GATE Instrumentation (IN): Signals and Systems (sampling, Fourier series, Laplace and z-transforms, LTI systems, DFT and FFT), previous-year questions solved
Every signals and systems question from the last five GATE Instrumentation Engineering papers, solved: fundamental periods and Nyquist rates, Fourier series symmetry, time shifting and reversal, Laplace and z-transforms with regions of convergence, linearity, time invariance and eigenfunctions of LTI systems, difference-equation filters, the bilinear transform, FFT operation counts, circular convolution and time aliasing in the DFT. Each answer matches the official key and was recomputed by code, from the transform integral, a DFT or a simulated response, with the values printed in the paper.
Questions from official GATE IN papers (IIT Kharagpur 2022, IIT Kanpur 2023, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 19 Nov 2026 · 15 of 31 questions on this page
Periodic function of period 2 defined as -1-x and 1-x: which sine and cosine terms its Fourier series has
The Fourier series of this function contains ________
- (A)Both and where
- (B)Only where
- (C)Only where
- (D)Only where
Answer: (B) Official key: B
- Symmetry tells you which terms exist without any integration. An odd function, , has only sine terms. An even function, , has only a constant and cosine terms.
- The period is 2, so the fundamental angular frequency is and the harmonics are and .
- Test the symmetry. For , the point is in , so . The function is odd.
- So the constant term and every cosine term vanish. Only remains. Option (B).
- Check by integrating: , so , nonzero for every , and every integrates to zero.
Trap: seeing the straight lines and the jump at and assuming a mixture. The jump (from to ) does not break the odd symmetry. We integrated the cosine and sine coefficients numerically for up to 8: all cosine coefficients are below , and the sine coefficients start at , , (as predicts), so only option B holds.
Checking symmetry before computing a spectrum saves work in vibration and power analysis, and it tells you what to expect: an odd waveform has no DC component and no cosine content. Spectrum analysers show the same pattern as harmonic lines at multiples of the fundamental.
x(t) band-limited 100 Hz to 200 Hz, y(t) = x(2t - 5): which statement about y(t)'s band is always true
The statement that is always true is ________
- (A) is band-limited between 50 Hz and 100 Hz
- (B) is band-limited between 100 Hz and 200 Hz
- (C) is band-limited between 200 Hz and 400 Hz
- (D) is not band-limited
Answer: (C) Official key: C
- Two rules of the Fourier transform apply: a time delay multiplies the spectrum by a phase factor and leaves its magnitude alone, while compressing time by a factor stretches frequency by the same factor: .
- Write . The delay of 2.5 s only adds the phase and does not move any frequency.
- The factor 2 compresses time, so .
- is nonzero where Hz, that is Hz. Option (C): the band doubles.
Trap: halving the band (option A) because the time argument has a factor of 2 in it. Faster time means higher frequency, not lower. Checking it with cosines of 100, 150 and 200 Hz passed through and read from an FFT: the output frequencies are 200 Hz, 300 Hz and 400 Hz.
Playing a recording at double speed doubles every frequency in it, which is the same effect, and it is why a data-acquisition system that is replayed faster needs a higher sampling rate. When a signal is time-scaled, the anti-aliasing filter and the sampling rate must be re-checked against the new band.
Discrete-time signal x[n] = u[-n+5] - u[n+3]: smallest n for which x[n] = 0
Answer: Official key: -3 to -3
- The unit step switches on when its argument reaches 0. For a shifted and reversed step, find the value of at which the argument crosses 0.
- when , that is for (a left-sided step, equal to 0 for ).
- when , that is for (a right-sided step, equal to 0 for ).
- Combine the three regions. For : . For : . For : .
- The signal is zero from to , so the smallest with is .
Trap: reading the boundary as (where ) or , an off-by-one from the in the step definition. We evaluated the signal for every integer from to in code: , , , .
Reversed and shifted steps are the building blocks of windows and gating signals in sampled data systems, for example the interval during which an acquisition is enabled. Counting samples at the edges correctly decides whether a gate is one sample too long or too short.
y(t) = x(4t) where x(t) is periodic with fundamental period 100 s: fundamental period of y(t)
Answer: Official key: 25 to 25
- Multiplying the time variable by a factor compresses the signal in time by , so every period shrinks by the same factor: the period becomes .
