Back

Root locus and control design

Change poles, zeros and gain to check stability and the step response.

G(s) = 1 / [(s + 1)(s + 3)] · unity negative feedback

Static initial example (K=5). It becomes interactive when the calculation loads.

0 · no feedback gain100

1 ● Pole positions K=5

Left: decaying modesRight: growing modes-6.762.6-1.53.76-2.6Re(s) [1/s]Im

× open-loop pole ○ zero ● current pole

2 How the response changes 0–10 s

-0.0800.551.0810Output y (dotted: input 1)Time [s]

Computed with all current poles and zeros · vertical axis auto-scaled

Stable

All roots are on the left. Rightmost real part −2. y(10 s) ≈ 0.625.

Change poles, zeros and model

Presets replace the poles and zeros. K stays the same. Re-opening the guide does not change the model.

In the 3rd-order example the boundary is K=90. Move K and compare the poles that reach the right side with the response.

Open-loop poles (up to 4)
Real [1/s]Imag [rad/s]
Open-loop zeros (up to 3)
RealImag

Leave the real part empty to skip that row. An empty imaginary part means 0. Conjugate pairs are not added automatically. Non-conjugate input is treated as a complex model, with real and imaginary parts shown separately.

Automatic K sweep and a second K control

While playing, K of the current model increases and stops at 100. Speed is the on-screen sweep speed.

It does not play automatically.

Plot range and response time

The pole plot uses the same scale on both axes. It starts auto-fitted so the current poles are visible. The faint line is a sampled locus for K=0–200; points outside the range are omitted.

Drag the plot to pan. With the plot focused, arrow keys pan and +/− zoom.

Roots, stability and numbers in detail

Damping ratio ζ of the dominant root
0.707
|s| of the dominant root [rad/s]
2.83
Response overshoot (within the window) %
4.32
Time from which the response stays within 2% (in window) s
—
Real part of the rightmost root [1/s]
−2
Stability boundary K found by the scan
not found in 0–200

ζ and |s| describe one dominant root. Overshoot and settling of a general higher-order system are not decided by these two values alone. “Boundary not found” does not guarantee stability for every K.

URL and PNG for this setting

Formulas and model scope

Open loop G(s)=N(s)/D(s); with unity negative feedback the characteristic equation is D(s)+KN(s)=0. The unit-step output is computed from T(s)=KN(s)/(D(s)+KN(s)) with zero initial conditions.

If every characteristic root has a negative real part, the internal modes are asymptotically stable. Treat roots near the imaginary axis as a boundary, and also check repeated roots and cancellations. At K=0 or with pole–zero cancellation, an unstable internal mode can remain even if the output does not grow.

Roots and responses are floating-point numerical results. The response axis is auto-scaled; if values grow too large, the curve is drawn up to that point and a stop is shown. The number of digits does not indicate real-machine accuracy. The model is linear and time-invariant; saturation, dead time and nonlinearity are not handled.

Initial 2nd-order example: D=s²+4s+3, N=1, K=5. The roots are −2±2j, T=5/(s²+4s+8), and the final value is 5/8. Even if the response looks faster, also check oscillation and steady-state error.

Root-locus definition and pole/zero notation (MathWorks) · MIT: feedback and the root locus

A calculation example for comparing linear models. It does not guarantee the stability of a real machine.

Theory & Key Formulas

Start: K=0 → open-loop poles

End: K→∞ → open-loop zeros (remainder to ∞ along asymptotes)

Asymptote angles: $(2k+1) \cdot \dfrac{180°}{n-m}$, $k=0,1,...,n-m-1$

Centroid: $\sigma_a = \dfrac{\sum p_i - \sum z_i}{n-m}$

Stability condition: all closed-loop poles in left half-plane

What you can check with this calculator

Set open-loop poles and zeros and vary the gain K to draw the root locus in the s-plane in real time. Asymptotes, stability margin and step response are computed automatically.

