GATE Electronics (EC): Signals and Systems (LTI systems and convolution, Fourier series and transform, DTFT, Laplace and z-transform, sampling, DFT and digital filters), previous-year questions solved, part 4

Part 4 of 4 of the Signals and Systems hub: 9 more official GATE EC questions on Laplace and z-transform and sampling, DFT and digital filters, each solved step by step and checked against the official answer key. Part 1 covers signals and system properties, impulse response and convolution, Fourier series and DTFS and Fourier transform, DTFT and frequency response.

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 · 9 of 31 questions on this page

GATE 2024 · EC · Q.24MSQ · 1 markLaplace and z-transform

Causal H(z) = (2z^2+3)/((z+1/3)(z-1/3)): which of stable, minimum phase, h[0]=2, h[inf]=0 are true?

For a causal discrete-time LTI system with transfer function which of the following statements is/are true?
  1. (A)The system is stable.
  2. (B)The system is a minimum phase system.
  3. (C)The initial value of the impulse response is 2.
  4. (D)The final value of the impulse response is 0.

Answer: (A) and (C) and (D) Official key: A;C;D

  1. The poles are the roots of the denominator, . The system is causal, so the ROC is , outside the outermost pole.
  2. Stability: the ROC must contain the unit circle. does contain it, because both poles lie inside the unit circle. So (A) is true.
  3. Minimum phase needs all poles and all zeros inside the unit circle. The zeros are the roots of , that is , with magnitude . They are outside, so (B) is false.
  4. Initial value theorem for a causal sequence: . So (C) is true.
  5. Final value: a stable causal system has an impulse response that decays, because every pole is inside the unit circle. The impulse response is for even , zero for odd , and , so . So (D) is true.
  6. The true statements are (A), (C) and (D).

Trap: option (B). A stable system is not automatically minimum phase, because stability looks only at the poles and minimum phase also needs the zeros inside the unit circle. Here the zeros have magnitude . The code simulated the difference equation for the impulse response, checked , that is below and the closed form above for the first 30 even samples, and used np.roots to confirm the zero magnitude.

Where this shows up in industry

Minimum-phase versus non-minimum-phase is a real design constraint in digital equalisers and controllers: only a minimum-phase system has a stable, causal inverse, so a non-minimum-phase channel cannot be perfectly equalised by simply inverting it. Stability from pole radius is checked the same way in every IIR filter on a DSP.

GATE 2025 · EC · Q.27MSQ · 1 markLaplace and z-transform

Which statements about the z-transform ROC and the existence of the DTFT of x[n] are TRUE?

Let be a discrete-time signal whose -transform is .

Which of the following statements is/are TRUE?

  1. (A)The discrete-time Fourier transform (DTFT) of always exists
  2. (B)The region of convergence (RoC) of contains neither poles nor zeros
  3. (C)The discrete-time Fourier transform (DTFT) exists if the region of convergence (RoC) contains the unit circle
  4. (D)If , where is the unit impulse and is a scalar, then the region of convergence (RoC) is the entire -plane

Answer: (C) and (D) Official key: C;D

  1. (A) is false. The DTFT is and needs the sum to converge. For the partial sum at just counts the samples and grows without limit.
  2. (B) is false. The ROC never contains a pole, but it can contain zeros. For , has a zero at , and the ROC is every , which includes .
  3. (C) is true. If the ROC contains , then converges, so is absolutely summable and the DTFT exists and equals on the unit circle.
  4. (D) is true. has no terms, so the sum converges for every , including and . The ROC is the entire -plane.
  5. Answer: (C) and (D).

Trap: (A) is the common wrong pick, because every sequence seems to have a Fourier transform. The code summed at : 2000 for 2000 terms and 20000 for 20000 terms, so it diverges. It also checked (C) numerically on and (the partial sums matched and to within ), and used sympy to find the zero of at , inside its ROC.

Where this shows up in industry

Checking whether the ROC includes the unit circle is how a DSP engineer decides if a digital filter has a usable frequency response, which is the same as asking if it is BIBO stable. An IIR filter in audio firmware whose poles drift outside the unit circle because of coefficient rounding is the practical failure of exactly this condition.

GATE 2026 · EC · Q.13MCQ · 1 markLaplace and z-transform

Laplace transform of x(t) = u(t-2) convolved with t u(t): which expression?

The Laplace Transform of the signal is given by which of the following expressions?
[“” represents convolution operator]
  1. (A)
  2. (B)
  3. (C)
  4. (D)

Answer: (D) Official key: D

  1. Convolution in time becomes multiplication in the -domain: .
  2. A unit step delayed by gives (time-shift property: a delay of multiplies by ).
  3. The ramp gives .
  4. Product: : option (D).
  5. Check in the time domain: the convolution is , a parabola that starts at . Its transform is .

Trap: the delay is a factor , not a change of to . Option (B), , is option (D) times : the code found () at both test points, so (B) is simply wrong. Another slip is to multiply the two signals instead of convolving them: has transform , which is at against for (D). The code built and on a grid, convolved them with np.convolve, summed numerically at and , and only (D) matched at both points (sympy gave the same closed form).

