Complementary Root Locus

In the field of control systems, the root locus is a foundational graphical method used to visualize how a system's stability changes as a key parameter—gain—is varied. The standard analysis, however, typically explores only positive gain values ($K>0$), which correspond to conventional negative feedback. This leaves a critical question unanswered: what happens when the gain is negative, or when the feedback itself is positive? This knowledge gap limits our ability to predict the behavior of systems with inverted signals or those designed to be inherently self-reinforcing.

This article provides a comprehensive exploration of the ​​complementary root locus​​, which maps system behavior for negative gain values. It delves into the fundamental law governing this alternate locus—the 0-degree angle condition—and derives its unique sketching rules. The article also demonstrates the profound practical relevance of this theory in analyzing system instabilities, preventing failures from wiring errors, and intentionally designing systems like electronic oscillators that harness the power of positive feedback. This provides a more complete picture of pole-zero behavior for a wider range of dynamic systems.

In control systems, the root locus provides a graphical map of a system's pole trajectories as a parameter, the gain, is varied. These trajectories are not random; they are strictly governed by a fundamental rule called the ​​angle condition​​. The system's open-loop transfer function, $L(s)$, defines the characteristics of the locus, analogous to how a force field dictates the path of a particle.

For a standard negative feedback system, the dance of the poles is choreographed by a simple, elegant characteristic equation: $1 + K L(s) = 0$. Here, $K$ is our gain, a real number we can vary. Let's rearrange this to see the core physics at play:

$K L(s) = -1$

This is not just an algebraic statement; it's a vector equation in the complex plane. The number $-1$ is special. It has a magnitude of 1 and, more importantly, an angle of $180^\circ$ (or $\pi$ radians, or any odd multiple thereof). If our gain $K$ is a positive real number, as it is in the standard root locus, it has an angle of $0^\circ$. For the equation to hold, the angle of $K$ plus the angle of $L(s)$ must equal the angle of $-1$. This forces the complex number $L(s)$ to have an angle of $180^\circ$.

This is the famous ​​180-degree angle condition​​ for the standard root locus ($K>0$): $\angle L(s) = (2k+1) \cdot 180^\circ \quad (\text{for any integer } k)$

A point $s$ in the complex plane is only "allowed" to be on the root locus if the system's transfer function $L(s)$ at that point aims directly toward the negative real axis. This single rule dictates the entire shape of the familiar root locus plots.

But what happens if we turn the gain knob the other way? What if we explore negative values for $K$? This is the question that opens the door to the ​​complementary root locus​​. Let's look at our law of motion again: $K L(s) = -1$. We can rewrite it as:

$L(s) = -\frac{1}{K}$

If $K$ is a negative real number ($K < 0$), then the term $-\frac{1}{K}$ becomes a positive real number. And what is the angle of a positive real number? It is $0^\circ$ (or any integer multiple of $360^\circ$). This unveils the fundamental law for the complementary root locus: the ​​0-degree angle condition​​.

$\angle L(s) = k \cdot 360^\circ \quad (\text{for any integer } k)$

Herein lies a beautiful symmetry. The universe of all possible pole locations is governed by two complementary laws. One law dictates that $L(s)$ must point left (180°), creating the standard root locus. The other dictates that $L(s)$ must point right (0°), creating the complementary root locus. The two conditions are perfectly out of phase, separated by a simple $180^\circ$ shift. Every rule we derive for the complementary locus will flow from this simple, elegant change in perspective.

Now, let's take a detour into what seems like a completely different universe: a system with ​​positive feedback​​. Instead of correcting errors, this system reinforces them. Its characteristic equation is $1 - K L(s) = 0$, where we'll assume the gain $K$ is positive. Rearranging gives us:

$K L(s) = 1$

Since $K$ is positive, this equation demands that $L(s)$ must be a positive real number. And what is the angle condition for $L(s)$ to be a positive real number? It's precisely the 0-degree condition we just discovered!

