相干平均法测扬声器频率响应的偏差分析

Bias Analysis of the Coherent Averaging Technique for Loudspeaker Response Measurement

实际工程中,常常需要在非消声室环境下准确测试扬声器在自由场响应.常用的非消声室测量方法有脉冲FFT技术,时延谱技术和最长序列快速哈德曼变换技术等,但其测量精度都不是很高,尤其是对于低频测量.George and Francosi(2003)[1]WESPACVIII上介绍了一种不使用消声室的扬声器自由场响应的测量技术,即相干平均法,其能够在普通房间里很准确地得到扬声器的自由场响应.其原理在普通房间不同位置测得扬声器到话筒的传递函数,由于每次直达声是不变的而反射声是不同的,因此把得到不同位置的频率响应求平均值就能得到接近在自由场测得的扬声器频率响应.事实上应用这种测量方法也存在一定的偏差,将探讨这种测量方法的偏差大小及其影响因素.通过对只有一个反射面的情况进行分析,得到了在不同频率时的偏差,并给出了测量的低频极限.结论是该方法在低频下,偏差会很大,而且随着频率增加偏差会变小.同时也对此平均法作了一点改进,从而在相同测量次数的情况下,能够有效减小其低频误差,从而更加准确得到扬声器的频率响应.

In practical applications such as on-line loudspeaker measurement, it is usually needed to measure the loudspeaker responses in an ordinary room. The most popular techniques include impulse FFT technique, time delay spectrometry technique （TDS） and the maximum-length sequence technique （MLS）. But the precision of all these techniques is not high enough. In addition, the frequency response of loudspeaker at low frequencies can not be obtained through these techniques. George and Francois（2003 ） introduced a method at the Eighth Western Pacific Acoustics Conference named coherent averaging technique which selects or rejects contributions to a measurement by choosing some transmission paths to be time invariant and making others-the ones we wish to remove from the measurement-vary. This technique is capable of making measurements of loudspeaker responses in ordinary rooms without the need for a time windowing approach-and hence avoiding the limitations of such an approach-based on coherent averaging of repeated measurements in different parts of the room. Measurement effort and complexity are increased but simple rotation of the transducers appears to be a satisfactory process-especially for medium and high frequencies. However, there exists bias by using this method. In this paper the bias at different frequencies is analyzed when only one reflecting surface exists. It can be seen that the bias at low frequencies is bigger than that at high frequencies. Also, a modified method is introduced in this paper to get the response of loudspeaker with higher precision.