618 



BELL SYSTEM TECHNICAL JOURNAL 



the filters, so that the passed frequencies were reproduced at constant 

 level at all times. F'ilters available were: high pass 30, 40, 55, 75, 100, 

 125, 250, 375, 500, 750, 1000, 1500 cycles cut-off frequency; low pass 

 13,000, 10,500 8500, 7000, 5500, 4500, 3750, 3250, 2850, 2450, 1900, 

 1500, 1000 and 750 cycles cut-off frequency. All were composite struc- 

 tures giving sharp cut-offs and attenuations of 60 db or more in the at- 

 tenuated region. Representative attenuation characteristics are shown 

 in Fig. 2. 



The loud speaker was mounted in one corner of a third room, of 

 dimensions 18' X 27' X 15', semi-sound proof in construction and 

 exhibiting reverberation characteristics similar to those of the micro- 

 phone room. To cover the required frequency range, two reproducing 



FREQUENCY IN CYCLES PER SECOND 



Fig. 2 — Attenuation characteristics of four typical filters. 



units were employed, one for the four and one-half-octave range below 

 500 cycles, the other for the five-octave range above this frequency. 



The degree of confidence to be placed in the test results depends 

 upon the uniformity with which this range was reproduced. The 

 average overall reproduction-ratio characteristic of the system, shown 

 in Fig. 3, departs from uniformity only about d=2.5 db between 20 and 

 15,000 cycles. It represents the average for that part of the room 

 which may be called the "listening area," the directional characteristics 

 of the loud speaker not permitting uniform sound pressure throughout 

 the room at very high frequencies. At no point in this area did the 

 measured pressure at any frequency depart more than ±3.5 db from 

 the average curve. One assumption is involved. Because the meas- 

 urements were made by supplying "warbling" frequencies to the volt- 



