that is, the interval in which the acquisition of all the
connected sensors will be carried out, this time is given in milliseconds.

6. Lines 76 to 91 correspond to the output log names of the system. Such information can be printed in the system or saved
into the predefined log files by setting ‘‘true” lines 95 and 96.
7. Once the code its adjusted, connect the MCU (Microcontroller Unity) via USB to the computer and press the upload button
at the Arduino IDE, If the MCU terminal is checked, output messages are delivered showing errors or correct operation of
the system.
8. After program uploading, the MCU will return multiple connection data, and the output is shown in Listing 2.
Such information is included in the code to facilitate debugging for users. And it is the resulting sensor state of the
scanning and enabling of the sensors connected to the I2C bus and the multiplexer. The OK status of each sensor indicates
that it is connected and functional and ‘ER’ that it is disconnected or had a problem. The first two values (in brackets)
represent the time in milliseconds of the MCU and the RTC (Real Time Clock), in Unix format.

6
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

9. Finally, in Listing 3, a typical example of running output is shown, it can be seen that: the MCU publishes the value of each
variable, and their outputs are programmed within the code to facilitate debugging. Available gathered data is sent to the
cloud for its processing and application, and the output is presented every 10 s in this case, nevertheless, this parameter
can be adjusted in the code. The system acquires all sensors data within 2929 ms, this time can be affected by the type of
used SD memory, however, each sensor is marked with its respective timestamp to know the acquisition instant.

It is important to consider which web service will be used to send the data to the cloud. In this case, the Ubidots company
service was chosen. Since MQTT is used, if you want to change the operator, then you only have to update the code and
libraries according to the selected service provider. Once the device is connected to a Wi-Fi or GPRS network and is properly
located to acquire the data, it is only a matter to check the information on the web and using the data according to the appli-
cation. Note, in case the user wants to make calculations directly in the MCU, the program code should be changed.

7. Validation and characterization

The Fig. 3 shows a summary of the available bands for each of the multi-spectral sensors. The system collects 38 bands
where six are repeated bands, so there are 32 bands between 410 and 940 nm considering the central wavelength. In Fig. 4,
photographs of proposed system assembly are shown.
Fig. 5 shows the full-day solar acquisition response with the multispectral sensors locating the system behind a window
facing the exterior. The graphs 5a–5b and 5e–5f-6a correspond to the sensors AS7341 and AS7265 information respectively,
the AS7265 has three independent sensors, AS7265-1, AS7265-2, AS265-3, in the legends can see the specific wavelengths,
data were divided to facilitate observation. It is important to notice that the sensitivity of the detectors can be adjusted dig-
itally, which is an advantage to manage significant changes in the actual application. The purpose of this test is to perform an
acquisition in an unstable environment since the behavior of solar radiation is affected by multiple environmental, atmo-
spheric factors and the earth’s rotation.
Fig. 6 shows the full-day solar acquisition response mainly in the VIS and in the UV using the non-multi-spectral sensors
locating the system behind a window facing the exterior. Fig. 6c and 6d, shows the response similarity of the TSL22 and
TSL25 sensors, that correspond to the two TSL2591, which are recommended to be located far apart from each other in
the assembly to use them as a reference and to bring uniformity to the measurements. The system also gathers the UV
response, as shown in Figs. 6b and 6f, such sensors were added to the system mainly to monitor solar activity.
Fig. 7 shows the response of multi-spectral sensors tested under a controlled environment, and illuminated with a 4000 K
artificial LED light source for six hours. The purpose of this test was to show the stability of the measurements over time in a
stable environment. For that reason, the system was illuminated by an artificial light source, in an environment without
external influence from other sources.. The decrease in light output, which is consistent across all sensors at the beginning
of the test, is a result of the time taken for the LED driver to stabilize the current. It can be seen that the infrared component
is non-existent as expected and is presented as an application of the proposed system to estimate light efficiency.
Finally, in the Fig. 8, the response for the other sensors illuminated with the artificial light source LED is shown, this test
was carried out for 6 h. Again it can be observed that the response in the NIR is non-existent and additionally that the LED
source has a low level of UVA radiation and a very low level of UVB radiation. For indoor applications, one of the most impor-
tant uses of this system would be light efficiency analysis and certifying light stability over time. This mainly because some-
times drivers, the electrical network, or obsolescence, generate degradation in the quality of the light artificial sources that
should be considered.
The correlation matrix is used to establish the dependence between two variables, such dependence is a measure of the
relationship between independent variables between 1 and 1. In Table 4, the correlation matrix for some selected channels
is presented, the number of variables was reduced to simplify the presentation. The data correspond to the acquisition made
of a full day of Sun, the high correlation is the result of the spectral composition of the Sun that is distributed in all the mea-
sured channels including the NIR and UV channels.
7
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

