INTER-NOISE AUGUST 2007 ISTANBUL, TURKEY

Transcription

1 INTER-NOISE AUGUST 2007 ISTANBUL, TURKEY An evaluation method for single pass-by noise Sandro Guidati a, Sebastian Rossberg b HEAD acoustics GmbH Ebertstrasse 30a Herzogenrath GERMANY ABSTRACT The evaluation of vehicle noise is mostly done using the A-weighted sound pressure level, e. g. the L max for pass-by noise or the L EQ for community noise. Unfortunately, the db(a)- value does not reflect the annoyance effect of vehicle noise correctly. This was the motivation for investigations concerning additional methods for vehicle noise evaluation, using psychoacoustic parameters. Several pass-by situations were recorded with an artificial head. The recorded sounds were evaluated in listening tests. Therefore, the sounds have been arranged in groups and their annoyance was rated by test individuals on a 9- point scale. In parallel, main acoustic features of the pass-by noises were defined by expert hearing and interviewing the test individuals. Adequate psychoacoustic analyses were selected for the physical description of frequent named features. Different approaches to represent the subjective evaluations with a suitable combination of psychoacoustic descriptors were carried out. High correlations were gained with combinations of analyses where the Relative Approach Analysis (detecting noise patterns in the time and frequency domain) contributes the primary share, which indicates the importance of patterns for the evaluation of pass-by noise. The work was done in the framework of the EU project QCity and will be continued considering traffic flow noise. 1 INTRODUCTION The evaluation of vehicle exterior noise is mostly done using the A-weighted sound pressure level, e. g. the L max for vehicle pass-by noise or the L EQ for community noise. Unfortunately, the db(a)-value does not reflect the annoyance effect of vehicle noise correctly [1]. This was the motivation for investigations concerning additional methods for vehicle noise evaluation, using psychoacoustic parameters. The investigations were performed as part of the project Quiet City Transport (QCity) [2] which is an integrated EU project within the 6 th framework. The consortium is composed of 27 partners from 10 different European countries. The partners are consulting firms and research institutes working in the field of environmental acoustics and noise mapping, as well as manufacturers of road surfaces, tires, rails, mitigation measures etc. The main goal of QCity is to develop tools and measures to support the communities to fulfil the requirements of the EU Noise Directive 2002/49/EC [3]. The main focus lies on midterm actions that the communities could take on their own. For example passenger cars play a major role in inner-city noise generation. A reduction of their emission requires long-term research activities that only the manufacturers could conduct. But the communities could influence the composition of the a address: sandro.guidati@head-acoustics.de b address: sebastian.rossberg@head-acoustics.de

2 vehicle fleet in the city, for example by tax incentives for low-noise vehicles, traffic-flow regulation or establishing quiet zones. The implementation of these measures can be carried out within a shorter time frame and they will show effects relatively fast. Furthermore, they can reduce the noise propagation by applying noise barriers or low-noise road surfaces. An important contribution to inner-city noise is generated by rams, metros etc. Here the communities can directly influence noise generation by choosing low-noise trains and rails. Another main topic within QCity is calculating and further developing/enhancing noise maps. This includes, for example, algorithms for the automatic detection of noise hot-spots and a noise rating system to evaluate the impact of noise on the city dwellers based on the standard information generated by the noise maps. Noise maps are based on the L EQ. The spectral composition and time structure of the urban noise is almost not considered. These quantities are important for human noise perception [4]. Therefore another subproject within QCity is working on defining parameters based on psychoacoustic analyses and jury tests that permit evaluating the annoyance of city dwellers. In a first step, all individual noise events relevant to the typical urban noise are measured separately. The gathered time signals are then evaluated regarding human perception. In a subsequent phase, the single noise events are synthesized in a software tool to generate complex sound scenarios. With the use of traffic-flow simulations as input, the simulation tools allow auralizing the traffic flow and analyzing the resulting time signals regarding human perception. Therefore the effect of the traffic regulations and of the various noise mitigation measures, developed and evaluated within QCity, could be investigated. 2 LIBRARY OF VEHICLE NOISE The basis for the psychoacoustic evaluation and listening tests are sound recordings. Extensive measurements were performed on a large number of different vehicles. (16 passenger cars from mini to upper class, two vans, one SUV, two utilities). For each vehicle six typical inner city driving situations were recorded. constant 30 km/h constant 50 km/h medium acceleration from 30 km/h rolling with 50 km/h traffic light situation (deceleration-idle-acceleration) idle The exterior noise was recorded using an artificial head in 7.5 m distance and 1.2 m height (in compliance with ISO 362). Figure 1: Measurement of vehicle exterior noise

