RF & Wireless interview questions

10 questions, easiest first

1. 1You're testing a small wireless sensor board with a printed PCB antenna. Sitting on a table it gets great range, but the moment you pick it up and hold it in your hand, the range drops noticeably. Why does this happen, and what does it tell you about how the antenna is actually working?intro

   Answer this one, graded

   Or just show what a strong answer covers

   The hand sits in the antenna's near field, close enough that it's not just physically in the way -- it's electromagnetically coupled to the antenna. Human tissue has a high dielectric constant and is lossy, so it does two things: first, it shifts the antenna's resonant frequency and input impedance away from where the matching network was designed, causing a mismatch and reflected power (detuning); second, it directly absorbs some of the radiated energy as heat rather than letting it propagate (absorption loss). Both effects reduce the effective radiated power, which is why range drops. This is the same 'hand effect' phones deal with, and it's why antenna keepout zones and sometimes proximity sensors exist in handheld products.

   • Identifies detuning mechanism — States that the hand acts as a lossy dielectric near the antenna, shifting the resonant frequency and impedance match away from the design frequency.
   • Identifies absorption/loss mechanism — Explains that body tissue absorbs part of the radiated energy, reducing the power that actually leaves as a propagating wave (separate from detuning).
   • Distinguishes near-field coupling from simple blocking — Explains that the effect is a near-field electromagnetic interaction, not just physical shadowing or line-of-sight blockage.

   Where answers go wrong: Saying the hand just 'blocks the signal' like an opaque object, ignoring near-field electrical coupling. · Confusing this with multipath fading rather than a local impedance/detuning effect. · Not distinguishing between mismatch loss (detuning) and absorption loss as two separate contributors.

2. 2You're given a chip antenna datasheet that specifies a required ground-plane keep-out area (no copper) directly beneath and around the antenna, plus a minimum ground-plane length feeding into it. A teammate, tight on board space, wants to route ground copper pour right underneath the antenna to save room. What happens to the antenna's performance if they do this, and why does the ground plane matter at all for a small antenna like this?intro

   Common early layout mistake on space-constrained boards with integrated chip or PCB antennas.

   Answer this one, graded

   Or just show what a strong answer covers

   For a lot of small chip antennas, the metal element alone is far too small to radiate efficiently by itself -- the datasheet's specified ground plane acts as a counterpoise, and near-field currents on that copper are part of the actual radiating structure. That's why there's a keep-out zone and a minimum ground length spec. If you route ground pour under the antenna anyway, you disrupt those near-field currents: the resonant frequency shifts (usually detunes lower or higher depending on the added capacitive loading), the matching network no longer presents 50 ohms at the target band, and radiation efficiency drops. Ground length also matters because it sets the effective aperture, affecting both resonance and bandwidth. I'd push back and instead find space elsewhere on the board, or pick a smaller antenna variant rated for the space available. I'd confirm any suspected detuning with a VNA S11 sweep comparing before/after copper placement.

   • Ground plane as counterpoise — States that the PCB ground acts as part of the radiator (counterpoise), not merely a reference/return plane.
   • Detuning mechanism — Explains copper under the antenna disrupts near-field coupling, shifting resonant frequency and/or reducing radiation efficiency.
   • Ground length/bandwidth link — Notes that ground-plane length/size affects resonance, bandwidth, or gain pattern, not just presence/absence of copper.
   • Verification method — Proposes confirming impact via VNA S11/return-loss sweep before and after the change.

   Where answers go wrong: Treating the ground plane purely as an EMI/return-path concern unrelated to radiation. · Assuming more ground copper is always better grounding practice. · Not recognizing that small antennas rely on PCB ground as part of the radiating structure.

3. 3You connect a VNA to a 50-ohm antenna and measure its input impedance at the operating frequency as 25 + j0 ohms. Calculate the magnitude of the reflection coefficient and the VSWR, then tell me what fraction of your transmit power is actually being reflected back toward the source instead of radiated.easy

   Answer this one, graded

   Or just show what a strong answer covers

   Gamma = (25-50)/(25+50) = -25/75 = -0.333, so the magnitude is 0.333. VSWR = (1+0.333)/(1-0.333) = 1.333/0.667 = 2.0. Reflected power fraction is Gamma squared, about 0.111, so roughly 11% of the incident power bounces back toward the source instead of being delivered to the antenna -- that's a return loss of about 9.5 dB. A VSWR of 2:1 is workable but not great; you'd want to add a matching network to pull that impedance closer to 50 ohms, especially if there's a PA upstream that's sensitive to reflected power.

