Adaptive Noise Cancellation A Panacea M Evans R

Adaptive Noise Cancellation: A Panacea M. Evans, R. Adhikari

The Concept • 5000 Channels • 1 GW Channel • Signal = DARM - noise

Wiener Filter Block Toeplitz Input Signal (PEM) Covariance Matrix Norbert Wiener, MIT Cross Correlation Matrix

Wiener Filter Sim H 1 L 1 x 10 x 7 • • • H 1 & L 1 DARM (S 5) All Guralp & STS 2 Vault not used Requires ~9 GB RAM Levinson-Durbin

Adaptive Filtering / Neural Network Keenan Pepper Elena Gasparri LMS Algorithm (1 st chapter of textbooks)

ATNF

Filtered-X LMS Edwards, Thesis (declassified) Naval Research Postgraduate School

Applications • Seismic Noise (0 - 40 Hz) • Upconversion • Lock Acquisition • Bilinear (Crab shoulder) • Acoustic Noise (50 -900 Hz) • Magnetic (power lines) • H 1/H 2 correlation • MICH/PRC/SRC

From LMS to Filtered-X Noisy Signal • LMS Noise Witness FIR Adapt Cleaner Signal

From LMS to Filtered-X Total Control Signal • Injecting into a high gain servo loop Noise Witness FIR Adapt G Servo Control Signal

From LMS to Filtered-X • Separated Actuators Total Control Signal • Special case of “Filtered XLMS” Noise Witness FIR G A 2 A 1 C Adapt Servo Control Signal

At the 40 m • Test on input modecleaner Seismic Sensors MC Length FIR G MC 1 MC 2 C Adapt MC_L

At the 40 m

At the 40 m

And more !