Antilogarithmic Amplifier

In the world of analog electronics, amplifiers are typically associated with linear scaling—multiplying a signal by a constant factor. However, a more specialized and powerful class of circuit exists: the antilogarithmic amplifier. This device performs a fundamentally non-linear operation, generating an output that is an exponential function of its input. While this might seem like a niche mathematical curiosity, it is in fact a key building block for sophisticated analog systems. The central challenge this article addresses is understanding how this non-linear behavior is achieved, controlled, and ingeniously applied to solve complex problems without digital processors.

This article demystifies the antilogarithmic amplifier. In the first chapter, 'Principles and Mechanisms', we will explore the elegant circuit design that inverts the function of a logarithmic amplifier, delve into the physics of diodes and transistors that make it possible, and examine how to build practical, stable circuits by mitigating temperature effects and non-ideal component behavior. Subsequently, in 'Applications and Interdisciplinary Connections', we will uncover how these amplifiers become the heart of analog computers, performing multiplication, division, and root extraction, and how they are used to generate complex functions and control dynamic signals in fields ranging from video processing to audio synthesis. We begin by examining the core principles that allow a simple arrangement of components to compute a complex mathematical function.

Now that we have a feel for what antilogarithmic amplifiers do, let's take a look under the hood. How does one persuade a collection of transistors, resistors, and operational amplifiers (op-amps) to compute something as elegant as an exponential function? The answer, as is so often the case in physics and engineering, lies in a beautiful symmetry and a clever exploitation of the natural behavior of our components.

Imagine you have a machine that can perform a specific mathematical operation, say, calculating the logarithm of a number. How would you build a machine that does the exact opposite—an "un-logger" or an exponentiator? The most direct path is often to run the first machine in reverse. In electronics, this "reversal" has a wonderfully tangible meaning.

Consider a basic logarithmic amplifier. It typically consists of an input resistor, $R_{in}$, and a Bipolar Junction Transistor (BJT) placed in the feedback path of an op-amp. The input voltage, $V_{in}$, drives a current $I = V_{in} / R_{in}$ through the resistor. The op-amp, in its clever way, forces this current to flow through the BJT. Now, a BJT has a wonderful, inherent physical property: the voltage across it ($V_{out}$ in this case) is proportional to the logarithm of the current flowing through it. Voila, we have a log amplifier.

So, how do we build the inverse? We simply swap the roles of the linear and nonlinear components. We take the BJT (or a similar nonlinear device like a diode) and move it to the input. We take the resistor (our linear component) and place it in the feedback path. Now, the input voltage is directly applied to the BJT, which generates a current that is exponentially related to the input voltage. The op-amp ensures this exponential current flows through the feedback resistor, generating an output voltage that is a scaled version of that current. By simply swapping two components, we have inverted the circuit's mathematical function from logarithmic to antilogarithmic. It is a stunningly simple and powerful illustration of functional inversion embodied in physical hardware.

Let's build the simplest possible version of this idea using a diode. The setup is just as we described: a diode at the input, a resistor $R_f$ in the feedback loop of an op-amp, and the non-inverting input of the op-amp connected to ground.

Here’s how the magic happens. The op-amp works tirelessly to keep its inverting input at the same potential as its non-inverting input. Since the non-inverting input is grounded (0 V), the inverting input becomes a ​​virtual ground​​. This means the input voltage $V_{in}$ is applied directly across the diode.

The fundamental physics of a semiconductor diode dictates that the current flowing through it, $I_D$, is exponentially dependent on the voltage across it, $V_D$. This relationship is described by the Shockley diode equation:

$I_D = I_s \left(\exp\left(\frac{V_D}{V_T}\right) - 1\right)$

Here, $I_s$ is the tiny reverse saturation current (a property of the specific diode) and $V_T$ is the ​​thermal voltage​​ (about $26 \text{ mV}$ at room temperature), a physical parameter that is directly proportional to absolute temperature. Since $V_D = V_{in}$, the diode current is an exponential function of our input voltage.

