CHAPTER 20 / Signals and stability

System type: count integrators, then check the error

The number of integrators in a stable loop predicts its final error for step and ramp targets.

THE SHORT ANSWER

System type is the number of uncancelled integrators, poles at s = 0, in the open-loop transfer function L(s) of a unity negative-feedback loop. If the closed loop is stable, the type decides the steady-state error for polynomial references: type 0 keeps a finite step error, type 1 removes step error but keeps a finite ramp error, and type 2 removes ramp error as well.

E(s)=R(s)1+L(s),Kv=lim⁡s→0sL(s)E(s)=\frac{R(s)}{1+L(s)},\qquad K_v=\lim_{s\to0}sL(s)
WHAT YOU WILL LEARN

Compare a type-0 and type-1 loop using one stable, ideal unity-feedback example.

Routh checks whether a linear loop is stable. Once stability holds, system type helps predict final error for steps and ramps.

Review the previous lesson: Routh–Hurwitz criterion: stability without solving every root
START WITH AN EXAMPLE

First, picture it

A controller that can hold a fixed target may still lag behind a target that keeps moving. An integrator remembers accumulated error. System type counts how many uncancelled integrators appear in the complete loop, rather than counting all states or all poles.

THE IDEA

Build the idea step by step

For unity negative feedback, error equals the reference divided by one plus the loop transfer function. A step is a fixed target; a ramp rises at a constant rate. A stable type-0 loop generally retains step error. A stable type-1 loop can remove step error but retain ramp error. These conclusions assume the exact model, zero initial state, and no input limit. Check closed-loop stability before using the final-error rules.

Read the equations

E(s)=R(s)1+L(s)E(s)=\frac{R(s)}{1+L(s)}

For unity negative feedback, reference error depends on open-loop transfer function L. Final-value claims require the closed-loop pole conditions.

What the symbols mean

E(s)E(s)
Laplace transform of tracking error
R(s)R(s)
Laplace transform of reference
L(s)L(s)
open-loop transfer function
KvK_v
velocity error constant: the limit of sL(s) as s tends to zero

Substitute R=1/s for a unit step or R=1/s² for a unit ramp, and take the final value only after confirming that the closed loop is stable.

SYSTEM TYPE / the error that remains
r+−eL(s)yE = R/(1 + L)unit stepunit ramp1/31/212type 0: L = 2/(s+1)type 1: L = 2/sreference r(t)final error e∞ (t → ∞)

Computed responses of the worked example (gain 2, zero initial state). Step: type 0 keeps e∞ = 1/(1 + 2) = 0.333, type 1 reaches zero error. Ramp r = t: type 1 lags by 1/Kv = 0.50. The plots stop at 2.5 s; the stated errors are the limits as t → ∞.

WORKED EXAMPLE

A fixed target and a moving target

  1. Compare stable unity-feedback loops with gain 2: type 0 has no integrator; type 1 has one. Use zero initial state and unlimited input.

    L0(s)=2s+1,L1(s)=2sL_0(s)=\frac2{s+1},\quad L_1(s)=\frac2s
  2. For a unit step, type 0 retains one-third error while type 1 reaches zero error.

    estep,0=11+2=13,estep,1=0e_{\mathrm{step},0}=\frac1{1+2}=\frac13,\quad e_{\mathrm{step},1}=0
  3. For a unit-slope ramp, type 1 lags by one-half unit in the long run. Adding one integrator solves the fixed-target error but not this moving-target error.

    eramp,1=1Kv=12e_{\mathrm{ramp},1}=\frac1{K_v}=\frac12
REFERENCE

Reference tables

Steady-state error of a stable unity-feedback loop
System typeStep, r = 1(t)Ramp, r = tParabola, r = t²/2
Type 011+Kp\dfrac{1}{1+K_p}∞ (grows without bound)∞ (grows without bound)
Type 101Kv\dfrac{1}{K_v}∞ (grows without bound)
Type 2001Ka\dfrac{1}{K_a}

Error constants: Kp = lim L(s), Kv = lim sL(s), Ka = lim s²L(s) as s → 0. The table assumes unity negative feedback, a stable closed loop, zero initial state and no input limit.

