Sound quality evaluation of heavy-duty engines in free field conditions

Abstract: Annoyance response to vehicle noise is commonly reported. The impact reduction measures must therefore be measured scientifically. A scientific system for describing and defining annoyance is therefore required. There is furthermore a need to reduce the annoying components of the sound of vehicles. The engine of a vehicle is one of the predominant sources of vehicle noise which cause annoyance. The sounds of engines heard outside of engines at idling, running and accelerating conditions have been reported to cause annoyance throughout the world. This study of sound quality focuses on noise heard outside vehicles, termed external vehicle noise, and which is outside the interior of the car, in free field conditions. The objectives of the study are a) the development of new descriptors and annoyance prediction models at idling and running conditions, b) comparison of several rating methods, c) testing of several recording and reproduction techniques, d) testing of the long-term repeatability of annoyance judgement obtained in tests, and e) comparison of the combustion process of various fuels in terms of annoyance. An annoyance prediction model for idling engine sound was developed using multivariate analysis on the basis of a new parameter which is termed "Ear Reasoning Range". Two further psychoacoustic descriptors, loudness and impulsiveness were also defined in the model. The sound spectrum which caused least annoyance was obtained by subjective adjustments to an idling engine sound. The quality of the sound was validated by subjective judgements, by test persons, and by objective measurements. An annoyance index for heavy-duty engine noise was developed on the basis of three psychoacoustic descriptors: loudness, sharpness, and harmonic ratio. The model was validated internally and externally and gave good predictions of annoyance judgement of 6-cylinder inline engine noise. Three rating techniques known as the method of paired comparisons, the method of equal-appearing intervals and the method of successive intervals were investigated with respect to annoyance judgements of engine noise in free field conditions. The method of successive intervals was shown to be useful for annoyance judgements of engine noise for a large number of samples. The annoyance response to engine sounds recorded by an artificial head did not differ significantly from that recorded by an orthostereophonic technique. The variance in annoyance response for headphone reproduction was higher than that for loundspeaker reproduction. It was found that both recording techniques gave similar predictions of annoyance. Subjects preferred loudspeaker reproduction to headphone reproduction. There was no significant difference in annoyance judgements between experienced and naive listeners. Annoyance judgements of engine noise were shown to be consistent when subjects were asked to judge the same stimuli on two occasions 10-20 days apart. A comparison of multivariate analysis and neural networks for modeling annoyance of engine noise in free field conditions was investigated. Partial Least Squares (PLS) regression and Neural Networks were used to develop the annoyance prediction models on the basis of loudness, sharpness and harmonic ratio (rumble noise). PLS regression and Neural Networks gave similar prediction of annoyance for engine noise but neural networks gave greater prediction accuracy than did PLS regression. Annoyance judgements of engine sounds in respect to six different fuels were investigated by forty subjects. The annoyance judgements were found to be consistent with the annoyance index based on loudness, sharpness and harmonic ratio. Engine sounds due to ethanol fuel with 9% beraid were rated as least annoying whereas engine sounds due to a mixture of diesel and ethanol fuels were rated as most annoying.