Where this shows up in industry

Delays and cascades show up everywhere in real systems: a sensor reading that arrives late through a filter, an echo in an audio effect, a transport lag in a process loop. Treating the delay as a factor of e to the power of minus s times the delay and the cascade as a product is how engineers build their transfer functions block by block.

GATE 2026 · EC · Q.16MCQ · 1 markLaplace and z-transform

y[n] = 5/6 y[n-1] - 1/6 y[n-2] + x[n] with x[n] = d[n] - 1/3 d[n-1]: stable and causal?

The response of a discrete time system obeys the following relation:

The input to the system is .
Which of the following options is TRUE for ?
  1. (A)Stable and causal response
  2. (B)Stable and non-causal response
  3. (C)Unstable and causal response
  4. (D)Unstable and non-causal response

Answer: (A) Official key: A

  1. Take the -transform of the difference equation: . The characteristic polynomial factors as , so the system poles are at and .
  2. The input has .
  3. The zero of the input cancels the pole at : .
  4. The equation is a recursion in the past, run forward from rest, so the ROC is outside the outer pole, . Then : it is zero for (causal) and the sum is finite (stable, since the ROC includes the unit circle).
  5. Answer: stable and causal, option (A).

Trap: reading the polynomial as one in instead of . The roots of are and , outside the unit circle, which points to an unstable response. The poles of the equation as written are at and : inside. The code ran the recursion from rest and got exactly for , the sum of is , and the mis-read system with poles and reaches .

Where this shows up in industry

Placing the poles of a digital filter inside the unit circle is the first check in IIR design, and cancelling a pole with a matching zero is a standard way to remove an unwanted mode. In fixed-point firmware the cancellation is never exact, so designers keep poles away from the unit circle with some margin.

GATE 2022 · EC · Q.55NAT · 2 marksSampling, DFT and digital filters

8-point DFT applied twice to x = [1,0,0,0,2,0,0,0]: value of y[0]

For a vector , the -point discrete Fourier transform (DFT) is denoted by , where . Here, . If and , then the value of is _________ (rounded off to one decimal place).

Answer: Official key: 7.9 to 8.1

  1. First DFT: only and are nonzero, so . That is for even and for odd : .
  2. Second DFT at index : .
  3. Shortcut: applying the DFT twice gives . With , . (Note it is that matters, not : index maps to itself under .)

Trap: treating the second DFT as the inverse of the first and answering . The inverse carries a factor and a in the exponent, so DFT followed by DFT is times a time-reversed copy, not the original. The code applied np.fft.fft twice and also summed the definition directly, both gave , and the identity held on random complex vectors of length and . Another slip, stopping after one DFT, gives .

Where this shows up in industry

The factor of N (the transform length) and the index reversal are the reason FFT libraries define a separate inverse transform with its own scaling. In spectrum-analysis and OFDM firmware, forgetting that the forward FFT is not its own inverse is a classic source of a result that is the right shape but reversed and scaled by the transform length.

GATE 2023 · EC · Q.50MCQ · 2 marksSampling, DFT and digital filters

Two band-limited signals (B = 4 pi x 10^3 rad/s) on 4 pi x 10^3 and 12 pi x 10^3 carriers: Nyquist rate of y(t)?

Let and be two band-limited signals having bandwidth rad/s each. In the figure below, the Nyquist sampling frequency, in rad/s, required to sample , is
Block diagram. Band-limited signal x1 of t is multiplied by cos of 4 pi times 10 to the 3 times t. Band-limited signal x2 of t is multiplied by cos of 12 pi times 10 to the 3 times t. The two products are added to give y of tx1(t)x2(t)y(t)cos(4π×103t)cos(12π×103t)
  1. (A)
  2. (B)
  3. (C)
  4. (D)

Answer: (D) Official key: D

  1. Multiplying by shifts a spectrum to . A baseband signal that occupies ends up occupying .
  2. Upper branch: and , so the spectrum extends from to rad/s.
  3. Lower branch: , so the spectrum occupies to rad/s.
  4. The adder keeps both: is band-limited with highest frequency rad/s.
  5. Nyquist rate rad/s, option (D). That is kHz for a highest frequency of kHz.

Trap: option (C), , is , the Nyquist rate of one baseband signal, ignoring the carriers. The code built from random band-limited ( kHz, which is rad/s) and read the top spectral line off a kHz FFT: 8.00 kHz, so the Nyquist rate is 16.00 kHz. Sampling a test version of at kHz put every line where it belongs, while sampling at kHz folded 3 line(s) to wrong frequencies (aliasing). The other options (A) kHz and (B) kHz are not the Nyquist rate for kHz.

Where this shows up in industry

The sampling rate of a signal that sits on a carrier is set by the highest frequency present, not by the message bandwidth alone, which is why a direct-sampling radio receiver needs a much faster ADC than a baseband one. Frequency-division multiplexed channels added together, as in cable or telephone trunk systems, follow exactly this rule.