   • Correct reflection coefficient calculation — Computes Gamma = (25-50)/(25+50) = -1/3, giving |Gamma| = 0.333.
   • Correct VSWR calculation — Computes VSWR = (1+0.333)/(1-0.333) ≈ 2.0.
   • Correct reflected power fraction — States reflected power fraction = |Gamma|^2 ≈ 0.111, i.e. about 11% of power reflected, ~89% delivered forward (before further mismatch loss consideration).

   Where answers go wrong: Plugging impedance values in with the wrong sign convention or forgetting to take magnitude. · Confusing VSWR with return loss in dB without converting properly. · Assuming reflected power fraction equals |Gamma| instead of |Gamma| squared.

4. 4During a radiated emissions pre-scan, you see a narrowband spike well above the limit line at exactly 120 MHz -- which is the third harmonic of your microcontroller's 40 MHz crystal -- along with smaller spikes at other harmonics of that same clock. Walk me through how you'd confirm the source and the first three things you'd try to fix it.easy

   Answer this one, graded

   Or just show what a strong answer covers

   The fact that 120MHz is exactly 3x the 40MHz crystal is a strong clue that the clock network -- oscillator, crystal traces, or a clock buffer output -- is the source, since square-wave clocks are rich in odd harmonics. To confirm, I'd use a small near-field H-field probe connected to a spectrum analyzer and physically scan the board near the crystal, MCU clock pins, and any long clock traces, looking for where the 120MHz energy peaks; I could also try temporarily changing the clock source or disabling peripherals to see if the spike moves or disappears, which confirms causation rather than coincidence. For fixes, I'd first try shortening the clock trace and minimizing its loop area (tighter ground return), then check decoupling near the oscillator and add a series termination resistor to slow edge rates, and if needed add a small local shield or ferrite. Enabling spread-spectrum clocking is another option if the MCU supports it, since it spreads the energy across a wider bandwidth and lowers the peak.

   • Uses harmonic relationship to identify likely source — Explicitly reasons that 120MHz being 3x 40MHz points to the crystal/clock network (oscillator, trace, or clock buffer) as the source, not a coincidence.
   • Describes a confirmation method — Names a concrete diagnostic step such as near-field H-field probing along traces, or disabling/re-clocking the MCU to see if the spike disappears or shifts.
   • Gives at least two concrete, distinct mitigation attempts — Names specific fixes such as shortening/rerouting the clock trace to reduce loop area, adding a series resistor or ferrite, improving ground return path, adding a local shield, or enabling spread-spectrum clocking -- not vague 'add filtering'.

   Where answers go wrong: Jumping straight to a shielding can without confirming the source location first. · Ignoring that the harmonic number itself (3rd harmonic) is diagnostic information. · Proposing only one fix instead of a prioritized list of things to try.

5. 5You're probing a switching net on the bench with a standard oscilloscope probe, using the accessory ground clip with about a 15cm wire lead, trying to characterize fast ringing you suspect might also be radiating and could fail an emissions test. The scope shows large, fast ringing spikes at every switching edge. A near-field H-field probe swept close to the same trace shows much smaller ringing at a noticeably different apparent frequency. Which measurement do you trust, and what's likely wrong with the oscilloscope setup?easy

   A classic probing artifact that gets misdiagnosed as a real EMI problem.

   Answer this one, graded

   Or just show what a strong answer covers

   The long ground clip lead forms a small loop antenna in series with the probe tip's parasitic capacitance; at fast edge rates that loop's inductance resonates and produces ringing/overshoot that's largely a measurement artifact, not necessarily present on the actual net. This is a very common scope probing mistake. I'd trust the near-field probe more for judging what's actually happening on the board, since it's not subject to this particular loop resonance, though it has its own caveats around proximity and bandwidth. Before drawing any conclusion, I'd fix the scope measurement -- use a spring ground tip or ground ring right at the probe barrel to minimize loop area -- and re-measure. If the corrected scope trace still shows ringing that matches the near-field probe's frequency, then it's real and I'd look at trace inductance, gate drive, or snubbering; if it disappears, it was purely the probing artifact.

