GATE Electronics (EC): Communications (random processes, AM and FM, information theory, digital modulation, PCM and coding), previous-year questions solved

Every communications question from the last five GATE Electronics and Communication papers, solved: random processes and noise, AM and FM with receivers, entropy and channel capacity, digital modulation with matched filters and error probability, PCM and error-correcting codes. Each answer matches the official key and was recomputed by code: a simulated noise sequence, the FFT of a sampled modulated wave, a direct sum over the probabilities, a Monte Carlo bit-error count or an enumeration of the code words.

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 · 10 of 31 questions on this page

GATE 2023 · EC · Q.30MSQ · 1 markRandom processes and noise

Which of four candidate functions are valid power spectral densities of a real signal?

For a real signal, which of the following is/are valid power spectral density/densities?
  1. (A)
  2. (B)
  3. (C)Option C: rectangular spectrum of height 1 for positive frequencies and height -1 for negative frequencies-110SX(ω)ω
  4. (D)Option D: rectangular spectrum of height 1 for positive frequencies and 0 for negative frequencies10SX(ω)ω

Answer: (A) and (B) Official key: A,B

  1. The power spectral density of a real signal is the Fourier transform of its autocorrelation . For a real signal is real and even, so must be real and even: .
  2. The power in any band is for every band, so everywhere. And the total power must be finite.
  3. (A) : even, positive, area , finite. It is the transform of . Valid.
  4. (B) : even, never negative (a square times an exponential), and the Gaussian factor makes the area finite. Valid. The wiggles do not matter, only the three conditions.
  5. (C) negative for and not even (it is odd): not valid.
  6. (D) never negative, but zero for and for , so not even: its autocorrelation is complex, which a real signal cannot have. Not valid.
  7. Answer: (A) and (B).

Trap: ticking only (A) because (B) 'does not look like a PSD', or ticking (D) because it is non-negative. The code built for each candidate by numerical inverse Fourier integration and tested it. (A) and (B) give a real autocorrelation and a positive semi-definite Toeplitz matrix (smallest eigenvalues 0.11 and -1e-16, that is not negative beyond rounding error). (D) is non-negative (minimum 0) but its has an imaginary part up to 0.115, so it cannot belong to a real signal. (C) has minimum value -1 and a Toeplitz matrix whose smallest eigenvalue is -3.71, which is negative. The set of valid options came out as A and B only.

Where this shows up in industry

Spectrum analysers and Welch-method estimators always report a non-negative, even two-sided PSD (the familiar one-sided plot from DC to half the sampling rate is that PSD folded and doubled), so a negative or lopsided spectrum from a measurement is a sign of a processing bug. The same conditions decide whether a noise model for a receiver or a channel is physically allowed.

GATE 2023 · EC · Q.59NAT · 2 marksRandom processes and noise

White Gaussian noise of PSD 1/2 W/Hz through an LTI system with h(t) = e^(-t)u(t): output power

Let be a white Gaussian noise with power spectral density W/Hz. If is input to an LTI system with impulse response . The average power of the system output is ___________ W (rounded off to two decimal places).

Answer: Official key: 0.24 to 0.26

  1. The output PSD is , with W/Hz at all frequencies.
  2. is the transform of : , so .
  3. . With this is .
  4. Time-domain check (Parseval): .
  5. Output power .

Trap: integrating over in rad/s and forgetting the gives W, far outside the key. The code integrated symbolically in (frequency domain) and (time domain), both exactly , and then filtered samples of white Gaussian noise (variance with s, fixed seed) through the sampled . The measured output power was W, inside the official range to . The -integral without gives 1.57, which is outside it.

Where this shows up in industry

This is the noise-power calculation behind every receiver link budget. Thermal noise is flat, the front-end filter shapes it by , and the output power is the area under the shaped PSD. For a first-order low-pass this area equals the noise bandwidth, which is times the bandwidth.

GATE 2024 · EC · Q.20MCQ · 1 markRandom processes and noise

White Gaussian noise of PSD N0/2 through a first-order RC low-pass filter: output variance

A white Gaussian noise with zero mean and power spectral density , when applied to a first-order RC low pass filter produces an output . At a particular time , the variance of the random variable is ________.
  1. (A)
  2. (B)
  3. (C)
  4. (D)

Answer: (A) Official key: A

  1. The noise has zero mean and the filter is linear with , so has zero mean. Then the variance of equals its average power, which is the area under the output PSD.
  2. The RC low-pass filter has , so .
  3. Output PSD: . The power is .
  4. Substitute , so and . Then .
  5. Check in the time domain: and .
  6. Variance , so option (A).

