LUT Gain Compression

In the compression region, the LUT stores complex gain values greater than unity to pre-compensate for the PA's gain reduction. As the input envelope approaches P1dB and saturation, the predistorter applies progressively larger gain expansion to maintain a linear AM-AM transfer characteristic.

• Expansion ratios typically reach 2-5 dB for Class AB PAs
• Complex-gain LUTs simultaneously correct AM-PM conversion that peaks near compression
• Coefficient precision requirements increase in this region due to steep nonlinearity gradients

The compression region demands higher LUT granularity than the linear region because the PA's gain curve changes rapidly near saturation. Non-uniform spacing allocates more entries where the derivative of gain vs. input power is largest.

• Typical allocation: 60-70% of LUT entries dedicated to the top 20% of the dynamic range
• Companding functions (μ-law, A-law) compress the index mapping to concentrate resolution
• Uniform spacing in the compression region causes excessive interpolation error and spectral regrowth

Gain compression characteristics drift with junction temperature, creating a moving target for LUT adaptation. The compression knee shifts to lower power levels as temperature rises, requiring the LUT to track both short-term thermal memory (envelope-dependent heating) and long-term thermal memory (ambient changes).

• GaN PAs exhibit 0.01-0.03 dB/°C gain variation in compression
• Multi-dimensional LUTs index on both instantaneous power and averaged power history
• Adaptation rates in the compression region must be faster than thermal time constants

Small errors in compression-region LUT coefficients produce disproportionately large ACLR degradation because the PA operates at peak nonlinearity. A 1% coefficient error near saturation can cause more spectral regrowth than a 5% error in the linear region.

• Quantization noise in this region directly maps to adjacent channel leakage
• LMS adaptation step sizes are often reduced by 50% in compression entries to prevent oscillation
• Smoothing filters applied across adjacent compression entries prevent discontinuous gain transitions

The LUT compression region defines a hard ceiling on predistortion gain expansion to prevent driving the PA into hard saturation where linearization becomes impossible. Beyond this boundary, the PA's gain collapses irreversibly, and no predistortion can recover linearity.

• Maximum expansion is typically limited to 6-8 dB above small-signal gain
• Crest factor reduction (CFR) works in tandem to keep signal peaks within the correctable range
• The compression region upper bound corresponds to Psat minus 0.5-1 dB for most PAs

Linear interpolation between compression-region entries introduces systematic underestimation of the required gain expansion due to the concave-down shape of the inverse PA characteristic. Higher-order interpolation is often necessary.

• Quadratic or cubic interpolation reduces residual distortion by 3-6 dB vs. linear methods
• Interpolation error manifests as spectral regrowth shoulders at specific offset frequencies
• Hardware-efficient piecewise-parabolic interpolation balances accuracy against FPGA resource utilization

The amplitude-to-amplitude correction component stored in the LUT that directly addresses gain compression. As the PA enters saturation, the AM-AM curve flattens, requiring the LUT to apply expanding gain values. This correction is most aggressive in the compression region, where small input power increases yield diminishing output power gains. The AM-AM LUT entries in this zone typically exhibit gain values significantly above unity to linearize the saturated amplifier response.

A look-up table architecture that allocates higher entry density in regions of rapid gain compression. Since the PA's gain characteristic changes most dramatically near saturation, uniform LUT spacing wastes resolution in linear regions while undersampling the critical compression knee. Non-uniform LUTs concentrate entries where the derivative of gain is largest, maximizing correction accuracy per stored coefficient and reducing overall memory requirements.

The amplitude-to-phase correction that becomes severe in the gain compression region. As the PA approaches saturation, the input capacitance variation causes significant phase shift that must be counter-rotated by the LUT. This phase distortion often increases nonlinearly with drive level, requiring the LUT to store complex-valued coefficients that simultaneously correct amplitude compression and phase rotation at each index point.

Gain compression is the primary cause of adjacent channel leakage. When the PA saturates, the flattened gain curve generates intermodulation products that spill into neighboring frequency bands. The LUT's compression region must apply precise inverse nonlinearity to suppress this regrowth. Even small errors in LUT gain expansion values near saturation can cause significant ACLR degradation, making this region the most critical for maintaining regulatory compliance.

A post-processing technique applied across adjacent LUT entries to prevent discontinuous gain transitions at the compression boundary. Abrupt changes in LUT gain values between the linear and compression regions can generate high-frequency spectral artifacts. Smoothing algorithms apply low-pass filtering across LUT entries, ensuring the gain expansion curve is continuous and differentiable, which preserves spectral purity during the transition into saturation correction.

The PA's gain compression characteristic drifts with junction temperature, causing the saturation point to shift. As temperature increases, the compression knee typically moves to lower power levels, requiring the LUT to adapt its gain expansion region accordingly. Temperature-compensated LUTs store multiple compression profiles or apply real-time coefficient scaling based on thermal sensors to maintain linearization accuracy across operating conditions.