Active noise cancellation system development for agricultural tractor cabs

Abstract

This work describes the implementation process of an Active Noise Cancellation system for agricultural tractor cabs. The most interesting feature of this implementation is the use of general-purpose hardware, without the need of expensive and hard-to-program Digital Signal Processing (DSP) devices. In particular, the reference signals, provided by accelerometers properly placed on noise-generating parts, and the error feedback signals are collected by means of a USB interface. All signal processing, aimed at primary path estimation and anti-noise audio signal generation, is performed using Simulink running on a commercial mini PC. The Exponential Sine Sweep (ESS) method is adopted for the measurement of the secondary path from the cancellation loudspeakers to the error microphones. An adaptive Filtered-x Least Mean Square (FxLMS) algorithm determines the anti-noise audio signal to be emitted. The system has been installed and tested on a commercial agricultural tractor cabin mounted over electromagnetic shakers to emulate realistic operating conditions. Two error microphones are attached to the headrest, close to the driver’s ears. The resulting ANC system relies on the use of very small buffers (for audio/accelerometric data), with a latency comparable to that of more complex and expensive specific DSP systems used in this kind of applications. In terms of acoustic performance, a significant reduction of annoying peaks in the 200- 500 Hz range and a broadband noise reduction at lower frequencies are observed, thus improving the overall sound quality experience. In conclusion, the implementation of an effective ANC system, employing common audio devices and a relatively simple Simulink program, was obtained. This paves the way to straightforward experimentation (in Matlab/Simulink) of new ANC processing algorithms, allowing direct testing of simulation-based solutions, without the need of porting them to proprietary DSP-based platforms.