Nyquist Plot

In the study of dynamic systems, understanding the response to different input frequencies is paramount. While tools like Bode plots offer separate views of magnitude and phase, they lack a unified picture. How can we visualize a system's complete frequency response in a single, intuitive map and, more importantly, use that map to predict the stability of a feedback system? The Nyquist plot provides an elegant and powerful answer to this question, serving as a cornerstone of classical control theory for its deep geometric insights into system behavior. This article explores the theory and application of this indispensable tool. In the following sections, we will first uncover its "Principles and Mechanisms," detailing how the plot is constructed and how the mathematical Principle of the Argument gives rise to the famous Nyquist Stability Criterion. We will then explore "Applications and Interdisciplinary Connections," learning to read the plot for nuanced insights into system robustness and discovering its surprising utility in fields far beyond traditional control engineering, from digital systems to electrochemistry.

Imagine you're trying to describe a journey through a landscape. You could make two separate charts: one showing your altitude at each point, and another showing the compass direction you're facing. This is how many traditional methods, like Bode plots, analyze a system's response to different frequencies. They give you a magnitude plot (the altitude) and a phase plot (the direction). But what if you could combine these into a single, elegant map? A bird's-eye view of your entire journey?

This is precisely what a Nyquist plot does. It’s a map of a system's frequency response drawn on the rich landscape of the complex plane.

Let's say we have a linear system, described by its transfer function $H(s)$. When we feed it a pure sinusoidal input of frequency $\omega$, say $\cos(\omega t)$, the system eventually settles into a sinusoidal output of the same frequency. However, the output's amplitude will be scaled and its phase will be shifted. These two effects, the scaling and the shifting, are captured by a single complex number, $H(j\omega)$.

The ​​magnitude​​ of this complex number, $|H(j\omega)|$, tells us the amplitude scaling factor. The ​​phase angle​​, $\angle H(j\omega)$, tells us the phase shift. The Nyquist plot is simply the path, or locus, traced out by the tip of the vector $H(j\omega)$ as the frequency $\omega$ is swept from $0$ to $\infty$. Each point on this curve is a snapshot of the system's complete response at one specific frequency.

For instance, consider the simple system $H(s) = \frac{s+1}{s+2}$.

• At zero frequency ($\omega=0$), the response is $H(j0) = \frac{1}{2}$. This is the DC gain; the plot starts at the point $(\frac{1}{2}, 0)$ on the real axis.
• As the frequency $\omega$ increases towards infinity, the response $H(j\omega) = \frac{1+j\omega}{2+j\omega}$ gets closer and closer to $1$. The journey ends at the point $(1, 0)$.
• In between, the plot traces a perfect semicircle in the upper-half of the complex plane.

This simple curve tells us everything: the gain is always between $\frac{1}{2}$ and $1$, and the phase shift is always positive (a phase lead) but never more than a certain amount. The beauty of the Nyquist plot is its ability to present this wealth of information in a single, geometric picture.

Furthermore, for any physical system we can build (which will have real-valued coefficients in its model), an elegant symmetry emerges. The response to a negative frequency, $H(-j\omega)$, is simply the complex conjugate of the response to the corresponding positive frequency, $H(j\omega)^*$. This means the Nyquist plot for negative frequencies is a perfect mirror image of the positive-frequency plot, reflected across the real axis. This property, known as ​​conjugate symmetry​​, means we only need to compute the plot for $\omega \ge 0$; the other half is free.

A beautiful map is one thing, but can it guide us? The most critical question we can ask about a feedback control system is: Is it stable? Will a small disturbance die out, or will it grow uncontrollably, leading to catastrophic failure?

For a standard negative feedback system with an open-loop transfer function $L(s)$, the closed-loop response is given by $T(s) = \frac{L(s)}{1+L(s)}$. The system's behavior is dictated by the poles of this closed-loop function—the values of $s$ that make its denominator zero. Instability occurs if the characteristic equation, $1+L(s) = 0$, has any solutions in the "danger zone": the right half of the complex $s$-plane.

This is where the Nyquist plot transforms from a descriptive map into a powerful predictive tool. The stability question, "Are there any roots of $1+L(s)=0$ in the right-half plane?", can be answered by looking at the Nyquist plot of just $L(s)$.

