ROOM NOISES IN TELEPHONE LOCA TIONS 665 



greater in summer than in winter, the average difference being 4 or 5 db. 

 The spread in values obtained under summer conditions was less than 

 that found for the winter conditions. This was because the noises 

 which showed the least magnitude, when measured in winter, were 

 found to be higher under summer conditions, while the highest noises 

 measured failed to show an appreciable change with season. These 

 highest noises were largely caused by indoor machinery, and would not 

 be appreciably modified by outside sources. 



The average frequency composition of the noises measured under 

 both summer and winter conditions seemed to remain about the same 

 as far as could be determined. 



Selection of Typical Room Noise and Its Reproduction 



The data obtained have been used in determining the characteristics 

 of a typical room noise to be recorded on a phonograph record and re- 

 produced for use in laboratory tests. 



Since the data revealed no difference between the average frequency 

 composition of great and small noises, it has been possible to choose a 

 single recorded noise and to vary merely the amplitude of the repro- 

 duced noise, keeping its frequency makeup constant. 



The recording and reproduction of such a noise have presented pro- 

 blems, particularly from the point of view of naturalness. It has been 

 found difficult to reproduce a noise by simple means in such a way as to 

 give the illusion that the noise is real, not artificial. The requirements 

 for reproducing a noise which will be typical in its effect on the intelligi- 

 bility of speech transmitted over telephone circuit are, however, con- 

 siderably less severe than those for obtaining naturalness. Three main 

 factors seem to be involved in the problem. In the first place, room 

 noises often contain high-frequency components, undoubtedly includ- 

 ing some extremely high frequencies. These components, while they 

 are generally of low energy content, seem to contribute substantially 

 to the naturalness of the sounds. The effect of these components on 

 the intelligibility of speech transmitted over a telephone circuit would, 

 however, be much less than their contribution to the naturalness of the 

 noise, since the transmitted speech is generally limited to a band of not 

 more than 3000 cycles. The frequency band transmitted by the re- 

 cording and reproducing system was nearly twice this amount, being 

 limited both by the mechanical characteristics of the apparatus and by 

 the unavoidable noise generated in this apparatus, the amount of this 

 noise increasing as the band width increases. Second, room noises 

 emanate from a considerable number of sources located in different 

 positions, so that in order to reproduce them with complete fidelity 



