Neural Volterra DPD with Feature Networks for PA Linearization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital predistortion (DPD) technologies face challenges in efficiently compensating for non-linearities in transceiver components, particularly in high-performance wireless systems, due to the complexity and impracticality of Volterra-based and artificial neural network models.
Innovation Solution
A hybrid approach combining Volterra-based models with artificial neural networks, utilizing complex feature and envelope neural networks to transform and predistort input signals, reducing non-linearity in power amplifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Volterra-based models are used for digital predistortion, then compensation accuracy for non-linearities is improved, but device complexity and implementation practicality deteriorate
Solution Approach 1:
The patent segments the complex Volterra-based predistortion model into multiple simpler sub-models that process different signal components separately. Each sub-model handles a specific aspect of non-linearity compensation, making the overall system more manageable and implementable while maintaining high accuracy through coordinated operation of the segmented components.
Solution Approach 2:
The patent transforms the mathematical parameters of the Volterra model into optimized coefficients that can be efficiently computed in hardware. By changing the representation and computation method of model parameters, the system achieves high compensation accuracy with reduced computational complexity and improved practical implementability.
2Measurement precision
If artificial neural network models are used for digital predistortion, then compensation accuracy is improved, but device complexity and power consumption worsen
Solution Approach 1:
The patent extracts the essential non-linear compensation functionality from complex neural network models and implements it through simplified mathematical relationships. By taking out only the critical compensation functions and removing unnecessary neural network complexity, the system achieves accurate compensation with reduced device complexity and lower power consumption.
Solution Approach 2:
The patent replaces expensive, computationally intensive neural network computations with simpler, more efficient mathematical models that consume less power and require fewer hardware resources. This substitution maintains compensation accuracy while significantly reducing implementation complexity and power consumption.
3Reliability
If higher performance DPD is implemented, then non-linearity compensation is improved, but power consumption increases
Solution Approach 1:
The patent applies partial predistortion processing to the most critical signal components while using simpler processing for less critical components. By applying excessive precision only where necessary and adequate precision elsewhere, the system achieves effective non-linearity compensation with optimized power consumption across different signal processing stages.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
Aspects of this disclosure relate to digital compensators, such as digital predistortion systems. Digital predistortion systems disclosed herein use a neural Volterra approach. Such digital predistortion systems can include a feature processing path comprising a feature artificial neural network, an envelope processing path, multipliers configured to multiply respective output signals of the feature processing path and the envelope processing path, and a combiner configured to generate a combined output signal based on at least output signals of the multipliers. The combined output signal is a digitally predistorted version of an input signal.