8.2 Feedback Control Loops, PID Tuning, Frequency Response & Stability
Key Takeaways
- Feedback control architectures continuously compare setpoint against process variable to form error $e(t) = SP(t) - PV(t)$, adjusting manipulated variables via the closed-loop transfer function $T(s) = G(s) / [1 + G(s)H(s)]$.
- First-order systems exhibit exponential lag with a $2\%$ settling time of $4\tau$, whereas underdamped second-order systems ($0 < \zeta < 1$) are governed by natural frequency $\omega_n$ and damping ratio $\zeta$, determining peak overshoot $\%OS$ and peak time $T_p = \pi / \omega_d$.
- The three PID terms perform distinct dynamic duties: Proportional ($K_p$) accelerates transient rise but leaves steady-state offset (droop); Integral ($K_i$) eliminates steady-state error by increasing system type but adds phase lag ($-90^\circ$) and risk of windup; Derivative ($K_d$) provides phase lead damping ($+90^\circ$) to suppress overshoot.
- Ziegler-Nichols closed-loop tuning identifies the critical ultimate gain $K_u$ and ultimate oscillation period $P_u$ at marginal stability, setting baseline parameters $K_p = 0.60 K_u$, $T_i = 0.50 P_u$, and $T_d = 0.125 P_u$.
- Control valve trim selection balances installed loop gain: Equal percentage valves provide an exponential characteristic ($dQ/dx \propto Q$) that linearizes installed loop response against falling line pressure drop in hydronic piping networks.
Feedback Control Loops, PID Tuning, Frequency Response & Stability
Automatic control systems maintain critical process variables (fluid temperature, steam pressure, volumetric flow, liquid level, and rotating shaft speed) at desired setpoints despite external load disturbances and supply variations. In mechanical and thermal power systems—such as central chiller plants, steam boilers, HVAC air handling units, robotic actuators, and hydraulic governors—a thorough understanding of transfer functions, transient response specifications, PID loop tuning, and frequency-domain stability criteria is essential for safe, reliable, and energy-efficient operation.
1. Feedback Control Architecture & Block Diagram Algebra
A closed-loop control system continuously measures the process variable ($PV$), subtracts it from the reference setpoint ($SP$) to compute error $e(t) = SP(t) - PV(t)$, and modulates the final control element to drive error toward zero.
+---------------------------------------------------------------------------------------------------+
| CLOSED-LOOP FEEDBACK CONTROL ARCHITECTURE |
| |
| Setpoint R(s) Error E(s) Control U(s) Disturbance D(s) Output Y(s) |
| ------------->(+)--->[ CONTROLLER ]--->[ ACTUATOR / ]--->(+)--->[ PROCESS ]--------+---> |
| ^ - C(s) VALVE ] G_a(s) ^ G_p(s) | |
| | | | |
| +---------------------[ SENSOR H(s) ]<-----+------------------------+ |
+---------------------------------------------------------------------------------------------------+
Closed-Loop Transfer Function Derivation
Let $G(s) = C(s) G_a(s) G_p(s)$ represent the combined forward-path transfer function, and $H(s)$ represent the feedback sensor transfer function:
Solving for the closed-loop input-output relationship:
Where:
- $G(s)H(s) = \text{Open-loop transfer function (loop gain)}$
- $1 + G(s)H(s) = 0 = \text{Characteristic equation of the closed-loop system}$
- The roots of $1 + G(s)H(s) = 0$ are the closed-loop poles, which dictate stability and dynamic response.
- The disturbance-to-output transfer function is $T_D(s) = \frac{Y(s)}{D(s)} = \frac{G_p(s)}{1 + G(s)H(s)}$.
2. Dynamic System Response: First-Order & Second-Order Systems
+---------------------------------------------------------------------------------------------------+
| FIRST-ORDER VS SECOND-ORDER SYSTEMS |
| |
| SYSTEM ORDER GOVERNING ODE STANDARD TRANSFER FUNCTION |
| ------------- -------------------------------- ------------------------------------------ |
| First-Order tau * dy/dt + y = K * u(t) G(s) = K / (tau * s + 1) |
| Second-Order d^2y/dt^2 + 2*zeta*wn*dy/dt G(s) = wn^2 / (s^2 + 2*zeta*wn*s + wn^2) |
| + wn^2 * y = wn^2 * u(t) |
+---------------------------------------------------------------------------------------------------+
First-Order System Step Response
Thermal systems (stirred tanks, thermowell RTDs) and single-reservoir liquid levels behave as first-order lag systems characterized by process gain $K$ and time constant $\tau$:
+---------------------------------------------------------------------------------------------------+
| FIRST-ORDER STEP RESPONSE MILESTONES |
| |
| Time Elapsed Percent of Final Value (y / K) Significance |
| -------------- ------------------------------ ----------------------------------------- |
| t = 1 * tau 63.2% Standard definition of time constant |
| t = 2 * tau 86.5% Intermediate rise |
| t = 3 * tau 95.0% 5% Settling time threshold |
| t = 4 * tau 98.2% 2% Settling time threshold (Standard Ts) |
| t = 5 * tau 99.3% Steady state achieved |
+---------------------------------------------------------------------------------------------------+
- Rise Time ($t_r$): Time required to rise from $10%$ to $90%$ of final value: $t_r = \tau \ln(9) \approx 2.2 \tau$.
