256 BELL SYSTEM TECHNICAL JOURNAL 



For observing positions near the center of the auditorium the 

 observed angles agreed reasonably well with calculations based only 

 upon loudness differences. As the observer moved to one side, how- 

 ever, the virtual source shifted more rapidly toward the nearer loud 

 speaker than was predicted by the computations. This was true of 

 reproduction in the auditorium, both empty and with damping simu- 

 lating an audience, and outdoors on the roof. Computations and 

 experiment also show a change in apparent angle as the observer 

 moves from front to rear, but its magnitude is smaller than the error 

 of an individual localization observation. Consequently, observers in 

 different parts of the auditorium localize given points on the pick-up 

 stage at different virtual angles. 



Because the levels at the three microphones are not independent, 

 and because the desired contours depend upon the effects at the ears, 

 a 3-channel stage is not as simple to lay out as a 2-channel stage. For 

 a given observing position, however, a set of contour lines can be cal- 

 culated. The dashed lines at the right of Fig. 6 show four contours 

 thus calculated for the circuit condition of Fig. 1 and the observing 

 position previously mentioned. The addition of the center channel 

 reduces the virtual angle for any given position on the pick-up stage 

 by reducing the resultant loudness difference at the ears. Although 

 the 3-channel contours approach the 2-channel contours in shape at the 

 back of the stage, a given contour results in a greater virtual angle for 

 2- than for 3-channel reproduction. 



Similar effects were obtained experimentally. As in 2-channel 

 reproduction, movements of the caller could be simulated by manipu- 

 lation of the channel gains. From an observing standpoint the 3- 

 channel system was found to have an important advantage over the 

 2-channel system in that the shift of the virtual position for side 

 observing positions was smaller. 



Effects of Quality 



If the quality from the various loud speakers differs, the quality of 

 sound is important to localization. When the 2-channel microphones 

 were so arranged that one picked up direct sound and reverberation 

 while the other picked up mostly reverberation, the virtual source was 

 localized exactly in the "direct" loud speaker until the power from 

 the "reverberant" loud speaker was from 8 to 10 db greater. In gen- 

 eral, localization tends toward the channel giving most natural or 

 "closeup" reproduction, and this effect can be used to aid the loud- 

 ness differences in producing angular localization. 



