RF Power Amplifiers
The power amplifier (PA) is the final active stage in any transmitter — the stage that takes the small RF signal that has been generated, processed, and shaped upstream and raises it to the power level required for transmission. In a 100 W HF transceiver, that means amplifying a signal of perhaps 10–100 mW by a factor of 1000–10000 (30–40 dB of power gain) while maintaining the signal's shape, keeping harmonics and spurious outputs within regulatory limits, and doing it all efficiently enough that the heat generated does not destroy the transistors.
RF power amplifiers present unique engineering challenges that do not exist in small-signal amplifier design. At power levels above a few watts, thermal management becomes critical — a transistor that dissipates 50 W must be mounted on a heatsink large enough to keep its junction temperature below the rated maximum. Impedance matching is critical — the transistor's optimum load impedance is rarely 50 Ω and must be transformed by a matching network. Stability is critical — RF power transistors can oscillate parasitically and destroy themselves in milliseconds. And regulatory compliance is critical — every watt of harmonic power radiated is a potential interference source and an invitation for a regulatory notice.
- RF PAs vs Audio Amplifiers
- Class AB Broadband Linear PA
- Class C Tuned PA
- Transistor Selection — BJT vs MOSFET vs LDMOS
- Optimum Load Impedance
- Pi Network Output Matching
- Broadband Transformer Matching
- Heatsinking and Thermal Design
- Protection Circuits
- Pi Network Component Calculator
- Frequently Asked Questions
RF PAs vs Audio Amplifiers
RF power amplifiers share the same basic transistor physics as audio amplifiers, but the practical design differs in important ways:
| Parameter | Audio PA | RF PA |
|---|---|---|
| Frequency range | 20 Hz – 20 kHz | 1 MHz – multi-GHz (HF: 1.8–30 MHz; VHF: 30–300 MHz) |
| Transistor selection | Any power transistor with adequate fT | Must have fT ≫ operating frequency; RF-specific types (LDMOS, GaN HEMT, Si BJT RF types) |
| Output matching | Output transformer or complementary pair directly to speaker (4–8 Ω) | Pi network, L network, or broadband transformer to 50 Ω coaxial system |
| Harmonic filtering | Inherent bandwidth limitation — harmonics above 20 kHz are inaudible | Mandatory low-pass filter to meet regulatory harmonic limits |
| Parasitic oscillation risk | Low — audio frequencies are far from transistor fT | High — parasitic L and C in PCB traces and package become resonant at RF |
| Load variation | Speaker impedance relatively constant | Antenna SWR can vary widely; PA must tolerate mismatched loads without failure |
Class AB Broadband Linear PA
For SSB, CW, AM, and data modes, the power amplifier must be linear — it must reproduce the amplitude envelope of the modulated signal faithfully. This requires Class A or Class AB operation. Class A wastes too much power for kilowatt-level designs, so Class AB is the standard for all modern solid-state linear HF power amplifiers.
The Class AB broadband linear PA for HF typically uses:
- Push-pull topology (two matched transistors) for even-harmonic cancellation
- Broadband ferrite-core transformers at input and output — covers 1.8–30 MHz without retuning
- Class AB bias (idle current of 50–200 mA total) for low crossover distortion
- Low-pass filter at the output for harmonic suppression (mandatory under amateur regulations)
- ALC (automatic level control) feedback to prevent overdrive
The broadband design has a significant advantage over tuned designs: the operator can change bands without retuning the PA. The trade-off is that broadband transformers cannot resonate out the transistor's output capacitance (Coss), which limits efficiency compared to a tuned design. Broadband HF solid-state PAs typically achieve 50–60% efficiency at full output, compared to 65–75% for tuned designs.
Class C Tuned PA
For CW and FM modes where amplitude linearity is not required, Class C offers higher efficiency. In Class C, the transistor conducts for less than 180° of the RF cycle — it fires in short pulses at the peak of the drive signal. These pulses are very rich in harmonics, but the output is connected to a parallel LC tank circuit (pi network) tuned to the fundamental frequency. The tank circuit acts as a flywheel, sustaining sinusoidal oscillation between drive pulses and presenting a high impedance at the operating frequency while shunting harmonics to ground.
