Inferensys

Glossary

Adjacent Channel Leakage Ratio (ACLR)

The ratio of power leaking into adjacent frequency channels relative to the main channel power, a key metric for spectral regrowth compliance.
Compliance officer monitoring AI compliance agent on laptop, policy dashboards visible, modern WeWork desk setup.
SPECTRAL REGROWTH METRIC

What is Adjacent Channel Leakage Ratio (ACLR)?

ACLR quantifies the amount of power from a transmitter's main channel that spills into adjacent frequency channels, serving as the primary regulatory compliance metric for spectral regrowth caused by power amplifier nonlinearity.

Adjacent Channel Leakage Ratio (ACLR) is the ratio of the integrated power in a specified adjacent frequency channel to the integrated power in the transmitter's assigned main channel, expressed in dBc. It directly measures spectral regrowth—the unwanted broadening of a signal's bandwidth caused by nonlinear distortion in the power amplifier, particularly from AM-AM distortion and AM-PM conversion that generate intermodulation products outside the intended carrier.

ACLR is the definitive pass/fail metric for regulatory standards like 3GPP's transmit spectrum mask requirements. In mmWave digital predistortion systems, improving ACLR is the primary optimization target, as beamforming arrays with antenna crosstalk and active impedance mismatch exhibit channel-specific leakage that must be suppressed to prevent interference with adjacent carriers in dense 5G NR deployments.

SPECTRAL REGROWTH DRIVERS

Key Factors Influencing ACLR

Adjacent Channel Leakage Ratio degradation stems from multiple interacting physical and signal-level phenomena. Understanding these root causes is essential for effective linearization.

01

AM-AM Distortion

Amplitude-to-Amplitude nonlinearity compresses the signal envelope, generating spectral regrowth that spills power into adjacent channels.

  • Caused by gain compression near the amplifier's saturation point
  • Creates odd-order intermodulation products that fall directly into adjacent bands
  • Severity increases with higher Peak-to-Average Power Ratio (PAPR) signals like OFDM
  • Example: A 3 dB gain compression can degrade ACLR by 10-15 dB in a Class-AB amplifier
10-15 dB
Typical ACLR Degradation
02

AM-PM Conversion

Amplitude-to-Phase distortion introduces signal-dependent phase shifts that broaden the transmitted spectrum asymmetrically.

  • Phase shift varies with instantaneous envelope power
  • Produces upper and lower sideband asymmetry in the regrowth spectrum
  • Particularly problematic in GaN HEMT devices due to nonlinear input capacitance
  • Cannot be corrected by amplitude-only predistortion; requires complex-valued DPD
5-10°
Phase Deviation at P1dB
03

Memory Effects

Thermal and electrical memory causes the amplifier's transfer characteristic to depend on past signal values, not just the instantaneous input.

  • Short-term memory: Bias circuit impedance variations at the modulation envelope rate
  • Long-term memory: Substrate temperature fluctuations and trapping effects in GaN devices
  • Creates frequency-dependent nonlinearity that static DPD cannot correct
  • Memory polynomial models with 3-5 taps typically required for wideband signals
3-5
Memory Taps Required
04

Power Back-Off Level

Operating point relative to the amplifier's 1 dB compression point (P1dB) directly determines the inherent linearity and achievable ACLR.

  • Higher back-off improves linearity but reduces Power-Added Efficiency (PAE)
  • Modern 5G signals require 6-10 dB back-off without DPD for compliant ACLR
  • DPD enables operation at 2-4 dB back-off while maintaining -45 dBc ACLR
  • Trade-off: Every 1 dB reduction in back-off can improve PAE by 2-3 percentage points
2-4 dB
Back-Off with DPD
05

IQ Modulator Impairments

Quadrature modulator errors in the transmitter chain create additional distortion products that degrade ACLR even with a perfectly linear amplifier.

  • IQ gain imbalance creates an unwanted image signal
  • Quadrature skew (phase error) causes constellation distortion
  • LO leakage produces a carrier feedthrough spur
  • These impairments interact with PA nonlinearity, requiring joint compensation in the DPD signal path
3-5 dB
ACLR Penalty from IQ Errors
06

Signal Bandwidth and PAR

Wider bandwidths and higher Peak-to-Average Ratios stress the amplifier's frequency response flatness and dynamic range.

  • 5G NR signals with 100 MHz bandwidth and 10+ dB PAR demand wideband linearization
  • Frequency-dependent gain ripple across the band creates unequal distortion products
  • Higher PAR pushes the signal deeper into compression on peaks, generating more regrowth
  • DPD bandwidth must be 3-5x the signal bandwidth to capture third and fifth-order intermodulation
3-5x
DPD Bandwidth Multiplier
ACLR FUNDAMENTALS

Frequently Asked Questions

Clear, concise answers to the most common questions about Adjacent Channel Leakage Ratio, its measurement, and its critical role in spectral compliance and digital predistortion.

Adjacent Channel Leakage Ratio (ACLR) is the ratio of the filtered mean power centered on an assigned channel frequency to the filtered mean power centered on an adjacent channel frequency. It is the primary metric for quantifying spectral regrowth—the unwanted spread of a modulated signal's energy into neighboring frequency bands caused by the nonlinearity of a power amplifier (PA). ACLR is typically expressed in dBc (decibels relative to the carrier). A higher negative value indicates better linearity and less interference. For example, the 3GPP standard for 5G NR mandates specific ACLR limits (often -45 dBc or better) to prevent a transmitter from desensitizing receivers operating on adjacent channels. The measurement involves integrating the power spectral density over the assigned channel bandwidth and the adjacent channel bandwidth, making it a direct gauge of out-of-band emissions compliance.

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.