RF Amplifier and LNA Interview Questions
RF amplifier interviews mostly test trade-offs between noise figure, gain, linearity, matching, stability, and efficiency. The LNA sets the receiver's noise boundary but needs enough gain, and raising gain tends to cost linearity. For power amplifiers, bias closer to linear operation costs efficiency, so be ready to reason through each trade-off with numbers.
Practice these questions live in a RFIC mock interview with a verified engineer. Book a RFIC mock interview
An interviewer draws a receiver on the whiteboard: antenna, filter, LNA, mixer. Then they ask you to make the LNA gain higher, and wait. Do you say yes right away? What does that do to the noise, and what does it do to the mixer behind it?
Amplifier questions mostly test trade-offs between a handful of numbers. This guide draws on a Cadence workshop on LNA design (hosted by UCSB), a Skyworks power amplifier datasheet, a Keysight application note on power-added efficiency, and an open-access paper on a low-band CMOS LNA. The practice prompts are ours, not reported questions from any employer.
What is the LNA for?
The Cadence workshop describes the first receiver stage as the place that sets the noise boundary and provides enough gain to overcome the noise of later stages such as the mixer. It adds that an LNA should handle large signals without distortion and present good matching, which matters most when a band-select filter sits in front of it, because many filters are sensitive to how they are terminated.
That paragraph is most of the interview. Noise, gain, linearity, and match all pull against each other. The workshop lists typically acceptable values for a heterodyne LNA: 50 ohm ports, return loss of -15 dB, reverse isolation of 20 dB, a noise figure of 2 dB, and a stability factor above 1. Treat those as one example, not a spec for every band. For noise figure and two-tone definitions, our post on why RF interviews keep asking about noise figure and linearity covers them, so this guide stays on the amplifier.
Matching, stability, and S-parameters
The workshop notes that, per Friis, the first amplification stage mostly determines overall noise figure. It writes noise factor in terms of a minimum value and an optimum source admittance, the formal reason the best-noise source impedance is not automatically the best power match.
For stability, the same document defines maximum available gain using the stability factor K and ties maximum stable gain to the condition K > 1. A good verbal answer: "I would check K across frequency, including out of band, because a device can look fine at the carrier and oscillate elsewhere."
A common-source stage with inductive source degeneration is the usual CMOS example. The open-access paper describes its low-band LNA core as a cascode common-source stage with source inductive degeneration, and gives the resonant voltage gain as Vout/Vin = RL / (2 Ls w0). So gain rises with load resistance and falls with source inductance and frequency. As general textbook background (not from that paper), the degeneration inductor also sets the real part of the input impedance without a noisy resistor, so one component appears in the gain, match, and noise answers.
Linearity, bias, and what the paper chose
The same paper shows real trade-offs. It sets bias current per gain mode (design targets of 12 mA down to 1.2 mA), avoids limiting current through the cascode device because that would reduce the main device's drain-source voltage and hurt linearity, and adds an input attenuator in the low-gain path to improve IIP3. It reports a noise figure as low as 1.2 dB at 18 dB gain with IIP3 of -6 dBm, rising to 8 dBm at the lowest gain setting (one section of the paper states a different low-gain figure, so check before quoting). The shape is clear: more gain, less linearity.
What should you know about power amplifiers at a conceptual level?
Keysight's note says power-added efficiency measures how well a power amplifier converts supplied power to output, that peak efficiency sits between the linear and saturation regions, and that beyond a certain input power more drive mostly adds heat. It does not link compression to efficiency directly, so that connection is yours to explain.
Skyworks' SKY66293-21 datasheet shows the trade in a real part: PAE of 29% at +28 dBm, and ACLR below 50 dBc at that power with pre-distortion, for a 20 MHz LTE signal. It also describes on-chip active bias for gain tracking over temperature and voltage, and internal matching tuned for power and efficiency. Class A versus class AB questions usually come down to the same thing: bias closer to linear operation costs efficiency.
Study table
Practice prompts are ours.
| Topic | What the source says | Practice prompt |
|---|---|---|
| LNA role | First stage sets noise boundary and needs enough gain (Cadence) | Why can a lossy filter before the LNA hurt more than a worse mixer? |
| Matching | Noise optimum and power match come from different source admittances (Cadence) | When would you accept a worse S11 for a better noise figure? |
| Stability | Maximum stable gain tied to K > 1 (Cadence) | How would you check stability out of band? |
| Degeneration | Gain falls as source inductance rises (open-access LNA paper) | What does raising it do to gain, match, and noise? |
| PA efficiency | Peak efficiency sits between linear and saturation (Keysight) | Where would you back off for a modulated signal? |
What does the noise figure versus linearity trade-off look like in an example?
