MAPES GENÒMICS DEL CROMOSOMA

The first study of road noise cancellers is dated back in 1989, when a digital feedback controller was designed for reducing impact road noise [Costin and Elzinga (1989)]. The system was based on a headphone system mounted at the headrest of the driver. The feedback controller was designed with aproportional integral (PI) control theory that achieved 5-10 dB reduction between 20-60 Hz. Also a version of the controller with genelarised minimum variance theory was implemented that managed to cancel up to 20 dB in the same low frequency range. Another very interesting study in terms of control theory is found in [Brown(1995)], where a state space was used for developing a linear quadratic gaussian (LQG) optimal control for road noise cancellation with a single micro- phone mearusing the sound field in the cabin. Poor performance was reported in this study and the need for multiple input multiple output (MIMO) feedback design techniques was highlighted. As previously mentioned single channel feed- back controller for ARNC with fixed filter coefficients has only been developed by Honda Motors in a production line vehicle [Sano et al. (2000), Sano et al.

(2001)]. Yet, its effectiveness was limited to a single frequency of a road noise drum frequency at around 40 Hz. The main reason for the limited performance of feedback road noise controllers is the delay that is introduced by the physical

path between the loudspeaker and the microphone in the cabin that feeds back the sound signal to the controller.

A simulation study for road noise cancellation [Elliott and Sutton (1996)] showed that feedback systems are extremely sensitive to the acoustic plant delays, especially as the distance between the loudspeaker and the microphone increases. Adachi presented a more detailed design and analysis of feedback ANC [Adachi et al. (2001), Adachi (2003)]. The effect of the latency that is introduced by the acoustic plant is demonstrated under an ANC experiment with simulated road noise through loudspeakers in free field and varying distances between the control source and the cancellation position in the field.

In Elliott’s study was suggested that the maximum distance should not be more than 0.3 m [Elliott and Sutton (1996)], whereas Adachi proved within an experiment that 0.1 m is required in order to attenuate 10 dB(A) at 150 Hz. The successful demonstration of Adachi was also based on the use of an extremely low order system for modeling the acoustic plant of the free field thus a simple control filter. In practice however, as the distance from the loudspeaker to microphone increases in the vehicle’s cabin the complexity of the acoustic system between the transducer and the sensor increases. As a consequence, a high order control filter that compensates for the phase changes in the response of the physical system is usually required [Adachi(2003)]. As a solution to this problem, Cheer suggested a method that weights the control filtering according to the spatial distribution of the acoustic modes in the vehicle cabin [Cheer and Elliott (2012)]. However, the simulation findings were not particularly successful as the phase character- istics of the electro-acoustic path between the loudspeaker and microphone are always limiting factors for feedback strategies. As a further improvement of the suggested method, Cheer proposed an optimisation technique that calculates an

finite impulse response (FIR) filter matrix for a multichannel feedback ANC with the use of operational data [Cheer and Elliott(2013)]. A comparison study based on simulations of road noise was published recently demostrated that the MIMO feedback controller was effective from 80 Hz to 180 Hz, but with the trade-off of small enhancements at frequencies above the range [Cheer and Elliott (2014)]. As a continuation of Cheer’s work for multichannel feedback road noise control, the author performed a comparison between feedforward control with reference microphones inside the vehicle’s cabin and feedback control [Cheer and Elliott

(2015)], where good performance is obtained at the road rumble range, but above this band it has noticed that the coherence between the microphones is poor, due to other NVH attributes contributing to the sound field.

An interesting study in terms of hardware selection for feedback ANC sys- tems for vehicles was presented by Howard [Howard and Leclercq (2006)]. The

control system was based on a single channel feedback ANC implemented in a

field programmable analog array (FPAA) platform. The system was designed to cancel a very low frequency boom at 35 Hz, which created very high sound pressure levels in the cabin of a Holden Commodore station-wagon. Limited per- formance was reported as there were no attempts to optimise the loudspeaker and microphone placement.

Up until the point of writing this thesis, there have been no implementations of broadband feedback ANC for interior road noise in vehicles for the whole structureborne noise frequency range. Simulation results of new MIMO feedback ARNC design techniques have shown promising results up to 200 Hz [Cheer and Elliott (2013), Cheer and Elliott (2014)]. Recently, Honda demonstrated an interesting feedback technique of generating narrowband reference signals instead of using actual accelerometer sensors for removing some low frequencies road booms [Sakamoto and Inoue(2015)]. This method allowed Honda to use the active controller that is already integrated in their vehicles that is also capable of engine order cancellation.