Control System Analysis

Learning Objectives

By the end of this page, you will be able to:

• Write and interpret a transfer function $H(s) = Y(s)/X(s)$ for a simple electrical system.
• Read a Bode plot and extract gain crossover, phase crossover, gain margin, and phase margin.
• Compute time-domain performance metrics (rise time, overshoot, settling time, steady-state error) from a step response.
• Apply the Routh-Hurwitz criterion to determine whether a system is stable without finding the roots directly.
• Interpret a pole-zero plot and connect pole location to stability and time-domain behavior.

Quick Answer

Control system analysis is the set of mathematical tools engineers use to predict how a system will behave before building it — whether it will be stable, how fast it will respond, and how much it will overshoot. It matters because a system that looks fine on paper can oscillate uncontrollably or respond far too slowly once built, and re-engineering hardware after the fact is expensive. The two main lenses are the frequency domain (transfer functions, Bode plots, gain/phase margin) and the time domain (step response, rise time, overshoot, settling time), backed by rigorous stability tests like the Routh-Hurwitz criterion and pole-zero plots — a system is stable if and only if every pole of its transfer function lies in the left half of the complex s-plane.

Transfer Functions

Definition: A transfer function $H(s)$ describes the ratio of a system's output to its input in the Laplace (s-domain), assuming zero initial conditions: $H(s) = \frac{Y(s)}{X(s)}$ where $X(s)$ is the Laplace transform of the input and $Y(s)$ is the Laplace transform of the output.

Explanation: Working in the s-domain converts differential equations (hard to manipulate) into algebraic equations (easy to manipulate), and it lets engineers characterize a system once, then predict its response to any input by simple algebra rather than re-solving a differential equation each time.

Example: For a simple RC low-pass circuit with $R = 10\text{k}\Omega$ and $C = 100\ \mu\text{F}$, the transfer function is $H(s) = \frac{1}{1 + sRC} = \frac{1}{1 + s(10{,}000)(100\times10^{-6})} = \frac{1}{1 + s}$ (using $RC = 1$ second here) — this single expression tells you the circuit's response to a step, a sine wave of any frequency, or an impulse.

Real-world example: A DC motor's speed-control transfer function is often written $H(s) = K / (Js + b)$, where $K$ is a torque/electrical gain, $J$ is rotor inertia, and $b$ is damping — engineers use this exact form to design the surrounding speed-control loop.

Why it matters: Every other tool in this chapter — Bode plots, Routh-Hurwitz, pole-zero plots — is built directly on the transfer function. If you can derive $H(s)$ correctly, every subsequent analysis step is mechanical.

Common misunderstanding: Students think the transfer function describes the system for one specific input signal. It doesn't — it's a general model of the system itself, valid for any input, because it's derived from the system's differential equation, not from a particular input waveform.

Frequency Response and Bode Plots

Definition: Frequency response describes how a system's gain and phase shift vary as the frequency of a sinusoidal input changes. A Bode plot displays this as two graphs: magnitude (dB) versus frequency, and phase (degrees) versus frequency, both on a logarithmic frequency axis.

Explanation: From a Bode plot you read off four key numbers: the gain crossover frequency (where magnitude crosses 0 dB), the phase crossover frequency (where phase crosses −180°), the gain margin (how much additional gain, in dB, would push the system to instability at the phase crossover frequency), and the phase margin (how many degrees of additional phase lag, at the gain crossover frequency, would push the system to instability).

Example: If a system's phase crossover frequency is 10 rad/s and its gain at that frequency is −10 dB, the gain margin is 10 dB — meaning the loop gain could increase by 10 dB before becoming unstable.

Real-world example: Op-amp datasheets publish open-loop Bode plots specifically so designers can check phase margin before adding feedback resistors, guaranteeing the final amplifier circuit won't oscillate.

Why it matters: As a rule of thumb, a well-behaved practical control system should have a gain margin greater than about 6 dB and a phase margin greater than about 45° — these numbers give a safety buffer against real-world component tolerances and unmodeled effects like extra delay.

Common misunderstanding: Students think gain margin and phase margin measure the same kind of "safety" and can be traded off arbitrarily. They actually measure different failure mechanisms (excess gain at the phase-180° frequency, versus excess phase lag at the 0dB-gain frequency) and both should be checked — a system can pass one and fail the other.

Time-Domain Analysis

Definition: Time-domain analysis studies how a system's output evolves over time in response to a specific test input — most commonly a step input — and summarizes the response using standard metrics.

