Impact of P, I, D (and PI, PD, PID) on System Performance

Control engineers often interpret the PID controller by how each term changes key performance metrics: stability , steady-state error , overshoot , and settling time.

A standard PID law is: $u(t)=K_p\,e(t)+K_i\int_0^t e(\tau)\,d\tau+K_d\,\frac{de(t)}{dt}$ where $e(t)=r(t)-y(t)$ is the error, $u(t)$ is the control input, $K_p$ is proportional gain, $K_i$ integral gain, and $K_d$ derivative gain.

Key intuition (applies broadly, with details depending on plant dynamics):

• Proportional (P) mode mainly increases “responsiveness” by applying an immediate correction based on $e(t)$.
• Integral (I) mode drives steady-state error toward zero by accumulating error.
• Derivative (D) mode predicts error trends (damping), often reducing overshoot and improving transient behavior.

When combined:

• PI is common when steady-state accuracy matters but overshoot must be limited.
• PD is common when overshoot/damping is critical.
• PID combines transient damping (D) with steady-state elimination (I).

A useful way to visualize how these terms reshape the closed-loop dynamics is to think in terms of effective “phase lead/lag” contributions: P tends to change loop gain, I adds low-frequency gain (reducing steady-state error but risking oscillation), and D adds phase lead (often increasing damping but amplifying noise).

Terms and performance metrics you’ll track

To study “impact,” you need explicit metrics and test scenarios. Typical experiments include step changes in reference ($r(t)$) and disturbances entering at the plant input or output.

Performance metrics:

• Overshoot and undershoot
• Settling time (e.g., $\pm 2\%$ or $\pm 5\%$ band)
• Rise time
• [Steady-state error]{def="Error as $t\\to\\infty$ for step/ramp inputs"}
• Control effort (important for actuator limits)

These metrics shift differently under each control mode.

P (proportional gain $K_p$):

• Increasing $K_p$ usually decreases rise time and may increase overshoot.
• Too much $K_p$ can drive oscillations and reduce stability margin.

I (integral gain $K_i$):

• Increasing $K_i$ reduces steady-state error for step references (and eliminates constant bias errors).
• Too much $K_i$ can cause slow oscillations, overshoot, or instability due to “accumulation” (often discussed alongside integral windup).

D (derivative gain $K_d$):

• Increasing $K_d$ often increases damping and can reduce overshoot and improve settling time.
• Too much $K_d$ can amplify measurement noise because differentiation magnifies high-frequency components; this is related to noise amplification and “derivative kick” under certain implementations.

Combination effects:

• PI: removes steady-state error while relying on P for transient speed.
• PD: improves damping without integral-driven steady-state correction.
• PID: aims to balance both.

Impact of each control mode (P vs I vs D)

Proportional (P) mode

Let $K_i=K_d=0$. The controller is: $u(t)=K_p e(t)$ Effect patterns:

• [Higher $K_p$]{def="Larger proportional gain applied to current error"} increases loop gain and typically speeds up the response (shorter rise time).
• Overshoot often increases because the loop reacts strongly to error, potentially reducing damping.
• If the plant is near marginal stability or has lightly damped dynamics, excessive $K_p$ can produce sustained oscillations or instability.

For first-order plants, P tends to yield monotonic improvement in rise time with limited overshoot. For higher-order plants (especially with dominant poles that create oscillatory dynamics), P can cause overshoot and ringing.

What to look for in plots

• $y(t)$: faster rise, possibly larger overshoot.
• steady-state error: for a step reference, P typically leaves a nonzero steady-state error depending on plant type.
• $u(t)$: magnitude grows roughly with $K_p$ times the error peak.

Integral (I) mode

Let $K_p=K_d=0$: $u(t)=K_i\int_0^t e(\tau)\,d\tau$ Effect patterns:

• Integral action increases low-frequency loop gain. For step references, it drives the steady-state error toward zero (in ideal conditions).
• Transient response can become slower (or more oscillatory) if $K_i$ is too large, because integral keeps “pushing” even after error is reduced—accumulation lingers.
• With actuator saturation, the integral state can keep growing while the actuator is unable to respond, causing delayed overshoot when saturation ends: integral windup.

What to look for

• $y(t)$: improved final accuracy, but potentially slower settling or larger overshoot for large $K_i$.
• $u(t)$: often ramps or accumulates; watch for periods where $u(t)$ is saturated.
• Recovery behavior: after saturation, does the system overshoot or oscillate?

Derivative (D) mode

Let $K_p=K_i=0$: $u(t)=K_d\frac{de(t)}{dt}$ Effect patterns:

• Derivative adds damping by reacting to the rate of change of error.
• It can reduce overshoot and improve settling time by counteracting rapid changes.
• Two practical issues dominate:
  1. Noise amplification: differentiation increases high-frequency noise in the measured signal.
  2. Derivative kick: if derivative is computed from error that includes abrupt setpoint changes, $de/dt$ can spike at step references.

What to look for

• $y(t)$: reduced overshoot/ringing compared to P-only.
• $u(t)$: potentially large spikes at step times (especially with derivative kick).
• Sensitivity to noise: add realistic sensor noise; D may degrade performance without filtering.

Combined modes: PI, PD, PID

PI control ($K_d=0$)

$u(t)=K_p e(t)+K_i\int_0^t e(\tau)\,d\tau$

• PI is designed to improve steady-state accuracy over P alone.
• Typical trade-off: PI often reduces steady-state error but can increase overshoot or slow settling if $K_i$ is aggressive.
• Actuator saturation becomes critical: PI is the primary mode where integral windup shows up most.

PD control ($K_i=0$)

$u(t)=K_p e(t)+K_d\frac{de(t)}{dt}$

• PD improves damping and often reduces overshoot compared to P-only.
• PD does not inherently guarantee zero steady-state error for step references in many systems; plant “type” still matters.
• Noise and derivative kick must be handled carefully (filtering or derivative-on-measurement strategies are common in practice).

PID control ($\;K_p,K_i,K_d\\neq 0$)

$u(t)=K_p e(t)+K_i\int_0^t e(\tau)\,d\tau+K_d\frac{de(t)}{dt}$

• PID aims to combine:
  • P for responsiveness,
  • I for steady-state elimination,
  • D for damping/transient shaping.
• In practice, PID tuning is about balancing competing effects:
  • Too much I → windup and oscillation/overshoot.
  • Too much D → noise sensitivity and derivative kick spikes.
  • Too much P → reduced phase margin and oscillation.