Figure 2 - Spreading function for the masking effect

Transcription

1 Modified method to quantify diesel noise using masking effects Dongun YUN 1 ; Sang-Kwon LEE 2 ; Insoo JUNG 3 1 Inha University, Republic of Korea 2 Inha University, Republic of Korea 3 Hyundai motor group, Republic of Korea ABSTRACT Recently, a diesel engine becomes popular since it has great fuel consumption compared to a gasoline engine. However, a diesel engine generates diesel knocking noise during operating because of its high combustion force. The sound radiated from a diesel engine is high pressure sound and noisy. Knocking noise is not high level of sound pressure but annoying noise. Its characteristic is the amplitude and frequency modulated sound. In order to evaluate knocking sound a new index is required. Many of researches has been done for the development of knocking index. In this paper, a new knocking index is developed by considering time and frequency masking. This index is applied to the estimation of the perceived quality of knocking sound measured in anechoic chamber. For the not anechoic chamber condition, the vibration data correlated knocking sound is also used for quantification of the knocking sound. Keywords: diesel, knocking, combustion noise, temporal masking, frequency masking 1. INTRODUCTION Diesel knocking sound is one of major annoying sounds inside of car and influences on the consideration of purchase of the customers. This sound has been controlled by adjusting the inject time to control the cylinder pressure. This method yields the reduction of fuel efficiency. Therefore a new method is required to control the knocking sound. That is masking. Masking sound is developed in low frequency relatively. The sound with low frequency band affects that with high frequency band. Therefore a new method should be considered for the frequency masking [1]. Furthermore, the perception of diesel knocking sound at high speed is not sensitive because of the modulation frequency is changed and increased. This effect is temporal masking [2]. The conventional method [3] qualifying the diesel knocking sound doesn t consider both of frequency masking and temporal masking. In this paper, a new diesel knocking sound index is developed for the evaluation of diesel knocking sound objectively. This index is developed based on the frequency masking and time masking 2. DIESEL KNOCKING INDEX 2.1 Frequency masking In daily life, the masking effect very often occurs. For example, when you talk with someone in the café, if café is quiet, you can dialogize with small speech. However, if there are many people in the café, you can t dialogize with small speech. For overcoming this situation, you should speak louder. This is a classic example of simultaneous masking. The simultaneous masking is quantified by defining masking thresholds. These masking threshold are needed to determine whether a test signal is audible or not. Masked threshold always exist on the threshold which is in quiet situations. If frequencies of masker is different with frequencies of test signal, masked threshold is same with threshold which is in quiet situations 1 bentotoda@naver.com 2 sangkwon@inha.ac.kr 3 isjung@hyundai.com 5377

2 Figure 1 - I khz masker pattern of narrow band in frequency masking As mentioned, simultaneous masking means frequency masking [1]. Fig. 1 shows the masking patterns of a narrow band which has 1 khz center frequency and band width of 160Hz. The lowest curve presents the threshold of quite zone. Other curves mean masking patterns followed by level of noise of narrow band. For example, for a masking test signal, if a test signal of 2 khz has 40dB level, masker of 1 khz should have at least 80dB. The maskers at low frequency band have symmetric shape but those at high frequency have non symmetric shape above 40dB. This effect is non-linear upward spread of masking and depends on the level of center frequency. Fig. 2 shows the spreading function corresponding to the center frequencies of every bark band. Therefore, new thresholds are determined by considering this effect. When a test signal doesn t excess thresholds, the sound of the test signal is not audible. If the level at low frequency bark band is higher than that at the adjacent upper bark band, the sound in adjacent bark band is masked by the sound at low frequency bark. In the practical situation, there are many spreading function models. This paper uses model of Terhardt spreading function (2). A new diesel sound index is developed using this spreading function. Figure 2 - Spreading function for the masking effect 2.2 Temporal masking In case of the level of masker increasing gradually, there are transient parts. Test signals are completely masked later, but, at first not. It means that incomplete parts of masking effect exist. The complete part of masking effect makes the value of loudness of test signals decrease, but not completely. This effect is figured out in Fig. 3.This is temporal masking [1]. Temporal masking constitutes forward masking and backward masking. Forward masking is masking maskee behind masker in the time domain. This effect is valid for 200ms and it steeply decreases. Backward masking is masking maskee in front of masker in the time domain. This effect is valid for 50ms and it also steeply decreases. 5378