3 3 LISTENING TESTS The recorded sound files have been organised in groups and evaluated in listening tests by test persons regarding the perceived annoyance. The sets were organised with the following criteria: - one specific driving situation for different vehicles. Here the aim was the comparison of the different vehicles and the identification of parameters describing the specific driving situation. - one specific vehicle in different driving situations. Here the aim was the comparison of the driving situations and the identification of parameters describing the specific vehicle (or vehicle class). - various vehicles and driving conditions. Here the aim was to finally identify parameters valid for all vehicle (vehicle classes) and driving situations. A set used in a single listening test consisted of 4 to 8 pass-by noises. The test person was free to playback and to compare the sound files as often as wanted. The listening tests were performed in a quiet room with the use of headphones. The rating was done with a 9 point scale from 1 (not annoying) to 9 (highly annoying). The used software regroups the sound files automatically regarding the rating. Figure 2: SQuare-Software for listening tests A typical problem with listening tests is the expectation of the subjects. The sound of e. g. a sports car is highly attractive to some persons, because they associate the sound to the pleasure of driving a powerful car. Other subjects prefer a quiet sound and connect this sound to the experience of driving a high class vehicle resulting in negative ratings. Both evaluations are mainly based on the expectation for the interior noise while driving the car. The test persons will decide whether a sound is fulfilling his expectations ( good sound) or not ( bad sound). The aim of the investigation in QCity is the reduction of the annoyance of the city dwellers. For these people the best possible sound is total silence. Therefore, the vehicle noise is always perceived as disturbing. The goal of the engineer is to find ways to make the noise less annoying.

4 To get a coherent evaluation by all subjects it is very important to create a common context. The common context used for the listening tests was an everyday situation (person sitting at a bus stop, e. g. reading a book). After the test the subjects have been interviewed to gather information about the perception of the test itself and the parameters that influenced their individual judgements. 4 ANALYSES After the listening tests the most important acoustic features of the sound have been defined based on the interviews with the subjects and the opinion of acoustic experts. For the description of these features acoustic and psychoacoustic analyses have been selected. These analyses include the A-weighted sound pressure level (average and maximum) and the time variant loudness [5] (average, 1% (N 1 ), 5% (N 5 ) und 10% (N 10 ) percentile). Further parameters have been chosen to describe specific features. The most important feature is the engine noise, especially the typical pattern of Diesel engines. The Relative Approach Analysis (RA) has been proven in prior investigations to detect this specific noise feature very well. The Relative Approach Analysis simulates the ability of the human hearing to adapt to stationary sounds and to react on the other hand to variations and patterns in the time and frequency structure of a sound [6]. Further selected analyses are the sharpness (annoying high frequency share, e. g. from tires), tonality, booming (low frequency contributions) and sudden loudness increase. 5 STATISTICAL ANALYSIS & DEVELOPMENT OF METRIC The subjective evaluation of each set of sound files has been analysed using Principal Component Analysis (PCA). This technique allows for the detection of main common factors on which the judgement of the test persons is based. The main two or three factors have been considered for each sound set. There was always one dominant factor with over 60% variance explanation that has correlated very well with the RA. One could conclude that the engine noise has the strongest influence on the evaluation, although it might not be dominating the overall level. For example the driving situation constant 50 km/h has been judged up to three categories more annoying than the situation rolling with 50 km/h although the sound pressure levels show only small differences. High correlation with the first or second factor was also found for the parameters N 5 and sharpness. Using linear regression and optimisation the following metric has been found: Metric ~ RA N S 5. The metric has an averaged correlation of r = 0.91 with the subjective evaluations (figure 3), whereas the standard A-weighted maximum sound pressure level (according to ISO 362) has a correlation of r = 0.54 (figure 4). These results have been verified using new sound recordings. The annoyance effect of idle sounds cannot be represented by this metric due to the fact that evaluations of pass-by sounds and idle sounds cannot be compared. In idle the car is standing motionless in front of the test person. This is a completely different situation compared to a pass-by, where the sound is rapidly changing and vanishes after short time. Therefore a different context is leading to different criteria for the evaluation.

5 Figure 3: Correlation between subjective evaluation and metric 9 8 deviations r sd max > 1 SPL max % calculated SPL max - index subjective evaluations Figure 4: Correlation between subjective evaluation and the Lmax,A 6 TRAFFIC NOISE SCENARIOS The metric has been developed and validated for the psychoacoustic evaluation of pass-by noises of a single vehicle. Real traffic is composed of numerous vehicles. The resulting noise