The gain-2 loops of this lecture
LoopClosed-loop polesStep errorRamp errorParabola error
L0=2s+1L_0=\dfrac{2}{s+1}−31/3 ≈ 0.333∞ (grows without bound)∞ (grows without bound)
L1=2sL_1=\dfrac{2}{s}−200.5∞ (grows without bound)
L2=2(s+1)s2L_2=\dfrac{2(s+1)}{s^2}−1 ± j1000.5
More detail and model assumptions
A CLOSER LOOK

Step and ramp ask different questions

A unit step asks the output to hold 1. Its errors are 1/(1+K), 0 and 0 for type 0/1/2. A unit ramp asks the output to keep moving with slope 1. Type 0 has no finite limiting error, type 1 has error 1/K, and type 2 tends to zero. These results follow by substituting R=1/s or 1/s² into E. An integrator stores the correction needed to remove a particular persistent error.

TRY IT

Connect the example to the lab

Compare type 0/1/2 at K=2 with a step, then a ramp. Why can a loop track a fixed target but lag behind a moving one?

Open the concept bench
BE CAREFUL

More integrators are not automatically better. They change stability and required input too. A finite plot shows the last recorded error; the theoretical steady error describes the infinite-time limit. Slow responses may make those two values quite different.

Check your understanding+

In the stable type-1 example with gain 2, what happens to error for a fixed target and for a unit-slope ramp?

Fixed-target error tends to zero, but ramp error tends to 1/2. One integrator removes constant-target error under these assumptions; it does not remove every moving-target error.
PRACTICE PROBLEMS

Solve these before opening the solutions

Work each problem on paper first. Open the answer to check yours, then the full solution if you need the steps, and mark how it went.

  1. Two unity-feedback loops are given: (a) L(s) = 12/((s + 2)(s + 3)) and (b) L(s) = 8/(s(s + 4)). Check that each closed loop is stable, give its type, and find the steady-state error for a unit step and for a unit ramp.

    Show answer

    Answer(a) Type 0: step error 1/3, and the ramp error grows without limit. (b) Type 1: step error 0 and ramp error 1/Kv = 0.5. Both closed loops are stable.

    (a) estep=13, eramp→∞(b) estep=0, eramp=1Kv=0.5\mathrm{(a)}\ e_{\mathrm{step}}=\frac13,\ e_{\mathrm{ramp}}\to\infty\qquad \mathrm{(b)}\ e_{\mathrm{step}}=0,\ e_{\mathrm{ramp}}=\frac1{K_v}=0.5
    Show full solution

    Solution

    1. Check stability first. Both characteristic polynomials are quadratics with positive coefficients, so both pairs of roots lie in the left half-plane.

      (a) s2+5s+18: s≈−2.5±j3.43(b) s2+4s+8: s=−2±j2\mathrm{(a)}\ s^2+5s+18:\ s\approx-2.5\pm j3.43\qquad \mathrm{(b)}\ s^2+4s+8:\ s=-2\pm j2
    2. (a) has no integrator, so it is type 0. Its step error uses L(0) = 2.

      estep=11+L(0)=11+2=13e_{\mathrm{step}}=\frac{1}{1+L(0)}=\frac{1}{1+2}=\frac13
    3. For a ramp, the type-0 loop cannot keep up: its error limit is infinite.

      eramp=lim⁡s→0s⋅1s2⋅11+L(s)=lim⁡s→01s (1+L(s))=∞e_{\mathrm{ramp}}=\lim_{s\to0}s\cdot\frac{1}{s^2}\cdot\frac{1}{1+L(s)}=\lim_{s\to0}\frac{1}{s\,(1+L(s))}=\infty
    4. (b) has one integrator, so it is type 1: zero step error, and a finite ramp error set by Kv.

