The three basic types of power amplifier used in transmitters are linear, class C, and switching. Linear amplifiers provide an output signal that is an identical, enlarged replica of the input. Their output is directly proportional to their input, and they therefore faithfully reproduce an input, but at a higher power level. Most audio amplifiers are linear. Linear RF amplifiers are used to increase the power level of variable-amplitude RF signals such as low-level AM or SSB signals. Most modern digital modulation techniques such as spread spectrum, QAM, and orthogonal frequency division multiplex (OFDM) require linear amplification to retain the modulating signal information. Linear amplifiers are class A, AB, or B. The class of an amplifier indicates how it is biased.
Class A amplifiers
Class A amplifiers are biased so that they conduct continuously. The bias is set so that the input varies the collector (or drain) current over a linear region of the transistor’s characteristics. Thus its output is an amplified linear reproduction of the input. Usually, we say that the class A amplifier conducts for 360° of an input sine wave.
Class B amplifier
Class B amplifiers are biased at cutoff so that no collector current flows with zero input. The transistor conducts on only one-half, or 180°, of the sine wave input. This means that only one-half of the sine wave is amplified. Normally, two class B amplifiers are connected in a push-pull arrangement so that both the positive and negative alternations of the input are amplified.
Class AB linear amplifier
Class AB linear amplifiers are biased near cutoff with some continuous collector current flow. They conduct for more than 180° but less than 360° of the input. They too are used primarily in push-pull amplifiers and provide better linearity than class B amplifiers, but with less efficiency.
Class A amplifier
Class A amplifiers are linear but not very efficient. For that reason, they make poor power amplifiers. As a result, they are used primarily as small-signal voltage amplifiers or for low-power amplifications. The buffer amplifiers described previously are class A amplifiers.
Class B amplifier
Class B amplifiers are more efficient than class A amplifiers, because current flows for only a portion of the input signal, and they make good power amplifiers. However, they distort an input signal because they conduct for only one-half of the cycle. Therefore, special techniques are often used to eliminate or compensate for the distortion. For example, operating class B amplifiers in a push-pull configuration minimize the distortion.
Class C amplifier
Class C amplifiers conduct for even less than one-half of the sine wave input cycle, making them very efficient. The resulting highly distorted current pulse is used to ring a tuned circuit to create a continuous sine wave output. Class C amplifiers cannot be used to amplify varying-amplitude signals. They will clip off or otherwise distort an AM or SSB signal. However, FM signals do not vary in amplitude and can therefore be amplified with more efficient nonlinear class C amplifiers. This type of amplifier also makes a good frequency multiplier as harmonics are generated in the
Switching amplifiers act like on/off or digital switches. They effectively generate a square wave output. Such a distorted output is undesirable; however, by using high-Q tuned circuits in the output, the harmonics generated as part of the switching process can be easily filtered out. The on/off switching action is highly efficient because current during only one-half of the input cycle, and when it does, the voltage drop across the transistor is very low, resulting in low power dissipation. Switching amplifiers are designated class D, E, F, and S.
Linear amplifiers are used primarily in AM and SSB transmitters, and both low- and high-power versions are used. Some examples follow.
Class A Buffers
A simple class A buffer amplifier is shown in Fig. 8-21. This type of amplifier is used between the carrier oscillator and the final power amplifier to isolate the oscillator from the power amplifier load, which can change the oscillator frequency. It also provides a modest power increase to provide the driving power required by the final amplifier. Such circuits usually provide milliwatts of power and rarely more than 1 W. The carrier oscillator signal is capacitively coupled to the input. The bias is derived from R1, R2, and R3. The emitter resistor R3 is bypassed to provide maximum gain. The collector is tuned with a resonant LC circuit at the operating frequency. An inductively coupled secondary loop transfers power to the next stage.
High Power Linear Amplifier.
A high-power class A linear amplifier is shown in Fig. 8-22. A power MOSFET may also be used in this circuit with a few modifications. Base bias is supplied by a constant-current circuit that is temperature-compensated. The RF input from a 50-V source is connected to the base via an impedance-matching circuit made up of C1, C2, and L1. The output is matched to a 50-V load by the impedance-matching network made up of L2, L3, C3, and C4. When connected to a proper heat sink, the transistor can generate up to 100 W of power up to about 200 MHz. The amplifier is designed for a specific frequency that is set by the input and output tuned circuits. Class A amplifiers have a maximum efficiency of 50 percent. Thus only 50 percent of the dc power is converted to RF, with the remaining 50 percent being dissipated in the transistor. For 100-W RF output, the transistor dissipates 100 W. Efficiencies of less than 50 percent are typical.
