Closed-Loop Frequency Response

Hey students! 👋 Welcome to one of the most fascinating topics in control engineering - closed-loop frequency response analysis! In this lesson, we'll explore how feedback systems behave across different frequencies and discover the fundamental trade-offs that every control engineer must understand. By the end of this lesson, you'll be able to analyze closed-loop transfer functions, understand sensitivity functions, and appreciate the delicate balance between performance and robustness that makes control systems work in the real world. Get ready to unlock the secrets behind everything from your car's cruise control to spacecraft navigation systems! 🚀

Understanding Closed-Loop Transfer Functions

When we add feedback to a control system, we create what's called a closed-loop system. Think of it like learning to ride a bicycle - your brain constantly receives feedback from your balance sensors and makes tiny adjustments to keep you upright. In control engineering, we represent this mathematically using transfer functions.

The most fundamental closed-loop transfer function is the complementary sensitivity function, denoted as $T(s)$:

$$T(s) = \frac{PC(s)}{1 + PC(s)}$$

Where $P(s)$ is your plant (the system you're controlling) and $C(s)$ is your controller. This function tells us how well our system follows reference commands. For example, if you're designing a temperature controller for your home's heating system, $T(s)$ describes how quickly and accurately the actual temperature follows your desired temperature setting.

Another crucial transfer function is the sensitivity function, denoted as $S(s)$:

$$S(s) = \frac{1}{1 + PC(s)}$$

The sensitivity function is like the system's immune response - it tells us how well the closed-loop system rejects disturbances. In our heating example, $S(s)$ would describe how well your system maintains the desired temperature when someone opens a window on a cold day.

Here's something amazing: these two functions are related by the fundamental constraint $S(s) + T(s) = 1$. This simple equation reveals a profound truth about control systems - you can't have perfect tracking AND perfect disturbance rejection at the same time! 🤔

The Sensitivity Function: Your System's Defense Mechanism

Let's dive deeper into the sensitivity function $S(s) = \frac{1}{1 + PC(s)}$. This function is incredibly important because it quantifies how "sensitive" your closed-loop system is to disturbances and model uncertainties.

When $|S(jω)|$ is small at a particular frequency $ω$, it means your system provides excellent disturbance rejection at that frequency. Real-world applications demonstrate this beautifully:

Automotive cruise control: Modern systems achieve $|S(jω)| < 0.1$ (or -20 dB) at low frequencies, meaning they can reject 90% of disturbances like gentle hills or headwinds
Precision manufacturing: CNC machines often require $|S(jω)| < 0.01$ at frequencies below 10 Hz to maintain accuracy despite vibrations and thermal effects
Satellite attitude control: Space missions demand $|S(jω)| < 0.05$ at very low frequencies to maintain precise pointing despite solar radiation pressure

The sensitivity function also reveals something crucial about robustness. If your plant model has errors, the actual closed-loop transfer function becomes:

$$T_{actual} = \frac{P_{actual}C}{1 + P_{actual}C} = T_{nominal} \cdot \frac{1}{1 + S \cdot \Delta}$$

Where $\Delta$ represents the relative modeling error. This shows that small sensitivity leads to robust performance - your system will still work well even if your model isn't perfect! 💪

The Complementary Sensitivity Function: Tracking Performance

The complementary sensitivity function $T(s) = \frac{PC(s)}{1 + PC(s)}$ describes how well your system tracks reference inputs. Ideally, we'd want $T(s) = 1$ for all frequencies, meaning perfect tracking. However, real systems have limitations.

At low frequencies, we typically want $|T(jω)| ≈ 1$ for good steady-state tracking. For example, a DC motor position control system might achieve $|T(jω)| > 0.95$ for frequencies below 1 Hz, ensuring accurate positioning.

At high frequencies, we usually want $|T(jω)|$ to roll off (decrease) to avoid amplifying noise. A well-designed system might have $|T(jω)| < 0.1$ above 100 Hz to filter out sensor noise and prevent actuator wear.

The bandwidth of $T(s)$ - the frequency range where $|T(jω)| > 0.707$ (-3 dB) - determines how fast your system can respond. Fighter aircraft require bandwidths of 10-20 Hz for agile maneuvering, while large ships might only need 0.01 Hz bandwidth for course corrections.