In contrast to the correlation matrix in the Table 4, in Table 5, the matrix for the six-hour acquisition of the artificial light
source LED is presented. It can be seen that now the correlation with the UV and NIR channels is low, and the reason is the
spectral composition of the LED sources that have relatively low values in these bands concerning the response in the VIS.

1
0.9
0.8 0.8
415 0.7
445 450
0.6 480 0.6 500
515 550
0.5
555 570
0.4 590 0.4 600
630 650
680 0.3
0.2 0.2
0.1
0

450
500
550
570
600
650
415
445
480
515
555
590
630
680

(a) Bands AS7341 (b) Bands AS7262

1 1
410
435
460
0.8 0.8 485
510
610 535
0.6 680 0.6 560
730 585
760 610
0.4 810 0.4 645
860 680
705
0.2 0.2 730
760
810
0 0 860
610

680
730
760
810
860

410
435
460
485
510
535
560
585
610
645
680
705
730
760
810
860
900
940
900
940

(c) Bands AS7263 (d) Bands AS7265

Fig. 3. Bands available in multispectral sensors.

(a) Top view (b) Side view

Fig. 4. Photographs of the real device.

8
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

w415 w555
2500 w445 2500 w590
w480 w630
w515 w680
2000 wCLR 2000 wCLR
wNIR wNIR
1500 1500

1000 1000

500 500
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
Mar 13, 2021 Mar 13, 2021

(a) AS7341-1 (b) AS7341-2

1200 1000
w450 w610
w500 w680
1000 w550 w730
800
w570 w760
800 w600 w810
w650 600 w860

600
400
400

200
200

0 0
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
Mar 13, 2021 Mar 13, 2021

(c) AS7162 (d) AS7163

400 w410 w560
350
w435 w585
w460 300 w645
300 w485 w705
w510 250 w900
w535 w940
200
200
150

100 100

50

0 0
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

Mar 13, 2021 Mar 13, 2021

(e) AS7165-1 (f) AS7165-2

Fig. 5. Data acquired from the Sun in a full day.

Table 6 shows the normalized response on a scale from 0 to 100 of the cross-talk of the channels of the multispectral sen-
sors used in this work. To obtain the results, a broad spectrum source EQ-99, monochromators and a spectrometer were
used, the procedure can be seen in [23]. The FWHM of the monochromator is approximately 15 nm, it can be observed that
there is a level of contamination between nearby channels (highlighted in yellow), however, the highest response was
always given in the specific value of the channel (highlighted in green), that is It is important to clarify that the columns
are the values in which the monochromator was set to make a measurement and the same values of the available bands were
considered.

9
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

8 UV
w610
800 w680
w730
w760 6
600 w810
w860

4
400

200 2

0 0

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
Mar 13, 2021 Mar 13, 2021

(a) AS7265-3 (b) LT390

300
FULL 350 FULL
IR IR
250 300

200 250

200
150
150
100
100
50 50

0
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
Mar 13, 2021 Mar 13, 2021

(c) TSL2591-1 (d) TSL2591-2

500 VIS RUVA
500 RUVB

400
400

300
300

200 200

100 100
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00

Mar 13, 2021 Mar 13, 2021

(e) VML6030 (f) VML6075

Fig. 6. Data acquired from the Sun in a full day.