Because no current flows into the ideal op-amp's input, all of this diode current, $I_D$, has nowhere to go but through the feedback resistor $R_f$. The voltage across the resistor is, by Ohm's Law, $I_D \times R_f$. Since one end of the resistor is at the virtual ground (0 V) and the other is at the output terminal, $V_{out}$, we find that $V_{out} = -I_D R_f$.

Putting it all together, the transfer function of our circuit is:

$V_{out} = -R_f I_s \left(\exp\left(\frac{V_{in}}{V_T}\right) - 1\right)$

For any positive input voltage where $V_{in}$ is much larger than $V_T$, the exponential term dominates, and we get the desired behavior: $V_{out} \approx -R_f I_s \exp(V_{in}/V_T)$. The output voltage is indeed an exponential function of the input! The feedback resistor $R_f$ simply acts as a scaling factor, converting the exponential current into a voltage of the desired magnitude. If we use a BJT instead of a diode, or a PNP transistor instead of an NPN, the core principle remains the same, though the exact signs and parameters in the equation might change.

And what if $V_{in}$ is negative? The exponential term $\exp(V_{in}/V_T)$ becomes nearly zero, and the output voltage becomes a tiny, constant positive value: $V_{out} \approx R_f I_s$. The circuit effectively "shuts off" for negative inputs, a characteristic feature of this simple design.

Here is where we see the true elegance of this approach. What happens if we take a log amplifier and feed its output directly into an antilog amplifier? We are performing an operation and then immediately performing its inverse. Intuitively, we should get back what we started with. Let's see if the electronics agree.

​​Stage 1: Log Amplifier​​ $V_{mid} = -V_T \ln\left(\frac{V_{in}}{R_1 I_s}\right)$

​​Stage 2: Antilog Amplifier​​ (taking $V_{mid}$ as its input) $V_{out} = -R_f I_s \exp\left(-\frac{V_{mid}}{V_T}\right)$

Now, we substitute the expression for $V_{mid}$ into the second equation. The term in the exponent becomes:

$-\frac{V_{mid}}{V_T} = -\frac{1}{V_T} \left[-V_T \ln\left(\frac{V_{in}}{R_1 I_s}\right)\right] = \ln\left(\frac{V_{in}}{R_1 I_s}\right)$

The exponential function and the natural logarithm are perfect inverses. They annihilate each other!

$\exp\left(-\frac{V_{mid}}{V_T}\right) = \exp\left[\ln\left(\frac{V_{in}}{R_1 I_s}\right)\right] = \frac{V_{in}}{R_1 I_s}$

Substituting this back into the equation for $V_{out}$:

$V_{out} = -R_f I_s \left(\frac{V_{in}}{R_1 I_s}\right) = -\frac{R_f}{R_1} V_{in}$

This is a spectacular result. By cascading two nonlinear circuits, we have created a perfectly ​​linear​​ amplifier. More importantly, notice that the pesky, temperature-sensitive terms $V_T$ and $I_s$ have completely vanished! The final behavior depends only on the ratio of two resistors, $R_f$ and $R_1$, which can be manufactured with high precision and stability. This principle—using inverse functions to cancel out component non-idealities and temperature dependencies—is a cornerstone of high-precision analog circuit design.

The simple diode-based circuit we first analyzed has a major practical flaw: its output is directly proportional to $I_s$, the reverse saturation current. This parameter is not only tiny and difficult to control during manufacturing, but it also changes dramatically with temperature. A circuit whose behavior depends so strongly on $I_s$ is too unstable for most real-world applications.

The solution, once again, is one of ratio and cancellation. Instead of relying on a single device, practical antilog amplifiers use a matched pair of transistors and a stable ​​reference current​​, $I_{ref}$. The idea is to create an output that is proportional to the ratio of two currents, which cancels out the common $I_s$ term.

A stable reference current is generated using its own op-amp circuit. A highly stable DC voltage source, $V_{ref}$, is connected through a precision resistor, $R_{ref}$, to the inverting input of an op-amp. This op-amp forces the collector current of one of the transistors to be exactly equal to the current flowing through the resistor. Because of the virtual ground at the op-amp's input, this current is simply $I_{ref} = V_{ref} / R_{ref}$. Since $V_{ref}$ and $R_{ref}$ can be made very stable, $I_{ref}$ is a reliable, temperature-independent anchor. The final output of the antilog amplifier then becomes proportional to $I_{ref} \exp(V_{in}/V_T)$, a much more robust and predictable relationship.

