Canonical Multi-Band DPD with Reduced-Dimension LUTs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-band digital pre-distortion (DPD) systems face challenges in managing complexity and memory requirements due to the exponential growth of multi-dimensional look-up tables (LUTs) and polynomial terms with the number of frequency bands.
Innovation Solution
The implementation of multi-band DPD using a combination of reduced-dimension look-up tables (LUTs) and computation of reduced-dimension polynomial terms, which reduces the dimensionality of the LUTs and polynomial terms, thereby minimizing memory size and computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multi-dimensional LUTs and polynomial models are used for multi-band DPD, then linearization performance is improved, but memory size and computational complexity grow exponentially
Solution Approach 1:
The patent segments the multi-dimensional LUT into multiple one-dimensional LUTs, where each LUT corresponds to a specific frequency band. Instead of using a single multi-dimensional LUT that requires exponential memory, the system uses multiple 1D LUTs that can be computed independently, reducing the memory requirement from O(N^D) to O(D*N) where D is the number of bands and N is the LUT size per band.
Solution Approach 2:
The patent combines the outputs of multiple one-dimensional LUTs through polynomial terms to achieve the same linearization function as a multi-dimensional LUT. The polynomial terms merge the contributions from different frequency bands, allowing the system to achieve multi-band DPD performance without requiring a multi-dimensional data structure.
2Measurement precision
If multi-dimensional LUTs are used for multi-band DPD, then linearization accuracy is improved, but memory size increases exponentially
Solution Approach 1:
The patent divides the multi-dimensional LUT into multiple one-dimensional LUTs, each storing data for a single frequency band. This segmentation reduces the memory size from exponential O(N^D) to linear O(D*N), making the system practical for implementation while maintaining the ability to capture non-linear effects across multiple bands through the combination of these segmented LUTs via polynomial terms.
Data Source
AI summary
A system and method for multi-band digital pre-distortion (DPD) for a non-linear system. The system includes a DPD circuitry configured to perform multi-band DPD on a multi-band input signal to compensate for a non-linearity of a non-linear system. The multi-band input signal includes input signals of multiple frequency bands and the DPD circuitry is configured to perform DPD on an input signal of each frequency band per frequency band. The DPD circuitry is configured to perform the DPD using a combination of a look-up table (LUT) that evaluates a non-linear function and computation of terms of a non-linear polynomial of one or more variables representing the input signals of multiple frequency bands. Both the non-linear function and the non-linear polynomial are in a reduced dimension lower than a dimension of the multi-band input signal.