Tricky Electronics & Communication interview questions

75 questions with answers (ECE-426 to ECE-500). Try each one aloud before reading the answer.

Signals & Systems

ECE-426 · "Is y(t) = 2x(t) + 1 a linear system, since its graph is a straight line?"

Trap: A straight-line input-output graph is mistaken for linearity. The system fails homogeneity and gives a non-zero output for zero input, so it is not linear. It is incrementally linear (affine).

ECE-427 · "Can a signal be both strictly time-limited and strictly band-limited?"

Trap: It is assumed a short pulse can also have a finite spectrum. A non-zero signal cannot be both: a finite-duration signal has an infinite spectrum, and vice versa. Practical systems accept approximate limits.

ECE-428 · "Is the sum of two periodic signals always periodic?"

Trap: Periodicity is assumed to add. The sum is periodic only if the ratio of the periods is rational. cos(t) + cos(√2·t) is not periodic.

ECE-429 · "Can a real-time filter have zero phase shift at all frequencies?"

Trap: It is assumed good design can remove phase shift. A zero-phase filter has a symmetric impulse response around t = 0 and is non-causal. It is possible only offline, for example by forward-backward filtering.

ECE-430 · "H(s) = 1/(s − 2) has a right-half-plane pole, so the system must be unstable. True?"

Trap: Stability is judged from pole location alone. It depends on the ROC. If causal (ROC Re(s) > 2) it is unstable. If the ROC is Re(s) < 2, the system is anti-causal and stable.

ECE-431 · "Negative frequencies are just a mathematical artifact with no physical meaning. True?"

Trap: It is assumed negative frequencies never matter. For real signals the negative half mirrors the positive half. In complex baseband (I/Q), positive and negative frequencies are physically different signals above and below the carrier.

Analog Communication

ECE-432 · "With 100% AM modulation, most of the power is in the sidebands. True?"

Trap: Full modulation is assumed to put most power into the information. Even at μ = 1, the carrier holds two-thirds of the power and the sidebands only one-third.

ECE-433 · "FM bandwidth is simply 2Δf. True?"

Trap: The bandwidth is assumed to depend only on deviation. Carson's rule gives 2(Δf + fm). For narrowband FM with tiny Δf, bandwidth approaches 2fm, not zero.

ECE-434 · "The carrier in AM carries the information. True?"

Trap: The strongest component is assumed to carry the information. The carrier is independent of the message; all information is in the sidebands. The carrier only enables simple envelope detection.

ECE-435 · "Increasing FM deviation always improves output SNR. True?"

Trap: FM noise improvement is assumed unlimited. Larger deviation widens bandwidth, admitting more noise. Once the carrier-to-noise ratio falls below threshold, SNR collapses.

ECE-436 · "The IF filter in a superhet removes the image frequency. True?"

Trap: IF selectivity is assumed to reject everything unwanted. After mixing, the image lands exactly at the IF and cannot be separated. Images must be rejected by RF filtering before the mixer or by an image-reject mixer.

ECE-437 · "An SSB voice signal has a constant envelope like FM. True?"

Trap: A single-tone SSB test (a pure sinusoid) is generalized. With real speech the SSB envelope varies strongly, so SSB transmitters need linear amplifiers.

Digital Communication

ECE-438 · "QPSK has worse BER than BPSK because its points are closer. True?"

Trap: Symbol packing is confused with per-bit performance. With Gray coding, QPSK is two orthogonal BPSK channels and has the same BER versus Eb/N0, while carrying twice the bits per hertz.

ECE-439 · "Required bandwidth is set by the bit rate. True?"

Trap: Bit rate is confused with symbol rate. Bandwidth depends on symbol rate and pulse shaping, B = Rs(1 + α) passband. 16-QAM carries four times the bits of BPSK in the same bandwidth.

ECE-440 · "Manchester coding is more efficient than NRZ because it has no DC. True?"

Trap: Removing DC is mistaken for efficiency. Manchester puts a transition in every bit, roughly doubling the required bandwidth compared with NRZ. It trades bandwidth for clock recovery.