Commonly available RF power transistors have an upper power limit of several hundred watts. To produce more power, two or more devices can be connected in parallel, in a push-pull configuration, or in some combination. Power levels of up to several thousand watts are possible with these arrangements.
Class B Push-Pull Amplifier
A class B linear power amplifier using push-pull is shown in Fig. 8-23. The RF driving signal is applied to Q1 and Q2 through input transformer T1. It provides impedance-matching and base drive signals to Q1 and Q2 that are 180° out of phase. An output transformer T2 couples the power to the antenna or load. Bias is provided by R1 and D1.
For class B operation, Q1 and Q2 must be biased right at the cutoff point. The emitter base junction of a transistor will not conduct until about 0.6 to 0.8 V of forward bias is applied because of the built-in potential barrier. This effect causes the transistors to be naturally biased beyond cutoff, not right at it. A forward-biased silicon diode D1 has about 0.7 V across it, and this is used to put Q1 and Q2 right on the conduction threshold.
On the positive half-cycle of the RF input, the base of Q1 is positive and the base of Q2 is negative. The Q2 is cut off, but Q1 conducts, linearly amplifying the positive half-cycle. Collector current flows in the upper half of T2, which induces an output voltage in the secondary. On the negative half-cycle of the RF input, the base of Q1 is negative, so it is cut off. The base of Q2 is positive, so Q2 amplifies the negative half cycle. Current flows in Q2 and the lower half of T2, completing a full cycle. The power is split between the two transistors.
The circuit in Fig. 8-23 is an untuned broadband circuit that can amplify signals over a broad frequency range, typically from 2 to 30 MHz. A low-power AM or SSB signal is generated at the desired frequency and then applied to this power amplifier before being sent to the antenna. With push-pull circuits, power levels of up to 1 kW are possible.
Fig. 8-24 shows another push-pull RF power amplifier. It uses two power MOSFETs, can produce an output up to 1 kW over the 10- to 90-MHz range, and has a 12-dB power gain. The RF input driving power must be 63 W to produce the full 1-kW output. Toroidal transformers T1 and T2 are used at the input and output for impedance matching. They provide broadband operation over the 10- to 90-MHz range without tuning. The 20-nH chokes and 20-V resistors form neutralization circuits that provide out-of-phase feedback from output to input to prevent self-oscillation.
Class C Amplifier
The key circuit in most AM and FM transmitters is the class C amplifier. These amplifiers are used for power amplification in the form of drivers, frequency multipliers, and final amplifiers. Class C amplifiers are biased, so they conduct for less than 180° of the input. A class C amplifier typically has a conduction angle of 908 to 150°. Current flows through it in short pulses, and a resonant tuned circuit is used for complete signal amplification.
Fig. 8-25(a) shows one way of biasing a class C amplifier. The base of the transistor is simply connected to ground through a resistor. No external bias voltage is applied. An RF signal to be amplified is applied directly to the base. The transistor conducts on the positive half-cycles of the input wave and is cut off on the negative half-cycles. Although this sounds like a class B configuration, that is not the case. Recall that the emitter-base junction of a bipolar transistor has a forward voltage threshold of approximately 0.7 V. In other words, the emitter-base junction does not really conduct until the base is more positive than the emitter by 0.7 V. Because of this, the transistor has an inherent built-in reverse bias. When the input signal is applied, the collector current does not flow until the base is positive by 0.7 V. This is illustrated in Fig. 8-25(b). The result is that collector current flows through the transistor in positive pulses for less than the full 180° of the positive ac alternation.
In many low-power driver and multiplier stages, no special biasing provisions other than the inherent emitter-base junction voltage are required. The resistor between base and ground simply provides a load for the driving circuit. In some cases, a narrower conduction angle than that provided by the circuit in Fig. 8-25(a) must be used. In such cases, some form of bias must be applied. A simple way of supplying bias is with the RC network shown in Fig. 8-26(a).
Here the signal to be amplified is applied through capacitor C1. When the emitter-base junction conducts on the positive half-cycle, C1 charges to the peak of the applied voltage less the forward drop across the emitter-base junction. On the negative half-cycle of the input, the emitter-base junction is reverse biased, so the transistor does not conduct. During this time, however, capacitor C1 discharges through R1, producing a negative voltage across R1, which serves as a reverse bias on the transistor. By properly adjusting the time constant of R1 and C1, an average dc reverse-bias voltage can be established. The applied voltage causes the transistor to conduct, but only on the peaks. The higher the average dc bias voltage, the narrower the conduction angle and the shorter the duration of the collector current pulses. This method is referred to as signal bias.
