The Anatomy of an RF Transmitter Chain (and Why Every Block Matters)¶
An RF transmitter isn't just "amplify and send." It's a cascade of interdependent stages — each with its own specifications, failure modes, and trade-offs that ripple through the entire chain.
If you've ever wondered what goes into designing a transmitter from scratch, or if you do this work and want a clear reference for how the blocks interact, this article lays out the complete picture.
What an RF transmitter actually contains
A transmitter converts a baseband signal into a modulated RF signal at the required power level, frequency, and spectral purity for over-the-air transmission. That sounds simple. The implementation isn't.
A typical heterodyne transmitter chain contains six to eight functional blocks, each solving a different problem:
Baseband processing and DAC. The digital signal — whether it's a 5G NR waveform, a radar chirp, or a simple FSK data stream — gets converted to analog I/Q signals. The DAC's sampling rate and resolution set the upper bound on signal bandwidth and dynamic range. A 14-bit DAC running at 2.5 GSPS can generate signals with ~200 MHz of instantaneous bandwidth. Everything downstream is limited by what the DAC can produce.
Local oscillator (VCO + PLL). The frequency synthesizer generates the carrier frequency. A voltage-controlled oscillator (VCO) provides the tunable RF signal, locked to a reference crystal via a phase-locked loop (PLL). The phase noise of this block directly determines the transmitter's noise floor and modulation quality. A VCO with -110 dBc/Hz phase noise at 100 kHz offset versus -120 dBc/Hz is the difference between meeting and failing an EVM specification.
Mixer / upconverter. The mixer translates the baseband signal to the target RF frequency by combining it with the LO signal. This is where unwanted products appear — image frequencies, LO leakage, intermodulation spurs. A mixer with 35 dB of image rejection might be acceptable for a wideband radar but inadequate for a cellular transmitter that needs to meet stringent out-of-band emission masks.
Channel filter. Following upconversion, a bandpass filter is employed to suppress unwanted spectral artifacts, including image frequencies, LO leakage, and mixer-generated spurs. The design of this block is governed by a strict trade-off between insertion loss (IL), out-of-band rejection, and physical volume. For example, a SAW filter at 2.4 GHz provides an excellent balance for mobile applications, offering roughly 1.5 dB IL and 40 dB rejection in a sub-millimeter package. Conversely, for infrastructure where efficiency is paramount, a cavity filter can achieve superior performance — 60 dB rejection with only 0.3 dB IL — at the cost of a significantly larger mechanical footprint.
Driver amplifier. The driver stage brings the signal from the mixer output level (typically -10 to 0 dBm) up to the level needed to drive the power amplifier's input — usually +10 to +20 dBm. This stage needs to be linear enough to preserve modulation quality but efficient enough not to become a thermal problem. Gain flatness across the operating bandwidth matters here more than raw output power.
Power amplifier (PA). The final stage. This is where most of the DC power is consumed and where thermal design either works or doesn't. A PA for a 5G small cell might deliver +27 dBm (500 mW) at 3.5 GHz with 30% power-added efficiency. That means 1.2 W of heat in a package smaller than your thumbnail. PA selection involves navigating a multi-dimensional trade-off space: output power, efficiency, linearity (P1dB, OIP3), operating frequency, supply voltage, and thermal impedance.
Output matching and harmonic filter. The PA's output impedance is rarely 50 ohms — it's often in the 2-5 ohm range for high-power devices. An output matching network transforms this to 50 ohms for the antenna interface. A low-pass harmonic filter after the match ensures second and third harmonics meet regulatory limits. At 2.4 GHz, that means suppressing energy at 4.8 GHz and 7.2 GHz by typically 40-60 dB.
Isolator or circulator (optional). In systems where antenna VSWR can vary — mobile handsets near the user's head, or phased arrays with active scan — a circulator protects the PA from reflected power. Without it, an antenna mismatch can shift the PA's load line, degrade efficiency, or in extreme cases cause device failure.
