<p>The proliferation of efficient edge computing has enabled a paradigm shift of how we monitor and interpret urban air quality. Coupled with the dense spatiotemporal resolution realized from large-scale wireless sensor networks, we can achieve highly accurate realtime local inference of airborne pollutants. In this paper, we introduce a novel Deep Neural Network architecture targeted at latent time-series regression tasks from continuous, exogenous sensor measurements, based on the Transformer encoder scheme and designed for deployment on low-cost power-efficient edge processors.


The research were incorporated an extended cohort monitoring campaign, validation of an existing exposure model and development of a predictive model for COPD exacerbations evaluated against historical electronic health records.A miniature personal sensor unit were manufactured for the study from a prototype developed at the University of Cambridge. The units monitored GPS position, temperature, humidity, CO, NO, NO2, O3, PM10 and PM2.5.Three 6-month cohort monitoring campaigns were carried out, each including of 65 COPD patients.


Dataset used for IEEE Sensors Article: Calibration of CO, NO2 and O3 Using Airify: a Low-Cost Sensor Cluster for Air Quality Monitoring


This dataset is in support of my research paper 'Comparison of ElectroMagnetic Emissions & Harmonic Analysis of 20 HP Motor Controlled by 3L NPC Inverter '.

Preprint : https://doi.org/10.36227/techrxiv.19687041.v1

This is useful for manufacturers and r&d engineers for product costing. For more information, results, conclusions on this, pls read research paper.


This dataset is in support of my 4 Research papers, initially submitted to different journals

  1. 2
  2. 3
  3. 4
  4. 5
  5. 6

Related Reseach Papers :

  1. Novel ß-Bio Model (Mathematics Foundation)
  2. ß-Model of  (Preprint:      )
  3.           and Humans Body - Part I (Preprint:      )
  4.           and Humans Body - Part II (Preprint:      )

Read Me

(1) This is an open access ,so everything  can be downloaded after login (free signup). You have to click on 'Title'.

(2) Data which was  earlier uploaded in 2021 under this same DOI  'Electro-Magnetic Radiations and Human Body' is explained in ' Experimental Physical Recording’.  That data is as it is. Neither earlier  data is removed nor it is modified, it is as it was earlier submitted. No additions are even done.

(3)  The main paper which has my scientific analysis on 'Electro-Magnetic Radiations and Human Body'  is ‘ and Humans Body’. This paper is used as the foundation because of the accepted facts by WHO, ICNIRP, IARC, NIH,medical doctors, and biomedical engineers. In this paper, I have claimed and proved something.

(4) Zip do not contain any simulation project folder.

(5) Extra Libraries created, modified , other scripts , not shared, as very elementary for any graduate,degree holder, so only results given in research paper.

(6) For  details like model block diagram, parameters, analysis, interpretation, mathematical formulae used to obtain these results etc. please refer "Research Paper".

(7) Radiation patterns - If you expecting the patterns are something easy to understand or decode, but they cannot easily interpreted. For this, pls. refer either textbook or research paper.

(8) The mobile tower installation/distance parameters are also taken according to 'Ministry of Communications, Department of Telecommunications,GoI.

(9) All operating frequency ranges are not mentioned for each 2G,3G,4G,5G,6G. For complete operating frequencies, pls refer your country or search on net. For other details, pls see Research paper.

(10)  This work has undergone complete revisions, loss of data many times,  and many computer crashes. 

(11) This is the last version in those datasets. Only update will be related to ß-Bio models which I

(12)  All work is simple , on basic and elementary concepts, can be easily copied, remade and understood.

(13) The dataset has been checked by the 'Data or Code or model Inspector' before uploading.

(14)  If any problem in creating or copying, pls contact your university professor or board or any of the companies engineer.

(15) As such, No other question or email will be replied. I may have left completely R&D or other reason.


Dataset Files

All the following 25 folders are zipped.

1)  2G

  • 2G_800 is CDMA 800MHz or 0.8 GHz
  • 2G_900 is GSM 900MHZ or 0.9 GHz
  • 2G_1800 is GSM1800MHz or 1.8 GHz

2) 3G 

  • 3G_1900 is 1900  MHz or 1.9 GHz
  • 3G_2100 is 2100 MHz or 2.1 GHz

3)  4G

  • 4G_2300 is 2300 MHz or 2.3 GHz
  • 4G_2400 is 2400 MHz or 2.4 GHz
  • 4G_2600  is 2600 MHz or 2.6 GHz

4) Low/Mid 5G FR1

  • 5G_3300 is 3300 MHz or 3.3 GHz
  • 5G_3500 is 3500 MHz or 3.5 GHz
  • 5G_5200  is 5200 MHz or 5.2 GHz
  • 5G_5900  is 5900 MHz or 5.9 GHz
  • 5G_6000 is 6000 MHz or 6 GHz
  • 5G_6200  is 6200 MHz or 6.2 GHz

Here 5G_3500 is n78 C-Band but 5G_6000, 5G_6200 are TDD, n96, n102  UNII defined by  US FCC. For details, pls refer Research paper.