ECE-441 · "We sent 10⁹ bits with zero errors, so BER is below 10⁻⁹. True?"

Trap: Zero observed errors are treated as proof. To claim BER < 10⁻⁹ at 95% confidence with zero errors, about 3 × 10⁹ bits are needed (3/BER). 10⁹ error-free bits support only about 3 × 10⁻⁹.

ECE-442 · "A matched filter reproduces the transmitted pulse shape at its output. True?"

Trap: Matched is taken to mean replicating. The output is the pulse's autocorrelation, which peaks at the sampling instant. The filter maximizes SNR at that instant and does not preserve shape.

ECE-443 · "The eye's width shows the noise margin. True?"

Trap: The two eye dimensions are confused. Vertical opening is the noise margin; horizontal opening is the timing margin, reduced by jitter and ISI.

Information Theory & Coding

ECE-444 · "Can better coding transmit above Shannon capacity?"

Trap: Coding is assumed to beat any limit. Capacity is a theoretical bound; above it, error probability cannot be made small. Codes such as LDPC and turbo only come close to it.

ECE-445 · "With infinite bandwidth, capacity becomes infinite. True?"

Trap: C = B·log₂(1 + SNR) is read as growing with B. SNR = P/(N₀B) falls as B grows, so capacity approaches the limit 1.44·P/N₀.

ECE-446 · "A good enough compressor can shrink any file. True?"

Trap: Compression is assumed always possible. By counting (the pigeonhole principle), no lossless compressor can shrink every input; random or already-compressed data cannot be compressed further.

ECE-447 · "Adding error-correcting code always lowers BER. True?"

Trap: Redundancy is assumed always helpful. At fixed energy per information bit, each coded bit has less energy. Below a code's threshold Eb/N0, coded BER can be worse than uncoded.

ECE-448 · "Huffman coding always achieves the entropy exactly. True?"

Trap: Optimal is confused with reaching the entropy. Huffman is optimal among symbol-by-symbol prefix codes, but L = H only when all probabilities are powers of 1/2. Block or arithmetic coding gets closer.

ECE-449 · "A code with dmin = 4 can correct two errors. True?"

Trap: dmin/2 is used instead of ⌊(dmin − 1)/2⌋. It corrects one error and detects up to three, or corrects one and detects two simultaneously. Correcting two errors needs dmin ≥ 5.

ECE-450 · "Shannon's formula C = B·log₂(1 + SNR) applies to any channel. True?"

Trap: The formula is used beyond its assumptions. It is for a band-limited AWGN channel with Gaussian input and average power constraint. Fading, interference-limited or non-Gaussian channels need other expressions, such as ergodic or outage capacity.

EM Theory & Transmission Lines

ECE-451 · "In an EM wave, E and H peak at different times, like voltage and current in an inductor. True?"

Trap: Circuit phase behaviour is applied to plane waves. In a lossless medium E and H are in phase because η is real. Only in a lossy medium does H lag E, by up to 45° in a good conductor.

ECE-452 · "Electrons travel along a cable at nearly the speed of light. True?"

Trap: Signal speed is confused with electron speed. Electron drift velocity is tiny, around mm/s. The signal is an EM wave guided by the conductors, travelling at VF × c.

ECE-453 · "An ohmmeter across a 50 Ω coax should read 50 Ω. True?"

Trap: Characteristic impedance is confused with resistance. Z₀ = √(L/C) is the voltage-to-current ratio of a travelling wave. An unterminated cable reads open at DC.

ECE-454 · "Skin effect only matters at radio frequencies. True?"

Trap: Skin effect is assumed negligible at power frequencies. Copper's skin depth at 50 Hz is about 9 mm, so thick bus bars and large conductors carry current unevenly even at mains frequency.

ECE-455 · "A strong permanent magnet radiates EM waves. True?"

Trap: Any field is assumed to radiate. Radiation needs time-varying fields from accelerating charges or changing currents. A static magnetic or electric field stores energy but does not radiate.

ECE-456 · "An open-circuited line absorbs no power, so nothing is reflected. True?"