Check the formulas and key points
Key points
The root locus is the path of the closed-loop poles as the gain K goes from 0 to ∞. It starts at the open-loop poles at K=0 and heads to the open-loop zeros (the rest to infinity) as K→∞. The asymptote angles are (2k+1)·180°/(n−m) and the centroid is σ_a = (Σpᵢ − Σzᵢ)/(n−m) (n = number of poles, m = number of zeros).The system is stable when all closed-loop poles lie in the left half-plane.

About the Root Locus Control System Designer

Explained as a conversation
🙋
What exactly can I learn by looking at the "root locus" in this simulator?
🎓
Roughly speaking, you can see at a glance whether the system stays stable, oscillates or diverges as you raise the control gain $K$. For example, when you increase $K$ from 0 with the slider above, the yellow ● (closed-loop poles) start moving away from the × (open-loop poles). If they stay to the left of the imaginary axis the system is stable; if they move to the right it becomes unstable.
🙋
I see! Then why can I place the poles and zeros myself from the start?
🎓
Because in practice the characteristics of the object you want to control (the plant), such as a motor or a sensor, are expressed by poles and zeros. Modelling a motor's inertia and friction, for example, puts a pole at the origin. Move the poles in this tool and you can immediately see how the starting points of the locus change and how the range of $K$ that keeps the system stable shifts.
🙋
How are the "asymptotes" (the orange dashed lines in the root-locus plot) and the "breakaway points" where the locus leaves the real axis used in design?
🎓
Good question! The asymptotes show the directions the poles head towards when $K$ is very large. From them you can predict whether the poles enter the unstable region (right half-plane) at high gain. A breakaway point is where the locus leaves the real axis; adjusting $K$ around it makes oscillatory modes appear. Try adding one zero in the simulator – you will see the number and angles of the asymptotes change.

Frequently Asked Questions

Set the value with the gain-K slider (0–100): the closed-loop poles for that K are shown as yellow ● on the root-locus plot, and the dominant pole, damping ratio, ωn, overshoot, settling time and step response are updated. The animation at the top sweeps K automatically, and you can stop it at any value with its K slider. The root-locus plot is auto-fitted, and you can zoom in on details with the zoom buttons, the mouse wheel, dragging, or by entering the Re/Im range.
The asymptotes give the directions of the straight lines that the root locus approaches as the gain K grows. Their angles are set by the numbers of poles and zeros, and the centroid (their intersection on the real axis) represents the centre of the locus. Checking them lets you predict the stability tendency of the system at high gain.
The stability margin this tool shows is the critical gain K_crit: the smallest K at which the largest real part of the closed-loop poles reaches 0, found from the exact roots of the characteristic equation with bisection (for poles −1, −2, −4, K_crit=90). The ratio K_crit/K to the current K corresponds to the gain margin. The phase margin is not shown, so check it with a Bode-plot tool.
Enter values in the real and imaginary fields of "Open-loop poles" (up to 4) and "Open-loop zeros" (up to 3) in the left panel to add them; leave a field empty to disable (remove) it. Enter complex poles and zeros as conjugate pairs (e.g. −2+2j and −2−2j). Changes are reflected in the root locus immediately. You can also start from a preset (2nd order, 3rd order, with PD, integrator).

Real-World Applications

Servo motor control: To balance positioning accuracy and response speed, the root locus is used to determine a suitable range for the proportional gain $K_p$. From the locus you look for the region where the poles lie in the left half of the complex plane and have a moderate damping ratio.

Vibration suppression of robot arms: The structural flexibility (elasticity) of an arm is modelled as complex-conjugate poles. With the root-locus method you can visually choose a gain that damps these vibration modes sufficiently while also making the rigid-body response fast.

Active suspension in cars: Ride comfort (body vibration) and road holding (tyre motion) are in a trade-off. Several controller candidates with different pole and zero placements are compared on the root locus to choose quickly the design that gives the best balance.

Detailed design with CAE: In control design for mechanical systems, a tool like this is first used with the root-locus method to grasp the rough gain range and stability margin. The usual workflow then moves on to simulation and frequency-response analysis with higher-fidelity models in MATLAB/Simulink or Python's control library to refine the design.

