ML Digital Pre-Distortion for Power Amplifier Feedback Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Power amplifiers in communication systems suffer from non-linearity issues that cause spectral regrowth, adjacent channel interference, and in-band distortion, leading to degraded performance and efficiency, particularly in systems with high peak-to-average power ratio and wide bandwidths, where traditional digital pre-distortion methods struggle to accurately determine and update distortion parameters due to limited feedback capabilities and dynamic changes.
Innovation Solution
A machine learning-based digital pre-distortion system that utilizes a phased training approach, employing internal and external feedback models to emulate high-fidelity signals, enabling accurate determination and adaptation of pre-distortion parameters in real-world conditions, even without external test equipment, by using neural networks to mimic external feedback receivers and account for antenna loading and aging effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional digital pre-distortion methods are used, then device complexity is reduced, but measurement precision and manufacturing precision of distortion parameters deteriorate due to limited feedback capabilities and dynamic changes
Solution Approach 1:
The patent introduces an intermediary feedback model that mediates between the limited internal feedback receiver and the power amplifier. This feedback model processes the limited feedback signal to generate accurate distortion parameters, effectively acting as a bridge that compensates for the insufficient feedback capabilities while maintaining system simplicity.
Solution Approach 2:
The patent creates a virtual copy of the external feedback receiver through the feedback model. This copied functionality allows the system to determine distortion parameters with high precision as if an external receiver were present, without actually requiring the complex external hardware.
2Device complexity
If internal feedback receiver with limited bandwidth is used, then device complexity is reduced, but loss of information increases due to insufficient feedback signal quality
Solution Approach 1:
The patent implements an enhanced feedback mechanism where the limited feedback signal from the internal receiver is processed through a trained feedback model. This model reconstructs the full-bandwidth feedback information from the limited input, effectively recovering lost information while maintaining the simplicity of the internal receiver architecture.
Solution Approach 2:
The feedback model transforms the limited bandwidth feedback signal into comprehensive distortion parameters by changing the parameter representation. Through machine learning training, the model learns to extract and reconstruct full information content from reduced-dimensional input signals.
3Manufacturing precision
If machine learning based pre-distortion is implemented, then manufacturing precision of distortion parameters is improved, but device complexity increases due to additional processing requirements
Solution Approach 1:
The patent applies preliminary action by training the machine learning models offline before deployment. The feedback model and pre-distortion model are pre-trained using comprehensive datasets, so that during actual operation, they can quickly process signals with high precision without requiring complex real-time computations, thus reducing operational complexity.
Solution Approach 2:
The patent replaces complex mechanical signal processing systems with machine learning-based virtual processing. Instead of using additional physical hardware components to achieve high precision, the system uses trained neural networks that can be implemented in software, substituting physical complexity with computational intelligence.
4Ease of operation
If conventional pre-distortion calibration is performed, then ease of operation is maintained, but adaptability deteriorates due to inability to account for antenna loading and aging effects
Solution Approach 1:
The patent introduces dynamics by implementing adaptive models that can adjust to changing conditions. The feedback model and pre-distortion model are designed to be re-trained or fine-tuned based on observed performance degradation from aging or loading effects, allowing the system to adapt dynamically while maintaining ease of operation through automated processes.
Solution Approach 2:
The system implements self-service through automated model training and adaptation mechanisms. The feedback model automatically learns from the limited feedback signals and adjusts the pre-distortion parameters without requiring manual recalibration, enabling the system to adapt to aging and loading effects autonomously while keeping the operation simple for users.
Data Source
Figure 1~2
Figure 3~4
Figure 5~7
AI summary
Example embodiments relate to machine learning based digital pre-distortion for power amplifiers. A device may amplify a signal with a power amplifier and transmit the signal. The signal may be received by an internal feedback receiver of the device. The device may further comprise a first machine learning model configured to emulate an external feedback receiver and to generate an emulated feedback signal based on the internal feedback signal. The device may further comprise a second machine learning model configured to determine digital pre-distortion parameters for the power amplifier based on the emulated feedback signal. Apparatuses, methods, and computer programs are disclosed.