620 BELL SYSTEM TECHNICAL JOURNAL 



Professional musicians were employed in all tests with musical in- 

 struments. Usually they were seated as close to the microphone as 

 practicable, and the amplifiers were set so that the sounds were repro- 

 duced at natural loudness. The power capacity of the system did not 

 permit this loudness on the drums, cymbals, piano, trumpet, and 

 trombone. For these cases the performers were seated about 10 feet 

 from the microphone and the amplifier gain was reduced the necessary 

 amount. Speakers were seated with their lips 18 inches from the 

 microphone. The keys were shaken about four feet away. Hand- 

 clapping and footsteps were produced at a distance of 15 feet. 



The musicians were instructed to play their instruments "loud," as 

 listening tests showed that the widest frequency ranges were thereby 

 produced. Tests were made with the instruments played in their 

 several octave ranges or with their different techniques to insure 

 "boundary" results. In general the performers played repeated three 

 or four note scales, for the differences produced by the boundary filters 

 were too small to be detected regularly except on repeated music not 

 supplied by melodies. However, such a procedure would not be repre- 

 sentative for the piano, and regular player rolls were used in testing it. 

 One repeated over and over a 15 second passage emphasizing the notes 

 of fundamentals 32 to 800 cycles, the second similarly emphasized notes 

 of fundamentals 200 to 3500 cycles, while the third was a march cover- 

 ing the range 40 to 1500 cycles. 



Before the regular crew started work on each sound the engineers in 

 charge listened to the reproduction and picked out the playing tech- 

 niques that promised the widest frequency ranges. Tests always 

 started with a filter giving 90-100 per cent correct judgments and con- 

 tinued through successive cut-offs until the 50 per cent cut-off was 

 reached. Throughout both the preliminary and regular tests the ob- 

 servers made notes relative to quality changes produced by the filters, 

 and noises produced by the instruments. With the performers and 

 observers in readiness for an actual test the procedure was as follows: 

 The filter operator threw the main switch from neutral to the position 

 lighting the "A" lamp, which might be "filter-in" or "filter-out" as he 

 chose. The performer, seeing his signal lamp light, then played his 

 instrument for a period of 15 to 20 seconds as the operator switched 

 "A-B-A-B-neutral." Switching to neutral stopped the musician by 

 extinguishing his signal light, and gave the observers an opportunity to 

 check on their recording blanks the condition they believed to be 

 filtered. The process was repeated five times with a random order of 

 correspondence between A or B and "filter-in." When necessary a 

 filter was retested until practice effects were eliminated. Since there 