Why the cascade matters
Each block's specifications are set by what comes before and after it. You can't select a PA without knowing the driver's output power. You can't select the driver without knowing the mixer's output level and the PA's input requirement. You can't choose the filter without knowing the mixer's spur spectrum.
The gain chain. The total gain from DAC output to PA output has to hit a precise target — enough to reach the required transmit power, but not so much that any stage is driven into compression. A typical chain might need 50-60 dB of total gain, distributed across the driver (+15 dB), intermediate stages, and PA (+25 dB). If the mixer has 8 dB of conversion loss instead of the expected 6 dB, every downstream stage needs to compensate.
The linearity chain. Distortion compounds. Each nonlinear stage adds harmonics and intermodulation products. The PA is usually the dominant source of distortion, but a driver stage operating at P1dB will contribute its own. System linearity is roughly limited by the worst individual stage — but "worst" depends on both the device's inherent linearity and how hard you're driving it.
The noise chain. In a transmitter, the noise concern is different from a receiver. You're not worried about noise figure — you're worried about phase noise and broadband noise floors. Phase noise from the LO propagates through the mixer to every downstream stage. Broadband noise from the DAC and amplifier stages creates a noise pedestal around the carrier that can violate adjacent-channel emission limits.
The thermal chain. Power dissipation is concentrated in the PA and driver. A transmitter delivering +30 dBm (1 W) at 25% efficiency dissipates 3 W of heat. At 40% efficiency, it's 1.5 W. That factor-of-two difference in heat drives PCB layout decisions — copper pours, via arrays, thermal interface materials, and sometimes heatsinks.
Where component selection gets complicated
The challenge isn't finding a PA that meets your power spec, or a VCO at the right frequency. Parts exist. The challenge is finding a set of components that work together as a system.
Impedance matching between stages. Most RF components are characterized at 50 ohms. But real inter-stage interfaces often aren't 50 ohms — especially between the driver and PA. Mismatched impedance causes gain ripple, stability risks, and power loss. The matching network design depends on both the source and load impedance, which means you need S-parameter data for both components before you can design the interface.
Supply voltage compatibility. A 28 V GaN PA, a 5 V GaAs driver, and a 3.3 V VCO all need different supply rails. The power supply design and bias sequencing (some devices require specific power-up order to prevent latch-up) directly affect system reliability.
Package and footprint constraints. A QFN driver at 3x3 mm, a flange-mount PA at 10x10 mm, and a shielded VCO module at 5x5 mm all have different thermal pad requirements, grounding strategies, and clearance needs. The PCB layout is constrained by the component choices, and sometimes the layout constraints should inform the component choices.
Availability and lifecycle. An elegant design using a discontinued PA is worthless. Lead times for RF components routinely exceed 20 weeks. Designing with second-source options — and verifying they're actually pin-compatible and parametrically equivalent — is part of the job.
The sequential dependency problem
Traditional transmitter design is sequential. Establish system requirements. Derive block-level specifications. Select components. Design matching networks. Lay out the PCB. Test. When a late-stage problem requires a component change, much of this work gets revisited.
The dependency chain runs in both directions. System requirements flow down to component specs, but component availability and real-world performance constraints flow back up to system-level trade-offs. A design that requires +33 dBm output power at 28 GHz with 35% PAE might be theoretically achievable — but if no available PA meets that combination, the system spec has to flex.
This bidirectional dependency is what makes transmitter design iterative in practice, even when it looks sequential on paper.
The rest of this series goes deeper on each stage: VCO selection in Part 2, mixers and upconversion in Part 3, filter selection in Part 4, driver amplifiers in Part 5, power amplifiers in Part 6, link budget analysis in Part 7, and integration in Part 8.
What's the component selection decision that's given you the most trouble in a transmitter design? For me, it's usually the PA — the trade-offs between efficiency, linearity, and thermal management never seem to converge cleanly.
This is Part 1 of an 8-part series on RF transmitter design. Next: "VCO Selection: Frequency, Phase Noise, and Tuning Range."