This is a profound insight. A negative feedback system with a negative gain behaves identically to a positive feedback system with a positive gain, at least as far as the location of their poles is concerned. The complementary root locus isn't just a mathematical oddity; it is also the map for a whole class of physically distinct systems. Nature, in its elegance, uses the same mathematical blueprint for two seemingly opposite scenarios.

Armed with our new 0-degree "law," we can now explore the geography of this complementary world. The familiar rules for sketching a root locus are all derived from the angle condition, so we should expect them to change in predictable ways.

On the Real Axis: A Perfect Partition

The simplest place to explore is the real axis. For any point $\sigma$ on the real axis, the angle of $L(\sigma)$ can only be $0^\circ$ (if $L(\sigma) > 0$) or $180^\circ$ (if $L(\sigma) 0$). This means that any given segment of the real axis can belong to either the standard locus or the complementary locus, but never both. They form a perfect, non-overlapping partition of the entire real line.

The rule is wonderfully simple:

• ​​Standard Locus ($K > 0$):​​ A point on the real axis belongs to the locus if the total number of real poles and zeros to its right is ​​odd​​. This creates the necessary $180^\circ$ phase shift.
• ​​Complementary Locus ($K 0$):​​ A point on the real axis belongs to the locus if the total number of real poles and zeros to its right is ​​even​​ (including zero). This provides the required $0^\circ$ phase.

Consider a simple system with poles at $s=-1$ and $s=-3$.

• For the segment $(-3, -1)$, there is one pole to the right (at -1). One is an odd number, so this segment belongs to the standard ($K>0$) locus.
• For the segments $(-\infty, -3)$ and $(-1, \infty)$, the number of poles/zeros to the right is two and zero, respectively. Both are even numbers, so these segments belong to the complementary ($K0$) locus. The paths must exist where the "law" allows them.

Journey to Infinity: The Asymptotes

When a system has more poles ($n$) than zeros ($m$), some locus branches must travel to infinity. Asymptotes are the straight lines that guide these paths. For the standard locus, their angles are given by $\theta = \frac{(2k+1)\pi}{n-m}$, creating a starburst pattern that is symmetric but avoids the positive real axis.

For the complementary locus, the 0-degree condition leads to a new formula. For large $s$, the angle of $L(s)$ is approximately $(m-n)\theta$. Setting this equal to our new law, $2k\pi$, gives:

$\theta = \frac{2k\pi}{n-m}$

This creates a new pattern of asymptotes, one of which will always be along the positive real axis (for $k=0$). This is a crucial difference and a hint of the practical implications to come.

Leaving Home: Angles of Departure

Just as the branches have predestined directions to infinity, they must also leave their starting points—the complex poles—at specific angles. This ​​angle of departure​​ is also dictated by our master law. To find it, we imagine a test point infinitesimally close to a pole $p_1$ and demand it satisfy the angle condition:

$\sum (\text{angles from zeros to } p_1) - \sum (\text{angles from other poles to } p_1) - \theta_d = \text{Angle Condition}$

For the complementary locus, the right side of the equation is $k \cdot 360^\circ$ instead of $(2k+1) \cdot 180^\circ$. This simple change can completely redirect the initial path of a pole. A branch that might have curved into the stable left-half plane for $K>0$ could, with this new rule, be directed into the unstable right-half plane for $K0$.

Why do we spend time charting this "complementary" world? Because in the real world, "negative gain" simply means inverting a feedback signal. It's flipping a plus sign to a minus sign on a wiring diagram. Sometimes, this is done by mistake. Other times, as in positive feedback, it's intentional. The complementary root locus is the tool that tells us the consequences.

Often, the consequences are dire. Consider a system that is stable for all positive gains. It's a robust, well-behaved system. One might assume that it's safe all around. But the complementary root locus can reveal a hidden danger. For a system with the transfer function $L(s) = \frac{s-z}{s(s+p)}$ (with $z, p > 0$), the real-axis rule tells us that for $K0$, the locus exists on the positive real axis for all $s > z$. This means that as you increase the magnitude of the negative gain, a closed-loop pole will inevitably move onto the positive real axis and speed away towards positive infinity. The system is guaranteed to become unstable.

