Questions: Disturbance Rejection and Feedforward Control

Question 1

A room temperature control system uses a thermostat measuring indoor air temperature. A cold front arrives, dropping outdoor temperature sharply. What is the fundamental limitation of this pure feedback approach?

• A. The thermostat gain is too low to detect rapid temperature changes
• B. The feedback loop cannot act until the indoor temperature has already dropped, meaning the disturbance has already degraded the output
• C. Feedback control is inherently unstable when outdoor disturbances are fast
• D. The sensor cannot distinguish between setpoint changes and external disturbances

Answer: (B)

Feedback is inherently reactive: it measures the output, detects an error, and then corrects — but the error must already exist in the output before any correction begins. When the cold front hits, indoor temperature must first drop before the thermostat triggers heating. A feedforward controller using an outdoor temperature sensor would act the moment outdoor temperature falls, before any indoor effect occurs. This reactive delay is structural, not a tuning problem.

Question 2

A process has a large, fast disturbance that is directly measurable at its source. Which control architecture best handles this situation?

• A. Pure feedback with very high loop gain, since this minimizes steady-state error regardless of disturbance speed
• B. Pure feedforward, since it completely eliminates the disturbance before it affects the output with no model dependency
• C. Combined feedforward-feedback: feedforward quickly cancels the measured disturbance; feedback corrects residuals from model error
• D. Cascade control, which uses a secondary feedback loop to handle fast inner dynamics

Answer: (C)

Combined FF+FB is the standard answer for measurable disturbances. Feedforward provides fast, anticipatory rejection by acting before the disturbance reaches the output — but only as well as the plant model is accurate. Feedback corrects the residuals that model imperfection leaves behind, plus any unmeasured disturbance components. Pure feedforward fails with model error; pure feedback is reactive; high gain alone cannot eliminate reactive delay.

Question 3

A feedforward controller can reject unmeasured disturbances more effectively than a feedback controller, since it acts before they affect the output.

• T. True
• F. False

Answer: False

Feedforward requires that the disturbance be measurable — it acts by detecting the disturbance at its source and computing a compensating signal. If a disturbance is unmeasured, the feedforward controller has no information about it and cannot act. For unmeasured disturbances, feedback is the only option: it corrects reactively once the disturbance causes an output error. This is precisely why combining both architectures is superior to either alone.

Question 4

Feedback control is inherently reactive: it can only apply corrective action after a disturbance has already caused a detectable error in the output.

• T. True
• F. False

Answer: True

This is the defining structural limitation of feedback control. The sensor must detect an output deviation before the controller computes an error signal and applies a correction. This sequence means the output error exists before the correction arrives — the latency is irreducible, not a tuning artifact. Feedforward control was developed specifically to overcome this limitation for cases where disturbances can be sensed directly at their source.

Question 5

Why does the combined feedforward-feedback architecture outperform pure feedforward or pure feedback alone when rejecting a large, measurable disturbance?

Model answer: Feedforward alone is fast but fragile: it can only cancel disturbances it measures and can only do so as accurately as its plant model. Model error, parameter drift, and unmeasured disturbance components leave residuals that feedforward cannot address. Feedback alone is robust but reactive: it corrects any error eventually, but only after the disturbance has already affected the output. Combined, feedforward rapidly attenuates the bulk of the measured disturbance while feedback corrects the residuals — each mechanism compensates for the other's weakness.

The complementarity is the key insight: feedforward handles what is measurable and modeled; feedback handles what is not. High-performance industrial control always uses both wherever disturbances can be sensed, because neither alone achieves fast AND robust disturbance rejection.