Linearity can be an elusive goal. Definitions for linearity are somewhat relative in nature, and a number of operating characteristics can contribute to the quality of an active or passive component's linearity. At the semiconductor level, even the process can be refined to improve linearity performance. Improved linearity of high-frequency components has become of greater importance in recent years because of the increasing use of digital modulation formats, such as quadrature amplitude modulation (QaM), that depend on preserving the amplitude, frequency, phase, or some combination of the three that has been modulated onto the carrier's signal envelope to represent digital information. Understanding how linearity is defined is the first step in appreciating which components and devices can be said to be truly linear.

As the name implies, linearity in its simplest sense has to do with the capability of a component to produce an output signal that is an accurate representation of its input signal, albeit with some amount of gain or loss. If the input signal has an amplitude level that is within 1 dB of some value across a range of frequencies, than the output signal should also be within a similar 1 dB amplitude window, although typically at some higher or lower amplitude level. In the real world, however, transmission of signals through cables, amplifiers, filters, and other components is never quite so routine, especially in modern communications environments where a receiver is subjected to so many different RF/microwave signals from multiple wireless standards and transmissions. Any time two or more signals enter the input of a high-frequency component, there is an opportunity to generate intermodulation or mixing products, which can destroy the ideal "straight-line" transmission characteristics of a component.

Frequency mixers rely on this property of multiple signals to translate an input signal up or down in frequency. In a downconversion mixer, for example, ideally the RF input is a single tone as is the local oscillator (LO) input used to translate the RF signal to a lower-frequency intermediate-frequency (IF) signal. The two tones mix to produce a third tone that is the sum or difference of the first two tones. When more than two input tones are involved, unwanted and sometimes unpredictable (and difficult-to-filter) additional output signal result. Components that are deemed nonlinear in nature can be expected to produce significant levels of intermodulation distortion (IMD).

A number of different parameters are used to compare the linearity performance of components and devices, including the third-order intercept point and the second-order intercept point, referenced to a signal level at either the input or the output of the device under test (DUT). They are terms which relate to the amount of second- and third-order intermodulation products to be found at a DUT's input or output ports, with higher numbers representing lower amounts of intermodulation products.

In comparing different components for linearity performance, the same reference ports should be used. In the case of an amplifier, for example, its third-order intercept point (IP3) might appear on a data sheet referenced to either as the input port, the input third-order intercept point (IIP3), the output third-order intercept point (OIP3), or a combination of the preceeding. Amplifier specifiers typically need to know the amplifier's linear output power level, as well as the highest allowable input level to maintain linear performance. For a mixer, it is the input power that is of concern, and the IIP3 that is typically used to specify for compare mixer linearity. Higher values of IP3 and second-order intercept point, or IP2, mean better linearity.

Such linearity parameters are generally extrapolated and theoretical, especially for large-signal components such as power amplifiers, since the active devices used in mixers (diodes) and amplifiers (transistors) would typically saturate at high power levels. Still, IP3 and IP2 specifications provide a useful means of comparing different products for linearity performance.

Although generally designed for small-signal use, integrated circuits (ICs) are part of a system's linearity budget; such components as analog-to-digital converters (ADCs), transceivers, modulators, and demodulators should be evaluated for linearity. For example, model LT5575 from Linear Technology is a direct-conversion quadrature demodulator with a range of 800 to 2700 MHz. It is designed for working with the in-phase (I) and quadrature (Q) signal components of a QAM signal in receiver applications.

Because it is meant to handle the typically low levels of a receiver signal chain, the LT5575's IIP3 performance is not overly impressive, at +28 dBm at 900 MHz and +22.6 dBm at 1900 MHz, compared to a power amplifier. But when its IIP2 numbers of +54.1 dBm at 900 MHz and +60 dBm at 1900 MHzare compared to other demodulators, it is apparent that this device effectively suppresses second-order IMD within a receiver signal chain in order to maintain good signal sensitivity and minimal distortion of the I and Q states.

On the large-signal side, many factors determine the linearity of an amplifiernotably, its bias configuration. An amplifier circuit in which the active devices are always fed energy, such as a Class A amplifier, will provide much improved linearity compared to an amplifier with a switching power supply, such as a Class E design. Obviously, the essential tradeoff for amplifier linearity is a sacrifice in efficiency. The "always-on" Class A amplifier is much less efficient and consumes a great deal more power for a given output-power level than a Class E amplifier which is switched on and off according to the needs of the input signal. In addition, a technique known as "backing off" the power supply has been applied to amplifiers to run the devices at reduced bias levels in order to improve their linearity performance, although this also degrades efficiency.

Amplifiers are also evaluated for linearity using the carrier-to-intermodulation (C/I) ratio, which applies multiple input tones and measures the ratio of the desired output to the IMD products at the output; the adjacent-channel power ratio (ACPR), a measure of how much energy is produced outside a desired frequency band; and error vector magnitude (EVM), which shows distortion as signal vectors, such as I and Q vectors.

In designing a power amplifier for high linearity, there is no "secret transistor type" that can deliver unmatched levels of IP3 and IP2 performance. Many different types of solid-state devices have been developed in the last few decades in search of the ultimate transistor (see p. 98). But it is more typically the substrate material that has more to do with linearity performance than the device structure. High-frequency devices were long based on silicon, and later on gallium arsenide (GaAs).

In recent years, development of alternative epitaxial materials, such as silicon carbide (SiC) and gallium nitride (GaN) have given device designers opportunities to apply proven transistor structures, such as heterojunction bipolar transistors (HBTs) and high electron mobility transistors (HEMTs), to these high-performance materials with good results. SiC and GaN substrates have already yielded some excellent high-power devices with respectable linearity performance.

Recently, encouraged by the linearity performance it was seeing from its initial GaN process, RF Micro Devices modified its High Power GaN 1 process for higher linearity. The new process, the High Power GaN 2 process, is based on GaN epitaxial materials on SiC substrates, forming 0.5-m gate-length HEMT devices with +48-VDC CW operating voltages and +65-VDC pulsed operating voltages. The devices feature extremely high power densities to 8 W/mm of device periphery, and the process is said to deliver about 6-dB better linearity.