   • Identifies ground lead artifact — Explains the 15cm ground clip forms an inductive loop with probe tip capacitance that can ring and exaggerate or fabricate spikes not truly on the net.
   • Correct fix proposed — Recommends using a short ground spring/tip directly at the probe barrel to minimize loop area.
   • Explains near-field probe's relative immunity — Notes the near-field probe measures actual local field and is less prone to this specific artifact, while acknowledging its own limitations (bandwidth, coupling distance sensitivity).
   • Recommends re-verification — States they would redo the scope measurement correctly and compare frequencies/amplitudes before concluding anything about real EMI behavior.

   Where answers go wrong: Assuming the scope measurement must be more trustworthy since it's direct galvanic contact. · Not recognizing the ground lead inductance/probe capacitance resonance artifact. · Jumping straight to adding a snubber based on the uncorrected scope waveform.

6. 6You're deploying a LoRa sensor node: transmit power is +14 dBm, both the transmit and receive antennas have 2 dBi gain, the calculated free-space path loss at the deployment range is 145 dB, and the receiver sensitivity is -130 dBm. Your design requirement is at least 10 dB of fade margin. Compute whether this link closes, and by how much margin you're over or under the requirement.medium

   Answer this one, graded

   Or just show what a strong answer covers

   Received power = 14 + 2 + 2 - 145 = -127 dBm. Margin over sensitivity = -127 minus -130, which is 3 dB. Since the design requires at least 10 dB of fade margin and I only have 3 dB, this link does not reliably close -- it might work in clear line-of-sight conditions but will drop out under fading, foliage attenuation, or seasonal changes. To fix it I'd look at increasing Tx power if regulations allow, using a higher-gain antenna at the gateway, reducing data rate or increasing spreading factor to lower the effective sensitivity threshold, or shortening the range/adding a repeater node.

   • Correct received power calculation — Computes Prx = 14 + 2 + 2 - 145 = -127 dBm.
   • Correct raw margin over sensitivity — Computes margin = -127 - (-130) = 3 dB above sensitivity.
   • Correct conclusion against the 10 dB requirement — States that 3 dB of margin is less than the required 10 dB fade margin, so the link does not reliably close and needs more margin (higher Tx power, higher gain antenna, shorter range, or lower data rate/higher spreading factor).

   Where answers go wrong: Forgetting to add both antenna gains, or double-counting path loss direction. · Concluding the link 'closes' simply because received power is above the noise floor, without checking against the required fade margin. · Mixing up sensitivity (a power level) with SNR or noise figure.

7. 7Your 2.4GHz Wi-Fi module passes bench testing perfectly using an SMA cable to a reference antenna. On the actual production PCB with a small chip antenna mounted near a metal enclosure wall, range drops significantly, and a VNA sweep of S11 shows the resonance dip has shifted down to 2.2GHz instead of the target 2.44GHz. What's going on, and how would you fix it without redesigning the PCB layout?medium

   Answer this one, graded

   Or just show what a strong answer covers

   The bench test used a reference antenna over an SMA cable, which completely bypasses the actual chip antenna and its real-world mounting -- so it never exercised the failure mode. On the real board, the metal enclosure wall sits close enough to the antenna to be in its near field, and it couples with the antenna's current distribution, effectively loading it and shifting the resonant frequency down to 2.2GHz instead of 2.44GHz. That mismatch means most of the power is reflecting back rather than radiating, which explains the range drop, and the VNA sweep confirms it directly. Since I can't redo the layout, I'd retune the existing matching network -- adjust the L/C values while watching S11 on the VNA to pull the resonance back up to 2.44GHz -- and if there's any flexibility in mechanical clearance, I'd push the antenna keepout area further from the enclosure wall even slightly, since proximity is the root driver of the shift.

