RF Amplifier Theory — Gain, P1dB, IP3, Noise Figure

// 01 — The Big Picture

What Does an Amplifier Actually Do?

An amplifier takes a small signal and makes it bigger — but that one-sentence description hides a mountain of interesting physics. The signal going in is tiny (maybe a few microvolts from an antenna). The signal coming out might be millions of times larger. The energy to do that doesn't come from nowhere — it comes from the DC power supply. The transistor is essentially a controlled tap on that supply, shaped by your input signal.

Think of it like a water tap. The signal on the input is a tiny finger turning the tap handle. The output is a torrent of water proportional to how far you turned the handle. The water comes from the mains — not from your finger. Your finger (the RF signal) controls, but doesn't supply, the power.

Figure 1 — The amplifier as a controlled tap. Input signal controls the output shape. DC supply provides the energy. An ideal amplifier adds gain without distorting the signal shape.

Gain — the one number everyone asks about

Gain (dB) = 10·log₁₀(P_out / P_in) · e.g. +20 dB = signal is 100× more powerful at the output
Gain (dB) = 20·log₁₀(V_out / V_in) · e.g. +20 dB = voltage is 10× larger

A 20 dB LNA on a −100 dBm antenna signal gives −80 dBm at its output. Still tiny — but now the mixer can work with it.

// 02 — The Power Characteristic

The Power Curve — Everything Lives Here

The most important plot in RF amplifier design is output power vs input power. Everything — gain, compression, P1dB, saturation — is read from this one curve. Understanding it deeply means you understand the amplifier.

Figure 2 — The RF amplifier power characteristic. Three distinct regions: linear (constant gain), compression (gain dropping) and saturation (output pinned). P1dB marks exactly where gain has dropped 1 dB from its small-signal value. Everything to the left of P1dB is "safe" operation for most applications.

The Linear Region — Where You Want to Be

In the linear region, doubling the input power doubles the output power. The gain is perfectly constant. A sine wave in gives a perfect sine wave out — just bigger. This is where receivers operate almost all the time, and where transmitters should operate for clean modulation.

Physically: The transistor is far from its limits. The signal swing is small relative to the DC bias point. The transistor behaves like a linear resistor controlled by the input — perfectly proportional.

Gain Compression — The Amplifier Starts Struggling

As input power increases, the transistor starts hitting the rails — it runs out of headroom on the positive or negative swings. The output can't follow the input perfectly anymore. The gain starts dropping. The amplifier is compressing.

Physically: Imagine pushing the transistor close to cutoff on one side and near saturation on the other. The peaks of the sine wave get clipped — the output looks more and more like a square wave. That's distortion.

Why compression is bad for signals: When an amplifier compresses, it generates new frequency components that weren't in the input — called intermodulation products. In a communication system, these land on adjacent channels and cause interference. Regulators (FCC, ETSI) set strict limits on how much of this you're allowed to produce.

P1dB — The Most Important Single Number on a Datasheet

P1dB (1 dB compression point) is the input (or output) power level where the amplifier's gain has dropped by exactly 1 dB from its small-signal value. It's the agreed-upon boundary between "working properly" and "starting to distort".

P1dB — Input and Output Referred

OP1dB = IP1dB + G − 1 dB · (output-referred = input-referred + gain − 1)

A 20 dB LNA with IP1dB = −20 dBm has OP1dB = −20 + 20 − 1 = −1 dBm
A PA with G = 30 dB and IP1dB = 0 dBm has OP1dB = 0 + 30 − 1 = +29 dBm ≈ 0.8 W

Physical meaning of P1dB: It's the amplifier equivalent of running a car engine at redline. You can do it briefly but it's stressful and the output isn't clean anymore. For linear amplifiers (like those used in 5G and WiFi), you typically back off 6–10 dB from P1dB to keep the signal clean — this is called "output back-off".

Saturation — The Wall

Past the compression region, the output power stops increasing no matter how hard you drive the input. The transistor is fully on or fully off at all times — producing a square wave regardless of what went in. This is Psat. Most PA datasheets quote Psat as the maximum output power of the device — but you never actually want to transmit at Psat for a modulated signal.

// 03 — Intermodulation & IP3

IP3 — The Intermodulation Intercept

P1dB measures one type of nonlinearity (single-tone gain compression). IP3 measures a more subtle and more dangerous type — what happens when two signals are present simultaneously. This is exactly what happens in a real radio — your desired signal plus every other transmitter in the neighbourhood.

When two sine waves at frequencies f₁ and f₂ pass through a nonlinear amplifier, the amplifier generates new frequencies at 2f₁−f₂ and 2f₂−f₁. These are called third-order intermodulation products (IM3). The danger: if f₁ and f₂ are close together (say 1 MHz apart), the IM3 products fall right on top of an adjacent channel — and you can't filter them out.