Of course, negative bias can also be supplied to a class C amplifier from a fixed dc supply voltage, as shown in Fig. 8-26(b). After the desired conduction angle is established, the value of the reverse voltage can be determined and applied to the base through the RFC. The incoming signal is then coupled to the base, causing the transistor to conduct on only the peaks of the positive input alternations. This is called external bias and requires a separate negative dc supply.
Another biasing method is shown in Fig. 8-26(c). As in the circuit shown in Fig. 8-26(a), the bias is derived from the signal. This arrangement is known as the self-bias method. When current flows in the transistor, a voltage is developed across R1. Capacitor C1 is charged and holds the voltage constant. This makes the emitter more positive than the base, which has the same effect as a negative voltage on the base. A strong input signal is required for proper operation. These circuits also work with an enhancement mode MOSFET.
Tuned Output Circuits
All class C amplifiers have some form of tuned circuit connected in the collector, as shown in Fig. 8-27. The primary purpose of this tuned circuit is to form the complete ac sine wave output. A parallel-tuned circuit rings,or oscillates, at its resonant frequency whenever it receives a dc pulse.
The pulse charges the capacitor, which, in turn, discharges into the inductor. The magnetic field in the inductor increases and then collapses, inducing a voltage which then recharges the capacitor in the opposite direction. This exchange of energy between the inductor and the capacitor, called the flywheel effect, produces a damped sine wave at the resonant frequency. If the resonant circuit receives a pulse of current every half-cycle, the voltage across the tuned circuit is a constant-amplitude sine wave at the resonant frequency. Even though the current flows through the transistor in short pulses, the class C amplifier output is a continuous sine wave.
Another way to look at the operation of a class C amplifier is to view the transistor as supplying a highly distorted pulse of power to the tuned circuit. According to Fourier theory, this distorted signal contains a fundamental sine wave plus both odd and even harmonics. The tuned circuit acts as a bandpass filter to select the fundamental sine wave contained in the distorted composite signal.
The tuned circuit in the collector is also used to filter out unwanted harmonics. The short pulses in a class C amplifier are made up of second, third, fourth, fifth, etc., harmonics. In a high-power transmitter, signals are radiated at these harmonic frequencies as well as at the fundamental resonant frequency. Such harmonic radiation can cause out-of-band interference, and the tuned circuit acts as a selective filter to eliminate these higher-order harmonics. If the Q of the tuned circuit is made high enough, the harmonics will be adequately suppressed.
The Q of the tuned circuit in the class C amplifier should be selected so that it provides adequate attenuation of the harmonics but also has sufficient bandwidth to pass the sidebands produced by the modulation process. Remember that the bandwidth and Q of a tuned circuit are related by the expression
If the Q of the tuned circuit is too high, then the bandwidth will be very narrow and some of the higher-frequency sidebands will be eliminated. This causes a form of frequency distortion called sideband clipping and may make some signals unintelligible or will at least limit the fidelity of reproduction.
One of the main reasons why class C amplifiers are preferred for RF power amplification over class A and class B amplifiers is their high efficiency. Remember, efficiency is the ratio of the output power to the input power. If all the generated power, the input power, is converted to output power, the efficiency is 100 percent. This doesn’t happen in the real world because of losses. But in a class C amplifier, more of the total power generated is applied to the load. Because the current flows for less than 180° of the ac input cycle, the average current in the transistor is fairly low; i.e., the power dissipated by the device is low. A class C amplifier functions almost as a transistor switch that is off for over 180° of the input cycle. The switch conducts for approximately 90 to 150° of the input cycle. During the time that it conducts, its emitter-to-collector resistance is low. Even though the peak current may be high, the total power dissipation is much less than that in class A and class B circuits. For this reason, more of the dc power is converted to RF energy and passed on to the load, usually an antenna. The efficiency of most class C amplifiers is in the 60 to 85 percent range.
The input power in a class C amplifier is the average power consumed by the circuit, which is simply the product of the supply voltage and the average collector current, or
For example, if the supply voltage is 13.5 V and the average dc collector current is 0.7 A, the input power is Pin=13.5(0.7)= 9.45 W
The output power is the power actually transmitted to the load. The amount of power depends upon the efficiency of the amplifier. The output power can be computed with the familiar power expression
where V is the RF output voltage at the collector of the amplifier and RL is the load impedance. When a class C amplifier is set up and operating properly, the peak-to-peak RF output voltage is two times the supply voltage, or 2VCC (see Fig. 8-27).