Trap: using instead of as the PSD value doubles the answer to , option (B). The code integrated the output PSD with sympy, integrated numerically, and ran a simulation of the filter on white noise of the stated PSD. All three give . Using lands exactly on option (B).

Where this shows up in industry

Receiver front ends are sized with exactly this calculation: the noise power at a filter output is the noise PSD times the filter's equivalent noise bandwidth, here per side of the PSD for a first-order low-pass. It is how a designer turns a thermal noise density into the rms noise voltage at an ADC input after an anti-alias RC.

GATE 2025 · EC · Q.42MCQ · 2 marksRandom processes and noise

f(t) = sum a_n p(t - nT), p a rectangle on [0, 0.5T], a_n independent zero-mean unit-variance: statements?

Consider a real-valued random process , where and is a positive integer. Here, for and otherwise. The coefficients are pairwise independent, zero-mean unit-variance random variables. Read the following statements about the random process and choose the correct option.
  1. The mean of the process is independent of time .
  2. The autocorrelation function is independent of time for all .
(Here, is the expectation operation.)
  1. (A)(i) is TRUE and (ii) is FALSE
  2. (B)Both (i) and (ii) are TRUE
  3. (C)Both (i) and (ii) are FALSE
  4. (D)(i) is FALSE and (ii) is TRUE

Answer: (A) Official key: A

  1. Mean: because each has zero mean. It is for every , so (i) is TRUE.
  2. Autocorrelation: . Independence and zero mean give for , and .
  3. So .
  4. Take : , which equals when lies in some and in the gaps . It depends on , so (ii) is FALSE.
  5. (i) TRUE and (ii) FALSE is option (A). The process is cyclostationary (statistics repeat with period ) but not wide-sense stationary.

Trap: option (B) comes from assuming that a zero-mean, uncorrelated sequence always gives a stationary process. That holds for the sequence, not for the pulse-shaped waveform, because the pulse switches on and off at fixed times. The code drew realisations of with and estimated the mean at six times: the largest magnitude was 0.0009 (statement (i) holds). The estimate of at the same six times was 1.000, 0.000, 1.001, 0.000, 0.995, 0.000: it is about inside a pulse and about in a gap, a difference far beyond the sampling error, so (ii) fails. The code then confirmed that exactly one pair, (A), matches both findings.

Where this shows up in industry

A digitally modulated signal is a pulse-shaped sequence like this one, and its statistics repeat every symbol period. That cyclostationary property is what lets a receiver recover symbol timing from the signal itself (spectral lines at the symbol rate appear after a squarer), and it is the reason PSD calculations for digital modulation average over one period.

GATE 2026 · EC · Q.38MCQ · 2 marksRandom processes and noise

X uniform on (-1, 1), N standard normal: correlation of X with Y = X + N and with Z = X^2 + N

Let , , and be random variables. The variables and are independent of each other. is uniformly distributed between and ; follows Normal distribution with zero mean and unity variance.
and are defined as, and .
Which of the following pairs represents the values of correlation between and and that between and ?
  1. (A) and
  2. (B) and
  3. (C) and
  4. (D) and

Answer: (A) Official key: A

  1. The correlation of two random variables is . The density of is on , so and . has and is independent of .
  2. .
  3. . is an odd function and the density of is even, so . Hence .
  4. The pair is and , option (A).

Trap: option (C), and , comes from writing as -type moment, that is, losing the odd power. The code integrated the moments with sympy (, , ) and ran random samples with a fixed seed: the sample means of and were and . Only (A) fits both. Also note the key reads 'correlation' as of the product: the correlation coefficient of and would be (sample value ), which is not among the options. and are uncorrelated (covariance ) although depends on .

Where this shows up in industry

Y = X + N is the basic received-signal model: the signal X plus independent noise N. Because noise has zero mean and is independent, the cross-correlation of the received waveform with the clean signal equals the signal's own power, which is why correlators and matched filters can pull a signal out of noise. A zero correlation with a nonlinear function of X, as here, is also a reminder that uncorrelated does not mean independent.

