GATE Electrical: Control Systems (Bode, Nyquist, compensators, state space), previous-year questions solved
Part 2 of the GATE EE Control Systems questions from the official 2022-2026 papers: transfer functions read off Bode plots, ramp error from the Bode slope, Nyquist encirclements, lead, PI and PD design for a target phase margin, and state-space poles, zeros and similarity transforms. Every solution matches the official answer key and notes the common traps. Block diagrams, time response and stability are on part 1.
Questions from official GATE EE papers (IIT Kharagpur 2022, IIT Kanpur 2023, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 16 Nov 2026 · 13 questions
First-order stable system with flat Bode magnitude and phase going to −180°: what are its poles and zeros?
- (A)one LHP pole and one RHP zero at the same frequency.
- (B)one LHP pole and one LHP zero at the same frequency.
- (C)two LHP poles and one RHP zero.
- (D)two RHP poles and one LHP zero.
Answer: (A) Official key: A
- First order and stable: exactly one pole, in the left half plane. That rules out (C) and (D).
- Constant magnitude: the zero must sit at the same frequency as the pole, so for every if the zero is at .
- Phase: the LHP pole contributes ; a RHP zero also contributes , so the total goes to . An LHP zero would contribute and cancel the pole completely (that is option (B), a constant).
- One LHP pole and one RHP zero at the same frequency: option (A), the all-pass filter .
We checked a representative of each option numerically from 0.001 to 1000 rad/s: only has a flat magnitude and phase from to .
All-pass behaviour is a non-minimum-phase effect: the system responds in the 'wrong' direction first. It shows up in boost converters, in steam drums (swell) and as a Padé approximation of a pure time delay in controller design, and it limits how fast any loop around it can be tuned.
Flat 8 dB magnitude and phase falling linearly to −60° at 1 rad/s: what is the transfer function?
- (A)
- (B)
- (C)
- (D)
Answer: (D) Official key: D
- A flat magnitude with a phase that falls in proportion to (a straight line on a linear frequency axis) is a gain times a pure delay: , with and .
- Gain: dB, so .
- Delay: rad, so s.
- : option (D).
Distractors: (C) has the right delay in degrees ( is not ) but the wrong gain; (B) has a term, so its magnitude would roll off; (A) has a delay of 0.032 s, which gives only at 1 rad/s. Note the phase plot uses a linear axis while the magnitude uses .
Transport delays are everywhere in real plants: conveyor belts, long pipes in process plants, network latency in remote control, and the computation delay of a digital controller. Because a delay adds phase lag that grows with frequency, it sets a hard limit on loop bandwidth, which is why Smith predictors exist.
Asymptotic magnitude falls at −20 dB/decade to 0 dB at ω0, then stays flat: which G(s)?
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- A slope of dB/decade from the lowest frequencies means a pole at the origin (an integrator).
- The slope changes by dB/decade at to become flat, so there is a zero at .
- The line reaches 0 dB exactly at : the integrator is . So : option (A).
(B) is a high-pass filter (+20 dB/decade below ); (C) and (D) have a flat 0 dB magnitude at every frequency. Checked numerically: only (A) has slope dB/decade well below and 0 dB well above it. This is exactly a PI controller, with , .
This is the Bode shape of a PI controller: infinite gain at DC (no steady-state error) and plain proportional gain above the corner. Drive and converter engineers place the PI zero a few times below the crossover frequency so the integrator's phase lag does not eat into the phase margin.
Bode slope −20 dB/decade with gain crossover at 10 rad/s: what is the steady-state error for a unit ramp?
Answer: 0.10 Official key: 0.09 to 0.11
- Constant dB/decade slope and constant phase at low frequency: there (a type-1 system).
- The asymptote crosses 0 dB at 10 rad/s: , so .
- Velocity error constant .
- Unit-ramp error .
For a type-1 system the low-frequency asymptote always crosses 0 dB at , so you can read straight off the plot. That trick is the whole question.
Ramp tracking error is what a radar antenna, a CNC axis following a contour or a solar tracker shows when following a steadily moving target. Designers read off the Bode plot to check that the following error stays within tolerance at the maximum feed rate.
G(s) = 1/(s(s+1)) with the contour bulging left round the origin: how many times is −1 encircled?
- (A)0
- (B)1
- (C)2
- (D)3
Answer: (B) Official key: B
- Nyquist criterion: , where is the number of open-loop poles inside the contour, the number of closed-loop poles inside it and the number of clockwise encirclements of .
- The small detour goes into the left half plane, so the pole at is inside the contour: .
- Closed loop: , roots . Neither is inside (neither is in the right half plane or at the origin): .
- : one anticlockwise encirclement. , option (B).
We traced the drawn contour numerically and counted the winding of round the origin: turn (anticlockwise). With the usual detour into the right half plane instead, and : the closed loop is the same stable system either way. The choice of detour changes the count, not the conclusion.
The Nyquist plot is the tool of choice when the plant has delays, right-half-plane poles or measured (not modelled) frequency response: frequency-response analysers inject a sweep into a running power supply or drive and the stability margins are read off the measured curve.
G(s)H(s) = (3s+5)/(s−1): where does the infinite semicircle of the Nyquist contour map?
- (A)
- (B)
- (C)
- (D)
Answer: (C) Official key: C
- On the infinite arc, with .
- for every .
- The whole arc collapses to the single point : option (C).