This is our invented example, with made-up numbers, to practice cascade reasoning. A filter with 1 dB loss feeds an LNA (1.5 dB noise figure, IIP3 of -5 dBm) and then a mixer (10 dB noise figure, IIP3 of +5 dBm). Ignore filter linearity. (The cascaded IIP3 formula is in our noise figure and linearity post.)
- With 15 dB LNA gain, Friis gives about 3.3 dB total noise figure. Referred to the LNA input, the cascade IIP3 comes out near -11.2 dBm.
- Drop the LNA gain to 10 dB. Noise figure becomes about 4.6 dB, because the mixer's noise now counts for more.
- Linearity improves to about -8.0 dBm, because the mixer sees less signal.
- So the gain change bought roughly 3 dB of linearity at a cost of about 1.3 dB of noise figure. Whether that is a good trade depends on the blockers the system must survive, so ask about them.
For a receiver end to end, see our RFIC design interview guide, the comparison of AMS and RFIC roles, and the Skyworks and Qorvo front-end prep post. Weighing RF against links? See wireline versus wireless careers. Synthesizer questions often follow, so try PLL and clocking questions.
How to practice
Work the cascade above with your own numbers until you can do it aloud in two minutes. Then change the filter loss to 2 dB, or make the mixer passive. For a second pair of ears, book a mock interview on MockVise and have an RF engineer push on your follow-ups.
Sources: Cadence SpectreRF LNA workshop (hosted by UCSB), Skyworks SKY66293-21 datasheet, Keysight PAE measurement note, Low-band multi-gain LNA paper (PMC). This article provides general technical preparation guidance, not an official process or reported company questions. Datasheet and paper figures are examples and may change, so check the current documents, and confirm the interview format with your recruiter. MockVise is not affiliated with or endorsed by Cadence, UCSB, Skyworks, Keysight, or the authors of the cited paper.FAQ
Q: What LNA specs are typically acceptable? The Cadence workshop lists typical values for a heterodyne LNA: 50 ohm ports, return loss of -15 dB, reverse isolation of 20 dB, a noise figure of 2 dB, and a stability factor above 1. Treat those as one example, not a spec for every band.
Q: Why is the best-noise source impedance not automatically the best power match? Noise factor is written in terms of a minimum value and an optimum source admittance. The noise optimum and the power match come from different source admittances.
Q: What happens if I lower the LNA gain in a cascade? In the invented example, dropping the LNA gain from 15 dB to 10 dB moves total noise figure from about 3.3 dB to about 4.6 dB, while cascade linearity improves from about -11.2 dBm to about -8.0 dBm. Whether that is a good trade depends on the blockers the system must survive.
Q: How should I think about power amplifier efficiency? Power-added efficiency measures how well a power amplifier converts supplied power to output, and peak efficiency sits between the linear and saturation regions. Beyond a certain input power, more drive mostly adds heat.
Related guides
RFIC Design Interview Guide (2026)
The essential RFIC interview guide: Smith charts, S-parameters, LNAs, mixers, VCOs, noise figure, and linearity, with clear answers for radio-frequency IC roles.
Why Every RF Interview Still Asks About Noise Figure, Linearity, and Two-Tone Tests
Noise figure, IIP3, and the two-tone test show up in nearly every RF interview loop. Here's what each one is actually checking for, beyond the textbook definition.
AMS vs RFIC Engineer: How the Interviews Differ
Analog/mixed-signal and RFIC roles share deep analog roots but diverge sharply in the interview. Here's what each tests, the questions unique to each, and how to choose.
Hardware Security Engineer vs Design Engineer
Compare hardware security and RTL or physical design roles by daily work, interview probes, skills, and questions to ask, grounded in public OpenTitan, MITRE, and industry sources.
Practice with engineers who've run these interviews
Book a 1-on-1 mock interview with verified experts from Intel, NVIDIA, Qualcomm, and Apple. Practice these questions live in a RFIC mock interview with a verified engineer.
Find a RFIC expertAlready running these interviews at work? Get paid to run them on MockVise, anonymously and on your own hours. Become an expert