5) High 5G  FR2

  • 5G_26000  is 26000 MHz or 26 GHz
  • 5G_28000  is 28000 MHz or 28 GHz
  • 5G_39000  is 39000 MHz or 39 GHz
  • 5G_41000  is 41000 MHz or 41 GHz
  • 5G_47000 is 47000 MHz or 47 GHz

6) 6G

  • 6G_90000  is 90,000 MHz or 90 GHz
  • 6G_150000 is 150 GHz
  • 6G_220000 is 220 GHz
  • 6G_500000 is 500 GHz
  • 6G_750000 is  750 GHz
  • 6G_1100000 is 1100 GHz, that is, 1.1 Terahertz (THz)

8) Each of the above zip has following datasets. The plots, images can be seen in IEEE CodeOcean DOI.

9) 3G has addition dataset

10) Following datasets are based on ß-Bio


 Experimental Physical Recording

The folder 'PhysicalRecording_2021.zip ' has recordings of Magnetic fields in the year 2021 measured using  Magnetic Sensor,  mobile app(software) and mobile phone

  •  Recordings.zip     
  •  14.mp4     
  •  327uT at 0:19/00:20   .   At 0:19/0:20 of the recording, 327 uT reading
  •  11uT at 0:04/0:05     .  At 0:04/0:05 of the recording, 11 uT reading
  •  5gproof.zip has screenshots from wifi detection   
  •  479uT at 0:42/0:44  .  At  0:42/0:44  of the recording, 479 uT reading

 Area: Delhi,NCR,India 

  • Physical Magnetic Sensor(hardware)

                  Resolution of the sensor is 0.0976 uT    &   Maximum Range of the sensor : 3000.0044 uT

  • Physical orientation and angular velocity  Sensor  (hardware)

                Resolution of the sensor is 0.0012216975 rad/s  &   Maximum Range of the sensor : 34.90549 rad/s

  •  Physical Proximity Sensor (hardware)

                    Resolution of the sensor : 1.0 cm    &   Maximum Range of the sensor : 5.0 cm

  •  Physical Gravity Sensor (hardware)

                    Resolution of the sensor :  0.01 m/s2  & Maximum Range of the sensor :156.98999 m/s2

Experimental Result   -   Lowest Recorded Reading : 11 uT

Highest Recorded Reading : 479 uT

Around 300 uT was measured anywhere, if nearby has 5G equipment ( fluctuates to 50 uT then 111, then 200 , 286,  ...) .   More details in paper.

Reading of 479 was measured, as few people were feeling unwell and when I checked, it was 420 uT, stationary and fluctuating to it around but that is not recorded.   So after some time, this was recorded.

 But later, this reading went to below 200 uT ? And even from 30 uT to 150 uT ,  how come

Experimental Result   - 24 April 2022, See Corona cases, rising, reading which was 29uT to 150uT is 243.95 uT



For scripts of IEEE Codeocean (Rstudio & Matlab). To see colored plots and images, pls. read details given in ReadMe.txt.

  • Capsule : Plots of EM Fields in 2G              , DOI :
  • Capsule : Plots of EM Fields in 3G            , DOI :
  • Capsule : Plots of EM Fields in 4G             , DOI :
  • Capsule : Plots of EM Fields in 5G             , DOI :
  • Capsule : Plots of EM Fields in 6G             , DOI :

Paper Citing : If want to cite this in paper etc. ,please refer DoI and/or this url.

Funding: There are no funders for this submission. The  author has himself fully self-financed (for his passion).I expect all these papers, would be nice Shroud for the passion and the price paid.

Acknowledgement : The author has generated this on Linux and had even used IEEE partner- Code Ocean - Python,C, Matlab ,Cloud Workstation, Jupyter Notebook,Rstudio,stata,julia,Tensorflow, pandas,trial (evaluation) of many proprietary softwares. No paid research, personal R&D work with no support, wastage of time in self teaching.Few gave trial (evaluation) sw with 2-5 months with even willing for 3-6 months further extension but didnt accepted hire contract request (the names cannot be disclosed & word of acknowledging expired in duration). No industry or academic will use their time only doing this work, even if given free unless financed or top MNC.  The author does not have any special name to be acknowledged.


This dataset is in support of my research paper 'ElectroMagnetic Fields in Wireless Charging of Electric Vehicles '.

Preprint :

This is useful for industries, manufacturers,doctors,environmentalists, who are curious to see and know.


