Users normally place a great deal of value in the high linearity of an amplifier and its corresponding low distortion. Audio amplifiers, for example, are quoted with many linearity specifications, such as total harmonic distortion (THD) or 1%, 0.1% or even less. Even the outputs of amplifiers with internal topologies which are inherently nonlinear, such as Class AB, B, C, and D, are filtered and compensated to drive their nonlinearity down to single-digit percentages and sometimes lower.

Where is a non-linear amplifier needed?

There are three primary reasons for needing a non-linear amplifier:

1) the signal of interest has a known, unavoidable curve or nonlinearity of its own, so the amplifier must compensate for it;

2) or, the signal may sometimes become so large that it overloads the input of the next stage. The resulting amplifier must then have a recovery time during as it comes out of saturation, during which no signal-handling is possible. In extreme cases, there may even be permanent physical damage to channel front end due to the signal overload;

3) finally, the signal naturally spans a very wide dynamic amplitude range. This is often the case an RF wireless, radar, or optical signals, which often range over 100 dB and more, but also applies to some test and measurement instrumentation signals.

What are the types of nonlinear amplifiers?

There are three commonly used types, each targeting one of the above characteristics:

1) for signals with a known nonlinearity, such as a thermocouple, the amplifier is designed with a corresponding curve or a series of “breakpoints” at which is gain factor is changed; this is discussed here and here.

2) for the signal with an amplitude which may exceed the range of the channel, a limiting amplifier is used. This amplifier clips and therefore deliberately “clamps” the output to a safe maximum value as the input exceeds the allowed maximum value, via a limiting diode or other circuit technique (therefore, this is also known as a “clipper” or “clipping” amplifier. A hard-limiting amplifier implements this clamping function with a sudden, sharp “knee” (transition) in the transfer function, Figure 1a, while a soft-limiting amplifier does this with a soft knee as the input approaches the maximum allowed value, so as to reduce the unintended distortion and frequency splatter that results from sharply limiting the channel gain, Figure 1b.

 

3) For signals with extremely wide dynamic range, there are two possible approaches:

• ●the simplest is way to use a variable gain amplifier (VGA) or programmable gain amplifier (PGA), see here. These have their gain adjusted automatically via a closed-loop circuit which also measures some signal characteristic, such as its RMS (root mean square) or maximum value, or by direction of a system processor. The VGA/PGA approach works for some applications, but is generally not suitable for signals with more than about 10 dB of dynamic range.
• ●For situations where the dynamic range is greater than that which can handled by the VGA/PGA, the logarithmic (log) amplifier is the solution. This amplifier implements a simple transfer function:

V_OUT = K ln (V_IN) where K is a scaling factor.

Despite the simple equation, it is difficult to build a log amplifier which performs accurately over decades.

How is a log amplifier actually implemented?

There are three basic architectures which may be used to produce log amps, beginning with the basic diode log amp. Each approach makes use of the basic, well-known logarithmic relation between current and voltage in a diode: the voltage across a silicon diode is proportional to the logarithm of the current through it. If a diode is placed in the feedback path of an inverting op-amp, the output voltage will be proportional to the log of the input current, Figure 1.

This basic log amp is only practical to about 40 dB range, due to the non-ideal characteristics of the diode. To extend the log amp to greater ranges, both the successive detection log amp and the “true log amp” are used. Each is based on a series of cascaded log amplifiers arranged in stepped stages, and each stage handles a subset of the total range, with seamless, nearly error-free transition handoffs between stages.