Inferensys

Glossary

Gain Compression

The deviation from linear gain at high input drive levels where the amplifier's incremental gain decreases, typically quantified by the 1-dB compression point marking the onset of significant nonlinear behavior.
ML engineer working on model compression and quantization, laptop showing performance benchmarks, technical workspace.
NONLINEAR DISTORTION

What is Gain Compression?

Gain compression defines the point where an amplifier's linear relationship between input and output power breaks down, marking the onset of distortion critical to wireless system performance.

Gain compression is the deviation from linear gain in a power amplifier where the incremental gain decreases as the input drive level increases, typically quantified by the 1-dB compression point (P1dB) marking the onset of significant nonlinear behavior. This phenomenon occurs when the amplifier's active device approaches its saturation region, causing the output power to no longer increase proportionally with the input power.

The P1dB metric defines the output power at which the actual gain has dropped by exactly 1 dB from the ideal small-signal linear gain, serving as a critical boundary between linear and nonlinear operation. In Doherty amplifier architectures, the carrier amplifier's soft compression characteristic, particularly in GaN HEMT devices, interacts with the peaking amplifier's turn-on profile to shape the overall linearity-efficiency trade-off that digital predistortion systems must subsequently correct.

NONLINEAR BEHAVIOR

Key Characteristics of Gain Compression

Gain compression defines the operating boundary between linear and nonlinear amplifier behavior. Understanding its characteristics is essential for setting power back-off levels and designing effective linearization strategies.

01

The 1-dB Compression Point (P1dB)

The P1dB is the most widely used figure of merit for quantifying the onset of gain compression. It is defined as the output power level at which the amplifier's actual gain has dropped by exactly 1 dB relative to its ideal linear (small-signal) gain.

  • Significance: P1dB marks the transition from quasi-linear to strongly nonlinear operation.
  • Measurement: Determined by sweeping input power and plotting the deviation from linear gain.
  • Design Rule: Amplifiers handling modulated signals are typically operated at an average power 6-12 dB below P1dB (back-off) to meet linearity specs.
  • Relationship: P1dB is closely related to, but typically 0.2-0.5 dB below, the saturated output power (Psat) in most solid-state amplifiers.
1 dB
Gain Deviation at P1dB
6-12 dB
Typical Back-Off Range
02

Soft vs. Hard Compression Characteristics

The shape of the gain compression curve varies significantly by transistor technology and bias class, directly impacting the complexity of digital predistortion (DPD) required.

  • Soft Compression: Exhibited by GaN HEMT and LDMOS devices. The gain rolls off gradually over several dB of input power. This smooth nonlinearity is inherently more amenable to polynomial-based DPD models.
  • Hard Compression: Exhibited by some bipolar and CMOS PAs near saturation. The gain drops abruptly, creating sharp nonlinearities and higher-order spectral regrowth that demand more complex, high-order DPD correction.
  • Class-AB Bias: Typically produces a softer compression knee than deep Class-AB or Class-B operation.
GaN HEMT
Soft Compression Technology
03

AM-AM and AM-PM Conversion at Compression

Gain compression is fundamentally an AM-AM distortion (amplitude-to-amplitude modulation), but it is always accompanied by AM-PM distortion (amplitude-to-phase modulation) that worsens near compression.

  • AM-AM: The output envelope amplitude no longer scales linearly with the input envelope. The gain curve flattens.
  • AM-PM: The insertion phase of the amplifier changes as a function of the instantaneous input power. Phase shifts can exceed 10-20 degrees per dB of gain compression in some devices.
  • EVM Impact: The combined AM-AM and AM-PM distortion severely degrades Error Vector Magnitude (EVM), particularly for high-order QAM modulation schemes.
  • DPD Requirement: An effective predistorter must correct both amplitude and phase nonlinearities simultaneously.
10-20°/dB
Typical AM-PM Conversion
04

Spectral Regrowth and ACLR Degradation

When an amplifier is driven into gain compression, the nonlinear transfer function causes spectral regrowth—the spreading of signal energy into adjacent frequency channels.

  • Mechanism: The nonlinearity generates intermodulation products between the spectral components of the modulated signal, filling the adjacent channels with distortion.
  • ACLR Limit: Regulatory bodies specify strict Adjacent Channel Leakage Ratio (ACLR) limits (e.g., -45 dBc for 3GPP). Operating too close to P1dB without linearization will violate these masks.
  • Back-Off Relationship: Without DPD, achieving -45 dBc ACLR for an LTE/5G signal may require 8-12 dB of output back-off from P1dB, severely degrading efficiency.
  • DPD Benefit: Digital predistortion can suppress spectral regrowth by 15-25 dB, allowing operation much closer to P1dB.
-45 dBc
Typical ACLR Limit
15-25 dB
DPD ACLR Improvement
05

Gain Compression in Doherty Amplifiers

In a Doherty power amplifier, gain compression behavior is more complex due to the interaction between the carrier and peaking amplifier stages.

  • Carrier Compression: The Class-AB carrier amplifier reaches its voltage saturation and begins to compress first, typically near the Doherty transition point (6-9 dB back-off).
  • Peaking Turn-On: The Class-C peaking amplifier turns on gradually, and its gain is initially low, contributing to a non-monotonic composite gain characteristic.
  • Composite Nonlinearity: The combined AM-AM profile often exhibits a 'soft knee' followed by a steeper compression as both amplifiers saturate, requiring a dual-branch DPD or a model that captures the combined nonlinearity.
  • Load Modulation Effect: The dynamic impedance presented to the carrier by the peaking amplifier's current injection alters the carrier's gain compression point as a function of instantaneous envelope power.
6-9 dB
Doherty Transition Back-Off
06

Memory Effects Near Compression

As an amplifier approaches gain compression, memory effects become more pronounced, meaning the output depends on the history of the input envelope, not just the instantaneous value.

  • Thermal Memory: Increased power dissipation near compression heats the transistor channel, causing slow gain and phase variations with sub-millisecond time constants.
  • Electrical Memory: Bias network impedances and envelope frequency components modulate the supply and gate voltages dynamically, creating frequency-dependent nonlinear behavior.
  • Trapping Effects: In GaN HEMTs, charge trapping and de-trapping at high drain voltages near compression introduce long-time-constant memory that is particularly challenging to linearize.
  • Modeling Requirement: Accurate DPD models for operation near P1dB must incorporate memory terms (e.g., Memory Polynomial or Volterra Series models) to capture these dynamic effects.
Sub-ms
Thermal Time Constant
GAIN COMPRESSION

Frequently Asked Questions

Clear, technically precise answers to the most common questions about gain compression in power amplifiers, its measurement, and its impact on linearization strategies.

Gain compression is the deviation from linear gain in a power amplifier where the incremental gain decreases as the input drive level increases, marking the onset of significant nonlinear behavior. It occurs when the amplifier's active device approaches its physical output power limits—the transistor's knee voltage and saturation region constrain the maximum voltage and current swing. As the input envelope magnitude increases, the instantaneous gain begins to roll off from the small-signal value, compressing the output waveform peaks. This nonlinear transfer characteristic generates AM-AM distortion and AM-PM distortion, producing spectral regrowth that degrades Adjacent Channel Leakage Ratio (ACLR) and Error Vector Magnitude (EVM). The compression behavior is fundamentally tied to the amplifier's bias class, load line, and semiconductor technology—GaN HEMT devices typically exhibit soft compression characteristics, while LDMOS devices may show sharper compression knees.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.