The Fundamental Trade-off: You Can't Have It All

Here comes the heart of control engineering - the fundamental trade-off between sensitivity and complementary sensitivity. Since $S(s) + T(s) = 1$, we have several important constraints:

Waterbed Effect: If you push down the sensitivity at one frequency (improving disturbance rejection), it must pop up somewhere else. This is mathematically expressed by Bode's integral theorem:

$$\int_0^∞ \ln|S(jω)| dω = π \sum \text{(unstable poles of } PC\text{)}$$

For stable open-loop systems, this integral equals zero - meaning any reduction in sensitivity at some frequencies must be compensated by increases elsewhere.

Performance vs. Robustness: At any frequency, we face the constraint:

If $|S(jω)|$ is small (good disturbance rejection), then $|T(jω)| ≈ 1$ (good tracking but potential noise amplification)
If $|T(jω)|$ is small (good noise rejection), then $|S(jω)| ≈ 1$ (poor disturbance rejection)

Real-world examples of this trade-off include:

Audio systems: High-end amplifiers use feedback to reduce distortion (small $S$) at audio frequencies, but this can make them sensitive to radio frequency interference
Economic systems: Central banks face similar trade-offs - aggressive monetary policy (high "controller gain") can reduce economic disturbances but may amplify the effects of model uncertainties
Biological systems: Your body's temperature regulation is incredibly robust to external temperature changes but responds slowly to internal heat generation during exercise

Frequency Domain Design: Shaping Your System's Response

Understanding these trade-offs allows engineers to "shape" the frequency response of $S(s)$ and $T(s)$ to meet specific requirements. This process, called loop shaping, involves designing the controller $C(s)$ to achieve desired closed-loop properties.

Common design objectives include:

Low-frequency performance: Make $|S(jω)|$ small below the bandwidth frequency to ensure good disturbance rejection and steady-state accuracy
High-frequency robustness: Ensure $|T(jω)|$ rolls off sufficiently fast to provide noise attenuation and robustness to high-frequency modeling errors
Bandwidth selection: Choose the crossover frequency (where $|PC(jω)| = 1$) to balance response speed with stability margins

Modern control systems often achieve remarkable performance by carefully managing these trade-offs. For instance, the Hubble Space Telescope maintains pointing accuracy of 0.007 arcseconds (equivalent to focusing on a dime 200 miles away) by using sophisticated frequency domain design techniques.

Conclusion

Closed-loop frequency response analysis reveals the fundamental nature of feedback control systems. The sensitivity function $S(s)$ and complementary sensitivity function $T(s)$ provide powerful tools for understanding system behavior, with their relationship $S(s) + T(s) = 1$ highlighting the unavoidable trade-offs in control design. By carefully shaping these functions in the frequency domain, engineers can design systems that balance tracking performance, disturbance rejection, noise attenuation, and robustness to meet real-world requirements. Understanding these concepts is essential for anyone working with feedback systems, from simple temperature controllers to complex aerospace applications.

Study Notes

• Complementary Sensitivity Function: $T(s) = \frac{PC(s)}{1 + PC(s)}$ - describes reference tracking performance

• Sensitivity Function: $S(s) = \frac{1}{1 + PC(s)}$ - describes disturbance rejection and robustness

• Fundamental Constraint: $S(s) + T(s) = 1$ - you cannot make both functions small simultaneously

• Waterbed Effect: Reducing sensitivity at one frequency increases it elsewhere (for stable systems: $\int_0^∞ \ln|S(jω)| dω = 0$)

• Low Frequency Design: Make $|S(jω)|$ small for good disturbance rejection and steady-state accuracy

• High Frequency Design: Make $|T(jω)|$ small for noise attenuation and robustness

• Bandwidth: Frequency range where $|T(jω)| > 0.707$ (-3 dB) - determines system response speed

• Trade-off Principle: Better disturbance rejection (small $S$) comes at the cost of potential noise amplification (large $T$)

• Robustness: Small sensitivity function leads to robust performance despite modeling errors

• Loop Shaping: Design process of shaping $S(s)$ and $T(s)$ frequency responses through controller design