Adjacent Channel Leakage Ratio (ACLR)

ACLR is defined as the ratio of the transmitted power within a specified assigned channel bandwidth to the power measured in an adjacent channel at a specified frequency offset. It is typically expressed in dBc (decibels relative to the carrier).

• Measurement Setup: Requires a spectrum analyzer with a root-raised-cosine (RRC) filter matched to the communication standard.
• Standard Offsets: 3GPP specifies ACLR at ±5 MHz and ±10 MHz offsets for a 5 MHz LTE carrier.
• Calculation: ACLR = 10 * log10(P_adjacent / P_carrier). A more negative value indicates better performance.

ACLR degradation is a direct consequence of spectral regrowth caused by intermodulation distortion (IMD) in the power amplifier. When a modulated signal passes through a nonlinear PA, the amplitude-to-amplitude (AM-AM) and amplitude-to-phase (AM-PM) distortions generate out-of-band spectral components.

• Third-order IMD products are the primary contributors to first-adjacent channel leakage.
• Fifth-order IMD products dominate the second-adjacent channel.
• A perfectly linear amplifier would have infinite ACLR; real amplifiers require power back-off or digital predistortion to meet specifications.

ACLR is a mandatory compliance metric enforced by regulatory bodies to prevent adjacent channel interference between different network operators. Failure to meet ACLR limits results in certification denial.

• 3GPP TS 36.104: Defines ACLR limits for LTE base stations (-45 dBc for adjacent channel).
• 3GPP TS 38.104: Specifies stricter ACLR requirements for 5G NR, especially for wideband carriers and carrier aggregation scenarios.
• FCC Part 24/27: U.S. regulations governing out-of-band emission limits for licensed spectrum.
• ETSI EN 301 908: European harmonized standard with equivalent ACLR requirements.

Digital predistortion optimization involves a fundamental trade-off between ACLR (out-of-band performance) and Error Vector Magnitude (EVM) (in-band performance). Aggressive linearization to improve ACLR can inadvertently degrade EVM by introducing in-band distortion.

• Over-linearization: Applying excessive predistortion correction can clip the signal peaks, raising the noise floor and worsening EVM.
• Joint Optimization: Advanced DPD algorithms use multi-objective cost functions that balance ACLR improvement against EVM degradation.
• System Margin: A well-designed DPD system achieves ACLR of -55 to -60 dBc while maintaining EVM below 1-2% for 256-QAM modulation.

As signal bandwidths expand to 100 MHz and beyond for 5G NR, maintaining ACLR becomes exponentially more difficult. The DPD system must linearize the PA over a much wider frequency range where memory effects and frequency-dependent gain variations are pronounced.

• Bandwidth Expansion Factor: The predistorted signal bandwidth is typically 3-5x the original signal bandwidth to capture and cancel IMD products.
• Observation Path Bandwidth: The feedback receiver must sample at rates sufficient to capture fifth-order distortion products without aliasing.
• Gain Flatness: PA gain variations across the wideband carrier degrade ACLR uniformity, requiring frequency-selective predistortion techniques.

Crest Factor Reduction (CFR) is a complementary technique to DPD that directly improves ACLR by reducing the peak-to-average power ratio (PAPR) of the transmitted signal. By clipping and filtering signal peaks before the PA, CFR reduces the instantaneous power excursions that drive the amplifier into deep compression.

• Peak Windowing: A CFR method that applies a smooth windowing function to clipped peaks to control spectral regrowth.
• Cascaded CFR+DPD: Industry-standard architecture where CFR reduces PAPR first, then DPD linearizes the remaining nonlinearity.
• Combined ACLR Gain: CFR alone can improve ACLR by 3-5 dB; combined with DPD, total ACLR improvement of 15-20 dB is achievable.

The physical phenomenon directly measured by ACLR. When a power amplifier operates near saturation, its nonlinear transfer function generates intermodulation products that spread signal energy into adjacent channels. This regrowth is the primary source of adjacent channel interference and must be suppressed to meet regulatory masks. The severity of regrowth depends on the peak-to-average power ratio (PAPR) of the input signal and the amplifier's AM-AM/AM-PM characteristics.

The mathematical root cause of ACLR degradation. When multiple frequency components pass through a nonlinear device, they generate spurious products at sum and difference frequencies. Third-order intermodulation products (IMD3) fall closest to the original carriers and are the dominant contributors to adjacent channel leakage. IMD is characterized by the third-order intercept point (IP3), a figure of merit for amplifier linearity.

The in-band counterpart to ACLR. While ACLR quantifies out-of-band spectral leakage, EVM measures in-band signal quality degradation caused by the same nonlinearity. A predistorter must balance both metrics: aggressive linearization improves ACLR but may introduce in-band distortion if overdriven. EVM is expressed as a percentage of the ideal constellation point deviation and is critical for modulation schemes like 256-QAM and 1024-QAM.

A complementary technique to digital predistortion for improving ACLR. CFR reduces the peak-to-average power ratio of the transmitted signal before it reaches the power amplifier, allowing the PA to operate with less back-off and reduced nonlinearity. Common algorithms include peak windowing and pulse cancellation. CFR and DPD work in tandem: CFR handles extreme peaks, while DPD corrects the residual nonlinearity across the signal envelope.

The regulatory limits that define acceptable ACLR performance. Standards bodies like 3GPP and FCC specify spectral emission masks that dictate the maximum permissible power in adjacent and alternate channels. These masks vary by frequency band, channel bandwidth, and operating mode. Compliance requires ACLR typically exceeding 45 dB for the adjacent channel and 60 dB for the alternate channel in 5G NR base stations.

The workhorse behavioral model for capturing the nonlinear dynamics that degrade ACLR. A memory polynomial extends the classical polynomial model by including delayed envelope terms to account for memory effects caused by bias networks, thermal dynamics, and trapping phenomena. The model coefficients are extracted from input-output measurements and directly form the basis of many indirect learning architecture predistorters.