   • Identifies the metal enclosure as the detuning cause — Explains that the nearby metal wall couples with the antenna's near field, shifting its effective resonant frequency downward (or explains the physical mechanism of parasitic coupling/lowered resonant frequency).
   • Explains why bench testing missed it — Notes that the SMA/reference antenna bench setup bypassed the actual chip antenna and its real mechanical environment, so the detuning only shows up in final assembly -- this isn't a component failure, it's an integration effect.
   • Proposes a concrete non-layout fix — Proposes retuning the matching network (adjusting L/C values) using the VNA to re-center S11 at 2.44GHz, and/or increasing keepout clearance from the enclosure if mechanically possible.

   Where answers go wrong: Assuming the chip antenna itself is defective rather than considering the mounting environment. · Proposing a full antenna or layout redesign when the prompt explicitly asks for a fix without one. · Not explaining why the bench test with the reference antenna failed to catch this.

8. 8A teammate wants to double the range of a wireless device and proposes simply increasing transmit power by 6 dB, reasoning that this should roughly double the range. Using the free-space path loss model, derive whether that intuition holds, and then explain why the actual range increase you'd measure in a real office deployment is often smaller than what the free-space math predicts.medium

   Tests whether the candidate can connect a logarithmic propagation model to a practical design decision.

   Answer this one, graded

   Or just show what a strong answer covers

   Free-space path loss in dB is proportional to 20*log10(d), so doubling distance costs 20*log10(2) ≈ 6dB more loss. That means in an idealized free-space model, adding 6dB more transmit power exactly compensates for doubling the distance -- the teammate's intuition is actually correct there. But real indoor or office environments don't follow the n=2 free-space exponent; walls, furniture, and multipath typically push the effective path-loss exponent to 3-4, meaning loss grows faster than 20dB per decade. So the same 6dB power bump buys distinctly less than 2x range indoors, maybe closer to 1.3-1.6x depending on the environment. I'd also flag that adding 6dB of TX power costs battery life and might run into regulatory EIRP limits, so before touching power I'd consider cheaper levers like antenna placement, gain, or improving receiver sensitivity.

   • Derives distance-to-dB relationship — Shows or states that FSPL depends on 20*log10(d), so doubling distance costs about 6dB more loss (20*log10(2)≈6.02dB).
   • Correct ideal-case conclusion — Concludes that under ideal free-space propagation, 6dB more power does approximately compensate for doubling distance -- the intuition is basically right there.
   • Explains real-world divergence — Explains indoor path-loss exponents exceed 2 due to walls/multipath, so the same 6dB buys measurably less than double the range in practice.
   • Notes power tradeoffs — Mentions battery drain or regulatory EIRP limits as a cost of simply raising TX power.

   Where answers go wrong: Reasoning linearly (power doubles so range doubles) without invoking the logarithmic path loss relationship. · Forgetting path-loss exponent differs indoors vs free space. · Ignoring practical costs (battery, regulatory limits) of just cranking TX power.

9. 9You're designing a battery-powered wireless sensor node that must: (1) stay under a regulatory EIRP limit of 20 dBm for its band, (2) maintain enough link margin to reach a gateway 2km away through foliage, and (3) pass radiated emissions testing for harmonics of its power amplifier -- all while board space constraints force the antenna to sit close to noisy digital traces. Walk me through the tradeoffs between output power, antenna gain, matching network loss, and EMC filtering, and tell me what you'd prioritize first and why.hard

   Answer this one, graded

   Or just show what a strong answer covers

   First, I'd recognize that EIRP is a hard regulatory ceiling, so Tx power and antenna gain aren't independent -- if I use a higher-gain antenna to help close the 2km link, I have to back off conducted power by the same amount in dB, and vice versa; there's no free lunch there. Second, the matching network is doing double duty: any insertion loss it has directly subtracts from my link margin, but if I design part of it as a low-pass or harmonic filter stage, that same loss can be spent deliberately to attenuate PA harmonics before they reach the antenna, buying EMC margin at the cost of range. Third, I'd treat the antenna's proximity to noisy digital traces as a separate problem from PA harmonics -- that's likely to show up as broadband or clock-harmonic-driven emissions coupled through the near field, and the fix there is layout, ground plane integrity, or shielding rather than RF filtering. Given that the EIRP limit and the EMC pass/fail test are both non-negotiable gates -- you don't ship if you fail either -- I'd prioritize hitting compliance and the EIRP ceiling first, and treat the 2km link margin as the variable to recover elsewhere: better antenna placement or a slightly higher gain antenna traded against conducted power, a more sensitive gateway receiver, or a lower data rate/higher spreading factor that trades throughput for sensitivity. I wouldn't try to brute-force more Tx power through the ceiling.