- Settling Time ($T_s$): Time required to remain within $\pm 2%$ of steady state: $T_s \approx 4 \tau$.
Second-Order System Dynamics
Mechanical mass-spring-damper mechanisms ($m \ddot{x} + c \dot{x} + k x = F$) and hydraulic actuators exhibit second-order dynamics characterized by the undamped natural frequency ($\omega_n$) and the damping ratio ($\zeta$):
+---------------------------------------------------------------------------------------------------+
| SECOND-ORDER DAMPING REGIMES & POLE PLOTS |
| |
| DAMPING REGIME DAMPING RATIO (zeta) POLE LOCATIONS RESPONSE CHARACTER |
| ------------------ -------------------- -------------------------- ------------------- |
| Undamped zeta = 0 Pure imaginary: ±j*wn Sustained oscillation |
| Underdamped 0 < zeta < 1 Complex conjugate: -s ± j*wd Damped oscillations |
| Critically Damped zeta = 1 Real & repeated: -wn Fastest non-oscill. |
| Overdamped zeta > 1 Two distinct real poles Sluggish exponential |
+---------------------------------------------------------------------------------------------------+
+---------------------------------------------------------------------------------------------------+
| UNDERDAMPED SECOND-ORDER STEP RESPONSE METRICS |
| |
| Output y(t) |
| ^ Peak Time (Tp) |
| | | |
| | +-----+ <--- Maximum Value y_max |
| | / \ |
| 1.0 + - - - - - - - - -/- - - - -\ - - - - - - - - - - - - - - - + - - Steady State Target |
| | / \ +---+ | |
| | / \ / \ | ±2% Error Band |
| | / +-+ +-------------------+ |
| | / | |
| 0 +-------------+-----------------------------------------------+---------> Time (t) |
| |<--- Tr --->| |<-- Ts -->| |
+---------------------------------------------------------------------------------------------------+
For an underdamped second-order step response ($0 < \zeta < 1$):
- Damped Natural Frequency ($\omega_d$):
- Peak Time ($T_p$): Time required to reach the maximum first overshoot peak:
- Percent Overshoot ($%OS$):
- Settling Time ($T_s$, 2% criterion):
3. The PID Controller Algorithm
The Proportional-Integral-Derivative (PID) controller is the most ubiquitous feedback algorithm in industrial automation.
+---------------------------------------------------------------------------------------------------+
| THE PARALLEL PID CONTROLLER STRUCTURE |
| |
| +---> [ PROPORTIONAL: K_p * e(t) ] --------+ |
| | | |
| Error e(t) = SP - PV --+---> [ INTEGRAL: K_i * ∫ e(t) dt ] ---->(+)---> Manipulated Variable |
| | | u(t) |
| +---> [ DERIVATIVE: K_d * de(t)/dt ] ------+ |
+---------------------------------------------------------------------------------------------------+
Mathematical Formulations
In the time domain, the parallel PID equation is:
In the standard (ISA / Ideal) format with integral reset time $T_i$ and derivative rate time $T_d$:
Where $K_i = K_p / T_i$ and $K_d = K_p \cdot T_d$.
In the Laplace $s$-domain:
+---------------------------------------------------------------------------------------------------+
| PID CONTROL TERM CONTRIBUTIONS |
| |
| TERM FUNCTION & ACTION BENEFITS DRAWBACKS / RISKS |
| ------------- -------------------------------- -------------------- ------------------- |
| Proportional u_P = K_p * e(t) Speeds response, Leaves steady-state |
| (P) Directly proportional to error reduces error offset ('droop') |
| |
| Integral u_I = K_i * ∫ e(t) dt Eliminates steady- Adds phase lag (-90°),|
| (I) Accumulates past error over time state offset (type +1) increases overshoot, |
| causes windup |
| |
| Derivative u_D = K_d * de(t)/dt Adds phase lead (+90°),Amplifies high- |
| (D) Anticipates future error trend damps oscillations, frequency noise, |
| improves stability ineffective on pure lag
+---------------------------------------------------------------------------------------------------+
[!WARNING] Integral Windup & Anti-Windup Solutions: When an actuator reaches physical saturation (e.g., a chilled water control valve is $100%$ fully open), error cannot be quickly cleared. The integral term continues accumulating past error, driving the internal control calculation far beyond $100%$. When the setpoint is finally reached, the valve remains stuck wide open for a long period while the integral "unwinds," causing extreme temperature undershoot. Industrial controllers use anti-windup clamping to freeze integration whenever the actuator reaches its output limit.