Trap: Zero load current is confused with zero reflection. At an open, Γ = +1: all incident power is reflected, creating a full standing wave with a voltage maximum at the open end.

Antennas & Wave Propagation

ECE-457 · "Why do phones lose signal in elevators?"

Trap: The signal is assumed to be blocked by distance or height. The metal car and shaft act as a Faraday cage, reflecting and attenuating RF strongly. Gaps or in-shaft antennas restore some coverage.

ECE-458 · "Why are antennas often λ/4 or λ/2 long?"

Trap: These lengths are assumed arbitrary conventions. They are resonant: the input reactance is near zero and the radiation resistance (about 73 Ω for λ/2, 36 Ω for λ/4 monopole) matches common feed lines, giving efficient radiation.

ECE-459 · "A 10 dBi antenna multiplies the transmitter power by ten. True?"

Trap: Gain is mistaken for amplification. A passive antenna creates no power; it concentrates radiation into certain directions at the expense of others. Total radiated power is still at most the input power.

ECE-460 · "Higher frequencies always have more path loss. True?"

Trap: The f² term in FSPL is taken as a property of propagation itself. It comes from the shrinking aperture of a fixed-gain antenna. With fixed-size dishes at both ends, gain rises as f², so the link improves with frequency in free space.

ECE-461 · "A longer dipole always radiates better. True?"

Trap: More wire is assumed to mean more radiation. Beyond about 1.25λ the pattern splits into multiple lobes and the broadside maximum disappears. Non-resonant lengths also have large reactance that is hard to match.

ECE-462 · "The ionosphere reflects all radio frequencies like a mirror. True?"

Trap: Ionospheric return is treated as reflection off a surface. It is gradual refraction that depends on frequency and electron density. Frequencies above the MUF pass through, which is why VHF and satellite links work.

ECE-463 · "The Friis equation works for two antennas 10 cm apart. True?"

Trap: Friis is applied everywhere. It assumes far-field conditions (R ≥ 2D²/λ and R ≫ λ). In the near field, coupling is reactive and the 1/R² law does not hold.

Microwave Engineering

ECE-464 · "A waveguide can carry DC power like coax. True?"

Trap: A metal tube is assumed to conduct like a cable. A hollow waveguide has a single conductor and no TEM mode, so it acts as a high-pass filter and cannot pass DC or anything below cutoff.

ECE-465 · "Phase velocity in a waveguide exceeds c, so information travels faster than light. True?"

Trap: Phase velocity is confused with signal velocity. vp > c, but energy and information travel at the group velocity vg < c, with vp·vg = c² for an air-filled guide.

ECE-466 · "S₂₁ = +0.5 dB was measured on a passive filter. Is the filter good?"

Trap: The number is accepted without checking. A passive network cannot have gain, so |S₂₁| > 0 dB indicates a calibration, reference plane or cable-flexing error. Recalibrate before trusting the data.

ECE-467 · "A lossless, reciprocal three-port can be matched at all ports. True?"

Trap: It is assumed a T-junction can be made perfect. Such a network is impossible. The Wilkinson divider adds a resistor (making it lossy), and the circulator uses ferrite (making it non-reciprocal).

ECE-468 · "Microwave ovens use 2.45 GHz because it is the resonant frequency of water. True?"

Trap: Heating is assumed to be molecular resonance. Liquid water's dielectric loss is broad, with no sharp resonance at 2.45 GHz. The frequency is an ISM band chosen for regulation, penetration depth and cost.

ECE-469 · "A larger waveguide always has lower loss, so oversize is better. True?"

Trap: Lower wall loss is assumed to be the only effect. An oversized guide supports higher-order modes, which cause mode conversion, dispersion and unpredictable behaviour unless carefully controlled.

Digital Signal Processing

ECE-470 · "If Nyquist says 2× is enough, why sample at 2.5× or more?"

Trap: The theorem's ideal brick-wall filter is assumed. Real anti-aliasing and reconstruction filters need a transition band, so the sampling rate must exceed 2·f_max with margin. CD audio uses 44.1 kHz for a 20 kHz band.

