Q10Control System

(a) Lead-Lag Compensation Network

A lead-lag compensator combines the complementary benefits of a lead compensation network and a lag compensation network into a single combined compensator, addressing both transient-response/stability requirements and steady-state accuracy requirements simultaneously, which neither a lead network alone nor a lag network alone can typically achieve together as effectively.

Lead network: has a transfer function of the form Gc(s) = (1+aTs)/(1+Ts), with a>1, placing a zero (at s=-1/(aT)) closer to the origin (i.e., at a lower frequency) than its associated pole (at s=-1/T). A lead network introduces a positive (phase-advancing) phase angle over a specific frequency range, which, when inserted in the forward path of a control system, increases the phase margin at the gain crossover frequency (improving relative stability and reducing overshoot/settling time in the transient response), and also tends to increase the system's bandwidth (yielding a faster overall response speed) — the primary purpose of a lead network is therefore to improve transient response and stability margin.

Lag network: has a transfer function of the form Gc(s) = (1+bTs)/(1+Ts), with b<1, placing a pole closer to the origin than its associated zero. A lag network introduces a negative (phase-lagging) phase angle, but its primary purpose is instead to increase the low-frequency (DC) open-loop gain substantially (improving steady-state tracking accuracy and reducing steady-state error for step/ramp inputs) while deliberately placing its corner frequencies well below the system's gain crossover frequency, so that its associated phase lag has largely settled and become negligible again by the time the gain crossover frequency is reached, minimizing any adverse impact on the system's phase margin/transient response.

Combined lead-lag network: Gc(s) = [(1+aT1s)/(1+T1s)] × [(1+bT2s)/(1+T2s)] with a>1 (lead section) and b<1 (lag section), typically designed so the lag section operates at low frequencies (improving steady-state accuracy) while the lead section operates near the gain crossover frequency (improving phase margin and transient response), allowing the combined compensator to simultaneously achieve the improved steady-state accuracy that a lag network alone provides together with the improved transient response and stability margin that a lead network alone provides — a combination of benefits that is often essential for meeting demanding control system specifications that neither individual compensator type could satisfy alone.

(b) PID Controllers

A PID (Proportional-Integral-Derivative) controller is the most widely used control law in industrial control systems, combining three distinct control actions, each computed from the instantaneous error signal e(t) = r(t)-c(t) (the difference between the reference input and the actual measured output), to generate the overall control signal u(t):

Proportional (P) action: Kp·e(t), producing a control signal directly proportional to the instantaneous error — provides fast response to changes in error, but a purely proportional controller generally leaves a residual steady-state error for step-type reference inputs applied to plants without an inherent integrator, and using too high a proportional gain alone can lead to excessive overshoot or even instability.

Integral (I) action: Ki∫e(τ)dτ, producing a control signal proportional to the accumulated (integrated) history of past error — since the integral term continues to grow as long as any nonzero error persists, it drives the steady-state error to exactly zero for step reference inputs (assuming a stable closed-loop system), at the cost of potentially slowing down the transient response and introducing a risk of increased overshoot/oscillation if the integral gain is too large (a phenomenon sometimes complicated further by 'integrator windup' if the controller output saturates a physical actuator).

Derivative (D) action: Kd·de(t)/dt, producing a control signal proportional to the instantaneous rate of change of error — since it responds to the trend/rate of the error rather than its instantaneous or accumulated value, it provides an anticipatory, damping effect, improving transient response by reducing overshoot and settling time, though it can amplify high-frequency measurement noise present in the error signal if not carefully filtered.

Combined PID action: the three actions together provide a well-rounded, generally applicable control law — proportional action gives immediate, direct response proportional to present error; integral action eliminates steady-state error by responding to accumulated past error; and derivative action improves transient response and damping by responding to the predicted future trend of the error — and the three gains Kp, Ki, Kd are typically tuned together (using methods such as Ziegler-Nichols tuning rules, or more modern optimization-based tuning techniques) to achieve the specific desired balance of response speed, overshoot, and steady-state accuracy for a given plant and application, making the PID controller the most widely deployed control algorithm across industrial process control, robotics, and countless other control system applications due to its conceptual simplicity, general applicability, and well-established, mature tuning methodology.

Transfer-Function Form of the PID Controller and Its Pole-Zero Structure

Taking the Laplace transform of the PID control law (with zero initial conditions) gives the controller's transfer function directly in the s-domain:

This reveals that a PID controller is, structurally, a single pole at the origin (contributed by the integral term, which is precisely what drives the steady-state error to zero for step inputs, since it provides infinite DC gain) together with two zeros (the roots of the numerator quadratic Kds²+Kps+Ki), whose locations can be independently placed in the s-plane by appropriate choice of Kp, Ki, Kd to reshape the root locus of the compensated open-loop system and thereby achieve a desired closed-loop pole location. This pole-zero view directly connects the PID controller to the lead-lag compensator discussed in part (a): a PD (proportional-derivative) sub-controller (Kp+Kds = Kd(s+Kp/Kd)) behaves similarly to a lead network (adding a real zero that advances phase and improves transient response/stability margin), while a PI (proportional-integral) sub-controller (Kp+Ki/s = Kp(s+Ki/Kp)/s) behaves similarly to a lag network (adding a pole at the origin for perfect steady-state tracking, together with a nearby zero that limits the associated phase lag) — so that a full PID controller can be understood as an idealized combination of lead and lag compensation action realized directly in terms of three tunable gains, rather than in terms of separate corner-frequency time constants T1, T2 as in the classical lead-lag network of part (a).

Practical Considerations: Derivative Kick and Integrator Windup

Two well-known practical implementation issues are commonly discussed alongside the idealized PID law given above. First, since pure differentiation of the error signal amplifies high-frequency sensor noise (as differentiation is a high-pass operation), practical PID implementations typically replace the ideal derivative term Kds with a band-limited (filtered) derivative Kds/(1+sTf) for some small filter time constant Tf, preventing noise amplification while still providing the desired anticipatory damping action at the frequencies of interest. Second, integrator windup occurs when the controller output saturates a physical actuator (e.g., a valve fully open or a motor at maximum voltage) while a persistent error remains, causing the integral term to continue accumulating (winding up) well beyond the value needed, which then causes a large overshoot and delayed recovery once the error finally changes sign, since the excess accumulated integral action must first unwind before the controller output can respond correctly — practical PID implementations therefore commonly include anti-windup logic (such as clamping or back-calculation schemes) that temporarily halts or corrects integral accumulation whenever the actuator is saturated, illustrating that real-world PID implementation requires several practical refinements beyond the idealized control law u(t)=Kpe(t)+Ki∫e(τ)dτ+Kde'(t) given above.