4. PID Controller Tuning Methodologies
+---------------------------------------------------------------------------------------------------+
| ZIEGLER-NICHOLS PID TUNING RULES |
| |
| 1. CLOSED-LOOP ULTIMATE GAIN METHOD |
| - Set K_i = 0, K_d = 0. |
| - Increase K_p until sustained, constant-amplitude oscillations occur at Ultimate Gain K_u. |
| - Measure the Ultimate Period P_u (time between wave peaks in seconds). |
| |
| CONTROLLER TYPE K_p T_i T_d |
| -------------------- -------------------- -------------------- ------------------ |
| P Controller 0.50 * K_u -- -- |
| PI Controller 0.45 * K_u P_u / 1.20 -- |
| Classic PID 0.60 * K_u 0.50 * P_u 0.125 * P_u |
| Pessen Integral Rule 0.70 * K_u 0.40 * P_u 0.150 * P_u |
| |
| 2. OPEN-LOOP REACTION CURVE METHOD (Step Input Δu) |
| - Measure Process Gain K_0 = Δy / Δu, Dead Time L (delay), and Time Constant T. |
| |
| CONTROLLER TYPE K_p T_i T_d |
| -------------------- -------------------- -------------------- ------------------ |
| P Controller T / (K_0 * L) -- -- |
| PI Controller 0.9 * T / (K_0 * L) 3.33 * L -- |
| Classic PID 1.2 * T / (K_0 * L) 2.00 * L 0.50 * L |
+---------------------------------------------------------------------------------------------------+
5. Dynamic Stability Analysis: Routh-Hurwitz & Frequency Response (Bode)
A dynamic system is Bounded-Input Bounded-Output (BIBO) stable if and only if all poles of its closed-loop transfer function lie strictly in the open left-half of the complex $s$-plane ($\text{Re}(s) < 0$).
Routh-Hurwitz Stability Criterion
Given the characteristic polynomial:
- Necessary Condition: All coefficients $a_n, a_{n-1}, \dots, a_0$ must be non-zero and share the same algebraic sign.
- Sufficient Condition (Routh Array): Construct the Routh array table. The number of roots with positive real parts (unstable right-half plane poles) equals the number of sign changes in the first column of the Routh array.
+---------------------------------------------------------------------------------------------------+
| ROUTH ARRAY CONSTRUCTION TABLE |
| |
| s^n | a_n a_{n-2} a_{n-4} ... |
| s^{n-1} | a_{n-1} a_{n-3} a_{n-5} ... |
| s^{n-2} | b_1 b_2 b_3 ... |
| s^{n-3} | c_1 c_2 ... |
| ... | |
| |
| Where: b_1 = (a_{n-1}*a_{n-2} - a_n*a_{n-3}) / a_{n-1} |
| b_2 = (a_{n-1}*a_{n-4} - a_n*a_{n-5}) / a_{n-1} |
| c_1 = (b_1*a_{n-3} - a_{n-1}*b_2) / b_1 |
+---------------------------------------------------------------------------------------------------+
Frequency Response: Bode Plot Stability Margins
Frequency response evaluates the open-loop transfer function $G(j\omega)H(j\omega)$ across frequency $\omega$.
+---------------------------------------------------------------------------------------------------+
| BODE PLOT STABILITY DEFINITIONS |
| |
| CRITICAL FREQUENCY DEFINITION STABILITY MARGIN |
| ----------------------------- ------------------------------ ---------------------------- |
| Gain Crossover Freq (w_gc) |G(j*w)H(j*w)| = 1.0 (0 dB) Phase Margin: |
| PM = 180° + ∠G(j*w_gc)H(j*w_gc)|
| |
| Phase Crossover Freq (w_pc) ∠G(j*w)H(j*w) = -180° Gain Margin: |
| GM = -20*log10|GH(j*w_pc)| dB |
+---------------------------------------------------------------------------------------------------+
- Stable System: $PM > 0^\circ$ and $GM > 0\text{ dB}$ (i.e., $\omega_{gc} < \omega_{pc}$).