GATE 2024 · EC · Q.59NAT · 2 marksSampling, DFT and digital filters

y[n] = F{F{F{F{x[n]}}}} for x[n] = {1, 2, 1, 3} with N-point DFT F: value of Y[0]?

The relationship between any -length sequence and its corresponding -point discrete Fourier transform is defined as . Another sequence is formed as below . For the sequence , the value of is _______.

Answer: Official key: 112 to 112

  1. The -point DFT is . Applying it twice gives , because is when and otherwise.
  2. Applying it four times: (the index reversal happens twice and cancels). With : .
  3. is the DFT of , so .
  4. Check from the sum of : .

Trap: reading the question as asking for instead of . That gives , which is outside the official key of . Another slip is stopping after two transforms, which gives for the first sample and for the sum. The code applied np.fft.fft four times to , confirmed , and .

Where this shows up in industry

The fact that the DFT of a DFT returns the original sequence reversed and scaled by the length is why one FFT routine can compute both the forward and inverse transform: the inverse is a forward FFT, a conjugation or index reversal, and a division by the length. Real FFT libraries use exactly this trick.

GATE 2025 · EC · Q.38MCQ · 2 marksSampling, DFT and digital filters

Band-limited f(t) sampled at t = tau + nTs, then an ideal lowpass with gain Ts: what is the output?

Consider a continuous-time finite-energy signal whose Fourier transform vanishes outside the frequency interval , where is in rad/sec.

The signal is uniformly sampled to obtain . Here,

with being the Dirac impulse, , and . The sampled signal is passed through an ideal lowpass filter with cutoff frequency and passband gain .

The output of the filter is given by ________________.

  1. (A) if
  2. (B) if
  3. (C) if
  4. (D) if

Answer: (A) Official key: A

  1. The sampling comb is a train of impulses at . Its Fourier transform is with . Multiplying in time is convolving in frequency, so .
  2. The shift only puts a phase factor on each image. The term has factor , so the baseband copy is exactly with no delay.
  3. The images centred at do not overlap the baseband copy if , that is , or (the Nyquist condition).
  4. The lowpass has gain up to (its impulse response transforms to a rectangle of height ). It keeps only the baseband copy and cancels the factor , so the output spectrum is and the output is .
  5. Answer: if , option (A).

Trap: option (B), . People assume that sampling late by delays the output by . It does not: the filter adds the sample values at the matching times , so the original waveform is rebuilt in place. The code sampled with and on (, 12001 samples) and applied the stated filter. For (below ) the largest error against was 6.9e-13, against 0.22. For (between and ) the error against grew to 0.53 from aliasing and against was 0.75, so (C) fails, and in (D) was off by 0.60.

Where this shows up in industry

Real ADCs never sample exactly at , and the sample clock has an arbitrary phase relative to the signal; the reconstruction is still exact as long as the sample rate is above twice the signal bandwidth. A DAC reconstruction filter after a codec (for example in digital audio) is the practical version of the ideal lowpass here, with gain set so the output level matches the input.

GATE 2026 · EC · Q.15MCQ · 1 markSampling, DFT and digital filters

cos(20 pi t) and cos(100 pi t) sampled at 40 Hz: which statement about the first ten samples is true?

Two analog signals and ( in second), are sampled at a rate , where , and .
The first ten samples (starting from ) are considered for the analysis.
Which of the following statements is TRUE?
  1. (A)All of the first three samples of are greater than the corresponding samples of .
  2. (B)All of the last three samples of are greater than the corresponding samples of .
  3. (C)All of the samples of are greater than the corresponding samples of .
  4. (D)All of the fourth to seventh samples of are equal to the corresponding samples of .

Answer: (D) Official key: D

  1. The sampling instants are for .
  2. . The tone is , below , so it is sampled correctly.
  3. . The tone is , above , so it aliases: , exactly the frequency of .
  4. So the two sample sequences are identical, term by term: (the pattern repeats). In particular the fourth to seventh samples of equal those of : option (D).
  5. Options (A), (B) and (C) each say that one sequence is greater than the other, which needs a strict inequality. Equal samples are never greater, so all three are false.

Trap: treating as a different signal because its frequency is five times higher, and then picking (A), (B) or (C) after comparing the two continuous curves. After sampling at there is nothing to tell them apart. The code sampled both cosines, found the largest difference over the first ten samples to be below , and took a -sample FFT of : its peak is at , the aliased frequency, not at .

Where this shows up in industry

This is why an ADC is always preceded by an anti-aliasing low-pass filter: once a tone above half the sampling rate is sampled it looks exactly like a lower-frequency tone, and no digital processing afterwards can tell the two apart.

From GATE prep to a signal-processing or wireless career

Talk to a counsellor.

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.

Talk to a career counsellor

Talk to a career counsellor

Free 20-minute call. We'll map the right course to where you want to work.

By submitting you agree to our Terms and Privacy Policy.