Time Response of Control Systems for GATE: Damping, Overshoot, Settling Time and Steady-State Error
Time response for GATE: first order systems, second order standard form, damping ratio, overshoot, peak and settling time, steady-state error and practice questions.
In the block diagram post we learnt to squeeze any feedback system down to one closed-loop transfer function. That was the bookkeeping. Now comes the part that tells you whether the system is any good: how does the output actually move when you give it a command?
Think of a lift. Press the button and you want it to reach your floor quickly, without shooting past and bouncing, and to stop exactly level with the floor. Those three wishes, speed, little overshoot and small final error, are exactly what GATE measures with the time response formulas in this post. They come up every single year in both EE and EC, and once you know the standard forms, most questions are one line of arithmetic.
The test input: a unit step
We judge a system by its response to a unit step, a sudden jump from 0 to 1, because it's the harshest ordinary command you can give. Its Laplace transform is $1/s$, so the output is
where $T(s)$ is the closed-loop transfer function. Everything below comes from taking the inverse Laplace transform of this.
First order systems
A first order system has one pole:
Its step response is the curve we met in the RC circuit post:
Only one number, the time constant $\tau$, describes it. After $\tau$ seconds the output has covered 63.2% of the way. It never overshoots, and the useful time measures are
A smaller $\tau$ means a pole further to the left, at $s = -1/\tau$, and a faster system.
Second order systems: the standard form
Most GATE questions are about second order systems, because they're the simplest ones that can overshoot and ring. Every one of them is compared with this standard form:
Two numbers describe it:
$\omegan$, the natural frequency, the frequency it would oscillate at with no damping at all. $\zeta$ (zeta), the damping ratio, which decides how much it overshoots.
The poles sit at
where $\omegad$ is the damped frequency, the rate at which the output actually rings. Picture where these poles sit and the whole topic falls into place:
Four kinds of damping
The value of $\zeta$ sorts every second order system into one of four families:
Damping ratio Name Poles Step response ------------ $\zeta = 0$ undamped on the $j\omega$ axis oscillates forever $0 1$ overdamped two different real poles slow, no overshoot
Here's what that looks like, all with $\omegan = 1$ rad/s:
The lightly damped curve shoots past the final value by more than half and takes a long time to calm down. Most real controllers are tuned somewhere around $\zeta = 0.4$ to $0.8$, which trades a little overshoot for a much quicker rise.
The time-domain specifications
For an underdamped system ($0 < \zeta < 1$), these five formulas answer almost every GATE question:
Notice what each one depends on. Overshoot depends only on $\zeta$, so it's set by the angle $\theta$ in Fig. 1. Settling time depends only on $\zeta\omegan$, the distance of the poles from the imaginary axis. Peak time depends only on $\omegad$, their height.
Worked example
A unity feedback system has open-loop transfer function
Find the overshoot, peak time, rise time and settling time.
The method is always the same:
Step 1. With unity feedback, $T = G/(1 + G)$:
Step 2. Match with $s^2 + 2\zeta\omegan s + \omegan^2$. So $\omegan^2 = 25$, giving $\omegan = 5$ rad/s, and $2\zeta\omegan = 6$, giving $\zeta = 0.6$.
Step 3. Work out the rest:
That's every specification from one transfer function in under two minutes.
Steady-state error
The last wish on our lift list was to stop exactly at the floor. For a unity feedback system, the final error is
It depends on the input and on the type of the system, which is the number of poles of $G(s)$ at the origin, the number of pure integrators:
Input Error constant Type 0 Type 1 Type 2 --------------- Step, $1/s$ $Kp = \lim{s\to 0} G(s)$ $\dfrac{1}{1 + Kp}$ 0 0 Ramp, $1/s^2$ $Kv = \lim{s\to 0} s\,G(s)$ $\infty$ $\dfrac{1}{Kv}$ 0 Parabola, $1/s^3$ $Ka = \lim{s\to 0} s^2 G(s)$ $\infty$ $\infty$ $\dfrac{1}{Ka}$
Our worked example has one integrator, the $s$ in the denominator, so it's type 1. It follows a step with zero error. For a ramp, $Kv = 25/6 \approx 4.17$, so it lags behind by $e{ss} = 1/Kv = 0.24$.
Common traps in GATE questions
These are the mistakes I see most often.
Using the open-loop transfer function. The standard form, and every formula above, applies to the closed-loop transfer function. Close the loop first. Forgetting to normalise. The $s^2$ term must have coefficient 1 before you match. Divide through if it doesn't. Mixing up ωn and ωd. Peak time and rise time use $\omegad$. Settling time uses $\zeta\omegan$. The wrong settling band. $4/(\zeta\omegan)$ is for the 2% band and $3/(\zeta\omegan)$ is for 5%. Read which one the question wants. Thinking overshoot depends on ωn. It doesn't. Two systems with the same $\zeta$ overshoot by the same percentage, however fast they are. Getting the system type wrong. Count poles at the origin in the open-loop $G(s)$, not in the closed-loop transfer function. Using the error table with non-unity feedback. The table assumes $H = 1$. Otherwise, go back to the full limit formula.
Practice questions
Give yourself about two minutes each.
A system has $T(s) = \dfrac{1}{2s + 1}$. Find its time constant, its 2% settling time, and its output at $t = 2$ s for a unit step. A second order system has $\zeta = 0.5$. What is its percentage overshoot? For $T(s) = \dfrac{16}{s^2 + 4s + 16}$, find $\omegan$, $\zeta$, the peak time and the 2% settling time. Describe the step response of $T(s) = \dfrac{9}{s^2 + 9}$. A unity feedback system has $G(s) = \dfrac{K}{s(s + 4)}$. What value of $K$ gives $\zeta = 0.5$? A unity feedback system has $G(s) = \dfrac{10}{(s + 1)(s + 2)}$. Find the steady-state error for a unit step.
Answers
$\tau = 2$ s, $ts = 4\tau = 8$ s, and $c(2) = 1 - e^{-1} \approx 0.632$. $e^{-0.5\pi/0.866} = e^{-1.814} \approx 0.163$, so about 16.3%. $\omegan = 4$ rad/s and $2\zeta \times 4 = 4$ gives $\zeta = 0.5$. Then $\omegad = 4 \times 0.866 = 3.46$ rad/s, $tp = \pi/3.46 \approx 0.91$ s, and $ts = 4/(0.5 \times 4) = 2$ s. $\zeta = 0$, so it's undamped. The output oscillates between 0 and 2 at 3 rad/s and never settles. The characteristic equation is $s^2 + 4s + K = 0$. Then $2\zeta\omegan = 4$ with $\zeta = 0.5$ gives $\omegan = 4$, so $K = \omegan^2 = 16$. It's type 0 with $Kp = 10/2 = 5$, so $e{ss} = 1/(1 + 5) \approx 0.167$.
If question 5 slowed you down, remember the trick: match the characteristic equation with $s^2 + 2\zeta\omegan s + \omegan^2$ term by term, and each unknown falls out. Next time we'll ask the big question every control engineer worries about: is the system stable at all? That's the Routh-Hurwitz criterion. Tell me in the comments which question took you longest.