- Marginally Stable: $PM = 0^\circ$ and $GM = 0\text{ dB}$ (sustained oscillations at $\omega_{pc}$).
- Unstable System: $PM < 0^\circ$ or $GM < 0\text{ dB}$ (closed-loop poles in RHP).
- Recommended Design Margins: Robust industrial loops target $PM \ge 45^\circ$ and $GM \ge 6\text{ to } 12\text{ dB}$ to prevent resonant hunting under load variations.
6. Final Control Elements: Modulating Control Valves
The final control element physically modulates fluid energy. In hydronic and steam networks, this is predominantly an automated control valve driven by a pneumatic diaphragm or electric actuator.
+---------------------------------------------------------------------------------------------------+
| CONTROL VALVE INHERENT CHARACTERISTICS |
| |
| CHARACTERISTIC FLOW EQUATION (Inherent) APPLICATION & SELECTION CRITERIA |
| ---------------- ---------------------------------- ------------------------------------- |
| Linear Q / Q_max = x / x_max Constant valve Delta_P systems, liquid |
| dQ/dx = constant level loops, bypass control |
| |
| Equal Percentage Q / Q_max = R^( (x/x_max) - 1 ) Variable Delta_P piping systems, heat |
| dQ/dx = (ln R) * Q exchangers (compensates for falling dP) |
| |
| Quick Opening Rapid initial flow rise at low lift On/off service, safety relief, dump |
+---------------------------------------------------------------------------------------------------+
+---------------------------------------------------------------------------------------------------+
| VALVE INSTALLED VS INHERENT CHARACTERISTIC |
| |
| Flow Fraction (Q/Q_max) |
| 1.0 + Equal Percentage (Inherent) |
| | / |
| | Linear (Inherent) |
| | / / |
| | Quick Opening / / <--- Equal Percentage (Installed in |
| | / / / piping with high line losses becomes |
| | / / / nearly LINEAR installed!) |
| | / / / |
| 0.0 +------+-------------+---------------+------------------> Valve Stroke (x/x_max) |
| 0.0 1.0 |
+---------------------------------------------------------------------------------------------------+
Valve Flow Sizing Coefficient ($C_v$)
The standard flow coefficient $C_v$ defines the flow of $60^\circ\text{F}$ water in US gallons per minute ($\text{gpm}$) through a valve with a pressure drop of $1.0\text{ psi}$:
Where $Q$ is in $\text{gpm}$, $\Delta P$ is valve pressure drop in $\text{psi}$, and $SG$ is fluid specific gravity ($SG = 1.0$ for water at $60^\circ\text{F}$).
Cavitation and Flashing in Liquid Valves
- Cavitation: If static pressure drops below liquid vapor pressure ($P_{\text{vena contracta}} < P_v$) inside the valve throat and subsequently recovers downstream above $P_v$ ($P_2 > P_v$), vapor bubbles collapse violently against trim surfaces, creating extreme noise, vibration, and rapid pitting erosion.
- Flashing: If downstream recovery pressure remains below vapor pressure ($P_2 < P_v$), vapor bubbles do not collapse, resulting in high-velocity two-phase sonic choked flow.
7. Step-by-Step Worked Engineering Problems
Worked Problem 1: Second-Order Dynamic Response of a Valve Positioner
Problem: A pneumatic valve positioner is modeled as a standard second-order system with closed-loop transfer function:
Determine: (a) undamped natural frequency $\omega_n$, (b) damping ratio $\zeta$, (c) damped natural frequency $\omega_d$, (d) percent overshoot $%OS$, and (e) settling time $T_s$ (2% criterion).