A well-designed Class C stage can achieve 80–85% efficiency, meaning only 15–20% of the DC input power is dissipated as heat. This makes Class C attractive for CW transmitters and FM repeaters where power efficiency is important. Class C is also used in the driver stages of high-power AM transmitters (the modulated final stage is Class AB or Class B to preserve amplitude linearity).
The critical constraint of Class C is that it requires a tuned output network — there is no other way to reconstruct a sine wave from the short pulses of collector current. This means Class C PAs must be retuned when changing frequency, which is why modern multiband transceivers almost always use Class AB broadband designs even though they are less efficient.
Transistor Selection — BJT vs MOSFET vs LDMOS
| Device Type | Typical Applications | Advantages | Disadvantages |
|---|---|---|---|
| Silicon BJT (e.g. 2SC2782, MRF150) | HF PA up to ~30 MHz, 1–100 W | Well characterised, low cost, good linearity | Thermal runaway risk, lower fT than MOSFET at VHF, must match beta |
| RF MOSFET (e.g. IRF510, RD16HHF1) | HF/VHF PA, 1–50 W | No thermal runaway (positive temperature coefficient of RDS_on), easy to parallel, no gate drive current | High input capacitance (harder to drive), substrate diode can be damaged by reverse voltage transients |
| LDMOS (e.g. MRFE6VP61K25H) | HF/VHF/UHF PA, 100 W – 1 kW+ | Excellent gain, power density, linearity; designed for RF; available in pre-matched package | Expensive, requires stable DC bias circuit, ESD-sensitive gate |
| GaN HEMT (e.g. CGH40010) | VHF/UHF/microwave PA, 10–100 W | Very high power density, high voltage operation, excellent efficiency at microwave frequencies | Expensive, requires careful matching, depletion-mode (normally on) needs gate bias circuit |
For HF amateur radio construction, the IRF510 MOSFET is the most popular choice — inexpensive, widely available, and usable from 1.8–30 MHz at powers up to about 50 W with a 28–30 V supply. For commercial-grade 100 W HF transceivers, LDMOS devices have become the standard. For VHF and UHF (2 metres and 70 cm band), purpose-made RF MOSFETs or GaN devices are used.
Optimum Load Impedance
The output impedance of an RF power transistor at its operating power and supply voltage is not 50 Ω. It is determined by the supply voltage and the peak collector (or drain) current needed for the rated power output:
RL(opt) = (VCC − Vsat)² / (2 × Pout)
where Vsat is the transistor's collector-emitter saturation voltage (typically 1–3 V for BJTs, even lower for MOSFETs)
Example: VCC = 28 V, Vsat = 2 V, Pout = 100 W
RL(opt) = (28 − 2)² / (2 × 100) = 676 / 200 = 3.38 Ω
For a push-pull stage with two transistors, each transistor sees the same optimum load but the centre-tapped transformer means the effective collector-to-collector impedance is 4 × R_L(opt). This collector-to-collector impedance must be transformed to 50 Ω by the output matching network.
Pi Network Output Matching
The pi network (named for its resemblance to the Greek letter π) is the traditional output matching network for tuned RF power amplifiers. It consists of two shunt capacitors (C1 at the input side, C2 at the output side) and a series inductor (L) between them. The pi network has three important functions:
- Impedance transformation: It transforms the transistor's low optimum load impedance (often 3–10 Ω) to the antenna system's 50 Ω.
- Harmonic filtering: The pi network acts as a low-pass filter, attenuating harmonics. A well-designed pi network can provide 30–40 dB of 2nd and 3rd harmonic attenuation.
- Band selection: The variable capacitors C1 and C2 (the TUNE and LOAD controls on a classic tube amplifier, or their solid-state equivalents) allow the network to be tuned to any frequency within its design range.
