[Paper Review] Active control of scattered acoustic radiation: a real-time implementation

Emmanuel Friot, Régine Guillermin+1 more|arXiv (Cornell University)|Nov 18, 2005

Acoustic Wave Phenomena Research9 references18 citations

TL;DR

This paper presents a real-time active control system for reducing scattered acoustic noise from a 3D rigid parallelepiped using 14 loudspeakers and 18 microphones in an anechoic chamber. The method estimates scattered noise via linear filtering of total pressure measurements using pre-identified frequency response functions, then applies a modified Filtered-Error LMS algorithm to minimize scattered noise, achieving an average 10 dB reduction at error sensors—close to the optimal performance predicted by 3D Boundary Element Method simulations.

ABSTRACT

This paper presents an active noise control experiment designed to validate a real-time control strategy for reduction of the noise scattered from a three-dimensional body. The control algorithm relies on estimating the scattered noise by linear filtering of the total noise measured around the body; suitable filters are identified from off-line measurements. A modified Filtered-Error Least-Mean-Squares algorithm then leads to the adaptive filters which drive the secondary sources. The paper provides the numerical simulations using a Boundary Element Method which helped in designing a feasible experiment in an anechoic chamber with a limited number of control sources. Eventually a real-time pure-tone implementation with 14 ordinary loudspeakers and a large body is shown to yield on average a 10 dB reduction of the scattered noise at the error sensors, which is close to the optimum reduction predicted by the numerical simulations for the sensor arrangement.

Motivation & Objective

• To develop and validate a real-time active control strategy for reducing scattered acoustic noise from a 3D rigid body.
• To demonstrate that scattered noise can be controlled without prior knowledge of the incident wave or medium properties.
• To design a feasible experiment using off-the-shelf components (loudspeakers, microphones, DSP systems) with limited channel counts.
• To verify that the control strategy, based on linear filtering of total pressure, achieves performance close to theoretical optimum.
• To explore practical applications such as reducing wall reflections in anechoic chambers, despite challenges in scaling to high-channel-count scenarios.

Proposed method

• The control strategy estimates scattered noise by linearly filtering total acoustic pressure measured around the 3D body using pre-identified frequency response functions (FRFs).
• These FRFs map secondary source inputs to microphone outputs and are identified from in situ measurements with only the secondary sources active, embedding propagation and source dynamics into the controller design.
• A modified Filtered-Error Least-Mean-Squares (FELMS) algorithm adaptively updates filters driving the secondary sources based on the estimated scattered noise as error signals.
• Numerical simulations using a Boundary Element Method (BEM) were used to determine optimal sensor and actuator arrangements, confirming that 3 microphones per wavelength are needed for broadside scattering reduction.
• The real-time implementation used a multi-processor system (COMPARS) in a large anechoic chamber, with 16 outer microphones measuring total and scattered pressure for validation.
• Control was applied using an electric reference signal to avoid instability risks from acoustic feedback, especially since the scattered noise was only ~10% of the total noise level.

Experimental results

Research questions

• RQ1 Can a real-time active control system effectively reduce scattered acoustic noise from a 3D rigid body using ordinary loudspeakers and microphones?
• RQ2 To what extent can scattered noise be estimated and canceled using only total pressure measurements and pre-identified FRFs, without knowledge of the incident wave?
• RQ3 How does the number and arrangement of sensors and actuators affect the achievable noise reduction in a 3D configuration?
• RQ4 Can the control strategy achieve performance close to the theoretical optimum predicted by numerical simulations?
• RQ5 Is the method robust enough for practical applications such as reducing wall reflections in anechoic chambers?

Key findings

• An average 10 dB reduction in scattered noise was achieved at the 16 error microphones, which is close to the optimal reduction predicted by 3D Boundary Element Method simulations for the given sensor and actuator arrangement.
• The control system successfully reduced scattered noise by at least 5 dB at all error microphones, with reductions exceeding 20 dB at a few locations, indicating better performance where the scattered field was stronger.
• The total noise level at the error sensors was only slightly affected by control, confirming that the reduction was primarily in the scattered component, which constituted only about 10% of the total noise.
• The estimated error signal had a 30% error relative to the true scattered noise, but the control performance remained effective, suggesting robustness to estimation inaccuracies.
• The use of an electric reference signal prevented instability issues from acoustic feedback, and the results suggest that an acoustic reference could have been viable due to low feedback levels during control.
• The computational load was manageable—only twice that of a standard FELMS algorithm—indicating that broadband implementation is feasible in principle, though not pursued due to convergence challenges with random noise.