An Analyzer for the Voice Frequency Range
Catherine Moore, Aaron Curtis · Bell System Technical Journal · 1927
[Editorial Note: The frequency analyzers described in this paper and in the paper immediately following, demonstrate in an unusual manner how a single fundamental principle may be employed to accomplish quite dissimilar results. The analyzers described in both papers employ a resonating element of fixed frequency and translate the wave components under study to this frequency by heterodyning them with the output of variable frequency oscillators. In the analyzer described in the first paper, the wave components under study are translated to a higher frequency while in that described in the second paper the translation is downward to a lower frequency. In view of these differences in design it is desirable to call particular attention to the reasons which have led to the working out of the two designs. The analyzer discussed by Moore and Curtis has been so designed as to sweep through the voice frequency range to as high as 5,000 cycles by the manipulation of a single control. To accomplish this end, it was found desirable to heterodyne upward by employing a variable frequency oscillator of considerably higher frequency than 5,000 cycles. The frequency of this oscillator can be varied continuously throughout the range from about 11,000 cycles to 16,000 cycles, and the fixed frequency resonating element is tuned to about 11,000 cycles. As translation of the wave under study to a higher frequency range reduces the percentage separation of the various components, it was necessary to choose a very sharply tuned resonating element. This takes the form of a steel rod which is loosely coupled magnetically to a driving circuit at one end and a registering circuit at the other. As the modulator used to accomplish the heterodyning process produces many frequencies other than the first of the “sum” and “difference” terms, it has been necessary to choose the frequency ranges such that all undesired frequencies which can not be made extremely small will be well removed from the single difference frequency under observation. The analyzer described in the paper by Landeen is capable of working over the range from about 3,000 cycles to 100,000 cycles. A requirement of this design was that very high resolution be obtained. To assist in accomplishing this end, the frequencies under study are translated downward in the frequency scale to the resonating element which consists of a circuit tuned to 800 cycles. This downward translation increases the percentage difference of frequency separation of the components under study. Because of the great range of frequencies covered by the analyzer it is not possible to have sum and difference terms other than those of the second order fall outside of the range of sensitivity of the resonator. The modulator has therefore been so designed as to preclude formation in the higher order terms. To increase its discrimination, the analyzer makes use of two tuned circuits and amplifiers arranged in tandem and placed before the modulator. The frequency to which these circuits are tuned must of course be variable and is set to coincide with the component under study.]