DPD Feedback Sample Combining for Lower Training Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital predistortion systems face high computational complexity and sample buffering needs due to the trade-off between amplification linearity and power efficiency, particularly in wireless communication systems, which is exacerbated by high sampling rates and correlated feedback samples.
Innovation Solution
A system and method that employs under-sampling and sample combining techniques to reduce the number of feedback samples by integrating consecutive samples based on predetermined parameters, using an integrate and dump filter to compress feedback signals, thereby reducing model coefficient estimation complexity and buffering requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high sampling rates are used in digital predistortion feedback path, then measurement precision is improved, but device complexity and energy consumption increase
Solution Approach 1:
Multiple consecutive under-sampled feedback samples are combined into a single compressed sample using integration. This merging process reduces the number of samples from N to 1, thereby reducing computational complexity and device resources while preserving the essential statistical information needed for model coefficient estimation.
Solution Approach 2:
The feedback signal processing is segmented into two stages: first under-sampling to reduce rate, then combining/integrating consecutive samples to further compress. This segmentation allows the system to achieve low complexity while maintaining measurement precision through the integrated representation of multiple samples.
2Measurement precision
If high sampling rates are used in digital predistortion feedback path, then measurement precision is improved, but energy consumption increases
Solution Approach 1:
Multiple consecutive under-sampled feedback samples are merged into a single compressed sample through integration. This reduces the total number of samples processed, directly lowering the energy consumption of the processing circuitry while maintaining measurement precision through the integrated statistical representation.
3Measurement precision
If full feedback samples are processed, then model coefficient estimation accuracy is improved, but sample buffering needs increase
Solution Approach 1:
Consecutive under-sampled feedback samples are combined into compressed samples, reducing the quantity of samples that need to be buffered. The integration process consolidates multiple samples into one, decreasing buffering requirements while preserving the statistical information necessary for accurate model coefficient estimation.
4Measurement precision
If full feedback samples are processed, then model coefficient estimation accuracy is improved, but processing time increases
Solution Approach 1:
Multiple consecutive feedback samples are merged into compressed samples through integration, reducing the total number of samples that need to be processed for model coefficient estimation. This significantly decreases processing time while maintaining estimation accuracy by preserving the statistical characteristics of the original signal through the integrated representation.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Example embodiments provide a compression technique of feedback samples for digital predistortion. A system (100) may comprise a feedback receiver (116) configured to receive feedback signal of a power amplifier (112) output and determine a set of under-sampled samples based on the feedback signal; a compressing circuitry (130) configured to: obtain the under-sampled samples; and compress the under-sampled samples, wherein two or more consecutive under-sampled samples are combined into one or more single samples based on one or more predetermined parameters; and a model coefficient training circuitry (102) configured to receive the compressed under-sampled samples and determine model coefficients for digital predistortion based on the compressed under-sampled samples. A system and a method are disclosed.