      Kv=lim⁡s→0sL(s)=84=2,eramp=1Kv=0.5K_v=\lim_{s\to0}sL(s)=\frac84=2,\qquad e_{\mathrm{ramp}}=\frac{1}{K_v}=0.5
  2. A loop has L(s) = 20(s + 1)/(s(s + 2)(s + 5)), and the reference is a ramp of slope 3 rad/s. Confirm closed-loop stability with the Routh test, then find the steady-state tracking error.

    Show answer

    AnswerThe characteristic polynomial s³ + 7s² + 30s + 20 passes the Routh test. Kv = 2, so the error settles at 3/2 = 1.5 rad.

    Kv=2,ess=3Kv=1.5 radK_v=2,\qquad e_{\mathrm{ss}}=\frac{3}{K_v}=1.5\ \mathrm{rad}
    Show full solution

    Solution

    1. Form the characteristic polynomial.

      s(s+2)(s+5)+20(s+1)=s3+7s2+30s+20s(s+2)(s+5)+20(s+1)=s^3+7s^2+30s+20
    2. The first column 1, 7, (210 − 20)/7 ≈ 27.1, 20 is entirely positive, so the loop is stable.

      7⋅30−207≈27.1>0\frac{7\cdot30-20}{7}\approx27.1>0
    3. The loop has one integrator, so it is type 1. Find its velocity error constant.

      Kv=lim⁡s→0sL(s)=20⋅12⋅5=2K_v=\lim_{s\to0}sL(s)=\frac{20\cdot1}{2\cdot5}=2
    4. The ramp error scales with the slope.

      ess=lim⁡s→0s⋅3s2⋅11+L(s)=3Kv=1.5e_{\mathrm{ss}}=\lim_{s\to0}s\cdot\frac{3}{s^2}\cdot\frac{1}{1+L(s)}=\frac{3}{K_v}=1.5
  3. A designer wants to remove the ramp error of a type-1 loop by adding a second integrator. Compare (a) L(s) = 5/s² and (b) L(s) = 5(s + 0.5)/s². Check closed-loop stability for each. For the stable one, find the step and ramp errors, and the error for the parabolic reference r(t) = t²/2, whose transform is 1/s³.

    Show answer

    Answer(a) The poles are ±j2.24, so the loop is not stable and no final error exists. (b) The poles are −0.564 and −4.44: stable, with zero step and ramp errors and a parabolic error of 0.4.

    (a) s=±j2.24(b) s≈−0.564, −4.44;estep=eramp=0,epar=0.4\mathrm{(a)}\ s=\pm j2.24\qquad \mathrm{(b)}\ s\approx-0.564,\ -4.44;\quad e_{\mathrm{step}}=e_{\mathrm{ramp}}=0,\quad e_{\mathrm{par}}=0.4
    Show full solution

    Solution

    1. (a) The characteristic polynomial s² + 5 has no s term, so its roots sit on the imaginary axis and the error oscillates forever.

      s2+5=0 ⇒ s=±j5≈±j2.24s^2+5=0\ \Rightarrow\ s=\pm j\sqrt5\approx\pm j2.24
    2. (b) The zero at −0.5 supplies the missing s term.

      s2+5(s+0.5)=s2+5s+2.5=0 ⇒ s≈−0.564, −4.44s^2+5(s+0.5)=s^2+5s+2.5=0\ \Rightarrow\ s\approx-0.564,\ -4.44
    3. With two integrators, the step and ramp error limits are both zero.

      eramp=lim⁡s→01s (1+L(s))=lim⁡s→0ss2+5s+2.5=0e_{\mathrm{ramp}}=\lim_{s\to0}\frac{1}{s\,(1+L(s))}=\lim_{s\to0}\frac{s}{s^2+5s+2.5}=0
    4. For the parabola, the limit is finite.

      epar=lim⁡s→0s⋅1s3⋅11+L(s)=lim⁡s→01s2+s2L(s)=15×0.5=0.4e_{\mathrm{par}}=\lim_{s\to0}s\cdot\frac{1}{s^3}\cdot\frac{1}{1+L(s)}=\lim_{s\to0}\frac{1}{s^2+s^2L(s)}=\frac{1}{5\times0.5}=0.4
    5. A second integrator helps only if the loop stays stable. Here the zero at −0.5 is what makes that possible.