Any class C amplifier is capable of performing frequency multiplication if the tuned circuit in the collector resonates at some integer multiple of the input frequency. For example, a frequency doubler can be constructed by simply connecting a parallel-tuned circuit in the collector of a class C amplifier that resonates at twice the input frequency. When the collector current pulse occurs, it excites or rings the tuned circuit at twice the input frequency. A current pulse flows for every other cycle of the input. A tripler circuit is constructed in exactly the same way, except that the tuned circuit resonates at three times the input frequency, receiving one input pulse for every 3 cycles of oscillation it produces (see Fig. 8-28).
Multipliers can be constructed to increase the input frequency by any integer factor up to approximately 10. As the multiplication factor gets higher, the power output of the multiplier decreases. For most practical applications, the best result is obtained with multipliers of 2 and 3.
Another way to look at the operation of a class C frequency multiplier is to remember that the nonsinusoidal current pulse is rich in harmonics. Each time the pulse occurs, the second, third, fourth, fifth, and higher harmonics are generated. The purpose of the tuned circuit in the collector is to act as a filter to select the desired harmonic.
In many applications, a multiplication factor greater than that achievable with a single multiplier stage is required. In such cases, two or more multipliers are cascaded. Fig. 8-29 shows two multiplier examples. In the first case, multipliers of 2 and 3 are cascaded to produce an overall multiplication of 6. In the second, three multipliers provide an overall multiplication of 30. The total multiplication factor is the product of the multiplication factors of the individual stages.
A key specification for all RF power amplifiers, especially linear amplifiers, is their efficiency. Efficiency is simply the ratio of the amplifier power output (Po) to the total DC power (Pdc) used to produce the output, or:
Efficiency is the mathematical percentage of DC input power that is converted into RF power. The ideal, of course, is 100 percent, which cannot be achieved. Most amplifier designs emphasize good efficiency where possible. Good efficiency means less overall power consumption. Any power not converted to RF is lost as heat dissipated in the power transistors.
Another measure of efficiency is called power-added efficiency (PAE), which takes into consideration the amount of input power needed to drive a higher-power amplifier to maximum output.
Some high-power amplifiers, especially in the VHF/UHF/microwave range, require high driving power, which adds to the overall efficiency rating. For example, it may take 100 watts of drive power to obtain an output power of 1500 watts. The 100 watts of driving power is not insignificant. The PAE rating factors this in.
Linear power amplifiers are the least efficient of all the amplifiers. However, most of the newer digital modulation techniques, such as OFDM, QAM, code division multiple access (CDMA) spread spectrum, and others (to be discussed ) require linear amplification to retain all the modulating information. Poor efficiency is a major issue, especially in cellular base stations where cooling and electrical power costs are major operational considerations.
Class C and switching power amplifiers are the most efficient RF power amplifiers and are used where nonlinear amplification can be used, as with some types of AM, FM, and PM. Otherwise, several special techniques have been developed to improve the efficiency in linear power amplifiers.
Table 8-1 is a summary of the theoretical and practical efficiencies that can be obtained with the basic linear power amplifiers.
The actual efficiency depends on the design and, in most cases, the input power level. For example, class A and AB amplifiers are most efficient when operated at the maximum possible input level before distortion occurs. Less input results in lower efficiency and signifi cant power lost as heat.
Switching Power Amplifiers
As stated previously, the primary problem with RF power amplifiers is their inefficiency and high power dissipation. To generate RF power to transfer to the antenna, the amplifier must dissipate a considerable amount of power itself. For example, a class A power amplifier using a transistor conducts continuously. It is a linear amplifier whose conduction varies as the signal changes. Because of the continuous conduction, the class A amplifier generates a considerable amount of power that is not transferred to the load. Because of the high-power dissipation, the output power of a class A amplifier is generally limited. For that reason, class A amplifiers are normally used only in low-power transmitter stages.
To produce greater power output, class B amplifiers are used. Each transistor conducts for 180° of the carrier signal. Two transistors are used in a push-pull arrangement to form a complete carrier sine wave. Since each transistor is conducting for only 180° of any carrier cycle, the amount of power it dissipates is considerably less, and efficiencies of 70 to 75 percent are possible. Class C power amplifiers are even more efficient, since they conduct for less than 180° of the carrier signal, relying on the tuned circuit in the plate or collector to supply power to the load when they are not conducting. With current flowing for less than 180° of the cycle, class C amplifiers dissipate less power and can, therefore, transfer more power to the load. Efficiencies as high as about 85 percent can be achieved, and class C amplifiers are therefore the most widely used type in power amplifiers when the type of modulation permits.