The complementary root locus is therefore not just an academic exercise. It is a vital diagnostic tool. It completes the picture, showing us the full range of behaviors a system can exhibit as we vary its gain across all real numbers, both positive and negative. It reminds us that in the interconnected world of feedback, an action as simple as inverting a signal can flip the landscape of stability, turning a safe path into a treacherous one. By understanding its principles, we gain a more complete, more intuitive, and ultimately safer mastery of the systems we design.

In our journey so far, we have mapped out the intricate rules and geometric elegance of the root locus. We focused primarily on the "standard" case of negative feedback—the kind of feedback that corrects, stabilizes, and tames a system, like a thermostat cooling a room that gets too hot. But nature, in its infinite variety, does not always seek to correct. Sometimes, it reinforces. A small nudge becomes a big push; a tiny whisper grows into a deafening roar. This is the world of positive feedback.

What happens to our map when feedback turns from a stabilizing hand to an amplifying one? The landscape of the $s$-plane transforms entirely, and to navigate it, we need a new chart: the complementary root locus. This is the map for systems with positive feedback, or equivalently, those with a negative gain $K 0$ in a standard negative feedback configuration. The characteristic equation changes from $1 + K G(s) = 0$ to $1 - K_p G(s) = 0$ (where $K_p = -K$ is positive), and this simple sign flip changes everything. The angle condition shifts from an odd multiple of $\pi$ to an even one: $\angle G(s) = 2\ell\pi$. A seemingly small change, but one that opens up a world of new behaviors—some perilous, some wonderfully creative.

The most dramatic effect of positive feedback is its tendency to drive systems toward instability. Imagine trying to balance a broomstick on your finger. If the broom starts to fall to the left, and you move your hand further left, you are applying positive feedback. The result is obvious: the broom falls faster. Many sophisticated systems contain this same inherent instability.

A beautiful, modern example is a magnetic levitation (Maglev) train. An electromagnet pulls the train car upwards, counteracting gravity. But the closer the car gets to the magnet, the stronger the magnetic pull becomes—a classic positive feedback loop. Left to its own devices, the train would either slam into the magnet or fall to the ground. The only reason it works is because a very fast, sophisticated negative feedback controller is constantly making tiny adjustments. However, to design that controller, engineers must first understand the underlying unstable system. They need to know the limits. A critical question is: at what gain does the system's instability manifest as uncontrollable oscillations? The complementary root locus provides the answer, allowing engineers to find the precise gain value at which a system pole crosses into the right-half plane, signaling the onset of instability.

This isn't just a concern for exotic systems. In any complex control loop, a simple wiring error or a software bug can invert the sign of the feedback. A system meticulously designed for stability can suddenly find itself in a positive feedback scenario. The poles, which were meant to stay safely in the left-half plane, might begin a rapid march toward the right. Analyzing the complementary root locus can reveal at what negative gain value an otherwise stable system might become unstable, for instance, by having a pole arrive at the origin. A stable amplifier, for example, might be designed with a feedback network. If that feedback is inadvertently positive, the amplifier will only remain stable for a very limited range of gain. Beyond a certain threshold, it will become unstable. The Routh-Hurwitz criterion, applied to the characteristic equation for positive feedback, allows us to calculate this exact boundary of stability, a crucial safety parameter in circuit design.

While positive feedback can be destructive, it can also be the source of creation. When carefully controlled, it is the fundamental principle behind nearly every electronic oscillator, the heart of devices from radio transmitters to quartz watches. The goal of an oscillator is not to be stable, but to be perfectly marginally stable—to have its poles sit precisely on the imaginary axis, producing a pure, sustained sinusoidal signal.