This work contains data gathered by a series of sensors (PM 10, PM 2.5, temperature, relative humidity, and pressure) in the city of Turin in the north part of Italy (more precisely, at coordinates 45.041903N, 7.625850E). The data has been collected for a period of 5 months, from October 2018 to February 2019. The scope of the study was to address the calibration of low-cost particulate matter sensors and compare the readings against official measures provided by the Italian environmental agency (ARPA Piemonte).


A Densely-Deployed, High Sampling Rate, Open-Source Air Pollution Monitoring WSN

Documentation for the air pollution monitoring station developed at Politecnico di Torino by:
Edoardo Giusto, Mohammad Ghazi Vakili under the supervision of Prof. Bartolomeo Montrucchio.

System Overview

This section includes a description of our architecture from several points of view, going from the hardware and software architecture, to the communication protocols.

Hardware Architecture

We target the following key characteristics of our system:

  1. The rapid and easy prototyping capabilities,
  2. Flexibility in connection scenarios, and
  3. Cheapness but also dependability of components.

As each board has to include a limited number of modules, to facilitate our prototype development, we select the Raspberry Pi single-board computer as a monitoring board.
Due to our constraints in terms of cost, size and power consumption we select its Zero Wireless version based on the ARM11 microprocessor.

The basic operating principle of the system is the following. The data gathered from the sensors are stored in the MicroSD card of the RPi. At certain time intervals the RPi tries to connect to a Wi-Fi network and, if such a connection is established, it uploads the newly acquired data to a remote server.
The creation of the Wi-Fi network is achieved using a mobile phone set to operate as personal hot-spot, while on the remote server resides the database storing all the performed measurements.

Software Architecture

Wi-Fi connectivity was one of the requirements for the system, but at the same time, the system itself should have not to produce unnecessary electromagnetic noise, possibly impacting the operating ability of the host's appliances.
To reduce the time in which the Wi-Fi connection was active, the Linux OS was set to activate the specific interface at predefined time instants in order to connect to the portable hot-spot.
Once connected to the network, the system performed the following tasks:

  1. synchronization of the system and RTC clock with a remote Network Time Protocol (NTP) server,
  2. synchronization of the local samples directory with the remote directory residing on the server.
    The latter task is performed using the UNIX rsync utility, which has to be installed on both the machines.

To gather data from the sensors, a Python program has been implemented, which runs continuously with a separate process reading from each physical sensor plugged to the board and writing on the MicroSD card.
It has to be noted that for what concerns the PM sensors, since the UART communication had to take place using GPIOs, a Pigpiod deamon has been leveraged, to create digital serial ports over the Pi's pins.

The directories on the remote server are a simple copy of the MicroSD cards mounted on the boards.
Data in these directories have been inserted in a MySQL database.

Mechanical Design and Hardware Components

In order to easily stack more than one device together, a 3D printed modular case has been designed.
Several enclosing frames can be tied together using nuts and bolts, with the use of a single cap on top.
Figure shows the 3D board design, together with the final sensor and board configurations.

Each platform is equipped with 4 PM sensors (a good trade-off between size and redundancy), 1 Temperature (T) and Relative Humidity (HT) sensor and 1 Pressure (P) sensor.
As our target was to capture significant data sampling for the particulate matter we adopt the following sensors:

  1. The Honeywell HPMA115S0-XXX as PM sensor.
    As one of our targets was to evaluate these sensors' suitability for air pollution monitoring applications, we insert 4 instances of this sensor in every single platform.
    This sort of redundancy allows us to detect strange phenomena and to avoid several kind of malfunctions, making more stable the overall system.

  2. The DHT22 as temperature and relative humidity sensor.
    This is very widespread in prototyping applications, with several open-source implementation of its library, publicly available on the internet.

  3. The Bosch BME280 as a pressure sensor.
    This is a cheap but precise barometric pressure and temperature sensor which comes pre-soldered on a small PCB for easy prototyping.

The system also includes a Real Time Clock (RTC) module for the operating system to retrieve the correct time after a sudden power loss. The chosen device is the DS3231.
The DS3231 communicates via I2C interface and has native support in the Linux kernel.

As a last comment, notice that a Printed Circuit Board (PCB) was designed to facilitate connections and soldering of the various sensors and other components.


Create database

The database structure can be created using the scripts located in the mysql_insertion folder of the Dataset/SQL_Table repository.

mysql -u <user> [-h <host>] [-p] < create_db.sql

Load SQL data (SQL Format)

Data formated in SQL can be loaded using the mysql command mysql -u username -p WEATHER_STATION < db_whole_data.sql, and the db_whole_data.sql is available in the SQL_data/ folder of the Dataset directory.