 Taking into account the number of sensors in the system, the battery should only be considered as a backup to a power
failure and not to provide fully autonomy to the system as the battery capacity will not provide long-lasting power, and
there is no solar or additional source power to keep the Lipo Battery charged. Instead, the ESP32 monitors the charging
state and automatically shutdowns the charging stage to protect the battery when reaches the maximum capacity.
 The device is only suitable for indoor operation, weather conditions like rain, storms, or more aggressive environments
can affect its operation.

10
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

500 w415 500 w555

w445 w590
w480 w630
400 w515 400 w680
wCLR wCLR
wNIR wNIR
300 300

200 200

100 100
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
Mar 17, 2021 Mar 17, 2021

(a) AS7341-1 (b) AS7341-2

700 w450 400 w610
w500 w680
w550 350 w730
600 w570 w760
w600 300 w810
w650 w860
250
500
200

400 150
100
300 50
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
Mar 17, 2021 Mar 17, 2021

(c) AS7162 (d) AS7163

250
w410 w560
w435 w585
w460 150 w645
200
w485 w705
w510 w900
w535 w940
150
100

100
50
50
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

Mar 17, 2021 Mar 17, 2021

(e) AS7165-1 (f) AS7165-2

Fig. 7. Data acquired from the artificial light source (LED) in 6 h.

 It is recommended to locate the system in areas with a low circulation of people because light reflections can affect the
measurements.
 The system can store up to 8 GB of data in SD, the files are saved in separate files, one for each sensor and a system diag-
nostic file.
 The model presented in this work offers a large amount of information that can be useful in studies of architecture, light
engineering, psychology, among others.

11
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

2
350 UV
w610
w680
300 w730 1.5
w760
250 w810
w860
200 1

150

100 0.5

50
0

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
Mar 17, 2021 Mar 17, 2021

(a) AS7165-3 (b) LT390

FULL FULL
40 IR 50 IR

40
30

30
20
20

10
10
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30
Mar 17, 2021 Mar 17, 2021

(c) TSL2591-1 (d) TSL2591-2

200
235 VIS RUVA
RUVB

230
150
225

220 100

215

210 50
18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

18:30
19:00
19:30
20:00
20:30
21:00
21:30
22:00
22:30
23:00
23:30

Mar 17, 2021 Mar 17, 2021

(e) VEML6030 (f) VEML6075

Fig. 8. Data acquired from the artificial light source (LED) in 6 h.

 With the acquired data, trend models can be developed over time to make predictions, to interpolate values with their
mixture and to be able to predict some measurements in special applications, as is done with some commercial sensors
in the case of UV.
 The system does not withstand the environmental conditions outside, it is not waterproof. However, it could be consid-
ered as future work, improving the enclosure of the sensors for an outdoor intended device.

12
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

Table 4
Correlation matrix between channels Sun.

Table 5
Correlation matrix between channels, LED source.

Table 6
Crosstalk multispectral sensors.

13
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

8. Human and animal rights

No human or animal studies were conducted in this work.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.

Acknowledgments

This study were supported by the Sistemas de Control y Robótica (GSCR) Group COL0123701, at the Sistemas de Control y
Robótica Laboratory, attached to the Instituto Tecnológico Metropolitano.

