5.8 Feedback Sensitivity

Feedback sensitivity is a measure of how responsive a feedback system is to changes in its input, reference, or disturbance conditions, and specifically how those responses are modified by the presence and properties of the feedback loop. In control theory, sensitivity functions quantify the relationship between external perturbations and the system's deviation from its desired operating point, capturing both the benefits and the inherent limitations of feedback control in a precise mathematical form.

The sensitivity function S(s) of a unity feedback control system is defined as the transfer function from an external disturbance at the output to the actual output error. For a plant with transfer function P(s) and a controller with transfer function C(s), the sensitivity function is:

The magnitude of S(jω) at a given frequency ω describes how much of a disturbance at that frequency appears in the tracking error. When |S(jω)| is small, the feedback effectively suppresses disturbances at that frequency. When |S(jω)| is large, disturbances at that frequency are amplified relative to the open-loop system. The loop gain L(s) = P(s)C(s) determines the sensitivity: high loop gain at a given frequency means low sensitivity, and low loop gain means sensitivity approaches 1 (the open-loop value).

The complementary sensitivity function T(s) describes how well the closed-loop system tracks the reference:

The sensitivity and complementary sensitivity functions satisfy the fundamental identity:

This relation expresses a fundamental conservation: reducing sensitivity (improving disturbance rejection) at some frequencies necessarily increases complementary sensitivity (potentially degrading noise rejection) at others. The design of feedback controllers always involves navigating this trade-off.

The Bode sensitivity integral establishes a waterbed constraint on feedback sensitivity for stable closed-loop systems. For a system with right half-plane open-loop poles p_i (unstable poles), the integral of the log sensitivity over all frequencies satisfies:

This integral constraint means that if sensitivity is reduced below 1 in some frequency band (improving performance in that band), it must be increased above 1 in another band (creating worse performance there). Unstable open-loop poles increase the total "area" available for this distribution, making sensitivity management harder for unstable plants. This is a fundamental information-theoretic constraint: no controller, however sophisticated, can escape the waterbed effect.

The concept of feedback sensitivity extends beyond control engineering to any system that uses feedback to regulate its behavior. In biological sensory systems, sensitivity refers to the minimum stimulus intensity that produces a detectable response, and adaptation mechanisms continuously adjust sensitivity to maintain optimal responsiveness across a wide range of input levels. Dark adaptation in the visual system increases photoreceptor sensitivity by orders of magnitude when the ambient light level drops, enabling effective vision across a range of illumination spanning many decades.

Organizational feedback sensitivity describes how strongly an organization adjusts its behavior in response to feedback signals from its environment. High-sensitivity organizations detect small market shifts and customer feedback and respond rapidly. Low-sensitivity organizations are slow to register and respond to external signals, making them less adaptive but also less subject to instability driven by overreaction to transient noise. The optimal sensitivity depends on the rate of environmental change relative to the organization's response capabilities and the costs of under- versus over-reacting.

In ecological systems, predator sensitivities to prey population fluctuations determine the dynamics of food webs. Highly sensitive predators that rapidly track prey availability can stabilize prey populations through tight density-dependent regulation. Less sensitive predators may allow prey populations to fluctuate more widely, potentially producing boom-bust cycles. The sensitivity of different feedback pathways within an ecosystem influences its resilience and its response to perturbations such as invasive species or habitat change.