Adjacent Channel Leakage Ratio (ACLR)

ACLR is the ratio of the filtered mean power centered on the assigned channel frequency to the filtered mean power centered on an adjacent channel. It is typically expressed in dBc (decibels relative to the carrier). The measurement integrates power over a specified bandwidth using a root-raised-cosine filter, capturing the spectral regrowth that spills into neighboring channels due to transmitter non-linearity. For 3GPP LTE and 5G NR, ACLR is a conformance requirement, with typical limits of -33 dBc or better for the first adjacent channel.

ACLR degradation originates primarily in the power amplifier (PA) , the final active stage of a transmitter. When a high-PAPR signal drives the PA near its compression point, the AM-AM distortion (amplitude compression) and AM-PM distortion (phase shift) generate intermodulation products. These products cause spectral regrowth—a widening of the transmitted signal's bandwidth that leaks into adjacent channels. The specific shape and level of this regrowth is a unique, device-specific fingerprint caused by the PA's semiconductor physics and biasing.

While both ACLR and Error Vector Magnitude (EVM) quantify signal degradation from non-linearity, they measure different effects. EVM captures in-band distortion—the deviation of the actual constellation points from their ideal positions. ACLR captures out-of-band distortion—the unwanted emissions into adjacent spectrum. A transmitter can have excellent EVM but poor ACLR if the PA's non-linearity generates significant out-of-band products without severely corrupting the modulation symbols themselves.

Modern wideband signals expose PA memory effects, where the amplifier's output depends not only on the instantaneous input but also on previous symbols. These effects, caused by thermal dynamics, bias circuit impedance, and trapping effects in the semiconductor, create an asymmetric spectral regrowth pattern. The ACLR may differ between the upper and lower adjacent channels. Accurately modeling this asymmetry is essential for high-fidelity synthetic impairment generation using Volterra series or memory polynomial models.

In synthetic RF impairment generation, ACLR serves as a primary fidelity metric. A digital twin of a transmitter must produce signals whose ACLR matches the physical device within a tight tolerance (e.g., ±0.5 dB) across varying power levels and signal bandwidths. Discrepancies in ACLR between synthetic and real signals indicate that the PA non-linearity model—whether a Generalized Memory Polynomial or a neural network—is incorrectly parameterized, leading to a non-representative fingerprint.

Beyond validation, ACLR itself can be a discriminative feature for RF fingerprinting. Different transmitters of the same model exhibit subtle variations in ACLR due to manufacturing variances in their PA transistors, matching networks, and power supply regulation. A deep learning classifier can use the ratio of upper-to-lower ACLR and the shape of the spectral regrowth shoulder as identifying characteristics, particularly when combined with other impairments like I/Q imbalance and carrier frequency offset.

The primary physical cause of elevated ACLR. When a PA operates near saturation, its AM-AM and AM-PM distortion curves generate intermodulation products that spill into adjacent channels. Synthetic impairment models must accurately replicate these non-linear transfer functions to produce realistic spectral regrowth.

• AM-AM distortion: Amplitude compression at high input power
• AM-PM distortion: Unwanted phase shift proportional to input amplitude
• Memory effects: Thermal and electrical hysteresis that shape the distortion spectrum

A signal characteristic that directly drives ACLR performance. High-PAPR waveforms like OFDM force power amplifiers to operate with significant back-off to avoid clipping and spectral regrowth. Synthetic waveform generators must model the PAPR distribution of real signals to produce valid ACLR measurements.

• OFDM signals: Typically 8-13 dB PAPR, highly sensitive to PA non-linearity
• Constant-envelope modulations: GMSK, FSK exhibit low PAPR and inherently better ACLR
• Crest factor reduction: Real transmitters apply CFR algorithms that leave unique residual artifacts

DPD linearizes a PA by pre-inverting its distortion characteristics, but residual ACLR remains due to imperfect compensation. These residuals form a subtle, device-specific signature. Synthetic impairment pipelines must model DPD loop convergence errors to capture the fine-grained spectral structure that distinguishes one transmitter from another.

• Lookup-table DPD: Quantization of correction values creates step artifacts
• Memory polynomial DPD: Limited model order leaves uncorrected high-order terms
• Loop delay mismatch: Timing errors between feedback and forward paths degrade suppression

SFDR defines the usable dynamic range of a DAC before spurious tones contaminate the output spectrum. These spurs appear as discrete peaks in adjacent channels, directly degrading ACLR. Synthetic signal generation must model DAC quantization noise and clock feedthrough to produce realistic spurious content.

• Quantization spurs: Correlated errors from finite bit resolution
• Clock leakage: Mixer and DAC clock harmonics appearing in-band
• Image rejection: Incomplete suppression of mixer sidebands

A mathematical framework for modeling non-linear systems with memory, including the PA behaviors that produce ACLR. The Volterra series represents the output as a sum of multi-dimensional convolution integrals, capturing both instantaneous non-linearity and frequency-dependent memory effects.

• 1st-order kernel: Linear frequency response
• 3rd-order kernel: Dominant source of spectral regrowth and ACLR
• Pruned models: Simplified Volterra structures (memory polynomial, Wiener-Hammerstein) reduce computational complexity while preserving ACLR accuracy

Often used interchangeably with ACLR, though ACPR typically refers to the ratio measured at the PA output before the antenna, while ACLR may include antenna and filter effects. Both metrics quantify spectral containment. Synthetic impairment models must target specific ACPR/ACLR values to match real device specifications.

• 3GPP specifications: Define ACLR limits of -33 dBm or -36 dBm for base stations
• Measurement bandwidth: Typically 3.84 MHz for WCDMA, scaled for other RATs
• Integration method: RMS power measured over the specified adjacent channel bandwidth