References

[1] X. Li, F. Duarte, C. Ratti, Analyzing the obstruction effects of obstacles on light pollution caused by street lighting system in Cambridge, Massachusetts,
Environment and Planning B: Urban Analytics and City Science (2019) 1–15, https://doi.org/10.1177/2399808319861645.
[2] L.E. Särkkä, K. Jokinen, C.O. Ottosen, T. Kaukoranta, Effects of HPS and LED lighting on cucumber leaf photosynthesis, light quality penetration and
temperature in the canopy, plant morphology and yield, Agricultural and Food Science 26 (2) (2017) 101–109, https://doi.org/10.23986/afsci.60293.
[3] D.B. Moyano, M.S.J. Fernández, R.A.G. Lezcano, Towards a sustainable indoor lighting design: Effects of artificial light on the emotional state of
adolescents in the classroom, Sustainability (Switzerland) 12 (10). doi:10.3390/su12104263..
[4] J. Cadet, T. Douki, Formation of UV-induced DNA damage contributing to skin cancer development, Photochemical and Photobiological Sciences 17 (12)
(2018) 1816–1841, https://doi.org/10.1039/c7 pp00395a.
[5] T. Prevenslik, Cancer caused by UV radiation from nanoparticles in GM food, Integrative Cancer Science and Therapeutics 4 (1) (2017) 1–4, https://doi.
org/10.15761/icst.1000222.
[6] A.K. Sahu, K. Sudhakar, Effect of UV exposure on bimodal HDPE floats for floating solar application, Journal of Materials Research and Technology 8 (1)
(2019) 147–156, https://doi.org/10.1016/j.jmrt.2017.10.002.
[7] I. Velo-Gala, J.J. López-Peñalver, M. Sánchez-Polo, J. Rivera-Utrilla, Role of activated carbon surface chemistry in its photocatalytic activity and the
generation of oxidant radicals under UV or solar radiation, Applied Catalysis B: Environmental 207 (2017) 412–423, https://doi.org/10.1016/j.
apcatb.2017.02.028.
[8] N.N. Osborne, C. Núñez-Álvarez, S. del Olmo-Aguado, J. Merrayo-Lloves, Visual light effects on mitochondria: The potential implications in relation to
glaucoma, Mitochondrion 36 (2017) 29–35, https://doi.org/10.1016/j.mito.2016.11.009.
[9] T. Han, V.A. Vaganov, S. Cao, Q. Li, L. Ling, X. Cheng, L. Peng, C. Zhang, A.N. Yakovlev, Y. Zhong, M. Tu, Improving color rendering of LED lighting for the
growth of lettuce, Scientific Reports 7 (2017) 1–7, https://doi.org/10.1038/srep45944.
[10] S. Dutta Gupta (Ed.), Light Emitting Diodes for Agriculture, Springer, Singapore, Singapore, 2017, https://doi.org/10.1007/978-981-10-5807-3.
[11] M.G. Figueiro, R. Hamner, A. Bierman, M.S. Rea, Comparisons of three practical field devices used to measure personal light exposures and activity
levels, Lighting Research and Technology 45 (4) (2013) 421–434, https://doi.org/10.1177/1477153512450453.
[12] D. Dellal, E. Yee, S. Lathwal, H. Sikes, J. Gomez-Marquez, Low-cost plug and play photochemistry reactor, HardwareX 3 (November 2017) (2018) 1–9.
doi:10.1016/j.ohx.2017.11.001..
[13] J.S. Botero-Valencia, J. Valencia-Aguirre, D. Durmus, A low-cost IoT multi-spectral acquisition device, HardwareX 9. doi:10.1016/j.ohx.2021.e00173..
[14] J. Salazar-Vazquez, A. Mendez-Vazquez, A plug-and-play Hyperspectral Imaging Sensor using low-cost equipment (2020), https://doi.org/10.1016/j.
ohx.2019.e00087.
[15] A. Hermann, J. Chladek, D. Stepputtis, iFO (infrared Fish Observation) – An open source low-cost infrared underwater video system (2020).
doi:10.1016/j.ohx.2020.e00149..
[16] K. Laganovska, A. Zolotarjovs, M. Vázquez, K. Mc Donnell, J. Liepins, H. Ben-Yoav, V. Karitans, K. Smits, Portable low-cost open-source wireless