Furthermore, we can tailor the circuit's response with great flexibility. As we've seen, the feedback resistor $R_f$ acts as a linear gain control for the output voltage. But we can be even more subtle. By placing a voltage-divider or an inverting amplifier stage before the antilog circuit, we can scale the input voltage $V_{in}$ itself. This allows us to modify the exponent in the transfer function, for example, changing the response from $\exp(-V_{in}/V_T)$ to $\exp(-V_{in}/(2V_T))$ by simply halving a resistor value in the input stage.

Of course, our components are never truly ideal. The op-amp, the heart of our circuit, has its own subtle imperfections. Understanding these "ghosts in the machine" is crucial for precision design. Let's look at two: input offset voltage and input bias current.

An ​​input offset voltage​​ ($V_{OS}$) is a tiny, residual voltage difference that exists between the op-amp's inputs even when it should be zero. You might guess this would add a small DC error to our output. But in an exponential circuit, the effect is far more insidious. The offset voltage $V_{OS}$ adds directly to the voltage at the inverting input, which means the voltage across our BJT becomes $V_{in} - V_{OS}$. This term appears inside the exponential! The result is that the entire output is multiplied by a scaling factor:

$M = \exp\left(-\frac{V_{OS}}{V_T}\right)$

Because the thermal voltage $V_T$ is so small (about $26 \text{ mV}$), even a microvolt-level offset voltage can create a noticeable ​​multiplicative (gain) error​​. A $1 \text{ mV}$ offset, for instance, would cause a gain error of about $4\%$!

An ​​input bias current​​ ($I_B$) is a small current that must flow into the op-amp's input terminals for its internal transistors to operate. In our antilog circuit, this current gets added to the BJT's collector current at the summing junction (the inverting input). This current then flows through the feedback resistor $R_f$, producing an error voltage. Unlike the offset voltage, the bias current results in a simple ​​additive (offset) error​​ at the output:

$\Delta V_{out} = R_f I_B$

This reveals a fascinating duality: two different non-idealities of the op-amp manifest in completely different ways in the final output of our nonlinear circuit. One creates a gain error, the other an offset error. It is by understanding these subtle but fundamental mechanisms that an engineer transforms a theoretical concept into a working, high-precision instrument.

Now that we have seen the inner workings of the antilogarithmic amplifier, we might be tempted to ask, "What is it good for?" It seems like a rather specialized tool, producing an output that grows with dizzying speed. A linear amplifier scales a voltage; we understand that intuitively. But an exponential one? The answer, it turns out, is that when you pair this curious device with its inverse, the logarithmic amplifier, you unlock the door to a world of analog computation that is as elegant as it is powerful. You essentially build an electronic slide rule, capable of performing sophisticated mathematics not with gears and sliders, but with flowing currents and voltages.

Let's begin with a wonderfully simple and profound idea. In school, you learned the rules of logarithms: adding logs is like multiplying the original numbers, and subtracting logs is like dividing them. What if we could teach a circuit to do this?

Imagine we have two voltages, $V_1$ and $V_2$, and we want to find their ratio, $V_1/V_2$, without a digital computer. We can perform a beautiful three-step electronic waltz. First, we feed $V_1$ and $V_2$ into separate logarithmic amplifiers. Out come two new voltages, proportional to $\ln(V_1)$ and $\ln(V_2)$. Second, we send these two logarithmic signals into a simple difference amplifier, which, as its name suggests, subtracts its inputs. The output of this stage is now a voltage proportional to $\ln(V_1) - \ln(V_2)$, which is, of course, $\ln(V_1/V_2)$. The final, magical step is to take this resulting voltage and feed it into an antilogarithmic amplifier. The antilog function is the inverse of the logarithm, so it "un-logs" the signal, leaving us with a voltage directly proportional to the original ratio, $V_1/V_2$.

