A practical introduction to State of Polarisation (SoP) data analysis
Colab notebook: https://1url.cz/@IASPEI-SOP
With your Google Account you can play the notebook live, make your own copy, download it as *.ipynb and work in Jupyter locally.
For those without google accoun, a backup Jupyter Hub instance is avaiable at: https://1url.cz/@IASPEI-SOP-BACKUP, without local Jupyter, but a EDUGain identity is required.
Table of content
-
Theoretical Background & SoP Intro: Introduces State of Polarisation (SoP) in optical fibers, Stokes parameters, and the Poincaré sphere for visualization. It explains how external
events (like seismic waves) cause polarization transients.
-
Introduction to HDF5: Explains why the Hierarchical
Data Format (HDF5) is used for large sensor datasets, its
directory-like structure, and provides code
examples for creating and reading these files.
-
Time-series & Feature Engineering: Covers
fundamental techniques for processing sensor data, including lag
features, moving window statistics, Fourier Transforms (for frequency
domain analysis), and autocorrelation.
-
CESNET and SUBMERSE Data: This practical section
involves downloading real-world HDF5 datasets containing SoP and DAS
(Distributed Acoustic Sensing) records of various events like mechanical
knocks, heartbeat simulations, and rack door movements.
-
Practical Tasks - Finding Event Properties:
Demonstrates how to inspect specific signals, calculate event duration,
and compare intensities across different optical channels.
-
Autocorrelation on Dataset 3: Shows how to use
autocorrelation to identify periodic patterns (like a rack door opening
and closing repeatedly) and estimate event frequency.
-
Aligning SOP and DAS Data: A complex exercise in
synchronizing two different types of fiber sensing data to visualize a
real earthquake event recorded on a submarine cable near Svalbard.
Theoretical Background & SoP Intro: Introduces State of Polarisation (SoP) in optical fibers, Stokes parameters, and the Poincaré sphere for visualization. It explains how external
events (like seismic waves) cause polarization transients.
Introduction to HDF5: Explains why the Hierarchical
Data Format (HDF5) is used for large sensor datasets, its
directory-like structure, and provides code
examples for creating and reading these files.
Time-series & Feature Engineering: Covers
fundamental techniques for processing sensor data, including lag
features, moving window statistics, Fourier Transforms (for frequency
domain analysis), and autocorrelation.
CESNET and SUBMERSE Data: This practical section
involves downloading real-world HDF5 datasets containing SoP and DAS
(Distributed Acoustic Sensing) records of various events like mechanical
knocks, heartbeat simulations, and rack door movements.
Practical Tasks - Finding Event Properties:
Demonstrates how to inspect specific signals, calculate event duration,
and compare intensities across different optical channels.
Autocorrelation on Dataset 3: Shows how to use
autocorrelation to identify periodic patterns (like a rack door opening
and closing repeatedly) and estimate event frequency.
Aligning SOP and DAS Data: A complex exercise in synchronizing two different types of fiber sensing data to visualize a real earthquake event recorded on a submarine cable near Svalbard.