Why isn't distortion much of a concern in RF amplifiers?

Question

Class C and class E power amplifiers both distort the waveform yet they are preferred in RF applications. Why? I mean if the signal is distorted what will happen to the information which was in the signal? Although it seems that the frequency of the output signal is the same , so is it only frequency modulated signals that can be given as an input?

Answer 1

Welcome to EE.SE.

These amplifier classes are used in narrow band RF amplifiers with a tuned circuit on the output, known as a tank circuit. Although the active device only puts energy into the tank circuit over part of the RF cycle the tank circuit reconstitutes output over the full RF cycle.

They are used because they are efficient (in the engineering sense of RF power output/DC power consumed from the supply).

This does mean that these amplifier classes are limited to narrow band operation. They also have non-linear amplification so are typically used with constant amplitude signals.

A broadband RF amplifier will need to run in class A or, in push-pull operation, class AB or B and will typically have a transformer output for matching to the external load rather than a tank circuit. These amplifier classes can also offer linear amplification of variable amplitude signals.

Some further reading

Answer 2

Two remarks:

1. They do distort the waveform, but that only means that they add harmonics to it at multiples of the carrier frequency. And that is not very important. They can be filtered out, and the matching network and antenna will not transmit those harmonics very efficiently anyway.
2. Unfortunately, however, they also distort the modulation, i.e. the amplitude envelope of the waveform if AM is used, or both the amplitude and phase if it is a more complicated modulation scheme.
3. For that reason, they are not preferred, at least by far not in all cases. Often a slightly less efficient alternative like class AB is preferred. Or a more complicated but more linear and still efficient alternative, like the Doherty.

And while this distortion of the modulation, at frequencies close to the carrier, is of course large for these class C and E amplifiers, it still can to some extent be overcome by digital predistortion of the transmit signal before it enters the amplifier chain.

In the end it depends on the modulation scheme what class can be used. (For 1024 QAM it will definitely not be class C or E!)

Answer 3

Distortion, yes, but what distortion, and for what modulation?

Not all modulation schemes are susceptible to harmonic or intermodulation distortion!

And for those that are, there are ways around it.

Consider a harmonic distortion of a variable-frequency signal: the harmonics are simply filtered off (with sufficiently sharp filters -- we might need to cut the 2nd and 3rd harmonic quite aggressively, say >60dB within an octave), and the fundamental is the fundamental, no problem. Distortion does not affect the base frequency of the signal.

This permits FM and PM schemes to use class C and E amplifiers with no special considerations (beyond the filter).

With AM, we can cheat: we can drive the carrier into a class C amplifier, and vary its supply voltage instead, to vary output amplitude proportionally. Indeed it's advantageous that the amplifier output be saturated (clipping), so that amplitude varies proportionally with supply. Thus we get the carrier at high efficiency, and the signal power at a modest compromise: typically from a class AB/B amplifier, but if the signal is low bandwidth (as is the case for commercial AM broadcast for example), we can even use a high-efficiency class D amp as well!

Class D can also be used for direct AM output: the sidebands around Fsw are given by the variation in duty cycle, while higher harmonics and DC are filtered off. (Basically, the over-the-air case of symmetrical PWM as used in full-wave forward converters. A forward converter is transformer-coupled, so cannot be driven with LF/DC as from an asymmetric/complementary PWM waveform. The forward converter even rectifies its output to DC: basically a power AM detector.) Some pre-correction may be necessary to account for the duty cycle-to-carrier transfer function (it goes as sin(πd/2), I think?*), or other effects (due to precise timing of the circuit, filter response, etc.), but the amplifier itself is fine.

*I can't actually find a reference for this offhand, namely the Fourier series of such a square wave. Internet search has never been particularly good at finding reference data like this, but I'm shocked that WolframAlpha doesn't even understand it. I can solve it by hand but it isn't worth the effort just to make an example point. Go figure...

In general, we can describe a signal x(t) in the Hilbert domain, consisting of some phase and amplitude, evolving over time: \$a(t) \left( e^{i \phi(t)} + e^{-i \phi(t)} \right)\$, where \$\phi(t)\$ is a strictly increasing function and \$a(t) > 0\$. (Or perhaps a should be real, to avoid step changes in φ.) The phase rate \$d\phi(t)/dt = \omega(t)\$ is the instantaneous frequency; for a sine wave, it's constant, but introducing a signal via phase or frequency modulation adds that signal to \$\omega\$ directly.

Put another way: consider a QAM signal. We can plot the I and Q coordinates of the signal (with respect to some average carrier frequency or external reference tone); and for example, we might impose a grid on this coordinate system to represent code symbols of a digital signal. We can also plot this with polar coordinates, \$(I, Q) = (A cos \Phi, A sin \Phi)\$, and then A might be controlled by supply voltage or PWM or somesuch, and Φ by frequency or phase shift of the carrier.

Since these are equivalent (a transform exists), it is a general method that can be applied to any signal; there may be practical issues remaining, like say the amplifier's tuning or bandwidth varies with supply voltage, or the amplitude transfer function may be distorted or cut off at one or both extremes, or sloped inbetween (limited power range, linearity), etc. But these are simply equivalent expressions of amplifier nonlinearity, subject to optimization and tuning as for any other architecture.