+---------------------------------------------------------------------------------------------------+
| STEP-BY-STEP SECOND-ORDER SOLUTION |
| |
| STEP 1: Match to Standard Form: s^2 + 2*zeta*wn*s + wn^2 |
| wn^2 = 144 --> wn = 12.0 rad/s |
| 2 * zeta * wn = 14.4 --> 2 * zeta * (12.0) = 14.4 --> zeta = 14.4 / 24.0 = 0.60 |
| |
| STEP 2: Calculate Damped Natural Frequency |
| wd = wn * sqrt(1 - zeta^2) = 12.0 * sqrt(1 - 0.60^2) |
| wd = 12.0 * sqrt(1 - 0.36) = 12.0 * sqrt(0.64) = 12.0 * 0.80 = 9.60 rad/s |
| |
| STEP 3: Calculate Percent Overshoot (%OS) |
| %OS = 100 * exp [ -pi * zeta / sqrt(1 - zeta^2) ] |
| %OS = 100 * exp [ -pi * 0.60 / 0.80 ] = 100 * exp [ -2.35619 ] |
| %OS = 100 * 0.09478 = 9.48% |
| |
| STEP 4: Calculate 2% Settling Time (Ts) |
| Ts = 4 / (zeta * wn) = 4 / (0.60 * 12.0) = 4 / 7.20 = 0.556 seconds |
+---------------------------------------------------------------------------------------------------+
Worked Problem 2: Routh-Hurwitz Stability Range for Feedback Control
Problem: A closed-loop temperature control loop has the characteristic polynomial $s^3 + 8 s^2 + 15 s + K = 0$. Determine: (a) the range of controller gain $K > 0$ for which the closed-loop system remains stable, and (b) the frequency of oscillation when $K$ is set to the marginally stable ultimate gain $K_u$.
+---------------------------------------------------------------------------------------------------+
| STEP-BY-STEP ROUTH-HURWITZ SOLUTION |
| |
| STEP 1: Construct the Routh Array |
| s^3 | 1 15 |
| s^2 | 8 K |
| s^1 | b_1 0 |
| s^0 | c_1 0 |
| |
| STEP 2: Compute First-Column Terms |
| b_1 = ( (8 * 15) - (1 * K) ) / 8 = (120 - K) / 8 |
| c_1 = ( (b_1 * K) - (8 * 0) ) / b_1 = K |
| |
| STEP 3: Enforce First-Column Positive Sign Condition for Stability |
| Condition 1: K > 0 |
| Condition 2: b_1 > 0 --> (120 - K) / 8 > 0 --> K < 120 |
| STABILITY RANGE: 0 < K < 120 |
| |
| STEP 4: Ultimate Gain and Oscillation Frequency at Marginal Stability |
| Ultimate Gain K_u = 120. |
| Row s^2 forms auxiliary equation: 8 * s^2 + K_u = 0 |
| 8 * s^2 + 120 = 0 --> s^2 = -15 --> s = ± j * sqrt(15) = ± j * 3.873 rad/s |
| Oscillation frequency w = sqrt(15) = 3.87 rad/s. |
+---------------------------------------------------------------------------------------------------+
8. Common Exam Traps & PE Pro-Tips
- Trap 1 — Confusing Open-Loop vs Closed-Loop Transfer Functions: Open loop is $G(s)H(s)$; closed loop is $G(s) / [1 + G(s)H(s)]$. Closed-loop stability is governed strictly by the roots of the denominator $1 + G(s)H(s) = 0$.
- Trap 2 — Equal Percentage vs. Linear Valve Selection: In water piping networks where pipe friction absorbs $50%$ or more of total system head loss at full flow, a linear valve behaves like a quick-opening valve in service (high distortion). Choosing an Equal Percentage valve compensates for falling available pressure drop, providing a linear installed flow characteristic.
- Trap 3 — Unit Mismatches in Natural Frequency ($f_n$ vs $\omega_n$): In second-order equations, $\omega_n$ is in $\text{rad/s}$. If an exam prompt requests frequency in Hertz ($\text{Hz}$), divide by $2\pi$: $f_n = \omega_n / (2\pi)$.
- Trap 4 — Negative Signs in Phase Margin Calculation: $PM = 180^\circ + \angle GH(j\omega_{gc})$. Because phase lag is negative (e.g., $-135^\circ$), $PM = 180^\circ + (-135^\circ) = +45^\circ$.
A hydronic cooling system experiences significant piping friction losses at high design flow rates, causing the available pressure drop across the modulating chilled water control valve to drop significantly as the valve opens. Which inherent valve trim characteristic should be specified to maintain linear installed control loop response?
An underdamped second-order mechanical actuator system has an undamped natural frequency of 20 rad/s and a damping ratio of 0.50. What is the approximate peak time at which maximum overshoot occurs following a unit step input?
During closed-loop Ziegler-Nichols tuning of a temperature process controller, integral and derivative actions are disabled. Proportional gain is increased until sustained, constant-amplitude sinusoidal oscillations occur at an ultimate gain of $K_u = 8.0\ \text{V/V}$, with a measured period between wave crests of $P_u = 12.0\text{ seconds}$. According to classic Ziegler-Nichols rules, what are the recommended PID controller parameters?
The open-loop transfer function of a unity feedback position control system is $G(s) = \frac{K}{s(s + 3)(s + 5)}$. Using the Routh-Hurwitz stability criterion, what is the maximum allowable value of gain $K$ before the closed-loop system becomes unstable?