Transfer Function Zeros

In the study of dynamic systems, the transfer function serves as a fundamental mathematical recipe, describing how a system transforms an input into an output. Much attention is given to the poles of this function—the roots of the denominator—which dictate the system's natural rhythms and stability. However, the roots of the numerator, known as ​​zeros​​, are equally crucial yet often more enigmatic. While poles define the frequencies a system naturally resonates with, zeros define the frequencies it actively seeks to silence. This article addresses the role and significance of these "anti-resonances," exploring the often counter-intuitive ways they shape system behavior.

This exploration will unfold across two main sections. First, in "Principles and Mechanisms," we will demystify what zeros are, examining their mathematical definition and their physical origins in mechanical and electrical systems. We will uncover how their location in the complex plane dictates critical aspects of the time response, such as the peculiar phenomenon of undershoot. Following this, the "Applications and Interdisciplinary Connections" section will showcase how engineers harness the power of zeros to sculpt signals, design precision filters, and implement advanced control strategies, revealing their unifying role across diverse fields from electronics to biology.

In our journey to understand the world through the language of mathematics, we often use a powerful tool called the ​​transfer function​​. Think of it as a system's recipe: it tells you exactly what output you'll get for any input you care to provide. This recipe is written as a fraction, $G(s) = \frac{N(s)}{D(s)}$, a ratio of two polynomials. The roots of the denominator, $D(s)$, are the famous ​​poles​​, which you can think of as the system's natural rhythms or resonances. They dictate the stability and the smooth, flowing character of the system's response.

But what about the numerator, $N(s)$? Its roots, called ​​zeros​​, are just as important, though perhaps more mysterious. If poles are the frequencies where a system wants to sing, zeros are the frequencies it wants to silence. They are the keys to understanding how a system can block, shape, and transform signals in ways that are both powerful and sometimes deeply counter-intuitive.

At its heart, a ​​finite zero​​ of a transfer function is a complex frequency, let's call it $s_z$, where the output of the system is zero, even if the input is not. Mathematically, it’s simply a root of the numerator polynomial, $N(s_z)=0$. For a simple transfer function like $G(s) = \frac{3s+1}{s^3+2s^2+s}$, finding the zero is straightforward algebra: the numerator $3s+1$ is zero when $s = -1/3$. This system has one finite zero at $s=-1/3$, alongside its three finite poles at $s=0$ and $s=-1$ (a double pole).

But this is just math. Where do these zeros come from in the real world? Why do some systems have them and others don't?

Let's look at a physical system, like a delicate instrument on a vibration-isolation platform. The platform is a classic mass-spring-damper system. If the floor vibrates (the input), the instrument itself moves (the output). The transfer function relating the floor's motion to the instrument's motion turns out to be $H(s) = \frac{bs+k}{ms^2+bs+k}$. The denominator, the famous characteristic equation $ms^2+bs+k=0$, gives the poles—the system's natural tendency to oscillate and decay. But look at the numerator, $bs+k$. It’s not just a constant! It has a root at $s = -k/b$. This zero arises from the physics of how the input forces are transmitted. The total force on the mass depends on a combination of the damping force (proportional to velocity, hence the $s$ term) and the spring force (proportional to position, hence the constant $k$). At the specific complex frequency $s = -k/b$, these two effects perfectly conspire to cancel each other out, so no net force from the ground motion is transmitted to the mass, and the output is zero.

The existence of a zero isn't just about the components in a system, but also about what you choose to measure as the output. Consider a simple series RLC circuit. Let's send a voltage in and see what comes out.

First, let's measure the voltage across the capacitor. This setup acts as a low-pass filter, and its transfer function is $H(s) = \frac{1}{s^2LC + sRC + 1}$. The numerator is just the number 1. It has no roots. This system has ​​no finite zeros​​.

Now, let's perform a simple change. In the exact same circuit, let's move our measurement probe and look at the voltage across the inductor instead. The circuit's internal workings haven't changed one bit. Yet, the transfer function becomes dramatically different: $H(s) = \frac{s^2LC}{s^2LC+sRC+1}$. Suddenly, the numerator is $s^2LC$, which has a ​​double zero at the origin​​ ($s=0$). Why the dramatic appearance of two zeros? At zero frequency (DC), the capacitor acts as an open circuit, blocking all current flow. With no current, the voltage across the inductor (which is proportional to the change in current) must be zero. The physics of the circuit creates a "transmission block" at $s=0$, and the mathematics reflects this with a zero. The fact that it's a double zero tells us this blocking effect is particularly strong. This powerful comparison shows us that zeros are not just abstract properties but are intimately tied to the path a signal takes from input to output.