The key insight is this: the equation $1+L(s)=0$ is identical to $L(s)=-1$. This means that if, for any frequency $s$ in the right-half plane, the function $L(s)$ happens to take on the value $-1$, our system has an unstable pole. The point $-1+j0$ in the complex plane thus becomes the ​​critical point​​. The geometry of the Nyquist plot relative to this single point holds the secret to the system's stability.

How can a plot that only uses inputs on the imaginary axis (the boundary) tell us about what's happening inside the entire right-half plane? This sounds like magic, but it is the consequence of a profound mathematical theorem called the ​​Principle of the Argument​​.

The intuition is wonderfully simple. Imagine you are walking along a closed path, and somewhere inside your path there is a tree. If you keep your dog on a leash, and the leash is tied to the tree, your dog (representing the value of a function) will run in a circle around the tree as you complete your path. The leash, representing the vector from the tree to your dog, will make one full $360^{\circ}$ turn. If the tree is outside your path, the leash will just wiggle back and forth, but it will have zero net rotation when you return to your starting point.

The Principle of the Argument is the formal version of this. It states that if we trace a closed contour in the $s$-plane (our path) and plot the corresponding values of a function $F(s)$ (the dog's path), the number of times the function's plot encircles the origin is equal to the number of zeros ($Z$) minus the number of poles ($P$) of the function $F(s)$ that lie inside our original contour.

For this principle to work, there's one crucial rule: our path cannot go directly through a zero or a pole. The function must be well-behaved (analytic) on the contour itself.

Now we can assemble all the pieces into the master tool: the ​​Nyquist Stability Criterion​​.

1. ​​The Function:​​ We choose our function to be $F(s) = 1+L(s)$. The zeros ($Z$) of this function are the closed-loop poles we are looking for. The poles ($P$) of this function are the same as the poles of our open-loop system, $L(s)$.

2. ​​The Contour:​​ To check the entire "danger zone," our path in the $s$-plane, the ​​Nyquist contour​​, must enclose the entire right-half plane. It does this by running up the imaginary axis from $-j\infty$ to $+j\infty$ and then taking a giant semicircular detour of infinite radius in the right-half plane to close the loop. If our open-loop function $L(s)$ has a pole right on the imaginary axis (like an integrator at $s=0$), we simply modify the contour to make a tiny semicircular detour around it to satisfy the "no poles on the path" rule.

3. ​​The Plot and The Rule:​​ The Argument Principle tells us about encirclements of the origin by the plot of $1+L(s)$. But an encirclement of the origin by $1+L(s)$ is identical to an encirclement of the point ​​-1​​ by the plot of $L(s)$. This is why $-1$ is the critical point!

Putting it all together, we get the famous Nyquist equation: $Z = N + P$ Here:

• $Z$ is the number of unstable poles in the ​​closed-loop​​ system. For stability, we absolutely require $Z=0$.
• $P$ is the number of unstable poles in the ​​open-loop​​ system $L(s)$, which we are assumed to know.
• $N$ is the net number of ​​counter-clockwise​​ encirclements of the critical point $-1+j0$ by the Nyquist plot of $L(s)$. (Note: $N$ is positive for counter-clockwise encirclements. Some texts define $N$ as clockwise, which changes the formula to $Z = P - N$).

For our closed-loop system to be stable ($Z=0$), the criterion demands that: $N = -P$ This single, powerful equation is the heart of the Nyquist analysis.

Let's see what this means in practice.

Case 1: The Open-Loop System is Stable ($P=0$)

This is the most common scenario. We start with a stable system and use feedback to improve its performance. Since $P=0$, the stability condition becomes $N = -0 = 0$.

This means: ​​If the open-loop system is stable, the closed-loop system is stable if and only if the Nyquist plot of $L(s)$ does not encircle the -1 point.​​

This is beautifully intuitive. The -1 point represents a condition of runaway feedback. If our system's frequency response, its Nyquist plot, keeps a safe distance and never loops around this critical point, the feedback system will remain stable.

Case 2: The Open-Loop System is Unstable ($P>0$)

What about inherently unstable systems, like a fighter jet or a magnetic levitation train? These systems have $P > 0$ unstable poles. This is where the Nyquist criterion reveals its true power, succeeding where simpler methods like Bode plots are insufficient.

Here, the stability condition is $N = -P$. Since $P$ is a positive integer, $N$ must be a negative integer. A negative number of counter-clockwise encirclements is equivalent to a positive number of ​​clockwise​​ encirclements.

