Multi-band Digital Predistortion with Reduced Basis Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital predistortion methods for power amplifiers face challenges in handling multi-band signals, as existing single-band models result in exponential growth of cross-terms, leading to increased complexity and distortion, particularly aliasing of high-order harmonics, necessitating higher bandwidth operation.
Innovation Solution
A digital predistortion system that applies weighed basis functions to multi-band inputs, using a single-band model with omitted cross-terms, specifically excluding those outside a predetermined frequency window or with low energy levels, to reduce distortion without increasing bandwidth, thereby using a reduced set of parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single-band model is used for multi-band signals, then the model complexity is reduced, but distortion (aliasing of high-order harmonics) increases
Solution Approach 1:
The patent extracts and removes specific cross-terms from the single-band model that correspond to frequency components outside the frequency window. By selectively eliminating these problematic terms (particularly high-order harmonics that would alias), the model maintains simplicity while reducing distortion in the bands of interest.
Solution Approach 2:
The patent applies local quality by treating different frequency regions differently - cross-terms within the frequency window are retained while those outside are removed. This selective approach optimizes the model for the specific multi-band signal characteristics rather than applying a uniform treatment to all frequency components.
2Object-generated harmful factors
If cross-terms are included in the single-band model for multi-band signals, then distortion is reduced, but the number of parameters increases exponentially
Solution Approach 1:
The patent extracts only the necessary cross-terms that fall within the frequency window, removing the exponential growth of parameters by eliminating cross-terms corresponding to frequencies outside the window. This selective extraction maintains essential distortion compensation while avoiding parameter explosion.
Solution Approach 2:
The patent applies partial action by including only those cross-terms that are relevant to the signal bands of interest, rather than implementing the full set of cross-terms that would be required for complete distortion compensation. This partial approach achieves sufficient linearization without the computational burden of the complete model.
3Object-generated harmful factors
If the DPD operates at increased bandwidth to handle multi-band signals, then distortion is avoided, but the bandwidth requirement increases
Solution Approach 1:
The patent removes cross-terms that would generate frequency components outside the desired bandwidth by excluding those corresponding to frequencies beyond the frequency window. This extraction allows the DPD to operate at the original bandwidth while still compensating for relevant distortion.
Solution Approach 2:
The patent changes the model parameters by selectively including or excluding specific cross-terms based on frequency window constraints. This parameter modification allows the system to maintain operation within the original bandwidth while achieving adequate distortion compensation for the multi-band signal.
Data Source
AI summary
Disclosed are implementations, including a method for digital predistortion of multiband signals that includes receiving an input signal comprising multiple signal portions in different frequency bands, the input signal configured to be processed by a transmit chain, of a power amplification system, comprising at least a power amplifier that produces output with non-linear distortions, with the non-linear distortions of the transmit chain being represented using a set of basis functions derived according to a single-band model of the non-linear distortions. The method further includes performing digital predistortion on signal components derived from the multiple signal portions of the input signal using a reduced set of the basis functions that excludes at least some basis functions for at least some cross-terms resulting from a full expansion of the single-band model applied to the multiple signal portions, to produce a digital predistorted signal provided to the transmit chain.