3 Figure 3 - Curve of temporal masking for the sound 3. OBJECTIVE EVALUATION WITH DIESEL KNOCKING INDEX The interior sound of a diesel car is recorded and its diesel knocking sound is evaluated objectively by using the diesel knocking index which is newly developed in the paper by considering the effect of time-frequency masking mentioned in session 2. In order to calculate the index output, we recorded the interior sound at two acceleration conditions and three steady state speeds. The loud knocking sound occurs at two pedal conditions of LTI and MTI during acceleration. LTI means driving conditions of light tip and MTI means middle tip condition during acceleration. In addition these, the three steady state speed of 1250rpm, 1500rpm and 1750rpm are used as a test condition. For the acceleration condition, the waveform of every 2 second are used for the calculation of diesel knocking sound using the new index. The sound data of 2 second is also used for three steady state conditions. According to these results the output of index is low at high speed because of the effect of temporal masking as we expected. Fig.4 shows the output of index versus engine speed.. Figure 4 - Index output for the diesel knocking sound index 4. SUBJECTIVE EVALUATION WITH JURY TEST In this session, the subject evaluation is proceed for the validation of the index results. For the subjective evaluation, 30 persons are participated. They are twenty or thirty years old. Before jury test, their ear condition are checked. The data used for the jury test are the same data used for the objective test which is index. The level of subjective rating is ranged from 0 to

4 Figure 5 - Subject rating for diesel knocking sound Reference signals are given to jury with the value of subjective rating of 0.8 point and 4.2 point. The maximum value of subjective ration is 5 and is corresponding to the knocking sound of commercial cars such as trucks. Fig.5 shows the results of subjective evaluation. According to these results, the value of subjective rating in steady state condition of 1250 rpm is similar to that of LTI condition. In case of 1500rpm and 1750rpm, the value of subjective rating is similar to these of MTI condition. Correlation between objective results and subjective results are calculated by using index output and subjective rating of jury test. Numerically, correlation coefficient is It is quiet high score, they have a high correlation and index is well validated. This index is can be used for the objective evaluation of diesel knocking sound. 5. OBJECTIVE EVALUATION USING VIBRATION DATA In practical situation the sound data is not available. In this case, we cannot obtain the sound data from anechoic chamber but can measure the vibration data at the cylinder block. Figure 6 Objective rate using vibration data 5380

5 The vibration data should be correlated to the sound date. Fig. 6 shows the index output which is calculated using vibration data which are measured on three different point of the cylinder block. The points is 3 positions. Vibration sensors are tri-axis accelerometers. We obtained 9 vibration data. The vehicle driving condition used for the acquisition of these vibration data is the LTI condition and MTI condition. We calculate the correlation between index output of these 9 data and that of the sound data. Table 1 listed the correlation coefficient form this test. According to table 1. Vibration data 2 has the highest correlation coefficient with sound data. The vibration 2 data is measured in y axis direction at the point A. Table 1 Correlation coefficient Vibration data 1 Vibration data 2 Vibration data Vibration data 4 Vibration data 5 Vibration data Vibration data 7 Vibration data 7 Vibration data CONCLUSIONS This study develops a new diesel knocking index. Because it is hard for conventional diesel knocking index to evaluate diesel knocking noise in condition of acceleration. This deadlock makes for engineer hard to estimate knocking sound. The diesel knocking index overcomes this shortcoming by employing time-frequency masking. This can be also used for the real car condition. ACKNOWLEDGEMENTS This work was supported by Mid-career Researcher Program through NRF of Korea grant funded by the MEST (No. 2015R1A2A1A ),". REFERENCES 1. E. Zwicker, H. Fastl. Psychoacoustics Facts and Models. Berlin,Germany: Springer-Verlag; Marina Bosi, Richard E. Goldberg. Introduction to digital audio coding and standards. Dordecht, Netherlands: Kluwer academic publishers; Markus Bodden, Ralf Heinrichs. Diesel Sound Quality analysis and evaluation. ForumAcusticum,