GATE Electronics (EC): Control Systems (block diagrams, time response, Routh and root locus, Bode, Nyquist, compensators, state space), previous-year questions solved
Every control systems question from the last five GATE Electronics and Communication papers, solved: block-diagram and signal-flow-graph reduction, time response and steady-state error, Routh-Hurwitz and root locus, Bode and Nyquist plots with gain and phase margins, compensators and state-space models. Each answer matches the official key and was recomputed by code: a simulated step response, the roots of the characteristic equation, a numeric frequency sweep or a matrix exponential.
Questions from official GATE EC papers (IIT Kharagpur 2022, IIT Kanpur 2023, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 17 Nov 2026 · 7 of 25 questions on this page
Block diagram with R(s), D(s), G(s), H(s) and a unity outer loop: find G1(s) and G2(s) in Y = G1 R + G2 D
and are given by
- (A) and
- (B) and
- (C) and
- (D) and
Answer: (A) Official key: A
- Name the output of the first junction (unity feedback from ), and the output of the second junction . The block gives .
- Substitute: .
- Collect the terms: .
- So , which gives , option (A).
- Mason's gain formula gives the same: the forward path from (or from ) to is . The loops are (outer unity loop) and (inner loop), and both pass through , so they touch each other and the forward path. and the path factor is .
- Both inputs see the same denominator because they enter ahead of the same block and the same two feedback loops.
Trap: options (B), (C) and (D) put where belongs. The inner feedback signal leaves , goes through , and then the loop is closed through , so the loop gain is , not . The code solved the junction equations in sympy and got exactly (and the same from Mason's formula). It then solved the node equations as a numeric linear system with and at : , while the denominator of option (B) gives .
This is the reference-and-disturbance structure of a cruise-control or motor-speed loop: is the set-point and is a load disturbance such as road gradient or load torque. Writing both transfer functions shows the designer how much of the disturbance reaches the output and sets how high the loop gain must be to suppress it.
Block diagram with G1 in a unity feedback loop and G2 in parallel: which signal flow graphs match?
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- Read the block diagram. The first summing junction gives the error , where is the output of , fed back with a minus sign. The signal also goes to (the pickoff point sits right after the junction). The output adds the two: .
- Solve: , so . Only is inside the loop; is a parallel path from the error signal.
- Graph (B): forward paths with gain and with gain . One loop, , with gain . Both paths touch it, so and . It matches.
- Graphs (A) and (D): the feedback branch is taken after and have been added, so the loop gain is and . Wrong: in the diagram the feedback comes from the output of alone.
- Graph (C): the feedback branch has gain instead of and also closes around , so . Wrong.
- Only graph (B) represents the system: option (B).
Trap: option (A) or (D) is chosen when the feedback is drawn around both parallel branches. The code applied Mason's gain formula to each graph (it enumerated every forward path and every loop itself and checked the result against a separate node-equation solve in sympy), and compared with the block-diagram algebra. Results: (A) , (B) , (C) , (D) . Only (B) equals the block diagram result, also confirmed numerically at , : .
Signal flow graphs and Mason's formula are used to reduce a multi-loop model, for example a motor speed loop with a feed-forward path, to one transfer function before tuning. A feed-forward path like G2 here (taken from the error and added at the output) does not change the loop poles, which are set by alone.
Two blocks 1/(s+4) with a second summer subtracting one from the other in the feedback: transfer function?
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- Name the signals. is the output of the first summer, the upper block output, the lower block output and the output of the second summer (the feedback signal).
- First summer: . Both blocks are fed from : and .
- Second summer: it adds and subtracts , so .
- The feedback signal is zero for every input, so the loop is effectively open. Then and .
- , option (B).
Trap: option (D), . It comes from treating the two blocks as a cascade inside a unity-feedback loop: . In the figure the lower block is not in series with the upper one: both take their input from the first summer, and the second summer cancels them. The code solved the four loop equations as a linear system at complex values of and compared with all four options (only (B) matches). Sympy gave and . It also confirmed that the cascade mistake gives exactly option (D), and that adding the lower block instead of subtracting it would give , which is not an option.
Signals that cancel exactly, such as a feed-forward term that subtracts a copy of the same disturbance, are used in motor and power-converter control to remove a known interaction before it reaches the loop. The first step in reducing any block diagram is to check what each summer actually adds and subtracts, since a loop whose feedback is zero has no feedback at all.
G1 = 10/(s^2+s+1) and G2 = 10/(s^2+s sqrt(10)+10): which step-response statements are true?
and
have unit step responses and , respectively. Which of the following statements is/are true?
- (A) and have the same percentage peak overshoot.