Load RAW data (CSV)

Data can be loaded using the python script sql_ins.py available in the mysql_insertion folder of the Dataset/SQL_Table repository.

python sql_ins.py <data_folder>

The script assumes the following folder structure:

* data_folder
|-- 01-board_table
|-- 02-unit_of_measure_table
|-- 03-param_type_table
|-- 04-board_config_table
|-- 05-physical_sensor_table
|-- 06-logical_sensor_table
|-- 07-board_sensor_connection_table
|-- 08-measure_table
|-- arpa
|-- mobility
|-- stations

Each folder contains a set of csv files. The script automatically loads data into the appropriate table and using the correct fields, which are specified as a list of parameters in the script. It is possible to edit the script to load only a subset of the folders.

System Usage

To replicate the experiments, the user should clone the raspberry pi image into a MicroSD (16-32 GB).
To do this, s/he can issue the command dd if=/path/to/image of=/path/of/microsd bs=4m on Linux.
The sampling scripts are run by a systemd unit automatically at system startup. The same systemd unit handles also the automatic respawn of the processes if some problems occur. The data are stored in the /home/alarm/ws/data directory, with filenames corresponding to the date of acquisition.

In order to upload these data to a database, it is possible to use the guide contained in the "database" directory.

In order to perform calibration and tests, it is recommended to take a look at the guide contained in the "analysis" directory. A Python class has been implemented to perform calibration of sensors against the ARPA reference ones. The resulting calibration can then be applied to a time window of choice.

3D Model

A 3D model of the case has been developed using SketchUp online software.
The resulting model is split in 5 different parts, each large enough to fit in our 3D printer (Makerbot Replicator 2X).
The model is stackable, meaning that several cases can be put on top of each other, with a single roof piece.

Printed Circuit Board

A PCB has been developed using KiCad software, so to create a hat for the RPi0 connecting all the sensors.

WS Analysis library documentation (v0.2)

The aim of this package is to provide fast and easy access and analysis to the Weather Station database. This package is located in the analysis directory, and it is compatible only with Python 3. Please follow the readme file for more information.

Directory Structure

├── 3D_Box
│   ├── Cap_v0_1stpart.skp
│   ├── Cap_v0_2dpart.skp
│   ├── ws_rpzero_noGPS_v1.skp
│   ├── ws_sensors_2d_half_v2.skp
│   └── ws_sensors_half_v2.skp
├── analysis
│   ├── arpa_station.json
│   ├── board.json
│   ├── example.py
│   ├── extract.py
│   ├── out.pdf
│   ├── requirements.txt
│   ├── ws_analysis
│   │   ├── __pycache__
│   │   │   └── ws_analysis.cpython-37.pyc
│   │   ├── rpt.txt
│   │   └── script_offset.py
│   ├── ws_analysis.md
│   ├── ws_analysis.pdf
│   ├── ws_analysis.py
│   └── ws_analysis.pyc
├── Dataset
│   ├── db_setup.html
│   ├── db_setup.md
│   ├── db_setup.pdf
│   ├── er_diagram.pdf
│   ├── mysql_insertion
│   │   ├── extract_to_file.py
│   │   ├── remove_duplicate.py
│   │   └── sql_ins.py
│   ├── SQL_Table
│   │   ├── create_db.sql
│   │   ├── create_measure_table.sql
│   │   └── load_data.sql
│   └── SQL_data
│      └── db_whole_data.sql.gz
├── PCB
│   └── WS_v2_output.tar.xz
├── readme.html
├── readme.md
├── readme.pdf
└── scripts
├── python
│   ├── csv
│   │   ├── arpa_retrieve.py
│   │   ├── filemerge.py
│   │   ├── gpx2geohash.py
│   │   ├── parse_csv.py
│   │   └── validation.py
│   └── mpu9250
│   └── gyro.py
└── README.md


This work presents a methodology of constructing three models respectively without blades, with straight blades and with curved blades, coupled for artificial simulated fog-haze environment with computational fluid dynamics (CFD), to predict the impact of the rotating blades on the flow velocities in the enclosed environment by simulation. Atmospheric flow characteristics and variation of flow velocities were analyzed, and the influences of different rotating blades on flow velocities were compared to get the related simulation results in three models. 


These videos show bottom and side views of the treatment by a He/O2 plasma jet of a solution containing ultra-pure water, potassium iodide and starch. As the plasma reaches the liquid, a purple filament characteristic of the triiodide ion/starch complex appears. This complex is formed by the reaction between iodide ion and a reactive oxygen and nitrogen species (RONS) such as ozone, hydrogen peroxide, hydroxide ion, nitric oxyde or nitrate. Therefore, the formation of these RONS in the liquid phase is temporally and spatially quantified.