Explanation: The four standard metrics are: rise time (time to go from about 10% to 90% of the final value, or sometimes to reach 90% depending on convention), overshoot (the percentage by which the response exceeds its final steady value before settling), settling time (time to stay within a tolerance band, commonly ±5%, of the final value), and steady-state error (the permanent gap between the final response and the target after transients die out).

Example: For a step response that rises to a peak of 1.15 (a target of 1.0), overshoot is $\frac{1.15-1.0}{1.0}\times100\% = 15\%$.

Real-world example: Elevator motion controllers are explicitly tuned for near-zero overshoot (passengers should not feel the car "bounce" past the floor) even at the cost of a slightly longer settling time.

Why it matters: These metrics translate an abstract transfer function into numbers that map directly onto real user experience and engineering specs ("settle within 2 seconds, overshoot under 10%") — this is the language used in design requirements.

Common misunderstanding: Students confuse rise time with settling time. Rise time only describes how fast the initial climb to near the target happens; settling time describes how long oscillations/overshoot take to die down afterward — a system can have a fast rise time and a slow settling time if it overshoots and rings for a while before stabilizing.

Stability Analysis: The Routh-Hurwitz Criterion

Definition: The Routh-Hurwitz criterion is an algebraic method that determines how many roots of a system's characteristic equation lie in the right half of the s-plane (unstable) without needing to solve for the roots directly.

Explanation: Given a characteristic equation $s^n + a_1 s^{n-1} + a_2 s^{n-2} + \dots + a_n = 0$, you build the Routh array from its coefficients. The system is stable if and only if every entry in the array's first column has the same sign (no sign changes); each sign change in the first column corresponds to one right-half-plane (unstable) root.

Example: For $H(s) = \dfrac{s^2+3s+2}{s^3+4s^2+5s+2}$, the characteristic equation is $s^3 + 4s^2 + 5s + 2 = 0$. Building the Routh array:

Row	Column 1	Column 2
$s^3$	1	5
$s^2$	4	2
$s^1$	$(4)(5)-(1)(2))/4 = 4.5$	0
$s^0$	2	—

The first column is $1, 4, 4.5, 2$ — all positive, no sign changes, so the system is stable.

Real-world example: Before building a physical prototype of a DC motor speed loop, engineers run the closed-loop characteristic equation through Routh-Hurwitz to catch a destabilizing gain choice on paper, avoiding a costly hardware failure.

Why it matters: Routh-Hurwitz gives a definite stability verdict from algebra alone, without factoring a potentially high-order polynomial — this makes it fast to check by hand or symbolically as a design parameter (like controller gain) is varied.

Common misunderstanding: Students think a "no sign changes in the first column" result only tells you the system is probably stable. It's actually an exact, provable criterion (given a correctly formed array) — the number of sign changes exactly equals the number of unstable (right-half-plane) roots, not an approximation.

Pole-Zero Plots

Definition: A pole-zero plot shows the locations, in the complex s-plane, of a transfer function's poles (roots of the denominator) and zeros (roots of the numerator).

Explanation: Every pole in the left half-plane (negative real part) contributes a decaying response; every pole in the right half-plane (positive real part) contributes a growing, unstable response; poles on the imaginary axis produce sustained oscillation. Zero locations reshape the frequency response and transient shape without directly determining stability.

Example: The transfer function from the Routh-Hurwitz example, $H(s) = \dfrac{s^2+3s+2}{s^3+4s^2+5s+2}$, has all three poles in the left half-plane (consistent with the stable verdict found above) and zeros at $s=-1, s=-2$.

Real-world example: Control engineers use "pole placement" design — deliberately choosing feedback gains so the closed-loop poles land at specific left-half-plane locations that guarantee a desired settling time and overshoot.

Why it matters: A pole-zero plot is a compact visual summary of exactly the stability and performance information the earlier tools compute numerically — many engineers check pole-zero plots as a fast sanity check before running detailed time- or frequency-domain analysis.

Common misunderstanding: Students think zeros affect stability the same way poles do. They don't — stability is determined entirely by pole locations (denominator roots); zero locations only shape the transient response and frequency characteristics, they never make an otherwise-stable system unstable.

Visual Learning

All four analysis tools start from the same transfer function and converge on the same verdict — they're different lenses on one underlying model, which is why an exam question can ask you to confirm one result (e.g., stability from Routh-Hurwitz) using another (e.g., pole locations).

