RF EngineerInterview Questions & Practice

Q1. How would you estimate receiver sensitivity from noise figure and bandwidth?

Strong answer: Sensitivity in dBm = −174 dBm/Hz + 10·log10(bandwidth in Hz) + noise figure + required SNR, plus any implementation margin. For example, a 20 MHz channel (+73 dB), a 7 dB noise figure, and a 10 dB required SNR give roughly −84 dBm. A strong answer derives it from the thermal noise floor rather than quoting a memorized number.

Weak answer: A weak answer gives a single number with no derivation or ignores bandwidth.

Q2. When would you choose a hybrid beamforming array over a fully digital one?

Strong answer: Fully digital beamforming needs an RF chain and data converters behind every element, which at mmWave array sizes and gigahertz bandwidths is prohibitive in power and cost — so hybrid arrays put analog phase shifters behind a few RF chains, trading per-element digital flexibility (multi-user, per-subcarrier processing) for feasibility. At sub-6 GHz with smaller arrays, fully digital usually wins.

Weak answer: A weak answer says hybrid is always cheaper without discussing performance or scaling limits.

Q3. What measurements do you trust first when an antenna detunes in the field?

Strong answer: Measure S11 in the installed condition first: a downward resonance shift indicates dielectric loading — a hand, the enclosure, condensation — rather than a broken antenna. Compare free-space against installed measurements, and check the matching network and feed before blaming the radiator.

Weak answer: A weak answer assumes the antenna is broken without checking the measurement chain or deployment environment.

Q4. A transmit chain meets output power but fails ACLR. Where do you look, and what tradeoff are you managing?

Strong answer: ACLR failure at rated power usually means the PA is being driven too deep into compression for the waveform’s PAPR, so the spectrum regrows into the adjacent channel. The tradeoff being managed is backoff versus efficiency: more backoff cleans up ACLR and burns efficiency, which is why digital predistortion exists. A strong answer checks DPD convergence, the waveform’s PAPR, and the measurement setup before touching hardware.

Weak answer: A weak answer reduces output power until ACLR passes without naming the efficiency cost or asking why the margin disappeared.

Q5. Walk through a link budget for a 100 m outdoor link and say where the margin usually goes.

Strong answer: A link budget chains transmit power, antenna gains, free-space and excess path loss, and receiver sensitivity derived from thermal noise, bandwidth, noise figure, and required SNR — then assigns an explicit fade margin on top. A strong answer names which term dominates the scenario: at 100 m outdoors it is rarely free-space loss alone, so the margin conversation is about multipath, blockage, and antenna placement.

Weak answer: A weak answer quotes a sensitivity number with no derivation and leaves no explicit fade margin, so the link works on paper and drops under multipath fading or blockage.

Q6. Why did direct-conversion receivers largely replace superheterodyne in modern radios, and what did the trade-off cost?

Strong answer: Direct conversion removes the IF chain and the image filter — the image problem disappears because the signal converts straight to baseband, and losing the large off-chip filter is what makes the receiver integrable. The price is a new impairment set parked at DC: LO self-mixing producing DC offsets, flicker noise sitting on the signal, and IQ gain/phase imbalance folding an image of the signal onto itself. Modern radios accept those because they are correctable in DSP; a superhet’s image filter is not integrable at any DSP budget.

Weak answer: A weak answer names the two architectures but cannot say what the image problem is or where the DC offset comes from.

Q7. What is the difference between antenna directivity, gain, and efficiency, and when does the distinction actually bite in a link?

Strong answer: Directivity describes how tightly the pattern concentrates radiated power and is pure geometry; gain is directivity multiplied by efficiency, so it folds in ohmic and mismatch losses; efficiency is the ratio that separates the two. The distinction bites in the link budget: range scales with gain, not directivity, so a high-directivity electrically small antenna running at 30% efficiency quietly costs more than 5 dB you will not see if you quote directivity. A strong answer also notes dBi versus dBd — a dipole reference sits 2.15 dB below isotropic.

Weak answer: A weak answer treats gain and directivity as synonyms and never mentions efficiency or the reference the dB figure is quoted against.

Q8. How do TRP and TIS differ from conducted power and receiver-sensitivity measurements, and why do device makers care?

Strong answer: TRP and TIS are over-the-air figures integrated over the full sphere in an anechoic chamber, so they capture the antenna, its placement, and hand, head, or enclosure detuning — what the device actually radiates and receives — while conducted measurements stop at the connector and miss every antenna-integration loss. A strong answer ties it to consequence: carriers gate device certification on TRP/TIS, so a board that passes conducted specs can still fail the radiated requirement once the antenna and mechanics are in the loop.

Weak answer: A weak answer conflates conducted power with radiated performance and never mentions the antenna-integration losses that over-the-air measurement exists to catch.

Q9. Your S21 measurement looks 2 dB worse than simulation. What do you rule out before blaming the design?

Strong answer: Rule out the measurement before the design: calibration plane and standards, cable and connector repeatability, fixture de-embedding, and temperature. Each of those can plausibly account for tenths of a dB to over a dB, so a strong answer quantifies them — and only after the measurement chain is exonerated treats the residual as a real model-to-hardware gap.

Weak answer: A weak answer re-simulates with tweaked parameters until the numbers match, which hides the real discrepancy instead of explaining it.