Radio-frequency engineering sits at the intersection of physics, math, and manufacturing. Whether you’re building IoT radios, Wi-Fi gateways, GNSS trackers, or industrial telemetry, early choices in topology, stack-up, and layout ripple through performance and certification. Teams that already excel at pcb design often find RF work both fascinating and unforgiving: tiny geometry changes shift impedance, grounding details decide noise floors, and a good or bad match can add or subtract tens of decibels from a link budget. This “Introduction to RF design” lays out the core ideas you will apply on every project, from propagation basics and decibels to filters, PAs/LNAs, mixers, and test strategy, then ties them back to simulation, layout, and real-world pitfalls that show up on chamber day.
Basics of RF waves and propagation
RF signals are time-varying electromagnetic fields. A sinusoid at frequency f has wavelength λ = c/ (n·f), where c is the speed of light and n is the medium index. In free space n≈1, so λ≈0.3 m at 1 GHz. As frequency rises, wavelength shrinks, so structures on the board stop behaving like “lumped” components and start acting as transmission lines. Power spreads with distance; in free space the Friis equation relates received power to transmitted power, antenna gains, frequency, and range. Real links add polarization mismatch, multipath, foliage, walls, and ground losses. Urban canyons cause fading; indoor paths show constructive and destructive interference with small position changes. Path loss exponents >2 are common indoors or near ground. Antenna height and ground clearance often help more than raw transmit power. Link budgets add transmit power, subtract feedline or switch losses, add antenna gains, subtract path loss, and compare received power to receiver sensitivity plus fade margin. A healthy budget includes 10–20 dB of headroom for fading and manufacturing variance. Human bodies detune small antennas; plastic housings shift resonance. Plan to measure the finished product, not just the bare board.
Decibel (dB) concepts and log scales
RF engineering speaks in decibels because they add when gains multiply and subtract when losses divide. Power ratio in dB is 10 log10(P2/P1). Voltage ratio in dB is 20 log10(V2/V1), assuming equal impedances. Absolute power uses dBm: P(dBm) = 10 log10(P/1 mW). A few anchors help mental math: +3 dB ≈ ×2 power, +10 dB = ×10, −3 dB ≈ ÷2, −10 dB = ÷10. Noise floors are often quoted as dBm/Hz; integrate over bandwidth to find total noise. Receiver sensitivity depends on noise density, noise figure (NF), bandwidth, and required SNR for the modulation. For example, thermal noise floor at room temperature is about −174 dBm/Hz; a 1 MHz bandwidth without front-end losses has −114 dBm noise. Add NF and SNR target to estimate the minimum signal level for a given BER/ PER.
Transmission lines and impedance matching
At RF, traces behave as transmission lines with characteristic impedance Z0 set by geometry and dielectric. Common Z0 values are 50 Ω single-ended and 100 Ω differential. Microstrip (outer layer) uses the reference of a nearby plane; stripline (internal layer) uses planes above and below, offering better isolation. Controlled impedance needs measured dielectric constant (Dk), loss tangent (Df), copper thickness, and precise width/spacing. Reflections occur when load impedance differs from Z0; the reflection coefficient Γ = (ZL − Z0)/(ZL + Z0). Standing waves waste power and distort signals. Matching networks use series/shunt inductors and capacitors (L, Π, T topologies) or short microstrip stubs to transform impedances. Smith charts visualize these transformations and are still the fastest way to see where you are and where to go. Keep stubs short compared to wavelength; at 2.4 GHz, a quarter-wave in FR-4 is only a few centimeters. On tiny boards, component parasitics dominate, so part libraries should include effective series resistance, self-resonant frequency, and Q at the target band. Good matching starts with a clean RF reference plane, continuous ground, and short return paths. Vias in series add inductance; via fences help confine fields. Keep the RF path straight, avoid right-angle bends (use 45° miter or curves), and place matching components tight to the device pin or antenna feed to avoid adding unintended line length. Solder mask over the feed can change impedance; check your fabricator’s mask thickness and Dk if your match is sensitive.
Antenna basics (gain, bandwidth, polarization)
Antennas turn current into radiation and vice versa. Gain (dBi) describes peak directivity relative to an isotropic radiator. Efficiency accounts for losses due to mismatch, conductor resistance, and nearby materials. Bandwidth is the frequency span where the antenna maintains acceptable return loss (often S11 < −10 dB) and radiation. Polarization describes the orientation of the electric field; a mismatch can cost many decibels. Small devices use chip antennas, PCB inverted-F antennas (IFAs), meanders, or stamped metal parts. External whips or SMA antennas increase volume but often improve link budgets. Ground plane length has a large impact on resonance and bandwidth; a quarter-wave monopole needs a counterpoise. Keep metal, batteries, displays, and ground pours away from the antenna’s keepout. Many chip antennas specify a reference ground and feed geometry; deviations must be retuned with a matching network. For multi-band systems, consider combo antennas or use antenna switches with separate feeds. Diversity and MIMO improve robustness and throughput; arrange antennas with spatial separation and orthogonal polarization when possible. Near enclosures, plastics shift resonance; final tuning should occur with the actual housing and cables installed.
