246 BELL SYSTEM TECHNICAL JOURNAL 



properties of the waves must be preserved then, and how are these 

 properties preserved by various arrangements of 2- and 3-channel 

 loudspeaker reproducing systems? To answer these questions, some 

 very simple localization tests have been made with such systems. 

 Perhaps attention can be focused more easily on their important 

 properties by considering briefly the results of these tests. 



Localization Afforded by Multichannel Systems 



In Fig. 1 is shown a diagram of the experimental set-up that was 

 used. The microphones, designated as LM (left), CM (center), and 

 Rj\I (right), were set on a "pick-up" stage that was marked out on 

 the floor of an acoustically treated room. The loud speakers, desig- 

 nated as LS, CS, and RS, were placed in the front end of the auditorium 

 at the Bell Telephone Laboratories and were concealed from view by a 

 curtain of theatrical gauze. The average position of a group of twelve 

 observers is indicated by the cross in the rear center part of the 

 auditorium. 



The object of the tests was to determine how a caller's position on 

 the pick-up stage compared with his apparent position as judged by 

 the group of observers in the auditorium listening to the reproduced 

 speech. Words were uttered from some 15 positions on the pick-up 

 stage in random order. The 9 positions shown in Fig. 1 were always 

 included in the 15, the remaining positions being introduced to mini- 

 mize memory effects. The reproducing system was switched off while 

 the caller moved from one position to the other. 



In the first series of tests, the majority of the observers had no 

 previous experience with the set-up. They simply were given a sheet 

 of coordinate paper with a single line ruled on it to indicate the line 

 of the gauze curtain and asked to locate the apparent position of the 

 caller with respect to this line. Following these tests, the observers 

 were permitted to listen to speech from various announced positions 

 on the pick-up stage. This gave them some notion of the approximate 

 outline of what might be called the "virtual" stage. These tests then 

 were repeated. As there was no significant difference in results, the 

 data from both tests have been averaged and are shown in Fig. 1. 



The small diagram at the top of Fig. 1 shows the caller's positions 

 with respect to the microphone positions on the pick-up stage. The 

 corresponding average apparent positions when reproduced are shown 

 with respect to the curtain line and the loudspeaker positions. The 

 type of reproduction is indicated symbolically to the right of the 

 apparent position diagrams. 



With 3-channel reproduction there is a reasonably good corre- 



