The first kind of transmission used a spark transmitter for Morse code. This was followed by the Poulson arc transmitter where the radiation from the sustained arc provided the continuous wave (CW) transmission. Then came CW transmission using the 30kc Alexanderson rotary alternator, and later a microphone was placed in the antenna lead to modulate the carrier.
Finally, the vacuum tube opened the gateway to relatively simple and easy RF carrier generation and concurrently adequate audio amplification. This brought with it the requirements for efficient methods of modulating the carrier wave to enable speech transmission. There developed a number of different modulating circuits. Most of these required large power tubes and corresponding large magnetics (chokes, transformers and other devices that tend to create magnetic fields) to handle high-power modulation.
The original attempts to transmit audio using modulated oscillators quickly proved inadequate because only relatively low power could be handled. Modulating the oscillator also led to other problems. Class A linear modulation with 30 percent efficiency was soon superseded the by class B amplifier, which in turn led to many modified class N systems in the effort to avoid the use of high-power modulating stages with expensive high-voltage power supplies, transformers and power tubes.
Throughout the years, a surprisingly large number of modulating systems were developed. Among these was a modified class C amplifier, which is cut off during the negative half cycle and its partial pulses added to another amplifier. The idea was excellent and ingenious, but proved to be difficult to operate in practice. This was partially due to the constantly changing operating load impedance problem. In 1936 a well-known engineer named William Doherty* devised an efficient operating system using this principle. This effectively made the Doherty linear amplifier a commercial proposition. In the following two or three years several other well known engineers such as Fred Terman, J. Woodyard and consulting engineer Jim Weldon of high-power AM transmitter and Continental Electronics fame postulated modifications that greatly improved circuit performance. Weldon took it a step further and developed a grounded grid circuit that considerably increased its efficiency.
When first developed, the Doherty amplifier gained a reputation for difficulty in adjustment and consistent operation. This was not so much the fault of the circuit but lack of knowledge of the principles and power flow by the operators. One of the major problems was a lack of understanding of the operating characteristics of the load and difficulty in obtaining a purely resistive output load.
Triode tubes were originally envisioned for the system, but as new tubes were developed the circuit was adapted to use the new tube types. The general theory of operation of the Doherty amplifier is shown in Figures 1 and 2.
Figure 1 shows a basic triode Doherty amplifier circuit. One tube is the carrier tube and the other is the peak tube. RL is the load across which the final RF power output is developed. This load is probably the cause of much of the misunderstanding and problems encountered in the Doherty amplifier. The problem occurs because two tubes feed power into a single load. Each source develops a voltage across the load, and when both sources feed power simultaneously the impedance will change depending on the relative powers.
Figure 2. The various waveforms created by the Doherty amplifier to yield an output. A is the input signal, B is the signal in saturation, C is the peak pulse and D is the resulting output. Click here to enlarge this image.
Figure 2 shows the modulation generation process. Figure 2A shows the input signal. The carrier tube is biased to operate in class B and is driven to saturation at carrier level as shown in Figure 2B. In this manner all the output ranging from peak negative modulation up to saturation level is produced. At the same time the peak tube is biased to operate in class C, and its drive is adjusted to produce absolute minimum plate current at carrier level. This results in the production of positive modulation peak pulses as shown in figure 2C.
Now we come to Figure 2D. Note that the waveform is shown as being composed of two pulses (2B with 2C added together). Also note that it is larger in amplitude than Figure 2A In fact, when the amplifier is properly adjusted it yields four times the power of Figure 2A. Although the average plate current of the peak tube is considerably lower than that of the carrier tube it is essential that its cathode emission be as great as the peak tube's. This is because under conditions of maximum power both tubes contribute the same amount of power; for this reason it is common to use the same tube type for each stage.
This is where many people have difficulty understanding the Doherty system. We mentioned earlier that a pi network was used. This has the valuable property of inverting the load and termination impedances. It also delays the signal by 90 degrees, which is not as useful in this instance. This calls for a phase-advancing network in a transmitter. This 90 phase shift is not a basic part of a Doherty amplifier but is merely an undesired by-product of the essential 90° network.
When two power sources feed a single load, complications result affecting the ultimate impedance value. This is where the phase shifting property of a network allows the Doherty amplifier to work. It is common to use the phase shifting property of a network to accomplish an impedance balance. If we merely paralleled the plate circuits of the peak and carrier tubes across the load the system would work only as long as the grid drive voltage on the carrier tube stayed between zero and carrier level (the peak tube is biased almost to cut-off).
When excitation goes above carrier level on the positive half cycle the output of the peak tube rises. This will cause the apparent load impedance to increase and the carrier tube output would decrease as the peak tube output increased. This would not provide a four-fold increase in carrier power.
This is where the pi network formed by the inductance L1 and capacitances C1 and C2 are used to invert the impedances seen by the peak and carrier tubes to produce the required varying load impedance value. Remember that a 90 degree network has the characteristic that an increase in the load impedance is matched by a reduction in the input impedance of the network. This operates in both directions so that a change in load impedance caused by the operation of one tube into the network will affect the output of the other tube.
But what about the 90-degree phase difference between the two voltages? The pulses won't match.
In an operational broadcast transmitter correction of the 90-degree phase shift lag is taken care of by means of a simple 90-degree phase advancing network and the peak tube input is synchronized with the carrier tube. This advancing network adds nothing to the performance of the Doherty amplifier, although without it the amplifier could not operate.
It is obvious from this rather brief description that the Doherty amplifier is a little different from the familiar amplifiers. Two networks are involved, one is a 90-degree lagging and the other a 90-degree leading. It is easy to see how a small adjustment of either of these networks will have considerable effect on the efficiency and operation of the amplifier. In normal use there should be no need to adjust the impedance of the load once tuned, except perhaps when installing new tubes. The Doherty amplifier is probably still to be found in use in a number of stations, although I can't help wondering how it would work with IBOC today.
E-mail Battison at
* Described in the Proceedings of the I.R.E., September 1936