- (B) and have the same steady-state value.
- (C) and have the same damped frequency of oscillation.
- (D) and have the same settling time.
Answer: (A) Official key: A
- Compare each denominator with .
- : , so rad/s, and , so . : , so rad/s, and , so .
- (A) Percentage overshoot depends only on : for both. True.
- (B) Steady-state value of the step response is : and . Different. False.
- (C) Damped frequency : rad/s for and rad/s for . Different. False.
- (D) settling time is about : for and for . Different. False.
- Only (A) is true: (A).
Trap: seeing the same-looking structure and the same damping ratio and concluding that everything matches, or reading the overshoot as a number that must move with the DC gain. Overshoot as a percentage of the final value ignores the gain of and the speed . The code integrated both step responses with RK4 and read the curves. Overshoot: and (formula ). Final value: and . Damped frequency from the peak-to-trough time: and rad/s. settling time: and . Only the overshoots agree.
Percentage overshoot is set by damping ratio alone, so a loop retuned to run faster by scaling every pole by the same factor keeps its overshoot and loses only settling time. That is how a servo or a regulator is sped up without changing how much it rings, as long as the actuator can supply the extra bandwidth.
Gain 10 then 1/(s(s+10)) with a unit step disturbance after the gain: steady-state error e(inf)?
Answer: Official key: -0.11 to -0.09
- With the error is . The gain block outputs , and the disturbance is added after it.
- The plant gives .
- Multiply out: , so . Its poles are and , both in the left half plane, so the final value theorem applies.
- Unit step: . .
- .
Trap: the plant has an integrator, so it is tempting to say the system is type 1 and the error to a step is zero, which gives . That is true for a step in (the code simulated the same loop with , and got ). But here the disturbance enters after the gain block and no integrator sits between the error and the disturbance point, so the loop only holds to of . The second slip is the sign: with gives , not , and is outside the key. The code solved the loop by sympy (), took the final value theorem (), and simulated with RK4 to ().
Disturbance rejection like this is what a cruise-control or motor-speed loop does when load torque or a hill suddenly changes. With only proportional gain ahead of the disturbance the speed settles off target by disturbance divided by gain, and adding an integrator in the controller, ahead of where the load enters, is what removes that offset.
Steady-state error to a ramp R(s) = alpha/s^2 in a unity feedback loop with k/(s(s+2))
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- The loop has unity feedback, so the error is with .
- The closed loop is stable for : the characteristic equation has both roots in the left half plane. So the final value theorem can be used.
- .
- Equivalently, the velocity error constant is and .
- , option (A).
Trap: option (B), , comes from taking , that is, reading as the gain without noticing that the pole is , not . At the factor is , not . The code integrated with RK4 (step ms, to s) and read : for it got against (option B would give ); for it got against ; for it got against . Only (A) matched all three.
A tracking servo that follows a moving target at constant speed, such as an antenna or a camera gimbal, lags the target by a fixed error equal to speed divided by the velocity error constant. Designers raise , or add an integrator to get a type-2 loop, when that lag is too large.
Unity feedback, G(s) = 6/(s(s+1)(s+2)), unit step input: what is the limit of e(t)?
- (A)
- (B)
- (C)
- (D) does not exist, is oscillatory
Answer: (D) Official key: D
- The error is with . The final value theorem gives a limit only if every pole of is in the open left half plane. So check stability first.
- Characteristic equation: , that is .
- Routh array: row : . Row : . Row : . Row : . The whole row is zero, so there are roots on the axis.
- Auxiliary equation from the row: , so . Dividing out gives , so the closed-loop poles are and .
- The poles at give a sustained oscillation in and in that never dies out. The final value theorem does not apply, and keeps oscillating about its mean.
- So does not exist and is oscillatory, option (D).
Trap: option (A) comes from the error-constant shortcut. has one pole at the origin (a type 1 system), so and . That shortcut is only valid when the closed loop is stable, and here it is not (it sits at the stability limit). The code simulated the closed loop exactly with a matrix exponential (step s, up to s) and read over s to s: it swings by 1.809 peak to peak around a mean of 0.00, and never settles. The oscillating term has amplitude 0.905 (computed from the residue at ). Option (B), , is with , the error for a unit ramp, not a unit step, and again it presumes a stable loop. None of the values in (A), (B) or (C) is a limit of the simulated .
Raising loop gain until the loop just oscillates is the Ziegler-Nichols way to find the ultimate gain of a PID loop, and the same Routh test tells a drive engineer how much gain a speed loop can take before it rings forever. The oscillation frequency of rad/s here is the phase-crossover frequency of the loop.