   • Recognizes EIRP as a fixed budget to split between power and gain — Explains that since EIRP is capped, the tradeoff between conducted Tx power and antenna gain is a zero-sum allocation under the same ceiling, not two independent knobs.
   • Identifies the matching network's dual role — Notes that matching network loss reduces link margin but can also serve as a harmonic filter, and that a low-pass or harmonic filter stage trades insertion loss for compliance headroom.
   • Addresses the antenna-proximity emissions coupling separately from the PA harmonics — Distinguishes emissions caused by digital trace coupling near the antenna from emissions caused by the PA's own harmonic content, since they require different fixes (layout/shielding vs filtering).
   • Gives a clear, justified prioritization — States a concrete priority order (e.g., regulatory EIRP and EMC compliance are non-negotiable pass/fail gates, so link margin is the variable to sacrifice or recover through other means like a higher SF/lower data rate, better antenna placement, or gateway sensitivity improvements) with reasoning, not just a list.

   Where answers go wrong: Treating EIRP, gain, and power as independently maximizable rather than recognizing the regulatory ceiling links them. · Only discussing the PA's own harmonics and ignoring digital-noise coupling from trace proximity as a separate emissions source. · Giving a prioritization with no justification, or refusing to commit to a priority order at all.

10. 10You've measured your antenna's input impedance on a VNA at 2.4GHz as 15 - j40 ohms -- noticeably mismatched and reactive relative to the 50-ohm system. You need to design a matching network to bring VSWR close to 1:1 at that frequency. Walk through how you'd choose and design an L-network for this, then discuss the tradeoff between matching bandwidth and network Q as you tighten the match -- why can't you get both a perfect match and wide bandwidth at the same time?hard

    A realistic bench task after a VNA sweep reveals a poorly matched antenna port.

    Answer this one, graded

    Or just show what a strong answer covers

    On the Smith chart, 15-j40 ohms normalized sits inside the unity-resistance circle, so a two-element L-network works: one element transforms the real part toward 50 ohms, the other cancels the remaining reactance. Since the antenna looks capacitive (-jX), I'd need an inductive element -- shunt or series depending on exactly which side of the unity circle the point falls -- to cancel that reactance while the other element does the resistance transformation; I'd solve the two design equations or just do it graphically on the Smith chart. The catch is that this network's loaded Q is set almost entirely by the resistance ratio, roughly Q ≈ sqrt(R_high/R_low − 1) ≈ sqrt(50/15 − 1) ≈ 1.5, and that Q determines how quickly the match degrades away from center frequency, since the reactive elements only cancel exactly at one frequency. If I wanted more bandwidth, I could use a multi-section network (pi or cascaded L-sections) to spread the transformation over more elements, but per the Bode-Fano limit there's a hard tradeoff: for a fixed reactive load, you can't achieve both an arbitrarily perfect match and arbitrarily wide bandwidth simultaneously -- pushing for tighter match at center frequency necessarily narrows the usable bandwidth.

    • L-network design approach — Describes using a two-element (series/shunt) network chosen by where the impedance lands on the Smith chart relative to the unity circle.
    • Reactance cancellation reasoning — Notes the antenna is capacitive (-j40) and an inductive element is needed to cancel that reactance as part of the network.
    • Loaded Q from impedance ratio — States or derives that network Q is set by roughly sqrt(R_high/R_low - 1), so the 15-to-50 ohm ratio inherently implies a moderately high Q.
    • Bandwidth-match tradeoff explained — Explains that reactive cancellation is frequency-specific, so tighter matches at center frequency come at the cost of narrower bandwidth, referencing a fundamental (Bode-Fano-type) limit rather than just 'it depends'.

    Where answers go wrong: Giving specific L/C component values without any topology reasoning. · Treating VSWR at center frequency as the whole story, ignoring in-band variation. · Assuming you can add arbitrarily many matching elements with no bandwidth or loss penalty. · Not recognizing the fundamental tradeoff and just saying 'use a wider network'.