Figure 3 — Two-tone intermodulation. Two signals at f₁ and f₂ enter the amplifier. The nonlinearity creates IM3 products at 2f₁−f₂ and 2f₂−f₁ — spaced exactly Δf away from the desired signals. You cannot filter them away because they sit in the adjacent channel.

Figure 4 — The IP3 intercept point. The fundamental output grows at slope 1 (dB for dB). The IM3 product grows at slope 3 (3 dB for every 1 dB of input). Extrapolate both lines until they cross — that intersection is the IP3 point. It's a fictional operating point (you'd never drive the amp that hard) but it's a consistent figure of merit that tells you how linear the device is.

The Rule of Thumb — and Why It Matters

P1dB vs IP3 — The Golden Ratio

IIP3 ≈ IP1dB + 9.6 dB (theoretical, for a memoryless 3rd-order nonlinearity)
In practice: IIP3 ≈ IP1dB + 8 to 12 dB (slightly varies by device)

Example: LNA with IP1dB = −20 dBm → IIP3 ≈ −10 dBm
Interpretation: If you need IIP3 = +10 dBm, you need an amplifier with P1dB ≈ 0 dBm at minimum.

Why does IM3 grow 3× faster? A nonlinear device produces output terms proportional to input¹ (fundamental), input² (second order), input³ (third order). The third-order term grows as input³ — so every 1 dB increase in input causes a 3 dB rise in the IM3 output. The slope difference of 2 dB/dB means they must eventually intersect — that's IP3.

// 04 — Noise Figure

Noise Figure — How Dirty Is the Amp?

Every amplifier adds its own noise on top of the signal. Noise Figure (NF) measures how much the amplifier degrades the signal-to-noise ratio. An ideal, noiseless amplifier would have NF = 0 dB. Every real amplifier has NF > 0 dB.

Think of it like a photocopier. You put in a document and get back an amplified copy — but the copier has added its own specks and smudges. The NF tells you how many extra smudges the copier added. A low NF amplifier is a clean copier.

Figure 5 — What noise figure means physically. The LNA amplifies the signal by 20 dB but adds its own noise — so noise goes up by 22 dB while signal only goes up by 20 dB. The SNR dropped by 2 dB. That 2 dB degradation is the noise figure.

The Gain vs NF Trade-off

Here's the cruel fact of RF amplifier design: increasing gain helps noise figure of the overall chain but hurts linearity. A high-gain LNA pushes more signal into the mixer, which reduces the mixer's NF contribution in the Friis chain — great. But that same high gain drives the mixer harder, reducing IIP3. You're always trading one for the other.

Parameter	More Gain	Less Gain
System NF	Better (later stage NF suppressed)	Worse (mixer/IF NF matters more)
System IIP3	Worse (more signal into nonlinear stages)	Better (less drive into mixer)
Dynamic Range	Narrower	Wider
Sensitivity	Better	Worse

The design problem: For a cellular handset, you need to receive signals from both a distant base station (weak, needs low NF) and a nearby base station (strong, needs good IIP3). The LNA gain choice is the single most critical design variable — it determines both sensitivity and blockers rejection simultaneously.

// 05 — Amplifier Types

Every Type of RF Amplifier

RF amplifiers are not one-size-fits-all. The application dictates the design completely. Here's every major type you'll encounter:

LNA — Low Noise Amplifier

The Receiver's First Line

Lives right at the antenna output — the most critical position in any receiver. Its job: amplify the tiny antenna signal before anything else touches it.

Designed for: Minimum NF above all else. Gain is secondary. Linearity is tertiary.
Typical NF: 0.5–3 dB
Typical gain: 12–25 dB
Typical IIP3: −10 to +5 dBm

Why it comes first: The Friis formula shows that the first stage dominates system NF. An LNA at the front with NF=1 dB and gain=20 dB means everything after it contributes 100× less to the total NF.

Real examples: Qorvo QPL9065 (0.55 dB NF, 5 GHz WiFi), Skyworks SE2632 (LTE handset)

PA — Power Amplifier

The Transmitter's Engine

Sits at the end of the transmit chain, driving the antenna with as much clean power as possible. Opposite design priorities to the LNA.

Designed for: Output power and efficiency above all else. NF irrelevant. Linearity critical for modern modulations.
Typical output power: +20 to +40 dBm (100 mW to 10 W)
Typical efficiency (PAE): 20–60%

The efficiency problem: A 2 W PA running at 30% PAE burns 6.7 W from the battery for every 2 W it radiates. The rest becomes heat. In a handset, the PA is the largest single consumer of battery power.