This "blocking" property is not just a curiosity; it's a phenomenally useful engineering tool. Imagine you are building a sensitive audio device, but the electrical wiring in the building is producing an annoying 60 Hz hum in your signal. How do you get rid of it? You build a filter designed to annihilate that one specific frequency.

A sustained sinusoidal oscillation at a frequency $f$ (in Hz) corresponds to the pair of points $s = \pm j\omega = \pm j2\pi f$ on the imaginary axis of the complex plane. To completely block a signal at this frequency, we need our system's transfer function, $H(s)$, to be zero at these points. So, we design a filter that has zeros precisely at the locations corresponding to the unwanted noise. For the 60 Hz hum, the angular frequency is $\omega = 2\pi \times 60 = 120\pi$ rad/s. By placing a pair of zeros in our filter's transfer function at $s = +j120\pi$ and $s = -j120\pi$, we create a "notch" in the frequency response. The filter will be transparent to other frequencies, but when the 60 Hz signal comes along, the filter's output at that frequency is zero. The hum vanishes. This is the principle behind the notch filters used everywhere from audio engineering to medical devices.

The location of zeros does more than just determine which frequencies are blocked. Their position in the complex plane has a profound and sometimes startling effect on the system's behavior over time. The complex plane is divided by the imaginary axis into two halves: the Left-Half Plane (LHP), where real parts are negative, and the Right-Half Plane (RHP), where real parts are positive. For poles, this division is a matter of life and death: poles in the LHP lead to stable systems, while poles in the RHP lead to instability.

What about zeros? A system is called ​​minimum phase​​ if all of its finite zeros lie in the LHP. If even one zero wanders into the RHP, the system is branded ​​non-minimum phase​​. This isn't about stability—a system can be perfectly stable with RHP zeros. Instead, it's about the "character" of the response.

RHP zeros have a strange effect on a system's phase, adding extra lag compared to a minimum-phase system with the same magnitude response. This extra phase lag translates into one of the most peculiar behaviors in dynamics: ​​inverse response​​ or ​​undershoot​​. Imagine you're steering a large ship. You turn the rudder to starboard (right), but the ship's bow first swings slightly to port (left) before beginning its long, slow turn to the right. That initial wrong-way movement is the signature of a non-minimum phase system. It happens because the RHP zero creates a conflict between a fast, initial response pushing the output in one direction and a slower, dominant response that eventually pushes it in the intended direction. For a system described by $G(s) = \frac{s^2 - s + 2}{s^2 + s + 2}$, the zeros are at $s = \frac{1}{2} \pm j\frac{\sqrt{7}}{2}$. Because their real part is positive, they lie in the RHP, and this system will exhibit that strange undershoot when given a step input.

So far, we've only talked about finite zeros. But what happens at extreme frequencies? What happens as $s$ flies off towards infinity? This behavior is governed by ​​zeros at infinity​​.

A transfer function like $G(s) = \frac{3s + 5}{2s^3 + 8s^2 + 7s + 1}$ is called "strictly proper" because the degree of the denominator polynomial ($n=3$) is greater than the degree of the numerator polynomial ($m=1$). This difference, $n-m$, tells us the number of zeros at infinity. In this case, there are $3-1=2$ zeros at infinity.

Each zero at infinity represents a pathway for high-frequency signals to be attenuated. A system with one zero at infinity will have its gain decrease proportionally to frequency at high frequencies. A system with two zeros at infinity will have its gain decrease with the square of the frequency, a much faster "roll-off." This is the very essence of a low-pass filter: it lets low frequencies pass but uses its zeros at infinity to squash high frequencies.

We've operated on a simple, powerful rule: poles are the roots of the denominator, and zeros are the roots of the numerator. But there is a crucial piece of fine print, a subtlety that reveals a deeper truth about the nature of systems. What if the numerator and denominator share a common factor? For instance, what if we have a transfer function like $G(s) = \frac{s-a}{(s-a)(s-b)}$?