Real-World Applications

• DC motor speed control — deriving $H(s) = K/(Js+b)$ and checking stability before tuning a PID speed loop.
• Aerospace navigation — Bode-plot-based margin checks on autopilot control loops before flight testing.
• Chemical process control — Routh-Hurwitz checks on reactor temperature loops to avoid runaway conditions.
• Automotive ABS — pole-zero placement used to guarantee fast, non-oscillatory wheel-slip correction.
• Power electronics — frequency-response analysis of voltage-regulator feedback loops to prevent oscillation under varying load.

Key Terms

Term	Definition
Transfer function $H(s)$	The ratio $Y(s)/X(s)$ of a system's Laplace-domain output to its input.
Bode plot	A pair of graphs (magnitude in dB, phase in degrees) versus frequency on a log scale.
Gain margin	How much additional gain (dB) a system can tolerate at its phase crossover frequency before becoming unstable.
Phase margin	How much additional phase lag a system can tolerate at its gain crossover frequency before becoming unstable.
Rise time	Time for the step response to rise from near 0% to near 100% of its final value.
Overshoot	Percentage by which the peak response exceeds the final steady-state value.
Settling time	Time for the response to remain within a specified tolerance band of its final value.
Routh-Hurwitz criterion	An algebraic stability test based on sign changes in the first column of the Routh array.
Pole-zero plot	A plot of a transfer function's poles and zeros in the complex s-plane.

Common Mistakes

Misconception 1: "A system with a stable-looking step response is definitely stable for all conditions." Why it's wrong: A single step response only tests one input and one operating point; parameter variation, disturbances, or nonlinearities elsewhere in the operating range can reveal instability that a single test misses. Correct understanding: Formal criteria like Routh-Hurwitz or pole-zero analysis prove stability analytically across the modeled range, and are why time-domain observation alone is not treated as sufficient proof.

Misconception 2: "Gain margin and phase margin are basically the same safety measure, so checking one is enough." Why it's wrong: Gain margin captures how much extra loop gain the system tolerates at the phase-crossover frequency, while phase margin captures how much extra phase lag it tolerates at the gain-crossover frequency — these fail under different real-world causes (component drift vs. added delay). Correct understanding: Robust designs check both margins; a system can have excellent gain margin and poor phase margin (or vice versa) and still be at real risk of instability.

Misconception 3: "Zeros in the right half-plane make a system unstable, just like poles do." Why it's wrong: Stability is determined solely by pole locations (the denominator's roots); a right-half-plane zero (a "non-minimum phase" zero) affects transient behavior — sometimes causing an initial response in the wrong direction — but does not by itself make the system unstable. Correct understanding: Only poles in the right half-plane indicate instability; right-half-plane zeros are a separate, more subtle design concern related to non-minimum-phase behavior.

Comparison and Connections

Analysis Tool	Domain	What It Tells You
Bode plot	Frequency	Gain margin, phase margin, bandwidth
Step response	Time	Rise time, overshoot, settling time, steady-state error
Routh-Hurwitz	Algebraic (from characteristic equation)	Exact stable/unstable verdict, no root-finding needed
Pole-zero plot	Complex s-plane	Visual link between pole location and time-domain behavior

Pole Location	Time-Domain Behavior
Left half-plane, real	Decaying exponential response
Left half-plane, complex pair	Decaying oscillation
Imaginary axis	Sustained (undamped) oscillation
Right half-plane	Growing, unstable response

Practice Questions

Recall

1. Write the general definition of a transfer function. Answer guidance: $H(s) = Y(s)/X(s)$, the ratio of the Laplace transform of the output to the Laplace transform of the input, assuming zero initial conditions.
2. What condition on the Routh array's first column indicates a stable system? Answer guidance: No sign changes in the first column; each sign change corresponds to one unstable (right-half-plane) root.

Understanding

3. Explain why engineers check both gain margin and phase margin rather than just one. Answer guidance: They measure different stability failure modes — gain margin covers excess loop gain at the phase-crossover frequency, phase margin covers excess phase lag at the gain-crossover frequency — so a design could pass one check and still be fragile against the other kind of real-world variation.
4. Why does a right-half-plane pole indicate instability while a right-half-plane zero does not? Answer guidance: Poles are roots of the denominator and directly set the exponential/oscillatory modes of the system's natural response; a right-half-plane pole produces a growing exponential term. Zeros are roots of the numerator and only reshape the response's transient details, not its fundamental stability.