General rule: the infinite arc maps to . That is 0 for strictly proper functions (more poles than zeros) and the ratio of leading coefficients for biproper ones like this. is the value at , a different point of the plot.
Biproper loop transfer functions (as many zeros as poles) come from lead compensators and PD action, which keep gain at high frequency. That is why practical PD controllers always add a roll-off filter: unlimited high-frequency gain amplifies sensor noise.
Lead compensator with maximum phase lead at 4 rad/s and 6 dB asymptotic gain there: what are a and β?
- (A)1, 16
- (B)2, 4
- (C)3, 5
- (D)2.66, 2.25
Answer: (B) Official key: B
- The zero is at and the pole at . Maximum phase lead occurs at their geometric mean: .
- The asymptotic magnitude rises at 20 dB/decade from , so at it is dB.
- , and . Option (B): 2, 4.
Check all four: (A) gives but dB; (C) gives rad/s; (D) gives but only dB. The exact (not asymptotic) gain at would be too, by symmetry, so 6 dB holds either way here.
Lead compensators add phase margin near crossover; the maximum lead ( for ) is placed right at the new crossover frequency. Power-supply designers do the same thing with the 'Type 2' and 'Type 3' compensator networks around an op-amp error amplifier.
Integral controller KI/s with tabulated G(jω): what KI gives a 50° phase margin?
| in rad/sec | Magnitude in dB | Phase in degrees |
|---|---|---|
| 0.5 | −7 | −40 |
| 1.0 | −10 | −80 |
| 2.0 | −18 | −130 |
| 10.0 | −40 | −200 |
Answer: 1.12 Official key: 1.11 to 1.13
- Loop transfer function . The integrator adds at every frequency.
- A 50° phase margin needs at the gain crossover, so : from the table, rad/s, where dB .
- Crossover means : , so .
Trap: picking the row with forgets the integrator's own . With the loop gain at 0.5 rad/s is exactly 1 and the phase is , which we recomputed from the table values.
Tuning from a measured frequency-response table is routine when no model exists: a network analyser gives magnitude and phase at a few frequencies, and the controller gain is chosen to hit a target phase margin (45° to 60° is typical) at crossover. Loop-tuning tools in drives and power supplies automate exactly this step.
PD controller (1 + KD s) for 1000√2/(s(s+10)²): what KD gives 45° phase margin at 10 rad/s?
Answer: 0.1 Official key: 0.1 to 0.1
- Plant at : , and .
- A 45° phase margin needs , so the controller must add : , i.e. and .
- Check the crossover: , so . 10 rad/s really is the gain crossover.
The was chosen so that both conditions (phase and magnitude) hold with one unknown. If the magnitude check had failed, no single could meet the specification.
Derivative action adds phase lead, which is how quadcopter attitude loops, hard-disk head positioners and many servo axes get their damping. Real PD controllers filter the derivative term () so that sensor noise is not amplified without limit.
State-space model with D = 1: at what input frequency is the steady-state output zero?
Answer: 2 rad/s Official key: 2 to 2
- . Here and .
- , so .
- The steady-state output to is . It is zero when : , rad/s.
The system has transmission zeros at : it is a notch filter. The system is stable (double pole at ), so the transient dies out and only the zero-output steady state remains. Checked by sweeping from 0.001 to 10 rad/s.
Notch filters built this way cancel a known disturbance frequency: 50 Hz mains pickup in sensor signals, or a mechanical resonance in a drive shaft or robot arm. Drive firmware places a notch at the measured resonance so the speed loop can be tuned faster.
A = [0 1; −2 −3]: what is the sum of the magnitudes of the poles?
Answer: 3 Official key: 3 to 3
- The poles are the eigenvalues of : .
- Poles at and . Sum of magnitudes 3.
Check that nothing cancels: has no zeros, so both eigenvalues really are poles of the transfer function. The matrix is in controllable canonical form, so the last row is just the characteristic polynomial's coefficients with the sign flipped.
Engineers read poles off a state matrix to check stability of everything from a battery thermal model to a quadrotor's linearised dynamics. In MATLAB this is eig(A); the canonical forms are what code generators and observer designs start from.
A = [a b; −a 0]: which a and b put the transfer-function poles at −2 ± j3?
- (A)
- (B)
- (C)
- (D)
Answer: (D) Official key: D
- .
- Required: .
- Match coefficients: ; . Option (D).
We also checked that no pole is cancelled by a zero of for this choice, so really are poles of the transfer function. Sign slips on give (A) or (C); option (B) has .
Choosing matrix entries to place poles is the idea behind state-feedback design (pole placement, Ackermann's formula): the gains of an inverted-pendulum robot or an active suspension are computed so the closed-loop eigenvalues land where the designer wants them.
New state z = T x with T = [2 1; −1 0]: what are the transformed A, B and C?
- (A)
- (B)
- (C)
- (D)
Answer: (C) Official key: C
- With : and .
- , so .
- , then .
- and . Option (C).
Option (A) is the classic slip: it uses and forgets the on the right. Quick check without full multiplication: a similarity transform keeps the trace and determinant of (trace 1, determinant 6). (C) has trace and determinant ; (A) has trace . We also confirmed both models give the same transfer function.
Changing state coordinates is how engineers move a model into the form a design method needs (controllable or modal canonical form) or into physically meaningful states. Field-oriented control of AC motors is a famous example: the Park transform turns three oscillating phase currents into two steady d-q currents a PI loop can regulate.