RF filters (bandpass, notch, low-pass)
Filters shape spectra, reduce out-of-band emissions, and protect receivers from blockers. Lumped-element filters (L, C) work when component Q is high and parts are well below self-resonance. At higher frequencies, distributed filters using microstrip stubs, hairpins, or coupled lines provide stable responses. Low-pass filters after PAs reduce harmonics. Bandpass filters in front of LNAs reject out-of-band interferers and improve linearity. Notch filters kill a specific local interferer (for instance, to coexist with cellular bands). Ceramic, SAW, and BAW filters offer sharp skirts in small footprints; they add insertion loss that must be budgeted in the noise and power chain. Design steps: pick response type (Butterworth for flat passband, Chebyshev for steep skirts with ripple, elliptic for aggressive stop bands), translate specs to element values, simulate with real parasitics, then place the network tightly with solid ground and minimal vias in the current loops. Keep the hot node away from noisy digital areas; shield if needed.
Power amplifiers and LNAs
The PA sets the top of the transmit chain. Key figures: output power at target EVM or ACPR, efficiency (PAE), linearity (IP3, ACLR), and stability. Many radios include an integrated PA, but external PAs help when higher power or band-specific linearity is needed. Pay attention to bias networks, supply decoupling, heat removal, and harmonic filtering. RF power density turns into heat; good thermal vias under the exposed pad and a solid copper plane help. LNAs sit at the front of the receiver. Their noise figure adds directly to the system noise temperature, so every tenth of a dB matters. Input matching must consider both noise match (for lowest NF) and power match (for maximum gain); optimum noise impedance may not be 50 Ω. Use a π pad when you need gentle loss to improve out-of-band stability, understanding it harms NF. Protect LNAs from strong transmitters with filters and a well-sequenced T/R path built from RF switches, duplexers, or circulators. For time-division systems, switch isolation and timing matter; for FDD systems, duplexer rejection and symmetry matter.
RF mixers and oscillators
Mixers shift frequency by multiplying the RF input with a local oscillator (LO), producing sum and difference terms. Important specs include conversion loss or gain, noise figure, third-order intercept (IIP3), port isolation, and LO drive level. Passive diode mixers handle high signals and deliver good linearity with conversion loss; active mixers provide conversion gain at the cost of headroom. Image rejection depends on architecture; IQ quadrature paths can cancel the unwanted sideband when balanced correctly. Oscillators set LO purity and frequency accuracy. Crystal oscillators and TCXOs offer excellent short-term stability; VCTCXOs add pull range for fine tune. PLL synthesizers generate agile LOs with selectable step sizes and loop bandwidths; phase noise close-in and at offsets dictates reciprocal mixing and EVM performance. Keep VCOs and PLL loop filters quiet: isolate analog supplies, place the loop filter close, and route control lines away from digital edges. Use clean references; jitter at the reference multiplies up.
Modulation schemes (AM, FM, QPSK)
Modulation puts information on a carrier. Amplitude modulation sets the envelope; frequency modulation varies instantaneous frequency; phase modulation moves the phase. Modern systems use digital schemes: FSK/GFSK in many ISM devices, QPSK and QAM in Wi-Fi and cellular, LoRa’s chirp spread spectrum for long range at low data rates, and OFDM to split spectrum into narrow subcarriers with equalization. Metrics include occupied bandwidth, EVM, spectral mask, and error rates. Higher-order QAM increases data rate but needs better SNR and linearity. Power amplifiers must operate in a region that maintains the required EVM and ACLR; envelope tracking or digital predistortion can help at higher powers. Constant-envelope schemes relax PA linearity but still need careful filtering to meet masks.
Simulation tools for RF design
Simulation saves spins, but it must reflect the real stack-up and component models. Circuit simulators (S-parameter, harmonic balance) analyze filters, matches, and active stages. 2.5D/3D EM tools model microstrip, stripline, ground slots, shields, and antenna housings. Small shifts in Dk or solder mask thickness change impedance; pull actual fab numbers. For antenna work, co-simulate the PCB with the enclosure and cables. Include vendor S-parameter files for switches, filters, and amplifiers. Use Smith chart optimization to hit target return loss across process corners. Validate with sensitivity sweeps for component tolerances and placement offsets. Keep simulation schematics under version control alongside your PCB library, and name models with frequency validity to avoid accidentally applying low-frequency models at high GHz.