Another way to achieve high efficiencies in power amplifiers is to use a switching amplifier. A switching amplifier is a transistor that is used as a switch and is either conducting or nonconducting. Both bipolar transistors and enhancement-mode MOSFETs are widely used in switching-amplifier applications. A bipolar transistor as a switch is either cut off or saturated. When it is cut off, no power is dissipated. When it is saturated, current flow is maximum, but the emitter-collector voltage is extremely low, usually less than 1 V. As a result, power dissipation is extremely low.
When enhancement-mode MOSFETs are used, the transistor is either cut off or turned on. In the cutoff state, no current flows, so no power is dissipated. When the transistor is conducting, its on-resistance between source and drain is usually very low again, no more than several ohms and typically far less than 1 V. As a result, power dissipation is extremely low even with high currents.
The use of switching power amplifiers permits efficiencies of over 90 percent. The current variations in a switching power amplifier are square waves and thus harmonics are generated. However, these are relatively easy to filter out by the use of tuned circuits and filters between the power amplifier and the antenna.
The three basic types of switching power amplifiers, class D, Class E, and Class S, were originally developed for high-power audio applications. But with the availability of high-power, high-frequency switching transistors, they are now widely used in radio transmitter design.
Class D Amplifiers
A class D amplifier uses a pair of transistors to produce a square wave current in a tuned circuit. Fig. 8-30 shows the basic configuration of a class D amplifier. Two switches are used to apply both positive and negative dc voltages to a load through the tuned circuit. When switch S1 is closed, S2 is open; when S2 is closed,S1 is open.
When S1 is closed, a positive dc voltage is applied to the load. When S2 is closed, a negative dc voltage is applied to the load. Thus the tuned circuit and load receive an ac square wave at the input.
The series resonant circuit has a very high Q. It is resonant at the carrier frequency. Since the input waveform is a square wave, it consists of a fundamental sine wave and odd harmonics. Because of the high Q of the tuned circuit, the odd harmonics are filtered out, leaving a fundamental sine wave across the load. With ideal switches, that is, no leakage current in the off state and no on resistance when conducting, the theoretical efficiency is 100 percent.
Fig. 8-31 shows a class D amplifier implemented with enhancement-mode MOSFETs. The carrier is applied to the MOSFET gates 180° out of phase by use of a transformer with a center-tapped secondary. When the input to the gate of Q1 is positive, the input to the gate of Q2 is negative. Thus Q1 conducts and Q2 is cut off. On the next half-cycle of the input, the gate to Q2 goes positive and the gate of Q1 goes negative. The Q2 conducts, applying a negative pulse to the tuned circuit. Recall that enhancement-mode MOSFETs are normally nonconducting until a gate voltage higher than a specific threshold value is applied, at which time the MOSFET conducts. The on resistance is very low. In practice, efficiencies of up to 90 percent can be achieved by using a circuit like that in Fig. 8-31.
Class E and F Amplifiers
In class E amplifiers, only a single transistor is used. Both bipolar and MOSFETs can be used, although the MOSFET is preferred because of its low drive requirements. Fig. 8-32 shows a typical class E RF amplifier. The carrier, which may initially be a sine wave, is applied to a shaping circuit that effectively converts it to a square wave. The carrier is usually frequency-modulated. The square wave carrier signal is then applied to the base of the class E bipolar power amplifier.
The Q1 switches off and on at the carrier rate. The signal at the collector is a square wave that is applied to a low-pass filter and tuned impedance-matching circuit made up of C1, C2, and L1. The
odd harmonics are filtered out, leaving a fundamental sine wave that is applied to the antenna. A high level of efficiency is achieved with this arrangement.
A class F amplifier is a variation of the class E amplifier. It contains an additional resonant network in the collector or drain circuit. This circuit, a lumped LC or even a tuned transmission line at microwave frequencies, is resonant at the second or third harmonic of the operating frequency. The result is a waveform at the collector (drain) that more closely resembles a square wave. The steeper waveform produces faster transistor switching and better efficiency.