Our first instinct is to cancel the $(s-a)$ term and declare that the system is simply $G(s) = \frac{1}{s-b}$, with one pole at $s=b$ and no zeros. This is mathematically correct for the input-output mapping. However, the cancelled factor $(s-a)$ represents a physical reality: a hidden "mode" within the system's internal machinery that is either disconnected from the input (​​uncontrollable​​) or invisible to the output (​​unobservable​​).

The most robust and fundamental definitions of poles and zeros come from the system's ​​state-space representation​​, a more detailed model of the internal dynamics. In this view, the poles are the eigenvalues of the system matrix $A$ for a minimal realization (one with no hidden modes), and the zeros are the frequencies where the entire system matrix loses rank, representing a fundamental blockage.

The process of canceling common factors in the transfer function is not just algebraic tidiness. It is the very procedure that guarantees we are analyzing the minimal, essential system. It strips away the hidden, decoupled parts to reveal the true input-output DNA. So, the rule is always to work with the ​​coprime​​ or "reduced" fraction. This ensures that every zero you find corresponds to a genuine transmission-blocking property of the system, not an algebraic ghost of a hidden, irrelevant part. Zeros, then, are more than just roots of a polynomial; they are fundamental descriptors of how a system transmits, or refuses to transmit, information about the world.

Having understood the principles of what transfer function zeros are, we can now embark on a more exciting journey: discovering what they do. If poles describe a system's natural tendencies—its inherent rhythms and modes of decay, like the notes a guitar string loves to sing—then zeros represent the opposite. Zeros are the system's "anti-resonances." They are the frequencies a system actively rejects, the notes it refuses to play. They are the mathematical signature of a system's ability to block, to nullify, and to shape its response in often surprising ways. This simple concept unlocks a profound understanding of phenomena across engineering, physics, and even biology.

Perhaps the most intuitive application of zeros is in the art of filtering. Imagine you have a signal contaminated with an unwanted frequency—a persistent 60 Hz hum from power lines, for instance. How do you get rid of it? You design a filter that has a zero placed precisely at that frequency.

A beautiful and straightforward example comes from digital signal processing. A simple digital filter, known as a Finite Impulse Response (FIR) filter, might have a transfer function like $H(z) = \frac{1}{4} + \frac{1}{2}z^{-1} + \frac{1}{4}z^{-2}$. This seemingly innocuous equation hides a powerful capability. With a bit of algebra, one can find that this filter has a "double zero" at the location $z=-1$ in the complex plane. For a digital system, this location corresponds to the highest possible frequency (the Nyquist frequency). This means the filter will completely annihilate any signal content at that frequency, acting as a perfect blocker. Many digital filters are, in essence, just carefully crafted polynomials whose roots—the zeros—are placed at frequencies we wish to eliminate.

This idea is not limited to the digital realm. In analog electronics, we can achieve the same feat with remarkable elegance. Consider a versatile circuit known as a state-variable filter. It can simultaneously provide low-pass, high-pass, and band-pass outputs from a single input. These outputs all share the same fundamental dynamics, meaning they have the same poles. However, by tapping the output voltage from different points in the circuit, we get different numerators in our transfer functions, and thus, different zeros. An ingenious application is to create a "notch" filter by simply adding the high-pass and low-pass outputs together. By carefully choosing the mixing ratio, we can position a pair of zeros right on the imaginary axis, for instance at $s = \pm j\omega_0$. The resulting filter has a frequency response that is flat almost everywhere, except for a sharp, deep "notch" at the frequency $\omega_0$, where it completely blocks the signal. This is like a sculptor precisely chipping away a single, undesirable sliver from a block of marble.

Sometimes, nature itself creates such filters. When a signal travels from a source to a receiver, it can take multiple paths. The main signal might arrive directly, while a secondary signal bounces off a nearby object, arriving slightly later as an echo. This time-domain phenomenon has a dramatic consequence in the frequency domain. The combination of the signal and its delayed echo creates a transfer function with a factor of $(1 - e^{-sT})$, where $T$ is the delay. This expression has an infinite number of zeros, all lined up periodically on the imaginary axis! This creates a "comb filter," which nullifies a whole series of frequencies. This is why an echo in a room can "color" the sound, or why "ghosting" on an old analog TV signal could distort the picture—you are witnessing the effect of zeros created by a multipath channel.