This log-operate-antilog recipe is astonishingly versatile. If we had used a summing amplifier in the middle step instead of a subtractor, we would have computed the product $V_1 V_2$. What if we summed the logs and then passed the result through an attenuator that halves the voltage before the antilog stage? We would be calculating $\exp(\frac{1}{2}(\ln(V_1) + \ln(V_2)))$, which simplifies to the geometric mean of the two inputs, $\sqrt{V_1 V_2}$. In an instant, with a handful of components, we have built a circuit that can multiply, divide, and find roots—a true analog computer.

The real artistry begins when we realize that the "operate" step in our log-operate-antilog sandwich doesn't have to be a simple sum or difference. Any linear operation we perform on the logarithmic signal will be transformed into a non-linear operation on the original signal after the antilog stage.

Suppose we take an input voltage $V_{in}$, convert it to $\ln(V_{in})$, and then simply pass it through an amplifier (or even a simple voltage divider) that scales it by a factor $\alpha$. The signal entering the final antilog stage is now $\alpha \ln(V_{in})$, which is equivalent to $\ln(V_{in}^{\alpha})$. The antilog amplifier then dutifully undoes the logarithm, and what emerges is a voltage proportional to $V_{in}^{\alpha}$. By simply changing the gain of that middle amplifier—perhaps with nothing more than the turn of a knob on a potentiometer—we can create a circuit that computes the square, the cube, the square root, or the cube root of the input voltage.

This is not just a mathematical curiosity. This very technique, known as "gamma correction," is fundamental to how we see images on screens. The way a cathode-ray tube (and many modern displays) translates voltage into brightness is inherently non-linear. Our eyes also perceive brightness non-linearly. To make the image on the screen appear natural, the video signal must be pre-corrected by applying a power-law function—a feat perfectly suited for a log-antilog circuit.

But why stop at simple powers? With enough ingenuity, we can synthesize far more complex and important functions. Imagine we want to build a circuit that produces a Gaussian curve, $V_{out} = A \exp(-B V_{in}^2)$, a shape that appears everywhere from statistics to quantum mechanics. It seems like a daunting task. Yet, it can be achieved by cleverly nesting our building blocks. First, we build a sub-circuit to compute $V_{in}^2$ using the power-law method. Then, we take the output of that circuit (which is now proportional to $-V_{in}^2$) and feed it into a final, standalone antilogarithmic amplifier. The result is a beautiful, clean Gaussian function, sculpted entirely from the exponential characteristic of the humble diode or transistor.

So far, we have mostly imagined our inputs as steady DC voltages. But in the real world, signals are dynamic; they wiggle and wave. How does an antilog amplifier treat a small AC signal, like a sound wave, that is riding on top of a larger DC voltage?

Because the amplifier's transfer function is a curve, not a straight line, its "gain" isn't constant. The steepness of the exponential curve changes at every point. This means that the amplification given to a small AC signal depends entirely on the DC level it's sitting on. If the DC voltage is high, the curve is very steep, and the small AC signal gets a huge boost. If the DC voltage is low, the curve is flatter, and the AC signal is amplified much less. This makes the antilog amplifier a natural voltage-controlled amplifier (VCA), a cornerstone of electronic music synthesizers and automatic gain control (AGC) systems in radio receivers.

However, this same non-linearity that gives us such powerful computational and control abilities can also be a menace. When you pass a pure sine wave through a perfectly linear system, you get a pure sine wave out. But when you pass it through a non-linear system like an antilog amplifier, the output is not just a scaled version of the input. The non-linearity distorts the wave, creating new frequencies—harmonics—that weren't there before. This phenomenon is called Total Harmonic Distortion (THD), and it's a critical measure of signal fidelity in audio systems and communication channels. The amount of distortion produced is, just like the small-signal gain, highly dependent on the DC operating point. An engineer using these circuits is therefore always engaged in a delicate balancing act: harnessing the non-linearity for computation and control, while simultaneously taming it to preserve the integrity of the signal.

From performing arithmetic to sculpting the fundamental curves of science and controlling the dynamics of audio signals, the antilogarithmic amplifier is far more than a niche component. It is a testament to the profound possibilities that arise when we master the fundamental physical laws governing our electronic components and combine them with the timeless rules of mathematics.