6 is therefore a synthesis of single pass-by events. As part of the QCity project traffic control measures are investigated. Complete traffic scenarios are calculated and compared by their impact on the noise maps. Again, the standard L EQ used for the noise map does not reflect sufficiently the annoyance of the traffic noise. Furthermore, for real traffic noises there are additional effects like masking that are not relevant for the single pass-by event. Therefore the developed metric must be revalidated in listening tests and if necessary adapted. The sound files used in the listening test must be chosen appropriately to cover the range of interesting effects. Within QCity the effect of traffic control measures and different tire/road noises has to be investigated. Therefore recordings of existing traffic situations (urban streets) are not sufficient. By combining existing technologies (Exterior Noise Synthesis, Sound Simulation Technology) vehicle sounds for arbitrary running conditions could be generated. Finally, by overlapping these single events arbitrary traffic scenarios are generated and auralized. The technology is described in the following. 7 EXTERIOR NOISE SYNTHESIS The sound of a vehicle is the sum of various sources. For the standard passenger car typical sources are: engine, intake, exhaust system, tires, gear box. To investigate the relevance of these single sources and to quantify their contribution to the overall sound a technology called Transfer Path Analysis (TPA) has been developed, first for the vehicle interior noise and in the last years also for the exterior noise. The contribution of a single source (component) depends on the radiated noise and the transfer path between the source and the interesting reference position (vehicle interior or exterior). Both quantities have to be measured. The vehicle is instrumented with sensors in the near field of each interesting source. The radiated noise is measured for the interesting running condition. The transfer path between the sources and the reference position is measured using a source with a known radiation characteristic. By combining the source signals with the respective transfer path the source contribution to the overall noise can be calculated. The summation of the contributions of all sources corresponds to a real measurement at the reference position. With the use of binaural transfer paths the standard TPA is extended to the Binaural Transfer Path Analysis (BTPA). Binaural time signals are absolutely necessary for listening test, since monaural signals do not correspond to the everyday experience of the test persons. This is even more important for the exterior noise. Here, the movement of the vehicle effect (speed, Dopplereffect) are important for the perception. For the generation of exterior noise signals, comparable to a standard pass-by measurement, the synthesised source signals are modified according to the instantaneous position and speed of the vehicle with respect to the reference position. 8 SOUND SIMULATION TECHNOLOGY The synthesis technology described above is limited to the analysis of measured running conditions. For the purpose of auralisation of arbitrary traffic flow scenarios arbitrary running conditions have to be available for a large number of vehicles. Furthermore the influence of different road surfaces and tires has to be investigated. To create this variety of sounds the Exterior Noise Synthesis is combined with the Sound Simulation Technology. Again, near field source signals and transfer paths are measured, but now for well defined running conditions: Slow run ups with various loads measured on a chassis dynamometer are used to describe the power train noise and coast downs on a street to describe the tire/road noise. Additionally the engine and tire RPM is recorded. The measurements are dynamically resampled according to the current RPM and divided in time slices for single RPM steps. The

7 time slices are analyzed regarding engine orders and incoherent noise. Both parts are modeled with filters (figure 5). Figure 5: Calculation of filters for the modelling of near field source signals For the generation of a sound file corresponding to an arbitrary running condition only the corresponding engine and wheel RPM and throttle time signals are needed. The power train and tire orders are synthesized using sweep signals. The incoherent noise is synthesized using white noise. Both signals are filtered according to the current RPM/throttle position. The sum of both signals corresponds to a near field recording of the corresponding source (figure 6). Figure 6: Resynthesis of source signals based on arbitrary speed/rpm/ and throttle time signals The synthesized source signals are filtered with the transfer paths and summated. The result corresponds to an artificial head recording of an arbitrary pass-by event. The events are superposed generating the interesting traffic noise auralisation, which can be used for the listening tests. Figure 7 shows an example for the Sound Simulation Technique. The left diagram shows the near field recording of a run up of an intake. The next two diagrams show the extracted order and noise signals. The last diagram shows the resynthesized signal.

8 Figure 7: Example for the resynthesis of an exhaust near field recording 9 CONCLUSION AND OUTLOOK As part of the European Research Project QCity a metric has been developed for the psychoacoustic evaluation of vehicle exterior noise. The combination of the Relative Approach Analysis with Loudness and Sharpness shows a significantly higher correlation with the results of the listening tests than the standard db(a). So far the metric has only been validated for single pass-by events. The main focus in QCity lies on the evaluation of complex traffic scenarios which consist in numerous single events. The adaptation and modification of the metric will be done with further listening tests. For these listening tests appropriate sound files are needed. It is not convenient to measure all interesting driving conditions for all types of vehicles. A technology has been presented that allows for the generation of exterior noise signals that correspond to arbitrary driving conditions. Apart of the listening tests this technology will be used to auralise the complex traffic scenarios generated and analysed within QCity. With this the decision makers get the opportunity to listen to the effects of traffic control and mitigation measures and to calculate the resulting annoyance metric. 10 ACKNOWLEDGEMENT The work has been supported by the European Commission under the Contract Number TIPA-CT REFERENCES [1] W. Probst, Noise mapping and action planning the European approach to reduce environmental noise. Proceedings of Inter-Noise 2006, Honolulu, Hawaii, USA [2] EU Project Quiet City Transport (QCITY), TIP4-CT ( [3] EU Directive: Directive 2002/49/EC of the European parliament and the Council of Ministers, 25 June 2002, relating to the assessment and management of environmental noise. Official Journal of the European Communities L 189/12 ( ) [4] B. Schulte-Fortkamp, B. M. Brooks, W. R. Bray: Soundscape - An approach to rely on human perception and expertise in the post-modern community noise era. Acoustics Today Volume 3, Issue 1 (January 2007) [5] current draft of the DIN 45631

9 [6] K. Genuit, Objective evaluation of acoustic-quality based on a Relative Approach, Inter-Noise 1996, 25th Anniversary Congress Liverpool, Conference Proceedings, Liverpool, UK [7] current draft of the DIN 45692