GATE 2023 · EC · Q.45MCQ · 2 marksAnalog modulation and receivers

FM signal through y = 2x + 5x^2: minimum carrier frequency to recover x(t), FM bandwidth B_T

Let a frequency modulated (FM) signal , where is a message signal of bandwidth . It is passed through a non-linear system with output . Let denote the FM bandwidth. The minimum value of required to recover from is
  1. (A)
  2. (B)
  3. (C)
  4. (D)

Answer: (B) Official key: B

  1. is a band-pass signal occupying (with and in the same frequency unit). We want to pull out of with a band-pass filter, so nothing else in may fall in that band.
  2. The squared term: . Squaring a signal convolves its spectrum with itself, so the width doubles: has a baseband part reaching up to , and a double-frequency part (an FM wave at ) spanning .
  3. Baseband part must stay below the wanted band: , so .
  4. Double-frequency part must stay above the wanted band: , so . Same condition.
  5. Minimum , option (B). At this value the bands just touch at their edges.

Trap: option (A), , reads the clearance as 'FM bandwidth plus message bandwidth', but the squaring doubles the occupied band, not the message band. The code built an FM line spectrum (, message tone ; the lines holding 99.8 percent of the power give and ), formed on a sampled grid, took its FFT and compared the band around the carrier with the spectrum of . Recovery failed for every carrier up to and including (the edges collide there) and worked from upwards. Option (A) gives , which fails. Options (C) and (D) give and , which work but are not the minimum.

Where this shows up in industry

Analog FM transmitters use exactly this effect on purpose. A non-linear stage such as a frequency multiplier generates harmonics of the FM wave, and a tuned band-pass circuit picks the one wanted, which also multiplies the frequency deviation by the harmonic order. In the other direction, the non-linearity of an RF power amplifier creates the same kind of extra bands, and the carrier must be far enough from them for the filter to separate them.

GATE 2024 · EC · Q.28NAT · 1 markAnalog modulation and receivers

AM output with carrier 50 W and sideband power one ninth of total: find B

An amplitude modulator has output (in Volts) . The carrier power normalized to resistance is Watts. The ratio of the total sideband power to the total power is . The value of (in Volts, rounded off to two decimal places) is _______ .

Answer: Official key: 2.49 to 2.51

  1. The three terms are a carrier at and two sidebands at and , each of amplitude .
  2. Power of a cosine of amplitude across is . Carrier: , so .
  3. Total sideband power: two lines, .
  4. Ratio: , so , , .
  5. .

Trap: counting only one sideband as the sideband power gives , so and , outside the key range. The code built with from a root-finder on the FFT line powers: the carrier line is and the two sideband lines carry one ninth of the total at . The one-sideband value was asserted to be outside to .

Where this shows up in industry

This is the carrier-versus-sideband power accounting used when an AM broadcast transmitter is rated: for a given carrier power the information-bearing sidebands carry only a fraction of the total power, which is why conventional AM is power-inefficient compared with DSB-SC or SSB.

GATE 2025 · EC · Q.14MCQ · 1 markAnalog modulation and receivers

FM phase 3 sin(2 pi f1 t) + 4 sin(6 pi f1 t), fc greater than 100 f1: peak frequency deviation?

Consider a frequency-modulated (FM) signal , where and are, respectively, the amplitude and frequency (in Hz) of the carrier waveform. The frequency is in Hz, and assume that . The peak frequency deviation of the FM signal in Hz is _______.
  1. (A)
  2. (B)
  3. (C)
  4. (D)

Answer: (A) Official key: A

  1. The instantaneous frequency is . The deviation from the carrier is with .
  2. Differentiate term by term: and . Dividing by : .
  3. Both cosines equal at , so the two terms add at their peaks: , option (A).
  4. The condition only makes sure that the instantaneous frequency stays positive and the carrier is far above the message tones.

Trap: option (B), , is the deviation of the second tone alone, the one that looks stronger because of its factor and the frequency. The first tone adds at the same instant. The code built the waveform with and , took the analytic signal by FFT, unwrapped its phase and read the instantaneous frequency: the peak deviation is 15.00 (the direct derivative formula gives 15.00 ). The second tone alone gives 12.00 , which is option (B).

Where this shows up in industry

Peak deviation sets the Carson bandwidth, , that an FM broadcast or two-way radio channel has to be given. A multi-tone message such as stereo audio plus a subcarrier has its deviations add at the worst instant, so the transmitter's deviation limiter is set on the sum, not on the strongest tone.