Class S Amplifiers
Class S amplifiers, which use switching techniques but with a scheme of pulse-width modulation, are found primarily in audio applications but have also been used in low- and medium-frequency RF amplifiers such as those used in AM broadcast transmitters. The low-level audio signal to be amplified is applied to a circuit called a pulse-width modulator. A carrier signal at a frequency 5 to 10 times the highest audio frequency to be amplified is also applied to the pulse-width modulator. At the output of the modulator is a series of constant-amplitude pulses whose pulse width or duration varies with the audio signal amplitude. These signals are then applied to a switching amplifier of the class D type. High power and efficiency are achieved because of the switching action. A low-pass filter is connected to the output of the switching amplifier to average and smooth the pulses back into the original audio signal waveform. A capacitor or low-pass filter across the speaker is usually sufficient. These amplifiers are usually referred to as class D amplifiers in audio applications. They are widely used in battery-powered portable units where battery life and efficiency are paramount.
RF Power Transistors
Although bipolar transistors are still used in some RF power amplifier (PA) designs, most new designs use FETs. They require less drive, and overall the circuitry is generally simpler. The newer models can achieve power outputs of several hundred watts to frequencies well into the GHz regions. The most widely used devices in modern RF PAs are LDMOS and GaN HEMT.
The laterally diffused MOS (LDMOS) FET is typically an n-type enhancement-mode MOSFET with extra-large elements to handle high power and heat. The geometry of the device is designed to reduce the drain to gate feedback capacitance to extend the high-frequency operating range. LDMOS FETs are widely used in radio transmitters for cellular base stations. Other RF uses are in radar and high-power transmitters for broadcast and two-way radios. They can accommodate drain supply voltages to 50 volts and power levels of 600 watts per device. LDMOS FETs can operate to frequencies of up to about 6 GHz.
A newer form of discrete power FET is one made from gallium nitride (GaN) instead of silicon. Those semiconductor materials permit the FET to operate at high drain voltages to 100 volts and drain currents to several amperes. GaN FETs are a type of metal-semiconductor junction FET or MESFET. They are called high-electron-mobility transistors (HEMT). Instead of the metal-semiconductor junction the gate junction of the MESFET, the HEMT uses different semiconductor materials for the gate and the channel. This is called a heterojunction. One common combination is GaAs for the channel and aluminum gallium arsenide (AlGaAs) for the gate. HEMTs are also made of GaN and are good power transistors. A variation called a pseudomorphic or pHEMT uses additional layers of different semiconductor materials, including compounds of indium (In). These layers are optimized to further speed electron transit, pushing their ability to amplify well into the millimeter-wave range of 30 to 100 GHz.
GaN RF power FETs are of the depletion-mode pHEMT type and handle heat better than silicon power transistors do. They can operate at power levels of tens of watts to frequencies of 30 GHz or more. They are used in radar, satellites, and cellular base stations.
Linear Broadband Power Amplifiers
The power amplifiers described so far in this chapter are narrowband amplifiers. They provide high-power output over a relatively small range of frequencies. The bandwidth of the signal to be amplified is set by the modulation method and the frequencies of the modulating signals. In so many applications, the total bandwidth is only a small percentage of the carrier frequency, making conventional LC resonant circuits practical. Some of the untuned push-pull power amplifiers described earlier (Figs. 8-23 and 8-24) do have a broader bandwidth up to several megahertz. However, some of the newer wireless systems now require much broader bandwidth. A good example is the code division multiple access (CDMA) cell phone standard. The CDMA system uses a modulation/multiplexing technique called spread spectrum that does, as its name implies, spread the signal out over a very wide frequency spectrum. The newer long-term evolution (LTE) cellular technology in modern 4G cell phone systems uses an advanced modulation technique called orthogonal frequency division multiplex (OFDM) along with quadrature amplitude modulation (QAM). Signal bandwidths of 5 to 20 MHz are common. Such complex modulation schemes require that the amplification be linear over a wide frequency range to ensure no amplitude, frequency, or phase distortion. Special amplification techniques have been developed to meet this need. Four common methods are discussed here.
The concept behind a feedforward amplifier is that the distortion produced by the power amplifier is isolated and then subtracted from the amplified signal, producing a nearly distortion-free output signal. Fig. 8-33 shows one common implementation of this idea. The wide-bandwidth signal to be amplified is fed to a power splitter that divides the signal into two equal-amplitude signals. A typical splitter may be a transformer like a device or even a resistive network. It maintains the constant impedance, usually 50V, but typically also introduces some attenuation. One-half of the signal is then amplified in a linear power amplifier similar to those broadband class AB amplifiers discussed earlier. A directional coupler is used next to tap off a small portion of the amplified signal that contains the original input information as well as harmonics resulting from distortion. A directional coupler is a simple device that picks up a small amount of the signal by inductive coupling. It is typically just a short copper line running adjacent to the signal line on a printed circuit board. At microwave frequencies, a directional coupler may be a more complex device with a coaxial structure. In any case, the sample of the amplified signal is also passed through a resistive attenuator to further reduce the signal level.
