The lowest resonance of a cone-type dynamic loudspeaker is one of the most important factors for the quality of sound radiated through it. A low speed analogue computer was used to make a real time simulation of a mechanical vibrating circuit of a loudspeaker, and the discriminal limens of sounds on changing the lowest resonant frequency f_0 and its sharpness Q_0 were obtained. The discriminal limens of the lowest resonant frequency below about 120c/s〜150c/s differe considerably for different source materials, while those over that frequency differ only a little in materials. The discriminal limens of f_0 below about 150c/s shown in level differences of the sound near f_0 are similar to those of Q_0, but differ very much above that frequency. From the results, a mechanism of the discrimination of f_0 below about 120c/s〜150c/s correspond to that of the detection of level difference in a certain band near f_0, and at higher f_0 it corresponds to the mechanism of the detection of frequency difference. It is shown that the reduction in feeling of bass tone with a rise of the lowest resonant frequency can be compensated for by an increase in Q_0. It interestingly corresponds to the hypothesis by Zwicker that the detection of level change is based on the same mechanism as for the detection of frequency change. The discriminal limen became smaller on reducing the higher cut-off frequency. From the results, the ratio of the sound power of the changed part to the over-all power is considered to be significant.
An Automatic Talking Apparatus using pre-recorded speech segments and its application to telephone numbers speech is treated from the point of view of realization of a practical model using a magnetic drum recorder. This research model, consisting of 20 channels on magnetic drum and an editing apparatus, enabled us to handle the separation and copulation of prerecorded speech segment. As the first step of the experiments, the quality grades of number speech of a simple case when a single segment is allocated to a digit, were measured on the "Roting Scale". This experiment showed that segments, of a cap-type pitch pattern which were discreetly uttered, sounded best and that the optimum gap between segments is about 50 milliseconds. As the next stage, we used more kinds of segments for each digit and measured the quality. As a result, the more kinds of segments we used, the better the quality of speech could be obtained. Three segments were sufficient for near natural utterance. And pitch movements were found to play a fairly important role in this system.
Many of the close-talking microphones studied, are heavy and of large size and are usually held in the hand. Therefore the sound quality and the output level are unsteady because it is difficult to hold the microphone in the hand at a constant distance from the mouth. This paper describes the design method for the close-talking microphone for broadcasting use without the above defects. This close-talking microphone is the pressure gradient microphone of the moving coil type. In order to keep the microphone at a constant distance from the mouth, the microphone is attached on the cap. Two acoustic pipes that are lengthened to be close to the mouth are attached to both sides of the diaphragm. The relationship between the equivalent network elements of this microphone and the desired frequency response, sensitivity and noise canceling effect, were analized. Furthermore, a concrete method for the application of these values of the equivalent network elements to the membrane or acoustic circuit was considered. The characteristics of the trial products were as follows. 1. Axial sensitivity at 2cm distance from a simple sound source -86. 5db(1, 000c/s) 2. Axial response at 2cm distance from a simple sound source 50〜15, 000c/s ±5db 3. Directional response pattern bidirectional 4. Output impedance 600Ω(1, 000c/s) 5. Dimensions Microphone unit about 30φ×25mm^3 acoustic pipe length 170mm 6. Weight(without the cap) 100gr.