Real examples: Qorvo QPA3064 (5G sub-6), Wolfspeed GaN PAs (base stations)

Driver Amplifier

The Stage Before the PA

The PA needs a certain input drive level — often +10 to +20 dBm — which the previous stage (usually a DAC or modulator) can't provide. The driver amplifier bridges this gap.

Designed for: Linearity and moderate output power. Must drive the PA input without compressing or adding distortion that degrades EVM.
Typical gain: 15–25 dB
Typical OP1dB: +20 to +28 dBm

Key spec: OIP3 and EVM performance at the required output power level. A driver with poor IIP3 corrupts the modulation before the PA even sees it.

VGA — Variable Gain Amplifier

The Gain Controller

A VGA has electronically adjustable gain — typically 30–50 dB of range controllable by a voltage or digital word. Used in AGC (Automatic Gain Control) loops to keep signal levels constant regardless of input amplitude.

Where you find it: Every receiver has one. The baseband processor detects if the signal is too strong or too weak and adjusts the VGA continuously.

Key challenge: Gain should change but NF and IIP3 should stay constant (or trade gracefully). Most VGAs get noisier at low gain settings — which is fine because at low gain you have a strong signal anyway.

Balanced Amplifier

Two Amps + Two Couplers

Two identical amplifier stages connected with 90° hybrid couplers at input and output. The reflections from each amplifier cancel at the input, giving perfect S11 = −∞ (ideally) across a wide bandwidth.

Advantage: Excellent input/output match over multi-octave bandwidth, graceful degradation if one stage fails.
Disadvantage: Uses twice the chip area and current. 3 dB of the output power goes into the termination load (not the signal path).

Where you find it: Wideband test equipment amplifiers, satellite transponders, radar front-ends.

Distributed Amplifier

The Ultra-Wideband Champion

Multiple transistors spaced along an artificial transmission line. The input signal travels down the input line, and each transistor adds a contribution that travels down the output line in phase.

Advantage: Enormous bandwidth — DC to 100+ GHz with a single design. The parasitic capacitances of the transistors become part of the transmission line.
Disadvantage: Low gain per stage, high DC power, large size.

Where you find it: Oscilloscope front-end amplifiers, microwave test instruments, EW (electronic warfare) systems needing instantaneous bandwidth of several octaves.

Power Amplifier Classes — A, AB, B, C, D, E, F

PA classes describe how the transistor is biased — specifically, what fraction of the RF cycle the transistor is conducting. This is the single biggest factor in PA efficiency.

Figure 6 — PA conduction angle. Class A conducts the full 360° of every cycle — always drawing current, always burning power regardless of signal level. Class B only conducts for 180° — much more efficient. Classes C, D, E, F push this further.

Class	Conduction Angle	Max Efficiency	Linearity	Best for
A	360°	50%	Excellent	Low-power linear apps, LNA topology
AB	180°–360°	50–78%	Good	Most RF PAs — practical sweet spot
B	180°	78.5%	Fair (crossover distortion)	Push-pull PAs, audio power amps
C	<180°	Up to 100%	Poor (heavy distortion)	FM transmitters, fixed-frequency CW radar
D	Switch mode	~90–95%	Nonlinear (switching)	Audio amps, DC-DC converters, EER systems
E	Switch mode	~96%	Nonlinear — needs linearisation	IoT, Bluetooth, constant-envelope signals
F	Switch + harmonics	~90%	Nonlinear	High-efficiency microwave PAs

The efficiency problem in 5G: 5G NR uses OFDM with a very high Peak-to-Average Power Ratio (PAPR) of 10–13 dB. This means the PA must handle peaks 10–20× the average power. A class AB PA optimised for average power efficiency still burns extra power for every peak — meaning real-world efficiency is far lower than the theoretical maximum. This is why 5G base stations run hot and why mmWave handsets have poor battery life.

// 06 — S-Parameters: What They Actually Mean

S-Parameters — Reading the Amplifier's Report Card

S-parameters are how amplifiers talk to you on a datasheet. Four numbers — S11, S21, S12, S22 — completely describe how a two-port device behaves at a given frequency. Once you understand what each one physically means, datasheets go from confusing to obvious.

The key concept: S-parameters describe what fraction of a wave is transmitted or reflected at each port when the other port is properly terminated (matched to 50 Ω). Think of them as a scorecard of waves.

Figure 7 — S-parameter wave definition. Incident wave a₁ arrives at Port 1. Some reflects back as b₁ (that ratio is S11). Some transmits through to Port 2 as b₂ (that ratio is S21). S22 and S12 are measured similarly with a wave incident at Port 2.

S11

Input Reflection — "What the amp rejects"

S11 is the ratio of the reflected wave at Port 1 to the incident wave at Port 1. In plain English: how much of the signal bounces back from the amplifier input.

A perfect match: S11 = 0 (−∞ dB) — nothing bounces back.
Typical LNA S11: −10 to −20 dB (1–10% of power reflected)
Poor match: S11 = −5 dB (30% of power wasted!)