This leads to a remarkable, almost paradoxical conclusion: ​​To stabilize an open-loop system with $P$ unstable poles, the Nyquist plot of the feedback system must encircle the critical point $-1$ exactly $P$ times in the clockwise direction!​​

The feedback controller must be designed to make the system's frequency response perform this precise and delicate dance around the critical point. The loop must be "unstable enough" in just the right way to cancel out the inherent instability of the original system. It is through this deep geometric insight that engineers can tame seemingly untamable systems, making rockets fly straight and levitating trains float on their tracks.

We have learned how to draw the Nyquist plot, tracing the journey of a system's transfer function, $L(s)$, through the complex plane as we probe it with frequencies from zero to infinity. But drawing the map is only the first step. The real adventure lies in reading it. This map, it turns out, is a kind of crystal ball. It doesn't just give a simple 'yes' or 'no' for stability; it tells a rich story about the system's character, its hidden vulnerabilities, its strange paradoxes, and its surprising resilience. In this section, we will explore some of these stories and discover that the language of the Nyquist plot is spoken not just by engineers, but by scientists in fields as diverse as digital signal processing and electrochemistry.

Imagine you are captaining a ship, and your Nyquist plot is the map of your journey relative to a treacherous reef located at the point $-1$. A stable system is one whose path does not encircle this reef. The classical measures of stability, gain margin and phase margin, are a bit like asking, "How far am I from the reef when I am sailing due west?" (phase crossover) or "How far am I from the reef when I am a specific distance from port?" (gain crossover). These are useful questions, but they are incomplete. What if the closest you get to the reef is when you are sailing northwest?

A more profound and honest question is: what is the minimum distance your ship ever gets to the reef, regardless of the direction? This is the true measure of safety, or what engineers call ​​robustness​​. On the Nyquist plot, this corresponds to the minimum Euclidean distance from any point on the locus of $L(j\omega)$ to the critical point $-1$. This distance, let's call it $m$, is defined as $m = \inf_{\omega} |1 + L(j\omega)|$. A large $m$ means you have a wide berth; a small $m$ means you are cutting it dangerously close, and any unexpected gust of wind or rogue wave (representing unmodeled dynamics or noise) could send you crashing into the rocks.

This geometric distance has a beautiful physical meaning. It is directly related to the ​​sensitivity function​​, $S(s) = \frac{1}{1+L(s)}$, which tells us how much an external disturbance is amplified by the feedback loop. The peak magnitude of the sensitivity, $\|S\|_{\infty}$, is simply the reciprocal of this minimum distance: $\|S\|_{\infty} = 1/m$. A Nyquist plot that skims past the $-1$ point implies a very small $m$, which means a very large peak sensitivity. Such a system is "twitchy" and fragile; it will overreact to noise and disturbances, even if it is technically stable.

This modern view of robustness reveals the naivety of relying solely on classical margins. Consider a system whose Nyquist plot never crosses the negative real axis for any frequency where its magnitude is greater than or equal to one. According to the classical definition, its gain margin is infinite! This sounds wonderfully robust. However, if the system is what's called "non-minimum-phase," its Nyquist plot might snake around and pass perilously close to the $-1$ point from above or below. The system might have an infinite gain margin and a decent phase margin, yet be extremely fragile, poised on the brink of instability. The minimum distance $m$, however, would be very small, correctly sounding the alarm. The Nyquist plot, when read with this deeper understanding, provides the true story of the system's resilience.

The Nyquist plot is at its most fascinating when it describes systems that don't follow simple, well-behaved rules. It reveals the personality of these "unruly" systems with perfect clarity.

​​The Treachery of "Bad" Zeros:​​ A system's poles describe its natural modes of response, like the notes an instrument can play. Its zeros, on the other hand, relate to how an input "plays" those notes. Most zeros are well-behaved ("minimum-phase"). But some systems possess "non-minimum-phase" zeros, which lie in the right-half of the complex plane. These zeros have a rebellious character. When you give the system a push, it initially lurches in the opposite direction before eventually moving toward its final destination. On the Nyquist plot, this contrary behavior manifests as an extra clockwise twist—an additional, unavoidable phase lag that pushes the curve toward the danger zone of $-1$. Stabilizing a system with these "bad" zeros is like trying to balance a broomstick with a mischievous weight on top that always leans the wrong way first. It's not impossible, but it is vastly more difficult, a fact the Nyquist plot makes geometrically obvious.

