Radio broadcast .. (1922-30)

How and Why It Is Better LINEAR POWER DETECTION imiiiiiiiiiiiiimiih By FREDERICK EMMONS TERMAN Stanford University f | ~\fio YEARS ago the typical radio set I employed a grid-leak grid-condenser • type of detector that was intended to operate with a radio-frequency input volt- age of 0.1 volt or less. In contrast with this, the new receivers of the 1929-1930 season practically all use a power detector, in which the radio-frequency signal applied to the detector has a magnitude of at least several volts. In advertising the manu- facturer makes this a big talking point. In some cases it is emphasized that a linear power detector is used. One may wonder whether this sudden change in detector practice is a real improvement, or whether it is another fad, such as the toroidal coils of several seasons ago, that is being used as a sales point, and will soon disappear. The answer to these questions is clear and decisive. Compared with the weak- signal detector of the past, the power detector introduces less distortion, is more efficient, has a tendency to reduce static, and also increases the selectivity. Any one of these features would be very desirable, but when the change from weak-signal to power detection gives benefits in all of these ways, there is very little question as to what is the best practice. Linear detectors and power detectors are very closely related in practice, since they usually go together. In the usual meaning of the term, "power" is used in connection with detection when the radio- frequency voltage applied to the detector is one or more volts, in contrast with a potential of 0.1 volt or less used with the weak-signal rectifier of the usual grid-leak grid-condenser type. Linear detection means that the detector output is pro- portional to the applied radio-frequency signal voltage. In the weak-signal detector the output is proportional to the square of the signal strength, so that the detector is proportionately more sensitive on the parts of the modulated signal when the signal voltage is at a maximum than when at a minimum. In contrast with this, the power detector as used in all of the present- day broadcast receivers has a character- istic that is approximately linear even though the manufacturers do not specifi- cally mention this point. The square-law characteristic of the weak-signal detector introduces distortion frequencies which were not present in the original signal. These distortion frequen- cies form a larger and larger percentage of the audio-frequency output as the degree of modulation of the transmitted signal is increased. The present trend in the design of broadcasting stations is to use much higher degrees of modulation than were possible a few years ago, and all of the newer stations can modulate up to 100 per cent., in contrast with a maximum possible figure of not over 50 per cent, of a few years ago. The distortion frequencies introduced by square-law action consist of second har- Professor Terman brings to light an interesting and hitherto unannounced advantage of linear detection. This is its tendency to eliminate an undesired weak signal in favor of a strong desired signal; a tendency which is not shared by square-law detectors and one which increases the apparent selectivity of a receiver. 0 0.5 1.0 1.5 2.0 2.5 3.0 CARRIER VOLTAGE APPLIED TO DETECTOR Fig. 1 — Audio-frequency output of a detector as a Junction of the input voltage. monies of the notes actually being trans- mitted and also all the possible sum and difference frequencies. Thus, if the sending station is simultaneously transmitting notes of 1000 and 1500 cycles, the output of the square-law detector, in addition to containing these desired frequencies, will also contain double-frequency components of 2000 and 3000 cycles and sun-and- difference frequencies of 2500 and 500 cycles. These distortion components may, under the most unfavorable conditions, be 25 per cent, as large as the desired com- ponents, making it apparent that weak- signal detection of signals that have a high degree of modulation will not give satis- factory results from the point of view of quality. In contrast with this square-law action, a detector which has a linear character- istic introduces no frequencies in the audio- frequency output that were not present in the original signal. In order to have dis- tortionless detection it is absolutely es- sential, therefore, that a linear character- istic be employed. The ordinary power detector of either the grid leak or C bias type has approximately such a character- istic and so gives substantially undistorted rectification. Another advantage of power detection in regard to quality is that in detection of large signals the audio-frequency output is so much greater than with the weak-signal rectifier that it is possible to use one less stage of audio-frequency amplification, thus dispensing with one of the audio- frequency transformers and with the dis- tortion which it introduces. The efficiency with which the ordinary power detector rectifies the radio-frequency voltage is very much greater than in the case of the weak-signal rectifier. Thus, if a weak-signal detector with 0.02-volt input is replaced by a grid-leak power detector having 0.5-yolt input the audio-frequency output is increased not 25 times, but nearly 80 to 100 times. As a result of this increased efficiency with large signals, a moderate increase of radio-frequency amplification before detection is equiva- lent to a much larger amount of audio- frequency amplification after rectification. In the usual radio set, increasing the radio- frequency amplification about ten times, to bring the normal detector input from 0.05 volt up to 0.50 volt, will make it possible to drop out one stage of audio- frequency amplification (which amplifies about 25 times) without reducing the out- put of the power tube. The performance of a typical detector for different input voltages is shown in Figs. 1 and 2. The audio-frequency output as a function of signal voltage applied to the detector is given in Fig. 1. For large signals this relation is obviously very nearly a straight line, while for small in- puts the output follows a curved path. This transition from a square-law to linear characteristic as the signal is increased is typical of all detectors. The efficiency of rectification for the case of Fig. 1 is shown in Fig 2. This efficiency is expressed in terms of the ratio of actual audio-frequency output to the audio-frequency output that would be obtained from a perfect rectifier. The most important features to observe in Fig 2 are the low efficiency with small signals, and the fact that the efficiency is relatively high and constant under conditions giving linear operation. It is also to be noted that grid-leak detection is much more efficient than plate detection when com- pared for the same signal. At the same time the C-biased power detector is always more efficient than the weak-signal grid-leak detector, and also has the very important advantage of a linear characteristic. Detectors with a linear characteristic are much less susceptible to interference from static than are those with square-law characteristics. This can be illustrated very easily by a concrete example in which it is assumed that a static crash 10 times as strong as the signal is present. In the square-law detector this static impulse produces an output that is 100 times as strong as the signal, while with the linear Fig. 2 — Efficiency of rectifica- tion as a function of the input voltage. • NOVEMBER 1929 49