- Show it directly. repeats every 100 s: . For we need the smallest with , that is .
- This holds when , so . The 100 s fundamental period of makes 25 s the smallest such , not just one of them.
Trap: multiplying instead of dividing, which gives 400 s. A slower signal has a longer period, a faster one a shorter period. We built a test signal with period 100 s, formed and searched upward in steps of 0.5 s for the first where at 400 test points: .
When a recorded signal is played back or resampled at four times the rate, every frequency component scales up by four and every period shrinks by four. The same scaling rule is why a sampled record must be stretched back correctly before comparing its cycle time with the original.
Nyquist sampling frequency of the signal x(t) = 10 sin^2(200 pi t)
Answer: Official key: 400 to 400
- The Nyquist sampling frequency is twice the highest frequency present in the signal: .
- The highest frequency is not obvious from . Use with : .
- The constant 5 is at 0 Hz. The cosine has angular frequency , so . Squaring a 100 Hz sine doubled its frequency to 200 Hz.
- .
Trap: reading as the signal frequency (100 Hz) and answering 200 Hz. The square doubles the frequency, so that sampling rate would alias. We sampled at 2 kHz for 1 s, took the DFT and found the highest component at 200 Hz.
Anti-aliasing is set from the highest frequency in the measured signal, and squaring nonlinearities such as a power or rectifier stage double it. A data-acquisition engineer must account for this before picking a sample rate and a front-end filter.
Fundamental period of the magnitude of x(t) = e^(j9t) + e^(j5t), t in seconds
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- The magnitude of a sum of two complex exponentials is easiest to find by pulling out the exponential at the average frequency: .
- The factor has magnitude exactly 1 for every , so .
- repeats every seconds. Taking the magnitude folds the negative half-cycles up: , so already repeats after half that time.
- The fundamental period of is (about 1.571 s). Option (B).
Trap: the period of itself. The two frequencies 9 rad/s and 5 rad/s have greatest common divisor 1 rad/s, so repeats only every , and option (A), s, is the period of before the magnitude is taken. We scanned the shift until held on a dense time grid and the first match was .
Beating between two close tones is how instruments read small frequency differences, for example in vibration analysis and in heterodyne and laser Doppler sensors. The envelope of the sum, its magnitude, repeats at a different rate from the carrier, and spotting that envelope period is the first step in reading such a signal.
20 kHz square wave through an ideal low-pass filter with 21 kHz cut-off: shape of the output
- (A)20 kHz sine wave
- (B)20 kHz square wave
- (C)20 kHz triangular wave
- (D)20 kHz saw-tooth wave
Answer: (A) Official key: A
- The Fourier series splits a periodic signal into a sum of sinusoids at the fundamental frequency and its harmonics. A square wave of amplitude that is symmetric about zero has only odd harmonics:
- For the components sit at , , and so on. There are no even harmonics, and no DC term.
- An ideal low-pass filter passes every frequency below its cut-off and removes everything above it. With a cut-off of , only the fundamental survives. The third harmonic and all the rest are removed.
- One sinusoid is left, with amplitude (for a unit square wave, about ). The output is a 20 kHz sine wave, option (A).
Trap: assuming that a filter that passes the signal frequency passes the signal shape. The sharp edges of the square wave are made of the higher harmonics, so removing them rounds the wave all the way into a sine. A triangular wave would keep a third harmonic at of the fundamental, a square wave would keep one at (we measured for the square wave), and a saw-tooth would keep the second harmonic. We took the DFT of one period of the square wave, deleted every harmonic above 21 kHz and tested each option against what remained: only the sine passes.
A low-pass filter ahead of a recorder or a converter is how instrumentation engineers recover a clean fundamental from a distorted or chopped signal, for example the excitation tone of a lock-in measurement. The same fact explains why a poorly filtered square-wave drive still radiates energy at three times its frequency.