In the world of feedback control, zeros take on a more active, dynamic role. They are not just for passively blocking signals but are used to proactively guide a system's behavior, making it faster, more stable, and more precise.

Consider the task of designing a controller that makes a system respond quickly to changes. A classic approach is the Proportional-Derivative (PD) controller. Its action depends on the current error (proportional part) and the rate of change of the error (derivative part). This "anticipatory" derivative action, which predicts where the error is headed, manifests in the transfer function as a zero. This zero adds "phase lead" to the system, which can be thought of as shining a flashlight further down a dark path. By seeing upcoming "turns" (changes in the reference signal) earlier, the system can react more swiftly and with less overshoot. A practical implementation of this idea is the lead compensator, a circuit or algorithm whose transfer function is defined by a carefully placed zero and pole to provide this phase-lead benefit over a targeted range of frequencies.

A crucial, and sometimes counter-intuitive, lesson in control theory is what happens to zeros when we "close the loop." If we take a system (the "plant") with a transfer function $G(s)$ and wrap a simple feedback controller around it, the zeros of the overall closed-loop system are, remarkably, the same as the zeros of the original plant, provided no cancellations occur. This tells us something profound: the inherent "blocking" characteristics of a system, its fundamental inability to transmit certain signal dynamics, are an immutable part of its identity. Feedback can move poles around to stabilize a system, but it cannot easily get rid of the plant's original zeros. An engineer must respect these intrinsic properties of the system they are trying to control.

The location of zeros reveals even more subtle truths about a system's nature. So far, we have mostly considered zeros in the left-half of the complex plane or on the imaginary axis. What happens if a zero wanders into the right-half plane? The consequences are fascinating and non-intuitive.

A system with right-half-plane zeros is called "non-minimum-phase." Its defining characteristic is often an "initial undershoot." Imagine trying to parallel park a car. To move the rear of the car towards the curb, you must first steer the front wheels away from it. The system (the car) initially moves in the opposite direction of the desired final outcome. This is the physical manifestation of a right-half-plane zero. Such systems are notoriously difficult to control. The zero acts as a fundamental limitation on performance. This behavior can be mathematically isolated. Any transfer function can be factored into a "minimum-phase" part (with all its poles and zeros in the left-half plane) and an "all-pass" part, which contains all the troublesome right-half-plane zeros. This all-pass component has a flat magnitude response—it lets all frequencies through equally—but it wreaks havoc on the phase, introducing the delay that causes the undershoot.

Where do such systems come from? They are more common than one might think. Remember our multipath channel with an echo, $H(s) = 1 + \alpha e^{-sT}$? It turns out this system is minimum-phase only if the echo is weaker than the primary signal ($|\alpha| 1$). If the echo is stronger ($|\alpha| > 1$), all the zeros move into the right-half plane, and the system becomes non-minimum-phase. Another critical source of non-minimum-phase behavior is time delay. A pure time delay, represented by the transcendental function $e^{-sT}$, is a nightmare for classical control analysis. A standard technique is to approximate it with a rational function of polynomials, a Padé approximation. This approximation replaces the transcendental function with a finite set of poles and zeros that mimic its behavior. Crucially, these approximations almost always feature zeros in the right-half plane, correctly capturing the non-minimum-phase nature inherent in any time delay.

These ideas are not confined to circuits and mechanics. They surface in the intricate feedback loops of life itself. A simplified model of the human glucose-insulin regulatory system, for example, can be described by a transfer function. Analyzing this model reveals not just poles, which describe the natural stability of blood sugar levels, but also a zero. The location of this zero influences the transient response—how quickly and smoothly a person's blood glucose returns to normal after an insulin dose. It shows that the language of poles and zeros is universal, providing a powerful framework for understanding dynamics wherever they appear.

From filtering out noise, to anticipating the future in a control loop, to revealing the strange "wrong-way" behavior of complex systems, transfer function zeros are far more than mathematical curiosities. They are a deep and unifying concept, a lens through which we can read the hidden story of how the world responds, reacts, and resonates.