Questions: Internal Model Principle and Integral Control Action

Question 1

A proportional-only (P) controller is applied to a Type 0 plant tracking a constant reference. What happens to steady-state error as the proportional gain increases toward infinity?

• A. Steady-state error goes to zero at some finite high gain value, solving the tracking problem
• B. Steady-state error approaches zero but never reaches it for finite gain; achieving zero error requires infinite gain, which causes instability
• C. Steady-state error is unaffected by gain — only controller structure (not gain) determines steady-state error
• D. Steady-state error decreases to a minimum then increases again due to oscillation from high gain

Answer: (B)

For a Type 0 system, the steady-state error to a step is 1/(1 + K_p), where K_p is the DC loop gain. As K_p increases, error decreases but asymptotically approaches zero — never reaching it at any finite gain. Setting K_p → ∞ achieves zero error in the limit but drives the system unstable. The internal model principle tells us the solution is structural: adding an integrator (pole at s = 0) achieves zero steady-state error at finite, stabilizable gain, not by pushing gain to its limit.

Question 2

A control system must track a ramp reference (linearly increasing with time) with zero steady-state error. According to the internal model principle, what must the controller contain?

• A. A single integrator — since ramps are just accumulated steps, one integration is sufficient
• B. A lead-lag compensator with high enough gain at low frequencies
• C. A double integrator (two poles at s = 0), because the ramp signal is generated by a double integrator
• D. A differentiator to respond to the constant rate of change in the ramp

Answer: (C)

The internal model principle states that to track a signal class with zero error, the controller must contain poles that match the generating dynamics of that signal. A constant (step) is generated by a single integrator. A ramp is generated by a double integrator (two poles at s = 0 in the Laplace domain). Therefore, to track a ramp with zero error, the controller must include a double integrator. A single integrator will track steps perfectly but still accumulate growing error on ramp references.

Question 3

A P-primarily controller with sufficiently high gain can achieve zero steady-state error to a step input.

• T. True
• F. False

Answer: False

For a Type 0 plant, steady-state error = 1/(1 + K_p). No matter how large K_p becomes, this expression never reaches zero at finite gain — it only approaches zero asymptotically. And infinite gain destabilizes the closed-loop system. Zero steady-state error to steps requires an integrator in the forward loop, not just high gain. The internal model principle makes this structural: no amount of tuning can substitute for the correct controller architecture.

Question 4

An integrator in the forward control loop guarantees zero steady-state error to constant (step) references because, at steady state, any nonzero error would cause the integrator's output to keep changing — contradicting the assumption of steady state.

• T. True
• F. False

Answer: True

This is the core intuition for why the internal model principle works. An integrator continuously accumulates its input (the error). If there were a nonzero constant error at steady state, the integrator's output would be ramping, meaning the control signal is still changing, meaning the plant output is still changing — which means the system has not actually reached steady state. The only state consistent with true steady state is error = 0. The integrator cannot 'rest' until the error is zero. This structural guarantee is independent of gain values.

Question 5

Explain intuitively why adding an integrator to a controller eliminates steady-state error to a constant reference, even without specifying the exact gain.

Model answer: An integrator accumulates the error over time. If there is any constant nonzero error at steady state, the integrator's output continuously ramps — either growing or decaying — which means the controller output is changing, which means the plant input is changing, which means the plant output is changing. But that means the system is not at steady state at all. The only value of error that allows the integrator to remain constant (i.e., allows true steady state) is zero. The structural property of the integrator makes zero error the only self-consistent steady-state condition, regardless of what the gain is.

This argument shows why correct structure is more fundamental than gain tuning. Gain selection affects transient behavior — speed of convergence, overshoot — but the steady-state property (zero error for steps) follows from the integrator's presence alone. The internal model principle generalizes this: the controller's pole structure commits it to handling entire classes of signals, not just specific magnitudes.