Fundamental period of x(t) = 4 sin(15 pi t) + 7 cos(4 pi t), t in seconds
- (A)15
- (B)60
- (C)30
- (D)2
Answer: (D) Official key: D
- A sum of periodic signals is periodic only if the periods have a common multiple. The fundamental period is the smallest such common multiple, the least common multiple (LCM) of the individual periods.
- A term or has period . For : . For : .
- Their frequencies are and . The fundamental frequency is the greatest common divisor of the two: , so .
- Check with the LCM of the periods: . In that time the first term completes cycles and the second completes cycles, both whole numbers. Answer: (D) 2 s.
Trap: picking 60 s or 30 s. Both are periods of , because they are multiples of 2 s, but neither is the smallest one, so neither is the fundamental. Option (A), 15 s, is not even a period: shifts the phase by , an odd multiple of , which flips the sign. We worked in exact fractions and tested each option as a period and as the smallest period: only 2 s is both.
When two instrument signals are mixed, for example a 7.5 Hz vibration and a 2 Hz pulsation on one pressure line, the combined waveform repeats at the greatest common divisor of the two frequencies. That sets the record length you must capture to see one full pattern, and the spacing of the lines in the spectrum.
Plot of y(t) = 1 - x(4 - t) from a given plotted signal x(t): which of four plots
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- Work in the order of the operations. is reversed in time (mirrored about the vertical axis) and then delayed by 4 s (shifted right). Then flips the vertical sign and adds 1.
- Find where each piece of lands. A point at on appears at . The signal lives on , so differs from 1 only for .
- from 5 down to 3 (the ramp from to 0) maps to from to : , from to 0. So , from 3 down to 1.
- from 3 down to (the level of 1) maps to from 1 to 5: , so .
- from to (the level of ) maps to from 5 to 6: .
- from to (the ramp from up to 0) maps to from 6 to 7: falls from 3 to 1. The result is 3 to 1 on , 0 on , 3 on and 3 to 1 on : plot (B).
Trap: shifting without reversing (plots C and D keep the original orientation, and so does the left-to-right order of the pieces) or forgetting the vertical flip and the . Note that outside the window the expression equals 1, while the printed plots draw their baseline at 0; the question is a match of shape and levels inside the window, and only (B) has the levels 3, 1, 0, 3, 1 there. We evaluated on a fine grid inside the window and compared it with each redrawn plot: only (B) agrees everywhere.
Reversing and shifting a recorded signal is the first step of convolution, and it is also how a time-aligned correlation is set up between a transmitted pulse and its echo in ultrasonic or radar level and flow measurement. Getting the flip and the delay in the right order is a common source of a delay error with the wrong sign.
Signal (t-1)^2 u(t-1): value of its Laplace transform X(s) at s = 1
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- Use the time-shift property: if , then .
- Here , whose transform is , and the shift is .
- .
- At : . Option (B).
Trap: using instead of , which gives (option A). The transform of is , so the factorial matters. We integrated numerically (the transform at ) and got 0.7358, matching only option B.
Delay and shift rules are used to model transport lag in process control, for example a reagent that takes a fixed time to travel down a pipe to a sensor. The shifted ramp shapes in this question are also how a delayed ramp setpoint is described in a controller.
Laplace transform of x(t) = e^{-3t} u(t - 5), with its region of convergence
- (A)
- (B)
- (C)
- (D)
Answer: (C) Official key: C
- Go to the definition: . The step is zero before , so the lower limit becomes 5.
- .
- The integral converges only if decays, that is . A right-sided signal has a region of convergence to the right of its pole at .
- Option (C). The same result follows from the shifting property: write , so and , which equals .
Trap: applying the time-shift rule to the step only. Option A, , is the transform of , where the exponential was shifted along with the step; here the exponential is not shifted, so the extra factor is needed. We integrated numerically at and compared it with each option's formula and region of convergence: only C matches.
A signal that switches on after a fixed delay, such as a transport lag in a pipe or a start-up trigger, has a Laplace transform with an exponential factor in s. Recognising that factor as a delay is what lets control engineers separate dead time from the dynamics of the process.