Measuring RF performance (VSWR, S-parameters)
A vector network analyzer (VNA) measures complex reflection (S11) and transmission (S21). Before measurements, calibrate using SOLT or TRL at the measurement plane; fixturing errors and cable flex cause misleading results. De-embed test coupons so the DUT reference is accurate. Return loss below −10 dB across the band is a common starting goal; many systems aim for −15 dB or better. Time-domain transforms of S-parameters expose discontinuities due to connectors, vias, and stubs. For power tests, a spectrum analyzer with a known reference and external attenuators captures output power, harmonics, and spurious emissions; for modulated signals, add a vector signal analyzer to measure EVM and ACLR. Noise figure measurements use the Y-factor method with a calibrated noise source. For antennas, a near-field scanner or anechoic chamber provides radiation patterns, gain, and efficiency. Cable loss and test lead placement matter; keep runs short, use quality connectors, and avoid moving cables after calibration. When testing finished products, measure in the final enclosure with batteries and displays live, because current draw and metal frames shift grounds and noise.
Transmission line layout and grounding on real boards
RF layout succeeds when return paths are short and predictable. Maintain a continuous reference plane under RF lines; avoid splits crossing the path. If you must cross a split, add a via fence and a stitching capacitor to provide a high-frequency bridge. Keep sensitive nodes away from fast digital traces and clocks; route orthogonally where paths cross to reduce coupling. Decouple every RF IC supply pin with a small ceramic capacitor placed tight to the pin, backed by a mid-value cap and, for PAs, bulk capacitance on the same layer where possible. Use multiple vias to ground at capacitor pads to cut inductance. Avoid long stubs on matching networks; a few extra millimeters of copper can retune a front end. Control current loops through compact placement, especially around switches, LNAs, and PAs. Isolate noisy DC-DC converters with filters and keep switch nodes far from antennas and feeds. When using RF shields, leave ground stitching around the perimeter at tight pitch and plan test access openings or removable lids. Shields help with emissions and immunity but also detune nearby antennas; re-measure after shield placement.
Power integrity, linearity, and thermal realities
RF stages react badly to supply noise. Buck converters need proper LC filters and damping to avoid peaking at switching harmonics that fall into RF bands. Place LDOs after buck converters for low noise rails feeding PLLs, LOs, and LNAs. Linearity metrics such as IP3, P1dB, and ACLR show how the chain handles blockers and modulated peaks; they degrade quickly with rail droop and poor biasing. Thermal rise shifts gain and center frequency; PAs drift and filters move. Model power dissipation, add thermal vias, and give hot parts copper spreads to the chassis where available. If the enclosure is plastic, integrate a small heat spreader or graphite sheet near the PA region. Measure temperature during long EVM tests to confirm stability at steady state.
Real-world RF design challenges
Component tolerance and PCB variation surprise many teams. A 1 pF capacitor at 2.4 GHz with ±0.1 pF tolerance is already a large swing in reactance. FR-4 Dk can vary by lot and with frequency. Solder voids under QFN thermal pads alter ground inductance and create gain or NF drift. Coexistence is another pain point; a nearby LTE uplink or Wi-Fi channel can compress your LNA. Good preselection filters and solid shielding help. Ground bounce from digital edges couples into the VCO control line through long traces; one stray millivolt of noise can broaden the spectrum. Cable-to-board transitions act as antennas unless clamped with ferrites and well-placed return vias. Battery wires form loops that pick up or radiate; twist pairs and place them against the chassis reference. Certification introduces limits for conducted and radiated spurs, occupied bandwidth, and out-of-band emissions. Final test in the enclosure is the only test that counts; an open-air board rarely matches the chamber result. Always budget time for two or three tuning cycles with your lab.
Step-by-step path for a first-time RF build
- Define bands, data rates, and regulatory targets. Build a link budget that includes antenna gains, cable and switch losses, and a 10–20 dB fade margin. 2) Choose parts: radio SoC or transceiver, external LNA/PA if needed, RF switches or duplexer, filters, and an antenna approach that fits the size and ground you truly have. 3) Simulate the input and output matches with realistic part Q and a board model. 4) Lock a controlled-impedance stack-up with your fabricator; pull actual Dk/Df and mask values. 5) Place RF parts tightly and straight; reserve an antenna keepout; add matching pads you might not populate; leave room for a shield. 6) Route short feeds with a continuous ground, via fence as needed, and avoid stubs. 7) Add test points and a coax test connector near key nodes for bring-up. 8) Order a solder stencil even for prototypes; paste volume variation dominates early failures. 9) On bring-up, verify supplies, clocks, and LO first; then measure S11, small-signal gain, and noise; finally run EVM and spectral mask. 10) Tune the match on the real board in its enclosure; update BOM and paste layers; repeat. 11) Pre-scan emissions with a near-field probe set and a spectrum analyzer; fix hotspots before the chamber.