Application

5. A transfer function has the characteristic equation $s^3 + 2s^2 + s + 2 = 0$. Build the Routh array and determine stability. Answer guidance: Row $s^3$: 1, 1. Row $s^2$: 2, 2. Row $s^1$: $(2\cdot1 - 1\cdot2)/2 = 0$ — this creates a special case (a zero in the first column), signaling marginal/borderline stability that requires the standard epsilon-substitution technique to resolve fully; students should recognize this as a boundary case, not a simple stable/unstable answer.
6. A step response shows a peak value of 1.2 for a target of 1.0, and settles within ±5% by 3 seconds. State the overshoot and describe what settling time tells you here. Answer guidance: Overshoot = $(1.2-1.0)/1.0 \times 100\% = 20\%$; settling time of 3 seconds tells you how long the transient oscillation/overshoot takes to die down to within 5% of the final target value.

Analysis

7. Two candidate controller designs give a closed-loop system phase margins of 50° and 8° respectively, with similar gain margins. Evaluate which design is preferable for a system that will be manufactured with ±10% component tolerances. Answer guidance: The 50° phase-margin design is strongly preferable — it has a much larger buffer against added phase lag from component variation or unmodeled dynamics, whereas 8° margin leaves almost no safety buffer and the system is likely to become unstable in a real manufactured unit.
8. A student claims that since all four poles of a transfer function lie in the left half-plane, the system's step response will have zero overshoot. Assess this claim. Answer guidance: The claim is false — left-half-plane poles guarantee the response decays (stability), but complex-conjugate pole pairs with a significant imaginary part still produce a decaying oscillation, which can include overshoot before settling; only real (non-oscillatory) left-half-plane poles guarantee a non-oscillatory, overshoot-free response.

FAQ

Q1: Why do we use the Laplace domain (transfer functions) instead of solving differential equations directly? Because Laplace transforms convert differentiation and integration into algebra, so combining and analyzing systems (in series, feedback, etc.) becomes straightforward multiplication and division of transfer functions instead of solving coupled differential equations.

Q2: What's a "good" phase margin and gain margin in practice? A common rule of thumb is a phase margin above roughly 45° and a gain margin above roughly 6 dB, giving a comfortable buffer against real-world component tolerances and unmodeled delays.

Q3: Can a system be stable but still perform poorly? Yes — a system can have all its poles safely in the left half-plane (stable) yet still have unacceptably slow settling time, large overshoot, or lingering steady-state error; stability is necessary but not sufficient for good performance.

Q4: What happens when the Routh array has a zero in the first column? It signals a special/marginal case that needs a workaround (such as replacing the zero with a small variable ε and taking a limit) to correctly determine stability — it typically indicates the system is at the boundary of stability.

Q5: How does this chapter connect to the next chapter on Control System Design? This chapter gives you the tools to evaluate whether a given system is stable and how well it performs; the Control System Design chapter uses those same tools in reverse — choosing controller parameters (like PID gains) specifically to achieve target stability margins and performance metrics.

Quick Revision

• Transfer function: $H(s)=Y(s)/X(s)$ — models the system for any input, not just one signal.
• Bode plot gives gain crossover, phase crossover, gain margin, and phase margin.
• Rule of thumb for robust stability: gain margin > 6 dB, phase margin > 45°.
• Time-domain metrics: rise time, overshoot, settling time, steady-state error — from the step response.
• Routh-Hurwitz: no sign changes in the Routh array's first column ⇒ stable; each sign change = one unstable pole.
• Pole-zero plot: left-half-plane poles ⇒ stable/decaying; right-half-plane poles ⇒ unstable/growing; imaginary-axis poles ⇒ sustained oscillation.
• Zeros shape the transient response but never determine stability by themselves.
• All four tools (transfer function, Bode, Routh-Hurwitz, pole-zero) analyze the same underlying model from different angles.
• Stability is necessary but not sufficient — a stable system can still have poor rise time, overshoot, or steady-state error.
• Common exam trap: a zero in the Routh array's first column signals a special marginal case, not automatic instability or stability.

Related Topics

Prerequisites: Feedback Control (loop components, gain, phase margin conceptually); Laplace transforms and basic complex numbers.

Related Topics: Control System Design (using these tools to choose controller parameters); Digital Control Systems (the Z-transform equivalent of these same ideas in discrete time).

Next Topics: Control System Design — turning analysis into synthesis: choosing PID gains and compensators to hit target stability margins and performance specs.