Look-Up Table (LUT)

The core mechanism maps the instantaneous signal envelope (magnitude or power) directly to a memory address. This non-parametric approach avoids complex polynomial calculations during transmission. The input signal's amplitude is quantized, and the corresponding address is used to fetch a pre-computed complex gain correction factor. This direct mapping ensures minimal latency, making it ideal for real-time, high-bandwidth applications where computational overhead must be strictly bounded.

Each LUT entry stores a complex-valued coefficient that simultaneously corrects for both amplitude distortion (AM-AM) and phase distortion (AM-PM). By multiplying the baseband signal by this complex gain, the predistorter expands the signal in the opposite direction of the power amplifier's compression curve. This single-step, complex multiplication is highly efficient in hardware, correcting the nonlinear rotation and magnitude compression introduced by the PA in one unified operation.

The finite number of table entries introduces quantization error. The spacing between entries, or LUT Granularity, directly impacts linearization performance. To mitigate this without exploding memory size, LUT Interpolation (linear or polynomial) is used to smooth the transition between discrete points. This creates a continuous correction function, suppressing spectral regrowth that would otherwise arise from abrupt coefficient jumps at the boundaries of quantized bins.

Static LUTs fail to track changes due to temperature, aging, or frequency shifts. Adaptive LUTs employ closed-loop algorithms like the LMS LUT Update to iteratively refine coefficients. The adaptation engine minimizes the error between the desired linear output and the actual PA output. The LUT Adaptation Rate is a critical design parameter, balancing the need to quickly track dynamic changes against the introduction of steady-state noise from overly aggressive updates.

Modern wideband signals induce memory effects in power amplifiers, where the current output depends on past inputs. A simple 1-D LUT is insufficient. LUT Memory Depth is introduced by using a multi-dimensional table indexed by the current envelope and one or more delayed envelope samples. This extends the LUT into a memory polynomial structure, allowing it to pre-correct for the frequency-dependent nonlinear behavior characteristic of high-power GaN and LDMOS amplifiers.

LUTs are inherently suited for FPGA and ASIC implementation. Techniques like LUT Compression and LUT Partitioning reduce the memory footprint for massive MIMO arrays. The Ping-Pong LUT architecture uses dual memory banks: one actively predistorting the transmit signal while the other is updated in the background. This ensures seamless, glitch-free coefficient switching, a critical requirement for live traffic in 5G base stations.

The process of mapping an input signal's instantaneous power or magnitude to a specific memory address within the predistortion look-up table. The input envelope is quantized into discrete bins, and the quantized value serves as the address pointer.

• Address calculation: address = floor(|x(n)| / step_size)
• Overflow protection: Inputs exceeding the maximum index must saturate to prevent memory access violations
• Companding: Non-linear quantization of the index space allocates more entries to the compression region where the PA characteristic changes rapidly

A mathematical technique for estimating predistortion values between discrete table entries to reduce quantization error and improve linearization accuracy. Without interpolation, the staircase approximation introduces spectral regrowth.

• Linear interpolation: Computes a weighted average between two adjacent entries based on the fractional address
• Quadratic interpolation: Uses three neighboring points for smoother correction, at higher computational cost
• Trade-off: Interpolation order directly impacts ACLR improvement versus hardware multiplier usage

A predistortion table architecture that stores a single complex-valued coefficient per entry to simultaneously correct both amplitude (AM-AM) and phase (AM-PM) distortion. Each entry contains an in-phase (I) and quadrature (Q) component.

• Single-table efficiency: One memory lookup provides the complete correction factor
• Multiplication: The input sample is multiplied by the complex gain to produce the predistorted output
• Alternative: Separate I and Q LUTs can be used when independent correction paths are required

The speed at which look-up table coefficients are updated, controlling the trade-off between tracking agility and steady-state noise in the linearization loop. The adaptation rate is typically governed by the step-size parameter μ in LMS-based update algorithms.

• Fast adaptation: Tracks rapid thermal transients and envelope-dependent bias modulation, but introduces coefficient jitter
• Slow adaptation: Provides stable, low-noise coefficients but cannot follow dynamic PA behavior changes
• Variable rate: Adaptation gain can be scheduled based on signal statistics or error magnitude

A dual-buffer memory architecture where one look-up table is actively used for predistortion while the other is being updated in the background. This ensures seamless, glitch-free switching between coefficient sets.

• Active table: Drives the predistorter datapath with stable coefficients
• Shadow table: Receives coefficient updates from the adaptation engine
• Atomic swap: A control signal toggles which buffer is active, synchronized to frame boundaries to prevent transient distortion

The distortion introduced by representing continuous predistortion functions with a finite number of discrete amplitude levels within the look-up table. Both the input indexing and the stored coefficient word length contribute.

• Address quantization: Coarse indexing creates step-wise approximation of the correction curve
• Coefficient quantization: Limited bit depth for stored I/Q values introduces granular noise
• Mitigation: Increasing LUT depth and coefficient width reduces error at the cost of memory and power