Physical picture: The amplifier input looks like an impedance. If that impedance isn't 50 Ω, the wave hits a "wall" and bounces. S11 measures the height of that wall. Also called input return loss (S11 in dB with a sign flip).

S21

Forward Gain — "What you bought it for"

S21 is the ratio of the wave coming out of Port 2 to the wave going into Port 1. This is simply the gain (or insertion loss).

S21 = +20 dB → amplifier gain of 20 dB ✓
S21 = −2 dB → filter with 2 dB insertion loss
S21 = −0.5 dB → nearly transparent (cable, connector)

Physical picture: The fraction of the wave that successfully travels from input to output through the device. For an amplifier this is >1 (it adds energy). For a passive device it's always <1 (some energy is lost).

S12

Reverse Isolation — "The leak backwards"

S12 is the signal that leaks backwards from Port 2 to Port 1. For an ideal amplifier, S12 = 0 — nothing leaks. In reality there's always some leakage.

Typical LNA S12: −20 to −40 dB
Bad S12: −10 dB (10% of output power leaks back to antenna)

Why it matters: The LO in a receiver is a strong oscillator. If it leaks backwards through S12 of the LNA, it radiates out of the antenna — making your receiver also a transmitter and potentially violating spectrum regulations. Good isolation (low S12) keeps the LO inside the radio.

S22

Output Reflection — "What bounces off the output"

S22 is the reflection at the output port — what fraction of a wave arriving at Port 2 reflects back. For an amplifier driving a 50 Ω load, you want S22 to be small.

Typical PA S22: −10 to −20 dB

Why it matters: If the PA output isn't well-matched to the antenna, some power bounces back into the PA. In a high-power system this reflected power can damage the transistor — this is why PAs have built-in protection circuits and why antenna VSWR matters so much.

Reading an Amplifier Datasheet S-Parameter Plot

On a real datasheet, S-parameters are plotted vs frequency. Here's how to interpret what you see:

Figure 8 — Typical LNA S-parameter plot (2–6 GHz). S21 (green) shows the gain is flat ~18 dB in the mid-band then rolls off at high frequency. S11 (red) shows the input is well-matched around 2.5 GHz (deep dip = good match). S22 (purple) shows output match. S12 (yellow) stays low — good isolation. Read this plot on every LNA/PA datasheet.

What you see on the plot	What it means physically	What's good
S21 flat and high	Amplifier has consistent gain across the band — signal comes out clean and amplified the same amount at every frequency	High and flat — ±1 dB across bandwidth
S21 rolling off at high f	The transistor runs out of bandwidth — gain falls at its fT. Normal for all transistors.	Know the 3 dB bandwidth and design within it
S11 deep dip at a frequency	At that frequency the input impedance is exactly 50 Ω — all power absorbed, none reflected	Deep dip (< −15 dB) at your operating frequency
S11 rising towards 0 dB	Input is poorly matched — significant power reflected. If S11 > 0 dB, the device is active (amplifying back)	Above −10 dB is poor match
S12 low and flat	Good reverse isolation — LO and noise don't leak backwards through the device	Below −20 dB is good
S12 rising at high frequency	Capacitive feedthrough in the transistor — increasing at high frequency. A sign the device may become unstable	Watch for stability issues when S12 rises

// 07 — Stability

Stability — Will It Oscillate?

An amplifier that oscillates is no longer an amplifier — it's a transmitter, and usually a badly-designed one radiating at a frequency nobody asked for. Stability is the single most important property after gain.

Physically: If any signal coming out of the output leaks back through S12 to the input, and gets re-amplified with enough gain to make it bigger than it started, the device will sustain its own oscillation. The loop gain must be <1 to prevent this.

Stability Criteria

Rollett K factor: K = (1 − |S11|² − |S22|² + |Δ|²) / (2·|S12·S21|) where Δ = S11·S22 − S12·S21

Unconditionally stable if K > 1 AND |Δ| < 1 at ALL frequencies
Conditionally stable if K < 1 at some frequency — stable only for certain source/load impedances
Potentially unstable — can oscillate with the wrong termination

Also check: μ factor = (1 − |S11|²) / (|S22 − Δ·S11*| + |S12·S21|) > 1 for unconditional stability

The most common stability mistake: An amplifier that is K>1 at its design frequency may still oscillate at a completely different frequency — often low frequency (100 MHz) or at the edge of the transistor's fT. Always plot K vs frequency from DC to several times fT. Many engineers have taped-up boards oscillating at 50 MHz while trying to build a 5 GHz amplifier.

Amplifier Stability Analyser → upload .s2p to compute K, Δ and μ → Plot S-parameters from Touchstone file →

// 08 — Try the Tools