Real causal finite-energy x(t): select all signals whose Fourier transform is purely imaginary
- (A)
- (B)
- (C)
- (D)
Answer: (B) and (C) Official key: B,C
- Use the symmetry rules of the Fourier transform . A real even signal has only the cosine part, so its transform is real. A real odd signal has only the sine part, so its transform is purely imaginary.
- is real and even (replacing by leaves it unchanged), so its transform is real. Option A fails.
- is real and odd (replacing by flips its sign), so its transform is purely imaginary. Option B passes.
- Multiplying a signal by multiplies its transform by , which turns a real transform into an imaginary one. So has an imaginary transform: option C passes.
- turns an imaginary transform into a real one (), so option D fails. Answer: (B) and (C).
Trap: assuming that the factor in the time signal gives an imaginary transform in every case. It does so only when the transform of the bracket was real. The causal condition on is not needed for this question, because the even and odd parts are built from and whatever is, as long as it is real. We took random real causal sequences, formed the four combinations, computed their Fourier transforms at eleven frequencies and tested whether the real part vanished: B and C do, A and D do not.
The even and odd parts of a recorded signal separate cosine content from sine content, which is how a lock-in amplifier splits a sensor signal into in-phase and quadrature components. The symmetry rules also let you check a numerical Fourier transform for errors, since a real even input must give a real output.
Region of convergence of the z-transform of x[n] = (0.2)^n for 0 <= n <= 7
- (A)all values of except
- (B)all values of
- (C)all values of except
- (D)all values of except
Answer: (C) Official key: C
- The region of convergence (ROC) is the set of for which stays finite.
- The sequence has only 8 non-zero samples, so the sum is finite: . A finite sum of finite terms converges for every except where a term itself blows up.
- The terms are powers of . Each is finite for any , but and its powers become infinite at . So must be excluded.
- At every with goes to 0 and , finite. Also at the sum is finite. Answer: (C) all except .
General rule: a finite-duration causal sequence has an ROC of the whole plane except . Trap: choosing (B) because the sum is finite, forgetting the negative powers of . We evaluated at , at and at : it is finite at the first and last and blows up near 0.
Finite-impulse-response (FIR) filters have exactly this kind of finite-length response, and their region of convergence covering every point except the origin is why they are always stable.
Signal x(t) for X(s) = 1/(s^2 + 13s + 42) with region of convergence -7 < Re{s} < -6
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- A Laplace transform does not fix its time signal until the region of convergence (ROC) is given. A pole to the left of the ROC belongs to a right-sided (causal) term, and a pole to the right of the ROC belongs to a left-sided term.
- Factorise: . Partial fractions: . The poles are at and , and the ROC lies between them.
- The pole at is to the left of the ROC, so is right-sided: (it converges for ).
- The pole at is to the right of the ROC, so is left-sided: (it converges for ). The pairs are for and for .
- . Answer: (B).
Check at : . Option (B) gives . Option (A) fails because does not converge at . We evaluated each option's integral with its own convergence condition: only (B) converges there and equals .
The same transfer function can describe a stable causal sensor or a non-causal signal, depending on the ROC. When you design a digital implementation of a continuous filter, the ROC is what tells you to pick the causal, stable version.
Laplace transform Z(s) of the convolution of u(t-2) - u(t-3) with e^(-4t) u(t)
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Convolution in time becomes multiplication in the Laplace domain: .
- is a rectangular pulse of height 1 from to . The transform of a unit step is and a delay by multiplies the transform by , so .
- , the transform of a decaying exponential with time constant .
- . Take out of the numerator: .
- : option (A).
Trap: option (D) has the right denominator but the wrong sign, since is negative for while the pulse response is positive. Options (B) and (C) have in place of , a factor that does not come from either signal. We computed by numerical integration and took its Laplace integral at and : 0.03773 and 0.00170, equal to (0.03773 and 0.00170). Only option (A) gives those values.
The output of a sensor or a filter driven by a pulse is the convolution of the pulse with the impulse response, and the product in the s domain is how it is handled by hand. A gated measurement window of fixed width, such as an integrating time on a counter or a rectangular averaging window, is exactly this pulse.