Different sampling frequencies have been recommended for different applications as digital audio technology is prevailing. An efficient sampling frequency converter has been required for a system which has to accomodate many units operating at different sampling frequencies. This paper describes the implementation of a practical sampling frequency converter using digital filters. In particular, the reduction of filter order is investigated for high fidelity transmission and practical hardware. Taking note that the order depends chiefly on transmission band width, the process of the order reduction is examined from a viewpoint of using the effective IIR filter and FIR filter combination. To obtain a good combination, a new 2nd order IIR filter is developed which has sharp cut-off response and approximately linear phase characteristic. The successful combination enables the filters to reduce its practical order(multiplication number)by1/3compared with conventional FIR filters. The converter incorporating the new filters is implemented by computer simulation. The computed results show that the converter is at satisfactory level of practical application, and has the advantages of high fidelity signal transmission and simplified circuits.
Digital filters have been applied to variable attenuation equalizers for tone-controlling of digitally-recorded music sources. This paper describes an implementation of a digital equalizer with low sensitivity coefficients. This equalizer uses versatile and simplified circuits for high quality signal processing. Several problems are examined from view points of the coefficient quantization effect and finite word length operation, which are inevitable in practical equalizers. Presented here is a new equalizer design procedure which incorporates the coefficient quantization algorithm for decreasing errors caused by finite-length words with small multiplication number. Actual models are also described and compared to conventional procedures. Two useful methods are proposed;one is for estimating the appropriate coefficient word length for the correct frequency response, the other is for estimating the operation length for maintaining high signal quality. Some examples of the estimation are also shown. It is found that the methods can be applied to general equalizer design. Computer simulations verify the usefulness of the methods and the feasibility of the equalizer implementation.
A new method of prediction for the noise propagation in a city area is proposed. In this method, buildings in a city area are assumed to be randomly distributed. Therefore, the characteristics of the noise propagation are easily given. By using the average height of the buildings and the perspective distance among them, the propagation losses are predicted without solving the wave-equation subject to the complex boundary conditions. In order to examine the results, the random distribution of the buildings is produced by the computer simulation, and the properties of the noise propagation are calculated. In addition, the scale model experiments are made on the distribution of the buildings. From the results of the computer simulation and the model experiments, the prediction method is proved to be appropriate.
This paper describes the experimental results of degenerate parametric amplification which are performed in free field. The amplitude and phase controlled sound source consists of twelve tweeters, which radiates a 10kHz powerful pump wave and a 5kHz weak signal wave synchronously and collinearly. SPLs of both waves are 142dB and 122dB re 20μPa at 1m, respectively. The pump level is not so high as shock is formed. The obtained data confirm that the amplification really exists and the parametric signal periodically varies in amplitude with the initial phase of the weak signal having a period of π. These results are in identical fashion with the plane progressive wave in a tube except for π/4 phase difference. Changes in beam pattern with the phase are also observed. We discuss the farfield behaviors of the parametric signal by using the modified Pestorius'algorithm.
In order to investigate sound image quality, white or band-limited noises with various degree of the cross-correlation coefficient are reproduced from two loud-speakers placed in anechoic or echoic chambers. Subjects are asked to make similarity judgments and some subjective evaluations of pairs of the noises. The experimental data are analyzed by Kruskal's multidimensional scaling(MDS)program. The analysis of the experimental data shows the followings:(1)sound image quality depends mostly on the width and distance of sound image, (2)the width of sound image depends on the absolute value of the cross-correlation coefficient, (3)the distance of sound image depends on the value itself of the cross-correlation coefficient, (4)with respect to physical and psychological factors governing sound image quality, there is no fundamental difference between anechoic and echoic chambers.