ECE-471 · "All FIR filters have linear phase. True?"

Trap: A property of most designed FIR filters is assumed for all. Only FIR filters with symmetric or antisymmetric coefficients have linear phase. Minimum-phase FIR filters, for example, do not.

ECE-472 · "A carefully designed IIR filter can have exactly linear phase. True?"

Trap: IIR flexibility is overestimated. A causal, stable IIR filter cannot have exactly linear phase. Approximate linear phase is possible with all-pass equalizers, and zero phase is possible offline with forward-backward filtering.

ECE-473 · "A higher-order filter is always better. True?"

Trap: Sharper response is assumed free. Higher order increases delay, computation and coefficient sensitivity, and for IIR filters the risk of instability and limit cycles. Use the lowest order meeting the specification.

ECE-474 · "If the signal of interest is below fs/2, no anti-aliasing filter is needed. True?"

Trap: Only the wanted signal is considered. Noise, interference and harmonics above fs/2 still fold into the band. An anti-aliasing filter is needed regardless.

ECE-475 · "The DFT of a pure real sinusoid is a single spike. True?"

Trap: The spectrum is assumed one-sided and leakage-free. A real sinusoid gives two lines at ±f (bins k and N − k), and they are single bins only if the record holds an integer number of cycles; otherwise leakage spreads energy.

ECE-476 · "Sampling below twice the highest frequency is always wrong. True?"

Trap: Nyquist is applied to the highest frequency rather than the bandwidth. A bandpass signal can be sampled at about twice its bandwidth, with intentional aliasing down to baseband, as in IF-sampling receivers.

Wireless & Cellular

ECE-477 · "Does higher frequency always mean better wireless performance?"

Trap: More spectrum is assumed to mean better everything. Higher bands offer wider bandwidth, but suffer more penetration loss, poorer diffraction and more rain and foliage loss, so range and coverage shrink.

ECE-478 · "Multipath is always harmful, so MIMO works best in pure line-of-sight. True?"

Trap: Multipath is seen only as fading. Spatial multiplexing needs rich scattering to give a high-rank channel. Pure line-of-sight with closely spaced antennas gives rank about 1, limiting MIMO to beamforming gain.

ECE-479 · "Full signal bars guarantee fast mobile data. True?"

Trap: Signal strength is confused with throughput. Data rate depends on SINR, cell load, scheduler share, carrier bandwidth and backhaul. A crowded cell with strong signal can still be slow.

ECE-480 · "5G means mmWave only. True?"

Trap: The most publicized band is taken as the whole system. 5G NR uses FR1 (sub-7 GHz, including low bands) for coverage and FR2 (mmWave above about 24 GHz) for capacity hotspots. Most 5G traffic runs on FR1.

ECE-481 · "CDMA has a hard limit on users like TDMA. True?"

Trap: A fixed slot count is assumed. CDMA is interference-limited with soft capacity: each extra user slightly raises interference and degrades everyone, until quality targets fail.

ECE-482 · "OFDM eliminates multipath fading. True?"

Trap: Avoiding ISI is confused with removing fading. OFDM turns a frequency-selective channel into many flat-fading subcarriers, but individual subcarriers still fade deeply. Coding, interleaving and diversity recover those bits.

ECE-483 · "Doubling transmit power doubles range. True?"

Trap: A linear power-range relation is assumed. With path loss ∝ dⁿ, range scales as 2^(1/n): √2 ≈ 1.41× in free space (n = 2) and only about 1.19× for n = 4.

Optical Communication

ECE-484 · "Light in fiber travels at the speed of light in vacuum. True?"

Trap: Light is assumed to always move at c. In glass it moves at c/n ≈ 2 × 10⁸ m/s, about 5 µs per km. Long fiber links therefore have significant latency.

ECE-485 · "Single-mode fiber is single-mode at every wavelength. True?"

Trap: Single-mode is assumed to be a fixed fiber property. It holds only above the cutoff wavelength (V < 2.405). At shorter wavelengths such as 850 nm, standard SMF carries several modes.