​​The Paradox of Conditional Stability:​​ Common sense might suggest that if a system is stable with a certain amount of feedback gain, it should remain stable if we reduce the gain. The Nyquist plot shows this is not always true. A system can be ​​conditionally stable​​: stable for very low gain, unstable for a medium range of gain, and then, miraculously, stable again for very high gain. The Nyquist plot explains this paradox with elegant simplicity. The plot of $L(s)/K$ stays fixed, while the critical point that matters is actually $-1/K$. As we increase the gain $K$ from a small value, the point $-1/K$ moves from far out on the negative real axis towards the origin. It might start outside a loop in the Nyquist plot (stable), then move inside the loop (unstable!), and finally, as $K$ becomes very large, it might exit the loop on the other side, returning the system to stability. The system is capricious: it dislikes a "lukewarm" controller but is happy with one that is either very gentle or very aggressive.

​​Taming the Beast: Stabilizing Unstable Systems:​​ Perhaps the most powerful demonstration of the Nyquist criterion is in stabilizing a system that is inherently unstable to begin with—like balancing a rocket on its column of thrust. For such a system, the open-loop transfer function $L(s)$ has $P$ poles in the right-half plane. The beginner's rule, "the plot must not encircle $-1$," is completely wrong here; in fact, it leads to disaster! The full Nyquist stability criterion, $Z = N + P$, tells us that to achieve stability ($Z=0$), we must have $N = -P$. This means the plot must encircle the critical point $-1$ exactly $P$ times in the clockwise direction. It is as if we must throw a lasso around the $-1$ point to tame the instability and drag the system's poles back into the stable left-half plane. For such a system, a controller that yields a positive phase margin but fails to produce the required encirclements still results in an unstable closed-loop system. The Nyquist plot is the only tool that can reliably guide the design of a controller for such a formidable challenge.

The beauty of a truly fundamental concept is its universality. The Nyquist plot, born from the mathematics of complex functions, is not confined to the domain of analog control systems. Its principles reappear, sometimes in disguise, across a remarkable range of scientific and technological fields.

​​The Digital World:​​ In the age of computers, control is often implemented digitally. The continuous variable $s$ is replaced by the discrete variable $z$, and the stability boundary is no longer the imaginary axis but the unit circle in the complex $z$-plane. Does this new setting require a whole new theory? Not at all. The Nyquist criterion adapts perfectly. The Nyquist plot is now the image of the unit circle, $z=e^{j\omega}$, as $\omega$ sweeps from $-\pi$ to $\pi$. The underlying rule, the Principle of the Argument, remains the same. Stability is still determined by counting encirclements of conundrums, with the number of required encirclements depending on the number of open-loop poles outside the unit circle. It is like translating a great play into another language: the setting and actors have changed, but the dramatic structure and the final outcome are governed by the same timeless rules.

​​The World of Chemistry:​​ The same graphical tool provides profound insights into the molecular world. In ​​Electrochemical Impedance Spectroscopy (EIS)​​, a chemist investigates a battery, fuel cell, or corroding metal by applying a small, oscillating voltage and measuring the resulting current. The ratio of voltage to current gives the impedance, a complex number that varies with frequency. Plotting this impedance in the complex plane for a range of frequencies generates... a Nyquist plot.

The shape of this plot is a fingerprint of the physical and chemical processes occurring at the electrode's surface. For example, a process limited purely by the diffusion of ions through the electrolyte produces a characteristic straight line with a constant phase angle of $-45^\circ$, known as ​​Warburg impedance​​. However, if a chemical reaction is coupled with the diffusion, the story changes. At low frequencies, the reaction has time to reach a steady state, and the plot curves back to the real axis, forming a semicircle. This signature, called ​​Gerischer impedance​​, is a dead giveaway that a reaction is taking place. By analyzing the shape of the Nyquist plot, an electrochemist can build a model of the unseen molecular machinery—quantifying reaction rates, diffusion coefficients, and other fundamental properties—all from a simple electrical measurement.

From balancing rockets to diagnosing batteries, the Nyquist plot provides a unified, geometric language for understanding the dynamic response of complex systems. It is a testament to the remarkable power and elegance of mathematical physics, revealing deep connections between seemingly disparate corners of the scientific world.