The lower output of the signal splitter is sent to a delay line circuit. A delay line is a low-pass filter or a section of a transmission line such as a coaxial cable that introduces a specific amount of delay to the signal. It may be a few nanoseconds up to several microseconds depending upon the frequency of operation and the type of power amplifier used. The delay is used to match the delay encountered by the upper input signal in the power amplifier. This delayed signal is then fed to a signal combiner along with the attenuated signal sample from the amplifier output. Amplitude and phase controls are usually provided in both signal paths to ensure that they are of the same amplitude and phase. The combiner may be resistive or a transformer like a device. In either case, it effectively subtracts the original signal from the amplified signal, leaving only the harmonic distortion.
The harmonic distortion is now amplified by another power amplifier with a power level equal to that of the upper signal power amplifier. The upper amplifier signal is passed through the directional coupler and into a delay line that compensates for the delay introduced by the lower error signal amplifier. Again, amplitude and phase controls are usually provided to adjust the power levels of the upper and lower signal levels, so they are equal. Finally, the error signal is now subtracted from the amplified combined signal in a signal coupler or combiner. This coupler, like the input splitter, is typically a transformer. The resulting output is the original amplified signal less the distortion.
Amplifiers like this are available with power levels of a few watts to over several hundred watts. The system is not perfect as the signal cancellations or subtractions are not precise because of amplitude and phase mismatches. Distortion in the lower power amplifier also contributes to the overall output. Yet with close adjustments, these differences can be minimized, thereby greatly improving the linearity of the amplifier over other types. The system is also inefficient because two power amplifiers are required. But the tradeoff is wide bandwidth and very low distortion.
Digital Predistortion Amplification
This method of amplification referred to as digital pre-distortion (DPD), uses a digital signal processing (DSP) technique to pre distort the signal in such a way that when amplified the amplifier distortion will offset or cancel the predistortion characteristics, leaving a distortion-free output signal. The amplified output signal is continuously monitored and used as feedback to the DSP so that the predistorion calculations can be changed on the fly to provide an inverse predistortion that perfectly matches the amplifier’s distortion.
Refer to Fig. 8-34 that shows a representative system. The digitized information signal, in serial format, is fed to a digital correction algorithm in a DSP. Inside the DSP are computing algorithms that feed corrective logic signals to the digital correction algorithm to modify the signal in such a way as to be an inverse match to the distortion produced by the power amplifier.
Once the corrective action has been taken on the baseband digital signal, it is sent to a modulator that produces the signal to be transmitted. The modulation is handled by the DSP chip itself rather than a separate modulator circuit. This modulated signal is then fed to a digital-to-analog converter (DAC) where it produces the desired analog signal to be transmitted.
Next, the DAC output is sent to a mixer along with a sine wave signal from an oscillator or frequency synthesizer. A mixer is similar to a low-level amplitude modulator or analog multiplier. Its output consists of the sum and difference of the DAC and synthesizer signals. In this application, the sum signal is selected by a filter making the mixer an upconverter. The synthesizer frequency is chosen so that the mixer output is at the desired operating frequency. Any modulation present is also contained on the mixer output.
The predistorted signal is then amplified in a highly linear class AB power amplifier (PA) and fed to the output antenna. Note in Fig. 8-34 that the output signal is sampled in a directional coupler, amplified, and sent to another mixer used as a down converter. The synthesizer also provides the second input to this mixer. The difference frequency is selected by a filter, and the result is sent to an analog-to-digital converter (ADC). The digital output of the ADC represents the amplified signal plus any distortion produced by the PA. The DSP uses this digital input to modify its algorithm to properly correct for the actual distortion. The signal is then modified by the digital correction algorithm in such a way that most of the distortion is canceled.
While the DPD method of broadband amplification is complex, it provides nearly distortion-free output. Only a single power amplifier is needed, making it more efficient than the feedforward method. Several semiconductor manufacturers make the predistortion circuits needed to make this work.