How is this achieved? We need to create a loop where a signal of a specific frequency, say $\omega_0$, is fed back to the input exactly in phase with itself (the $0^\circ$ condition). This reinforcement causes the signal at that frequency to grow, while all other frequencies die out. The complementary root locus is the perfect tool for this design problem. It shows us the path the poles take as we increase the gain in a positive feedback loop. We can then find the exact gain $K$ that pushes the poles onto the $j\omega$-axis.

For certain system configurations, this analysis can yield wonderfully elegant results. Consider a system with poles and zeros at specific locations on the real axis. By applying the $0^\circ$ angle condition, we can solve for the exact frequency $\omega$ where the locus crosses the imaginary axis. For one such system, this frequency turns out to be the geometric mean of its pole and zero locations, $\omega = \sqrt{ab}$. This is a beautiful illustration of how the abstract geometry of the $s$-plane connects directly to a physical property—the pitch of the generated tone.

The complementary root locus is more than just an analytical tool; it's a designer's drawing board. By understanding its rules, we can sculpt the behavior of a system, even an unstable one.

The sketching rules themselves provide immediate insight. For instance, the rule for which parts of the real axis belong to the locus is inverted: a point on the real axis is on the complementary locus if the total number of real poles and zeros to its right is even (including zero). This simple change in parity dramatically alters the paths the poles can take. A region that was a "no-fly zone" for negative feedback might now be the main highway for poles under positive feedback. The rule applies regardless of whether the system's other poles are real or complex, as only real-axis singularities affect the angle contributions along the real axis itself.

With these rules, we can begin to design. Suppose we want to modify the long-term behavior of a robotic actuator operating in an unusual mode corresponding to negative gain. The high-gain behavior is dictated by the asymptotes of the root locus. The intersection of these asymptotes, the centroid, tells us the general direction the poles are headed. Remarkably, the formula for the centroid is the same for both standard and complementary loci. This allows us to strategically add a compensator (for instance, a zero) to the system to shift the centroid, thereby steering the poles away from undesirable regions of the $s$-plane as the gain becomes large.

Of course, in any design process, we must verify our assumptions. If we hypothesize that a certain desirable pole location, say $s_0 = -1+j$, will give us the performance we want, is it even possible to get there? We can simply test the point $s_0$ against the $0^\circ$ angle condition. If the angles from all the open-loop poles and zeros sum to a multiple of $360^\circ$, the point is on the locus; if not, no amount of gain will ever place a pole there, and we must rethink our design. We must also be wary of subtleties. The mathematical condition for finding breakaway or break-in points on the real axis ($dK/ds = 0$) still applies, but we have an extra check: the point must also satisfy the even-parity rule. It is entirely possible to find a mathematical extremum for the gain $K$ that lies in a segment of the real axis where the locus for positive feedback simply cannot exist.

Stepping back, we see a beautiful symmetry. The standard and complementary root loci are not separate subjects but two faces of the same underlying reality of feedback. They are governed by the same principles of complex arithmetic, differing only by a phase of $180^\circ$.

This unity is revealed in curious and profound ways. Let's ask a thought-provoking question: What kind of system, with all its poles and zeros on the real axis, would have a complementary root locus that covers the entire real axis? For this to happen, the "even number of poles and zeros to the right" rule must hold for every point on the real axis. As we move from right to left along the axis, the count of singularities to our right only changes when we cross one. If the count is to remain even everywhere, it must start at zero (which is even) and every time it changes, it must change by an even number. This leads to a startling conclusion: the multiplicity of every single real pole and zero must be an even number. A system with a double pole, or a double zero, would satisfy this, but a system with any single pole or zero would not. This is not just a mathematical curiosity; it's a deep insight into how a system's fundamental structure—its pole-zero DNA—dictates its global behavior under feedback.

By studying both negative and positive feedback, we gain a complete and holistic understanding of how a system's poles dance across the complex plane. We learn to not only predict their motion but to guide it—to tame instabilities, to create oscillations, and to appreciate the profound and beautiful unity that governs the complex world of feedback control.