  4. A position loop has L(s) = K/(s(s + 4)(s + 6)). The specification asks for an error of at most 0.05 for a unit ramp. (a) Find the gain the specification requires. (b) Find the stable range of K with the Routh test. (c) Can a proportional gain meet the specification? What ramp error does K = 120 give, and what would have to change?

    Show answer

    AnswerThe specification needs K ≥ 480, but the loop is stable only for K < 240, so proportional gain alone cannot meet it. K = 120 gives a ramp error of 0.2. The controller needs more gain at low frequency without the same loss of stability, which is what dynamic compensation provides.

    Kv=K24≥20 ⇒ K≥480;Routh: 0<K<240;K=120: eramp=0.2K_v=\frac{K}{24}\ge20\ \Rightarrow\ K\ge480;\qquad \mathrm{Routh}:\ 0<K<240;\qquad K=120:\ e_{\mathrm{ramp}}=0.2
    Show full solution

    Solution

    1. The loop is type 1. Turn the ramp specification into a minimum Kv and hence a minimum K.

      Kv=lim⁡s→0sL(s)=K24,eramp=24K≤0.05 ⇒ K≥480K_v=\lim_{s\to0}sL(s)=\frac{K}{24},\qquad e_{\mathrm{ramp}}=\frac{24}{K}\le0.05\ \Rightarrow\ K\ge480
    2. Form the characteristic polynomial.

      s(s+4)(s+6)+K=s3+10s2+24s+Ks(s+4)(s+6)+K=s^3+10s^2+24s+K
    3. The cubic Routh condition ab > c gives the upper limit.

      10×24>K ⇒ 0<K<24010\times24>K\ \Rightarrow\ 0<K<240
    4. The two requirements do not overlap. Even close to K = 240 the error only falls toward 0.1, with no stability margin left. A practical gain such as K = 120 is stable, with poles −8.82 and −0.589 ± j3.64.

      K=120:eramp=24120=0.2K=120:\quad e_{\mathrm{ramp}}=\frac{24}{120}=0.2
    5. Gain alone trades error against stability. Meeting both needs a controller with more gain at low frequency than at the frequencies that decide stability, such as the lag compensator of a later lecture.

Mark each exercise after you check your answer. Marks stay in this browser.

KEY TERMS

Terms this lecture teaches

All terms in the glossary →
QUESTIONS STUDENTS ASK

Frequently asked questions

Is system type the same as system order?

No. Order counts all the poles of a model; type counts only the uncancelled poles at the origin of the open-loop transfer function. A third-order loop with one integrator is type 1.

What is the steady-state error of a type-1 system to a ramp input?

For a unit-slope ramp it is 1/Kv, where the velocity error constant Kv is the limit of sL(s) as s → 0. For the stable loop L(s) = 10/(s(s + 2)), Kv = 5, so in the long run the output lags the ramp by 0.2.

Why must I check stability before using the error constants?

The error constants come from the final-value theorem, which is valid only for a stable closed loop. An unstable loop still produces a number from the formula, but its error grows instead of settling, so that number means nothing.

Does adding an integrator always improve tracking?

No. Each integrator removes the steady error for one more class of reference but adds 90° of phase lag at every frequency, which erodes stability margins and can destabilize the loop. It also tends to demand more actuator effort during transients and cannot help once the actuator saturates.

Can a type-0 system ever track a step with zero error?

Not with feedback alone. A stable type-0 loop settles with step error 1/(1 + L(0)); with L(s) = 2/(s + 1) that is 1/3. Raising the gain shrinks the error but never removes it, and stability limits how far the gain can go.

Why the next lesson follows

We now shift from checking a chosen loop to designing one: root locus shows how gain moves closed-loop poles.

YOUR PROGRESS

Finished the reading and the exercises? Mark the lecture as done to tick it on the syllabus. Progress stays in this browser.

EIGENROOM · https://eigenroom.com/theory/system-type · Updated 2026-11-15