Skip to content
Tutorials Center
My learning

Topics

Lesson 13 of 37 ยท Time-domain Analysis

Steady-State Error and System Type

Compute steady-state error with the final value theorem, learn system type and the Kp, Kv and Ka error constants, and verify ramp tracking in Python.

After the transients die out, how close does the output get to the reference? The leftover difference is the steady-state error e_ss. For a thermostat it is the degrees you still miss by; for a radar tracker it is how far behind a moving target you lag. This lesson shows how to compute it from the loop transfer function and why adding integrators removes it.

The error formula

For the unity-feedback loop with loop gain L(s) = C(s) G(s), the error is E(s) = R(s) / (1 + L(s)). If the closed loop is stable, the final value theorem gives

Text
e_ss = lim (s -> 0) of s R(s) / (1 + L(s))

The stability condition matters. If the closed loop is unstable, the error does not settle and the formula is meaningless; check stability first.

System type

Write the loop gain as

Text
L(s) = K (1 + b1 s)(1 + b2 s)... / (s^N (1 + a1 s)(1 + a2 s)...)

The integer N, the number of pure integrators (poles at the origin) in L(s), is the system type. Type 0 has no integrator, type 1 has one, and so on.

Error constants

Three constants summarise the low-frequency gain of the loop:

Text
Kp = lim (s->0) L(s)              position error constant
Kv = lim (s->0) s L(s)            velocity error constant
Ka = lim (s->0) s^2 L(s)          acceleration error constant

The steady-state errors for unit inputs are:

Input Type 0 Type 1 Type 2
Step 1/(1 + Kp) 0 0
Ramp infinite 1/Kv 0
Parabola infinite infinite 1/Ka

System type in one line

Type 0 Constant error for a step.
Type 1 Tracks steps, lags ramps.
Type 2 Tracks ramps, lags parabolas.

Each integrator buys one more input degree of accuracy.

1 / 2

The reasoning is simple. An integrator in the loop keeps ramping its output as long as any error remains, so the only way to reach a steady state is for the error to be zero. A type-0 loop has no such mechanism and settles with a constant offset 1/(1 + Kp), which is the 0.048 we found for the heater in Open-Loop and Closed-Loop Control: Kp = 20 gives 1/21 = 0.0476.

Computing error constants numerically

You can approximate each limit by evaluating L(s) at a very small s.

Python Error constants and steady-state errors

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The large numbers in each row show which constant is finite: for type 0 it is Kp, for type 1 it is Kv (while Kp is huge), for type 2 it is Ka. The others are zero or infinite, matching the table.

Seeing the error in a simulation

Now simulate the type-1 loop L(s) = 10/(s(s+2)) with a ramp input r(t) = t and look at the final error. The closed loop is 10 / (s^2 + 2 s + 10), and Kv = 5, so e_ss = 0.2. scipy.signal.lsim computes responses to arbitrary inputs.

Python Tracking a ramp with a type-1 loop

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The ramp error settles at 0.2 while the step error goes to 0, exactly as the table says. Doubling the gain K would halve the ramp error, but also reduces damping, the usual trade-off.

Disturbances

Steady-state error from a constant disturbance depends on where the integrator sits. An integrator before the point where the disturbance enters (in the controller) cancels the offset; an integrator only in the plant after that point does not. This is why controllers add an integrator term: the I in PID, covered in Chapter 6.

Improving steady-state accuracy

  • Increase the loop gain K: reduces e_ss for a given type, but may lower damping or cause instability.
  • Add an integrator: raises the system type and can bring the step error to zero.
  • Add feedforward: inject a model-based signal so feedback has less to correct.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. A unity-feedback loop has L(s) = 10/(s (s + 2)). What is its type and Kv?

    There is one integrator, so the type is 1. Kv = lim s L(s) = 10/2 = 5.
  2. 2. A type-0 loop has Kp = 9. What is the steady-state error to a unit step?

    e_ss = 1/(1 + Kp) = 1/10 = 0.1.
  3. 3. Which system type tracks a ramp with zero steady-state error?

    A ramp needs two integrators for zero error: type 1 has a finite error 1/Kv, and type 2 has zero.
  4. 4. Before using the final value theorem for e_ss you must check that

    If the closed loop is unstable the error never settles, so the limit does not describe the real behaviour.

Summary

  • e_ss = lim s R(s) / (1 + L(s)) for a stable unity-feedback loop.
  • The system type is the number of integrators in L(s).
  • Error constants Kp, Kv, Ka give the errors for step, ramp and parabola inputs.
  • Each extra integrator removes one more kind of steady-state error, and raising gain reduces it further.
  • Always confirm closed-loop stability before applying the theorem.

Next: Dominant Poles and the Effect of Added Poles and Zeros