Skip to main content
Terrence Tak Lun Tsang

Dr Terrence Tak Lun Tsang

Research Associate

School of Physics and Astronomy

Overview

I am an experimentalist. My work focuses on optimizing seismic isolation systems in gravitational-wave detectors. Gravitational-wave detectors, such as LIGO, Virgo, and KAGRA, are ground-based kilometer-scale interferometers. The main optics of the interferometer are susceptible to external disturbances such as ground motion. Seismic noise at low frequencies (<10 Hz) can cause these optics to move excessively, causing the optics to misalign. Therefore, seismic noise must be mitigated for the gravitational-wave detector to function.

Seismic isolation systems in gravitational-wave detectors carry the optics and are used to reduce the motion of the main optics. They are equipped with sensors and actuators in feedback or feedforward configurations to achieve active damping or active isolation. However, using these active components can inject control noise at higher frequencies where the gravitational waves are detected. Therefore, the proper design of the control filters is required to achieve an optimal trade-off between seismic noise rejection and control noise injection.

During my PhD studies, I worked with the KAGRA vibration isolation subgroup and commissioned some of the KAGRA suspensions by implementing active control systems. Also, I developed a method to optimize several seismic isolation control filters, such as feedback controllers, complementary filters, and sensor correction filters, using H-infinity synthesis. I am currently stationed at LIGO Livingston and I will be developing new control methods that can improve seismic isolation performances in gravitational-wave detectors.

Publication

2022

Articles

Research

H-infinity optimization for seismic isolation

I am interested in optimizing active vibration isolation systems in gravitational-wave detectors. I am interested in control theory and I have developed a method to optimize control filters, including feedback controllers, complementary filters, and sensor correction filters, using H-infinity synthesis. Using H-infinity methods, one can specify frequency-dependent upper bounds for signals in a control system. With a clever selection of specifications, the optimized filter suppressed the signal of interest (displacement, sensor noises, etc) close to the lower bound at all frequencies. As opposed to manual designs, H-infinity methods allow us to design control filters according to the system characteristics and specifications themselves. Necessary features in the control filters, such as notches, roll-offs, and peaks, are naturally generated as a result of optimization. The synthesized control filters are optimized rather than tuned manually. This would help further improve the seismic isolation performance in gravitational-wave detectors.

Making seismic isolation systems adaptive

Unlike other detector noises, seismic noise is not static and optimal control laws can change from time to time. With the H-infinity method, we are only able to design the optimal control laws such that the systems are optimal on average. To further improve the seismic isolation performance, the system must react to seismic activities in real time. The lack of adaptivity in gravitational-wave detectors owes to the static filtering control architecture. Switching control filters in real-time is possible but this could lead to uncertainties due to unforeseen transient responses. There are several modern control methods that might solve the problem. For example, model predictive control takes sensor signals and optimizes the required control signals by foreseeing the future using models of the systems. Unlike H-infinity methods, these methods do not complement the static control architecture used in current gravitational-wave detectors. Therefore, these methods may not be implementable currently. I will be working on simulations and developing the necessary software for implementing new methods.

Kontrol

I have used Python throughout my career and I have written a Python package called Kontrol. The package is named after the Japanese detector KAGRA and control. Before I do anything to the real detector, I code and document the methods into Kontrol and use Kontrol to obtain the results for implementation. I did this with the hope that all of my results are reproducible and my methods would be available to everyone.

Biography

Employment

  • 2022-present: Research Associate, Cardiff University.

Education

  • 2018-2022: PhD in Physics, The Chinese University of Hong Kong.
  • 2014-2017: BEng in Mechanical Engineering, The University of Hong Kong.

Professional memberships

  • LIGO Scientific Collaboration (2022-present)
  • KAGRA Collaboration (2018-2022)