Common Misconceptions and Pitfalls

When you start using this tool there are a few points to keep in mind. First, it is a waste to "place poles and zeros only on the real axis". With poles only on the real axis the response does not oscillate and has no overshoot. But when a fast response is needed in practice, you inevitably have to accept moderate oscillation (damped oscillation) and aim for complex-conjugate poles. Try comparing in this tool how the step response changes between poles at −2 and −3 on the real axis and poles at −2±2j.

Next, the misconception that "any zero improves stability". Left-half-plane zeros (minimum-phase zeros) do speed up the response, but right-half-plane zeros are very hard to deal with. For example, with poles at −1 and −5 and a zero at +2 (right half-plane), raising K quickly drives the locus into the right half-plane and the system becomes unstable. This is called a non-minimum-phase system and is the cause of the inverse response seen in real machines.

Finally, do not think that "the root locus tells you everything". The root locus is ideal for visualising transient characteristics (stability, frequency, degree of damping), but it does not directly tell you disturbance-rejection performance or robustness (tolerance of model errors). In practice, after choosing a rough gain with the root locus, the rule is to check gain and phase margins with frequency-response analysis using Bode plots.

How to Use

  1. Enter the poles (up to 4) and zeros (up to 3) of the open-loop transfer function as real and imaginary parts. Enter complex poles as conjugate pairs (e.g. −10 and −5±j8)
  2. The root locus is drawn automatically in the s-plane, and the gain-K slider (0–100) shows where the closed-loop poles (●) lie for that K. Each branch runs from an open-loop pole to a zero (or to infinity). The view is auto-fitted and can be changed with the zoom buttons, the mouse wheel, dragging or the Re/Im range inputs
  3. The damping ratio ζ, natural frequency ωn, overshoot, settling time Ts and critical gain K_crit update in real time. Choosing a gain inside the stable region (left half-plane) lets you achieve the target response

Worked Example

For an industrial DC speed-control loop with open-loop poles p1=−1, p2=−2, p3=−4 (no zeros), at K=20 the dominant poles are −0.82±j2.13, giving a damping ratio ζ=0.358, natural frequency ωn=2.28 rad/s, overshoot 30.0% and settling time (±2%) Ts=4.89 s. The critical gain is K_crit≈90; beyond it the root locus enters the right half-plane and the system becomes unstable.

Practical Notes

  1. In complex higher-order systems (5th order and above) crossing points of the root locus can become multiple roots, and a Nyquist plot is needed alongside for detailed local stability assessment
  2. For control systems with dead time (e.g. conveyor-belt speed control with τ=0.15 s) the root-locus method cannot be applied directly; modelling with a Padé approximation is a prerequisite
  3. Designing for ζ>0.7 keeps overshoot below 5%, but the settling time grows, making it hard to also achieve a fast response
  4. When there are several controlled axes, as in magnetic bearings or robot joints, each axis has different dominant poles, so a multivariable gain setting that accounts for cross-axis coupling is essential

Standards & Assumptions

Basis/reference: classical control, Evans root locus. Characteristic equation 1+K·G(s)H(s)=0. Asymptote centroid σ_a=(Σp_i−Σz_j)/(n−m), asymptote angles (2k+1)·180°/(n−m), angle condition ∠G(s)H(s)=±180°.

Model assumptions: LTI, rational transfer function with real coefficients (complex poles in conjugate pairs). Standard locus for negative feedback with proportional gain K≥0. Dead time, nonlinearity and discretisation are out of scope.

Scope and limits: for education and analysis. Check case: poles −1, −2, −4 (no zeros) give σ_a=−2.33 and asymptote angles 60°/180°/300°, matching this tool. The closed-loop poles are computed directly as the roots of the characteristic equation D(s)+K·N(s)=0 (up to 4 poles). The locus is drawn by sweeping the gain logarithmically from 10⁻⁶ to 200, so branches leaving closely spaced poles can be seen by zooming in.