MULTICHANNEL MODULATION SYSTEM 25 



istic. To allow any compressor to be used with any expandor, all the silicon 

 elements are matched to a chosen standard unit, using selection and 

 resistance ''padding" methods. By such means, and by use of sufficient 

 loop gain in the feedback amplifier, very satisfactory overall linearity of the 

 system has been attained. 



At the output of the expandor the waveform of the multiplex signal is 

 essentially the same as that at the input of the compressor in the trans- 

 mitting terminal. The samples are distributed to their respective channel 

 destinations by a distributor Z>, resembling the collector described earlier. 

 The rotation rate is 8CC0 revolutions per second, but the duration of contact 

 on any one segment is only five microseconds instead of the possible full 

 twelfth of the 125-microsecond frame period. This effective narrowing of 

 the contact segments is done to allow the closure to occur well within the 

 interval in which the circuit is completed by switch C from the output of a 

 particular decoder. Each of the 12 segments of the distributor is provided 

 with a holding capacitor, which stores its allotted samples for the full 125- 

 microsecond frame period. The potential on any one of these capacitors 

 thus changes at 125-microsecond intervals from one quantized sample ampli- 

 tude to the next derived from the same original speech wave. 



This potential is of sufficient magnitude to require use of only a simple 

 single-stage triode amplifier for the output of each channel. Lengthening 

 the samples by holding, as described, helps to make this possible by causing 

 the amplifier to deliver useful power continuously instead of on a fractional 

 time basis. 



The only disadvantage of using lengthened pulses arises from an effect, 

 very similar to the "aperture effect" encountered in sound movies, which 

 introduces a curving slope across the audio gain-frequency characteristic of 

 the system. In the present case the gain drops about three decibels as the 

 frequency goes from the lowest to the highest value of interest. This slope 

 can be corrected by a simple equalizing network, as shown in Fig. 11. In 

 the present system the equalization is incorporated in the low-pass filter at 

 the input of each audio channel. This is preferable to equalizing at the 

 output, where power is at a premium. 



The outputs from the channel amplifiers are passed through 34C0-cycle 

 low-pass filters, identical with the input filters except for omission of the 

 equalization, and are delivered to standard voice-frequency circuits at the 

 same levels*' as are provided by a type /, K, or L carrier system. 



IV. Component Circuits 



Many of the circuit techniques used in the experimental system are con- 

 ventional, others^are more or less unfamiliar, and still others are believed to 

 be novel. In the following some of the more important building blocks are 



