The coefficient W can be updated as the spectrum pattern of the remaining error would be a flat shape against frequency, because the convolution in time domain was used. As a result, the noise which is at the peak frequency domain could be reduced. Accordingly, the noise at other frequency ranges could be reduced.
In order to make the desired remaining noise spectrum pattern, the filters were installed to both reference and error microphones. The comparison between the sound pressure level in rated power at 1m and the exhaust pipe outlet is shown in Fig.15. The solid line shows the sound pressure level when the speaker was off. The broken line shows the sound pressure level acted on by the active noise cancellation system without the signal filter. In that case, the reduction of exhaust noise at the low frequency range is larger than that at the high frequency range. Here, the Noise reduction level was only about 7 dB(A). The thick solid line shows the sound pressure level acted on by the AENC with the signal filter, which has larger signal reduction at the lower frequency range. By adding such a filter to both points of reference and error microphones, the noise reduction level was increased drastically. Finally, the exhaust noise was reduced more than 20dB(A) by the AENC. The spectrum of remaining noise could be controlled through the designed of an adequate signal filter.
5.2 Reference Signal
Many of AENC for practical usage adapted the rotation pulse for reference signals, or in other words, the disturbance signal5)〜7). The performance of noise reduction by the pulse method and microphone method was compared by elemental experiment. The comparison of the noise reduction level 30 seconds after the secondary speaker is turned on is shown in Fig.16. The noise reduction by microphone method was larger than that by pulse method. Also, the convergence of noise by microphone method was quicker than the other. As shown in Eq(1), (11), the coefficient of W would be updated by following the convolution of the reference signal. Therefore, by measuring the exhaust noise directly, the coefficient of W could be updated adequately and quickly. Also, even when the load of engine is changed at the same engine speed, the microphone could detect the change of original noise source. Hence, the microphone was adapted for the reference sensor. By installing the reference microphone at the upper stream of the passive silencer, the howling error could be prevented.
6. Experiment Result
6.1 Effect of Noise Reduction by AENC
The 4-cycle, water-cooled, turbo-charged, direct injection engine was used for the experiment. The rated power was 224kW/1800rpm.