spectrophotometer for fast and reliable measurements (2020). doi:10.1016/j.ohx.2020.e00108..
[17] J.S. Botero-Valencia, J. Valencia-Aguirre, D. Durmus, W. Davis, Multi-channel low-cost light spectrum measurement using a multilayer perceptron,
Energy and Buildings doi:10.1016/j.enbuild.2019.07.026..
[18] J.S. Botero-Valencia, J. Valencia-Aguirre, J.F. Vargas-Bonilla, Estimation of Light Source Color Rendition with Low-Cost Sensors Using Multilayer
Perceptron and Extreme Learning Machine, LEUKOS 00 (00) (2020) 1–11, https://doi.org/10.1080/15502724.2020.1755306.
[19] R. Bhagavathula, R.B. Gibbons, Light Levels for Parking Facilities Based on Empirical Evaluation of Visual Performance and User Perceptions, LEUKOS –
Journal of Illuminating Engineering Society of North America 16 (2) (2020) 115–136, https://doi.org/10.1080/15502724.2018.1551724.
[20] F. Feltrin, F. Leccese, P. Hanselaer, K.A. Smet, Impact of illumination correlated color temperature, background lightness, and painting color content on
color appearance and appreciation of paintings, LEUKOS – Journal of Illuminating Engineering Society of North America 16 (1) (2020) 25–44, https://
doi.org/10.1080/15502724.2018.1522261.
[21] E.G. Vicente, I. Arranz, D. Galarreta, P. Barrionuevo, M. Rodríguez-Rosa, S. Mar, J.A. Aparicio, B.M. Matesanz, Effect of different spectral power
distributions on mesopic visual performance with blue light-filtering intraocular lens, LEUKOS – Journal of Illuminating Engineering Society of North
America 00 (00) (2020) 1–15, https://doi.org/10.1080/15502724.2020.1757461.
[22] F. Li, Y. Chen, Y. Liu, D. Chen, Comparative in situ study of LEDs and HPS in road lighting, LEUKOS – Journal of Illuminating Engineering Society of North
America 8 (3) (2012) 205–214, https://doi.org/10.1582/LEUKOS.2012.08.03.003.
[23] J.S. Botero-Valencia, F.E. Lopez-Giraldo, J.F. Vargas-Bonilla, Characterization of photodectors using a monochromator and a broadband light source in
the XYZ color space, International Journal on Smart Sensing and Intelligent Systems 9 (2) (2016) 752–764, https://doi.org/10.21307/ijssis-2017-893,
URL:https://www.exeley.com/in_jour_smart_sensing_and_intelligent_systems/doi/10.21307/ijssis-2017-893.
[24] M.D. Ahmad, S.Z. Mohammad Noor, N.F. Abdul Rahman, Faranadia Abdul Haris, IEEE International Conference in Power Engineering Application
(ICPEA), Lux Meter Integrated with Internet of Things (IoT) and Data Storage (LMX20) (2021) 138–142, https://doi.org/10.1109/
ICPEA51500.2021.9417762.

14
J.S. Botero-Valencia and M. Mejia-Herrera HardwareX 10 (2021) e00236

J. S. Botero-Valencia Magister in Automation and Industrial Control, and PhD in Engineering, has experience in control systems
and robotics, specifically in the Internet of Things (IoT) and mobile robotics. He currently works as a Professor in the Department
of Mechatronics and Electromechanics of the Faculty of Engineering of the Metropolitan Technological Institute, and belongs to
the Laboratory of Control Systems and Robotics.

M. Mejía-Herrera, Mechatronic Engineer and MSc (e) in Automation and Industrial Control of Instituto Tecnológico
Metropolitano in Medellin Colombia, Member of the research group classification A1 ”Automática, Electrónica y Ciencias
Computacionales” since 2018 with experience in artificial vision, embedded systems, IoT and 3D printing for the development of
low-cost prototypes.

15