Millimeter wave QR based frequency selective surface (FSS)
In this work, a mmWave - QR based Frequency Selective Surface (FSS) for IoT Sensing Applications with a variety of configuration is proposed. The QR based FSS has a variety of resonant values and can be used as a filter due to very small band stop. Discovered the following FSS with unique S11 charac...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Theses and Dissertations |
Language: | English |
Published: |
2019
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/78718 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | In this work, a mmWave - QR based Frequency Selective Surface (FSS) for IoT Sensing Applications with a variety of configuration is proposed. The QR based FSS has a variety of resonant values and can be used as a filter due to very small band stop. Discovered the following FSS with unique S11 characteristics that can be used for various applications. M model - ports kept at 5 mm it is found that at 54.5GHz with a -13dB power drop as per time domain solver. Extended dipole model - ports kept at 10 mm it is found that at 49GHz with a -14dB power drop as per time domain solver. full QR code model - ports kept at 10 mm it is found that at 13GHz and 15GHz with a -22dB power drop as per time domain solver. QR code which is a link to the site: https://ntulearn.ntu.edu.sg/ - ports kept at 10 mm it is found that at 47.5GHz with a -20dB power drop as per time domain solver. It is proposed that these FSS can be suitably used for sensing and filtering applications in the unlicensed 57-63 GHZ spectrum and from 45GHz – 57GHz frequency band. The fabrication of the proposed structures was done to validate results.
Different settings of the receiver readings are noted and used to train the machine learning model to recognize the different responses that can be received for the same FSS. One such model Ensemble – Subspace discriminant has been trained to retrieve label with an accuracy of 84.2% for correct classification of data. This accuracy is observed for a data set of 5 models at different reader distance each with 1000 data points with 9 identified features. |
---|