ECE-486 · "A larger numerical aperture is always better. True?"

Trap: Better light gathering is assumed to be the only effect. Higher NA lets in more modes, increasing modal dispersion and reducing bandwidth. Long-haul fiber uses small NA.

ECE-487 · "Optical fiber has unlimited bandwidth. True?"

Trap: Huge capacity is mistaken for infinite capacity. Dispersion, nonlinearities, amplifier bandwidth and OSNR set a nonlinear Shannon limit on capacity per fiber.

ECE-488 · "Total internal reflection is lossless, so bending fiber does not matter. True?"

Trap: TIR is assumed to hold at any bend. Tight macrobends push rays below the critical angle so light leaks out. Bend loss is worse at longer wavelengths such as 1550 nm.

ECE-489 · "An APD always gives better receiver performance than a PIN. True?"

Trap: Gain is assumed to always help. APD gain also multiplies noise (excess noise factor), so there is an optimum gain. At high received power, or when shot noise dominates, a PIN is as good or better and simpler.

Data Communication & Networking

ECE-490 · "UDP is unreliable, so it is worse than TCP. True?"

Trap: Reliability is assumed always desirable. For real-time voice, video and gaming, late retransmitted data is useless, so UDP's low latency is better. Reliability can be added at the application layer if needed, as in QUIC.

ECE-491 · "Upgrading to more bandwidth always reduces latency. True?"

Trap: Bandwidth is confused with delay. More bandwidth cuts serialization and queueing delay, but propagation delay (about 5 µs/km in fiber) is unchanged. A transatlantic RTT stays tens of milliseconds at any speed.

ECE-492 · "A switch puts every port in one collision domain, like a hub. True?"

Trap: Switches are treated as hubs. Each switch port is its own collision domain, and full-duplex links have no collisions. All ports are still in one broadcast domain unless VLANs or routers separate them.

ECE-493 · "Every device's IP address is globally unique. True?"

Trap: IP addresses are assumed to identify devices globally. Private IPv4 ranges are reused everywhere behind NAT, and addresses change with DHCP. Global uniqueness applies only to public addresses at a given time.

ECE-494 · "A 1 Gbps Ethernet link moves files at 125 MB/s. True?"

Trap: Line rate is assumed to equal payload rate. Preamble, inter-frame gap, Ethernet, IP and TCP headers reduce maximum TCP goodput to about 940 Mbps (about 117 MB/s), and disks or protocols may limit it further.

ECE-495 · "Ping time is the one-way network delay. True?"

Trap: Ping is assumed to measure one direction. Ping reports round-trip time, including processing at the far end, and the two directions can differ. One-way delay needs synchronized clocks.

RF Circuits

ECE-496 · "A 2 dB cable loss before the LNA only reduces signal level, not noise figure. True?"

Trap: Passive loss is assumed not to add noise. A passive lossy element at room temperature has NF equal to its loss, so 2 dB before the LNA adds directly, about 2 dB, to system NF. Mount the LNA at the antenna.

ECE-497 · "Conjugate matching gives maximum power transfer and therefore maximum efficiency. True?"

Trap: Maximum power transfer is confused with maximum efficiency. At conjugate match, half the power is dissipated in the source, so efficiency is 50%. Power amplifiers use load-line matching instead.

ECE-498 · "More LNA gain always improves the receiver. True?"

Trap: Gain is assumed to be free. Beyond what is needed to suppress later-stage noise, extra gain drives the mixer and ADC into compression and intermodulation, lowering dynamic range.

ECE-499 · "A mixer's SSB and DSB noise figures are the same. True?"

Trap: Mixer noise figure is quoted without context. For a heterodyne mixer where the image band adds noise but no signal, the SSB noise figure is about 3 dB higher than the DSB value.

ECE-500 · "0 dBm means zero power. True?"

Trap: The logarithmic scale is read as linear. 0 dBm is 1 mW; negative dBm values are less than 1 mW but still positive. Zero power corresponds to −∞ dBm.

Other levels: Basic · Intermediate · Advanced · All Electronics & Communication questions