Digital Pre-Distortion Lookup Table Initialization Using Input Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional digital pre-distortion systems face challenges such as high complexity, hardware, and cost due to training processes, and poor linearization performance when initial lookup table values are inaccurate, especially in environments with unknown or changing distortion characteristics.
Innovation Solution
The method involves using actual input data samples to initialize and update the lookup table, avoiding training sequences and relying on blind settings, which allows for better adaptation to distortion characteristics and improved convergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a training process is performed to determine inverse gain values in the lookup table, then the linearization performance is improved, but the system complexity, hardware requirements, and cost increase considerably
Solution Approach 1:
The system uses actual input data samples to self-initialize and self-update the lookup table entries, eliminating the need for external training sequences or dedicated training processes. The DPD system adapts autonomously using its own operational data, thereby improving linearization performance without adding system complexity or hardware requirements.
2Measurement precision
If a training process is performed to determine inverse gain values, then the accuracy of distortion compensation is improved, but transmission delays increase due to the additional training phase
Solution Approach 1:
The lookup table entries are initialized and updated using actual input data samples before they are needed for distortion compensation. By continuously preparing and updating the lookup table in advance using incoming data samples, the system ensures accurate distortion compensation is ready when needed, eliminating transmission delays associated with training processes.
3Device complexity
If blind settings are used for initial lookup table values, then the system complexity is reduced, but the linearization performance deteriorates when distortion characteristics are unknown or changing
Solution Approach 1:
The system transitions from static blind settings to dynamic adaptation by continuously updating lookup table entries using actual input data samples. This dynamic approach allows the DPD system to adapt to changing distortion characteristics in real-time, maintaining high linearization performance without increasing system complexity, as the update mechanism uses the existing data flow.
4Reliability
If actual input data samples are used to initialize and update lookup table entries, then the linearization performance is improved without training processes, but the system must process and store additional data temporarily
Solution Approach 1:
The system updates only the specific lookup table entries that are actually accessed and needed, rather than storing or processing all possible data. By focusing updates on locally relevant entries based on actual input data samples, the system improves linearization performance while minimizing additional data storage requirements to only what is immediately necessary.
Data Source
AI summary
A system (100, FIG. 1) performs digital pre-distortion using gain values stored in a lookup table (150). A method for performing digital pre-distortion includes identifying (310, FIG. 3) a lookup table entry, based on input data, and updating the lookup table entry by writing an updated gain value into the lookup table entry. In an embodiment, update tracking information corresponding to the lookup table entry may be updated (324) to indicate that the lookup table entry has been updated. Another embodiment includes identifying (412, FIG. 4) consecutive lookup table entries based on input data, determining (413) whether the consecutive lookup table entries have been previously updated, and performing (414) a weighted interpolation process to produce an output gain value. A previous gain value (158, FIG. 1) is used in the weighted interpolation process when at least one of the consecutive lookup table entries has not been updated.