Envelope tracking (ET) is a technique that allows class A, AB, and B amplifiers to become more efficient. Linear amplifiers must be used to amplify the signals deploying advanced modulation techniques used in 3G and 4G cellular radios, such as CDMA spread spectrum, QAM, and OFDM. The ET method provides a way to adopt existing linear amplifiers but increases their efficiency, making them run cooler and use less power.
The advanced modulation methods have wide signal amplitude swings that must be preserved. This condition is generally known as a wide peak-to-average power ratio (PAPR). This leads to the problem inherent in class A amplifiers, where the efficiency is maximum when the signal levels are high and the output is near compression where the signals come closest to the amplifier supply voltage levels. Efficiency is high when the amplifier is near compression where peak clipping occurs. With a maximum undistorted signal level, the efficiency can approach the theoretical maximum of 50 percent. However, at lower input signal levels, efficiency drops to a very low level. Efficiencies may be below 10 percent, meaning that most of the dc supply power is dissipated as heat by
the RF power transistors.
The ET system tracks the amplitude or modulation envelope of the RF signal and uses that to control the dc power supply voltage to the amplifier. In other words, the signal amplitude modulates the dc supply voltage. The process is similar to the high-level amplitude modulation technique discussed in (Fig. 4-13). This allows the amplifier to always remain close to the compression point where the efficiency is greatest.
Fig. 8-35a shows the transmitter stages where the analog baseband signal is fed to the linear RF power amplifier. The power amplifier has a fixed power supply voltage, usually from a DC-DC converter/regulator. The resulting output signal is shown along with the power supply level. The red area indicates the power lost as heat.
Fig. 8-35b illustrates the ET method. The analog baseband signal also drives the power amplifier as before. However, note that the analog signal is passed through a diode rectifier similar to that used in AM demodulation. This rectifier circuit recovers the
the envelope of the signal and applies it to the ET power supply that supplies voltage to the PA. The dc supply voltage, therefore, tracks the modulation envelope, keeping the amplifier near compression and operating at maximum efficiency. The ET power supply is essentially a very wide bandwidth DC-DC converter that can adequately follow the signal amplitude and frequency changes. The common dc supply voltage can range from 0.5 to 5 volts.
The ET technique is implemented with special ET tracking and dc modulation circuits. These may be in IC form or be fully integrated into another IC. The ET technique is widely used in cellular base station power amplifiers as well as in cellphone handsets using 3G and 4G (LTE) digital wireless technologies. Practical efficiencies can be as high as 40 to 50 percent, delivering considerable power and heat savings in base stations and providing longer battery life and cooler operation in a cell phone. The ET technique is also widely used in combination with the DPD technique, described earlier, to provide even greater linearity and higher efficiencies.
The Doherty amplifier is a unique design using two amplifiers that work together to maintain linearity while improving efficiency. The design originated in 1936 as an attempt to bring greater power and efficiency to high-power vacuum tube shortwave radio amplifiers. The design has now been widely adopted in power amplifiers for multiband, multimode cellular base stations.
Fig. 8-36 shows the basic Doherty arrangement. The main carrier amplifier is a linear class AB circuit, and the peaking amplifier is usually a class C amplifier. Both amplifiers supply power to the load (antenna), creating what is referred to as a load pulling circuit whose impedance changes with signal levels. The carrier amplifier supplies power to the load at the lower input signal levels and the peaking amplifier switches in at the higher power levels.
The signal to be amplified is split equally by a power divider circuit. The signal goes directly to the carrier amplifi er and through a quarter wavelength (λ/4) transmission line that provides a 90-degree phase shift of the signal to the peaking amplifier. The transmission line is usually a short copper pattern on the amplifier printed circuit board (PCB). At low signal levels, the peaking amplifier bias keeps it cut off so that it does not supply power to the load. The carrier amplifier operates near compression to maintain high efficiency and supplies power to the load. At higher input levels, the peaking amplifier cuts in and supplies more power to the load.
The amplifiers deliver power to the load by a λ/4 transmission line used as an impedance converter and another λ/4 line used as an impedance transformer. These transmission lines combined provide dynamic load impedance adaption for the amplifiers to increase efficiency. The λ/4 lines provide a way to modulate the load impedance so as to ensure that the amplifiers operate to supply maximum power with the best efficiency.
Doherty amplifiers used today are made with GaN HEMT FETs because of their high-power, high-frequency capability. They operate in the 700 MHz to 3 GHz range. The typical 12 to 20 percent efficiency of a class AB linear amplifier is boosted to the 25 to 40 percent range with the Doherty configuration. Most Doherty amplifiers are combined with the DPD linearization technique described earlier to deliver increased efficiency.