Input impedance: Rin = RL(opt) of transistor
Output impedance: Rout = 50 Ω (antenna system)
Q factor (loaded): QL = typically 10–15 for HF (compromise between harmonic attenuation and bandwidth)
XC1 = Rin / QL → C1 = 1 / (2π × f × XC1)
XL = QL × Rin → L = XL / (2π × f)
XC2 = Rout / √(QL² × (Rin/Rout) − 1) → C2 = 1 / (2π × f × XC2)
Broadband Transformer Matching
Most modern solid-state HF transceivers use broadband ferrite-core transformers instead of tuned pi networks for band-switching convenience. A broadband transformer can cover 1.8–30 MHz in a single winding without retuning.
The impedance ratio of the transformer is set by the turns ratio squared:
For Rcc = 12.5 Ω (collector-to-collector) → 50 Ω (antenna):
(Nsec/Npri)² = 50/12.5 = 4 → Nsec/Npri = 2
A 1:2 turns ratio (2:4 turns, bifilar wound) achieves a 1:4 impedance transformation.
The ferrite toroid is wound with bifilar or trifilar wire so that the two halves of the push-pull stage couple equally to the common magnetic core. Type 61 ferrite (μ ≈ 125) is a common choice for broadband HF transformers covering 1.8–30 MHz. Type 43 ferrite works better at 1–10 MHz but has higher loss above 10 MHz. The transformer must handle the full output power: for a 100 W amplifier, the transformer must manage 100 W continuously without saturating or overheating its core.
After the broadband output transformer, a multi-band low-pass filter is used to suppress harmonics — typically a set of switched LC low-pass filters, one for each amateur band. These filters replace the harmonic attenuation that the pi network provides in tuned designs.
Heatsinking and Thermal Design
Thermal management is the most common cause of RF PA failure. At 65% efficiency with 100 W output, the PA dissipates 54 W as heat. Each transistor dissipates approximately 27 W. Junction temperature must remain below the rated maximum (typically 150–200°C for silicon). The thermal path is:
where:
θJC = junction-to-case thermal resistance (from datasheet, °C/W)
θCS = case-to-heatsink thermal resistance (typically 0.1–0.5 °C/W with heatsink compound)
θSA = heatsink-to-ambient thermal resistance (°C/W, depends on heatsink size and airflow)
Each transistor dissipates 27 W.
θ_JC = 1.5 °C/W (from transistor datasheet)
θ_CS = 0.2 °C/W (with thermal compound)
θ_SA = 2.0 °C/W (heatsink in natural convection)
T_ambient = 40°C (warm shack in summer)
T_junction = 40 + 27 × (1.5 + 0.2 + 2.0) = 40 + 27 × 3.7 = 40 + 99.9 = 139.9°C
This is below the 150°C maximum for most silicon transistors, but only just. Adding a small fan to improve convection can reduce θ_SA from 2.0 to 0.5 °C/W, reducing T_junction to 40 + 27×2.2 = 99.4°C — a much safer margin. This illustrates why fans are standard in most 100 W HF transceivers.
Protection Circuits
RF power amplifiers require protection against several failure modes:
SWR Protection
When the antenna load deviates from 50 Ω, reflected power returns to the PA. High SWR can present the transistors with an impedance much higher than their optimum load, causing the collector voltage to swing above the supply voltage (reactive stored energy in the mismatched line). This can exceed the transistor's VCEO or VDSS rating and cause immediate failure. SWR protection circuits sample the forward and reflected power (using a directional coupler) and reduce the ALC (automatic level control) voltage to back off the PA drive when SWR exceeds a safe threshold — typically 2:1 to 3:1.
Thermal Protection
A thermistor or temperature-sensing IC (e.g. LM35) mounted on the heatsink sends a signal to a protection circuit that reduces PA output or shuts down completely if heatsink temperature exceeds a set threshold — typically 70–80°C. Many modern transceivers also enable the cooling fan at reduced power levels and increase fan speed as temperature rises.
