A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Cascade compensation (series controller) provides loop shaping via Bode plots; feedback compensation (unity feedback plus compensator) separates error signal. Cascade excels at disturbance rejection; feedback excels at reference tracking and model uncertainty. Most systems use both: inner feedback loop + outer cascade compensator. Design must balance speed, bandwidth, robustness, and noise sensitivity.
You've studied pole placement and root locus — techniques for choosing where closed-loop poles should be. Compensation is the implementation question: how do you actually reshape the loop to put those poles there and meet performance specifications? The choice between cascade and feedback compensation is the central architectural decision before any detailed design begins.
Cascade compensation places a controller C(s) in series with the plant G(s) in the forward path. The open-loop transfer function becomes C(s)G(s), and you shape this product directly via Bode plots: add poles and zeros to adjust the gain crossover frequency, reshape phase near crossover, control low-frequency gain. A lead compensator adds phase near the crossover frequency, improving phase margin and speed of response. A lag compensator boosts low-frequency gain, reducing steady-state error without destabilizing the loop. Cascade design is conceptually clean and directly connected to Bode and root locus methods you already know. Its limitation is that it is essentially an open-loop shaping strategy with respect to disturbances — if the plant has parameter drift or disturbances entering at the output, cascade compensation cannot directly counteract them.
Feedback compensation (inner-loop or minor-loop feedback) wraps an additional feedback loop around part of the plant. The inner loop forces a subsystem to behave predictably regardless of parameter variation. A tachometer wrapped around a motor's mechanical dynamics, for instance, creates an inner velocity loop: high inner-loop gain makes the motor's speed response fast and insensitive to load variation. The outer cascade compensator then sees a well-conditioned inner-loop transfer function rather than the raw uncertain plant. Disturbances entering within the inner loop are rejected before they propagate to the outer loop — a capability cascade control cannot match.
The practical tradeoff has three dimensions. Disturbance rejection: if disturbances enter mid-plant, inner feedback addresses them at the source; cascade control cannot. Model uncertainty: high inner-loop gain suppresses parameter variation (robustness) but requires faster sensors and actuators. Noise sensitivity: high-bandwidth inner loops amplify sensor noise, so inner-loop bandwidth must stay below the frequency where sensor noise becomes significant. Most industrial control systems resolve these tradeoffs with a cascade-plus-inner-loop architecture: a fast inner feedback loop stabilizes and linearizes the plant, and an outer cascade compensator handles reference tracking and low-frequency performance. The two loops are designed separately, with the inner loop bandwidth at least 5–10× the outer loop bandwidth so they interact minimally.