GATE 2025 · EC · Q.28MSQ · 1 markAnalog modulation and receivers

Envelope detector: DSB-SC if m(t) greater than 0, and AM only if m(t) greater than 0: which is correct?

Consider a message signal which is bandlimited to , where is in Hz. Consider the following two modulation schemes for the message signal:
  • Double sideband-suppressed carrier (DSB-SC):
  • Amplitude modulation (AM):
Here, and are the amplitude and frequency (in Hz) of the carrier, respectively. In the case of AM, denotes the modulation index. Consider the following statements:
  1. An envelope detector can be used for demodulation in the DSB-SC scheme if for all .
  2. An envelope detector can be used for demodulation in the AM scheme only if for all .
Which of the following options is/are correct?
  1. (A)(i) is TRUE
  2. (B)(i) is FALSE
  3. (C)(ii) is TRUE
  4. (D)(ii) is FALSE

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

  1. An envelope detector outputs the magnitude of the slowly varying amplitude multiplying , that is .
  2. DSB-SC: the amplitude is . If for all , then , so the envelope is exactly the message and the detector recovers it. Statement (i) is TRUE, option (A). (If changes sign, the output is , which is distorted. That is why DSB-SC normally needs a coherent detector.)
  3. AM: the amplitude is . The detector follows it faithfully whenever , that is when for . The message itself may go negative: is not needed.
  4. So \"only if \" is wrong. Statement (ii) is FALSE, option (D). Correct options: (A) and (D).

Trap: option (C) comes from remembering that AM needs a positive envelope and then attaching that condition to instead of to . The code built the waveforms (, message at ) and took the envelope with the Hilbert transform. DSB-SC with gave an envelope matching to within 0.0000. DSB-SC with a sign-changing missed by 2.00 (it returned ). AM with , which is negative half of the time, and was recovered to within 0.0000, so (ii) fails. Over-modulation (, so goes negative) missed by 1.00, which shows the true condition is on . Testing each option against these results gave exactly (A) and (D).

Where this shows up in industry

Envelope detection is why AM broadcast radios can be built with a diode and a capacitor, and why AM transmitters are never allowed to over-modulate. DSB-SC and its single-sideband cousins save power but need a local oscillator locked to the carrier, which is the cost that cheap receivers avoid.

GATE 2026 · EC · Q.57MSQ · 2 marksAnalog modulation and receivers

Narrowband signal x(t) = A(t) cos[2 pi fc t + theta(t)], fc = 1000 fM: PSK, AM, noise, narrowband FM

Consider a real, narrowband signal where the maximum frequency components of and are and , respectively.
Which of the following statements is/are correct for ?
  1. (A) represents a PSK modulated signal for suitable choices of and .
  2. (B) represents an amplitude modulated signal for suitable choices of and .
  3. (C) represents a band-limited Gaussian noise process.
  4. (D) never represents a narrowband FM signal.

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

  1. The form is the general narrowband (band-pass) signal: an envelope and a phase that change slowly compared with the carrier at . Any particular modulation is a choice of and .
  2. (A) PSK: keep and let switch between and with the data. This fits the form, so (A) is true.
  3. (B) AM: keep and let . This fits the form, so (B) is true.
  4. (C) Narrowband noise: a band-pass Gaussian noise of bandwidth about around can always be written with (Rayleigh) and (uniform). So (C) is true.
  5. (D) Narrowband FM: take and with . That fits the form and is exactly a narrowband FM signal, so 'never' is false and (D) is false.
  6. Answer: (A), (B) and (C).

Trap: the tempting wrong pick is (D), on the thought that a signal with an amplitude function cannot be frequency modulated; but is allowed. The code built all four signals with ( Hz) and recovered and from the analytic signal (FFT Hilbert transform). PSK: envelope error , phase error . AM: the recovered envelope equals to . Band-limited Gaussian noise (band ): rebuilt from and matched to , sample kurtosis and skewness (Gaussian). Narrowband FM (): envelope constant to , phase to , and the carrier with its first sidebands holds of the power. So (A), (B) and (C) hold and (D) does not.

Where this shows up in industry

Treating every band-pass signal as an envelope and a phase around a carrier is how modems and SDR front ends work: the radio demodulates to in-phase and quadrature components and the DSP does the rest, whether the signal is AM, PSK or noise. The same description is what lets a simulation run on the slow complex envelope instead of the fast carrier.

From GATE prep to a telecom 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.