Overcurrent Protection
A current sensor in the PA supply rail shuts down or folds back the PA if drain/collector current exceeds the safe rating. This protects against drive overdrive (too much input signal) and against short-circuit conditions. Current foldback (where output power is progressively reduced as current increases) is preferred over hard shutdown because it handles sustained fault conditions gracefully without creating high-voltage transients.
Pi Network Component Calculator
Pi Network Output Matching Calculator
Design the C1, L, C2 component values for a pi network matching a transistor's optimum load impedance to a 50 Ω antenna system at a given frequency and Q factor. Assumes ideal (lossless) components.
Frequently Asked Questions
Why is the optimum load impedance of an RF power transistor much lower than 50 Ω?
The optimum load impedance is determined by the supply voltage and the output power required. R_L(opt) = (VCC − Vsat)² / (2 × Pout). With VCC = 28 V and 100 W output, R_L(opt) ≈ 3.4 Ω. At 50 V supply and 1 kW output, R_L(opt) ≈ 1.2 Ω. This very low impedance is why output matching networks are essential — without the pi network or transformer to step the impedance up to 50 Ω, almost all the transistor's output power would be reflected back rather than delivered to the antenna.
Why do RF MOSFETs not suffer from thermal runaway, while BJTs do?
BJT collector current increases with temperature (β rises with temperature, increasing IC, which increases temperature further — thermal runaway). MOSFET drain current decreases with temperature because the channel mobility falls as the lattice heats up. This means a hot MOSFET conducts less, not more — a self-stabilising negative feedback. Multiple MOSFETs can be paralleled without emitter-equalising resistors, and a single MOSFET can tolerate load mismatch better without risk of runaway. This property makes MOSFETs and LDMOS the preferred device for modern solid-state HF and VHF power amplifiers above about 50 W.
Why must a Class AB linear HF amplifier be followed by a low-pass filter?
Even a Class AB linear PA generates harmonics — typically 30–40 dB below the fundamental without filtering, due to the transistors' non-linearity and the push-pull topology's cancellation of even harmonics. Regulations (FCC Part 97, OFCOM, ITU) require that harmonics from amateur transmitters above 30 MHz be at least 43 dB below the carrier. A broadband Class AB PA with no harmonic filter may only achieve 30–35 dB harmonic suppression. A multi-section low-pass filter (typically Chebyshev or elliptic design, one per amateur band) adds 20–40 dB additional attenuation, ensuring compliance. In broadband solid-state transceivers, these are usually implemented as switched filter banks.
What happens to an RF PA when the antenna is disconnected during transmission?
When the antenna is disconnected, the PA sees a theoretically infinite SWR — an open circuit. The collector or drain voltage can swing to nearly 2×VCC due to the stored energy in the output circuit inductance. For a transistor with VCEO = 60 V running from a 28 V supply, a 2×VCC transient of 56 V is right at the breakdown limit and can immediately destroy the device. The solution is an SWR protection circuit that reduces PA drive within microseconds of detecting a high reflected power condition, and physically robust transistors with VCEO or VDSS ratings well above 2×VCC. Many commercial transceivers also include a protection relay that disconnects the antenna socket from the PA if SWR exceeds a safe level before the PA itself is switched on.
Why do some QRP (low power) amplifiers use the IRF510 MOSFET, and what are its limitations?
The IRF510 is an inexpensive, widely available power MOSFET originally designed for switching applications. Its characteristics — adequate gain at HF, 100 V drain-source rating, 5.6 A drain current, TO-220 package — make it workable as an RF power device in the 1.8–30 MHz range at powers up to about 20–40 W. It is a favourite in QRP transceiver projects (BITX, many homebrewed rigs) because it costs under a dollar. Its limitations are: relatively high output capacitance (Coss ≈ 200 pF) that reduces efficiency at higher HF frequencies; lower gain than purpose-made RF transistors (more drive power needed); and being a general-purpose switching MOSFET rather than an RF type, its gain and output capacitance are not characterised at RF frequencies in the datasheet. At frequencies above 14 MHz it works but efficiency and gain drop significantly.
Check Your Understanding
Answer each question and click Check to see the explanation.