Common rules of thumb
• Keep matching components within a few millimeters of the RF pin or antenna feed. • Every millimeter of 50 Ω microstrip at 2.4 GHz adds a measurable phase shift; measure before adding “helpful” meanders. • Two or more ground vias at every shunt capacitor reduce ESL. • Use short 0402 parts for high-frequency matches; 0603 may have lower self-resonance at GHz. • A 1 nH series inductor looks roughly jω·1nH; at 2.4 GHz that’s ~15 Ω of reactance. • A −10 dB return loss is ~90% power delivered; −15 dB is ~97%. • Keep PA output filters away from high-current digital rails and switch nodes.
Antenna selection and tuning workflow
Pick your antenna last only in marketing slides; in engineering, it’s picked early. For a compact product, evaluate a few chip antennas on your ground size using vendor reference boards. Place the feed at the position that gives the best bandwidth on your exact ground plane. Add a three-element π match footprint even if it starts as a zero-ohm link. If your device sits near a user’s hand or head, test both free space and “loaded” conditions with a phantom or human tester. Keep metal screws and battery tabs out of the near field; a 2–5 mm shift in hardware can kill gain. If you must meet both 2.4 GHz and sub-GHz, consider separate antennas or a combo unit with a proper isolation plan. When possible, route feedlines on the outer layer over a solid ground reference and avoid vias in the feed. After assembly, tune by adjusting the series/shunt values while watching S11 and total radiated power in a chamber. Lock the final values in your AVL and protect the area from rework that might shift part placement.
Coexistence and EMC
Products that mix radios with fast digital logic benefit from early EMC plans. Spread-spectrum clocks and careful grounding reduce emissions. Partition noisy regions and keep RF sections away from switch nodes and memory buses. Use differential routing for high-speed digital and route reference planes to keep return currents tight. Shield sensitive oscillators and mixers and use feed-through capacitors where cables enter or leave the RF region. Protect antenna feeds from ESD with RF-rated protectors that have low capacitance. If both Bluetooth and Wi-Fi share 2.4 GHz, consider time-sharing or vendor coexistence APIs to reduce collisions. Avoid DC-DC switching harmonics that land in your bands; sometimes a small adjustment of switching frequency gets you out of trouble.
Manufacturing and test hook-ups
A design that works on the bench but stalls in production isn’t finished. Give your assembler the paste layer, controlled-impedance notes, and shield heights. Plan for automated test: add a coax pad and pogo test points so the line can measure output power, frequency error, and modulation quality quickly. Provide golden units for AOI and RF calibration. Document calibration procedures for parts with tuning (e.g., VCTCXO pull, PA bias). For low-cost devices, limit per-unit test time by relying on statistical sampling for full S-parameter sweeps and running quick power/EVM checks on every unit.
Future of RF design
RF is moving toward higher integration and higher frequency. On-package antennas reduce feed loss. System-in-package radios pack transceiver, filters, and passives into a few square millimeters, which raises demands on power integrity and thermal paths. Millimeter-wave bands bring short wavelengths where even pad shapes matter; 3D EM simulation becomes routine. Wideband PAs and digital predistortion move downmarket as silicon improves. Machine-assisted tuning uses lab data to update matches and filters faster. Sustainability trends encourage repairable modules and part transparency; designers are revisiting material choices and disassembly paths. Test is getting smarter too: built-in sensors track temperature and supply noise to predict EVM drift before it fails limits. Teams that connect lab results back into their libraries and pcb design rules ship faster in each new product cycle.
Conclusion
Radio design rewards careful planning and disciplined measurement. Master the decibel so link budgets are second nature. Treat every RF trace as a transmission line with a real return path. Pick and tune an antenna with the actual ground size and enclosure in place. Protect receivers from blockers with the right filters and a clean power supply. Place and route with short loops, tight matches, and plenty of ground vias. Simulate with realistic models, then verify with calibrated instruments and repeat inside the final enclosure. When your measurements disagree with simulation, trust the instruments and hunt for parasitics in layout or packaging. With these habits, your next “Introduction to RF design” project will move from whiteboard to chamber with fewer surprises and a lot more margin in the places that matter.
