Dual Learned Model S-Parameter Estimation for High-Frequency Circuits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating S-parameters in high-frequency circuits face challenges in accuracy, especially when the frequency changes, and struggle to adapt to changes in circuit configuration.
Innovation Solution
The use of a dual learned model approach, where a first learned model estimates S-parameters at 0 Hz based on first information, and a second learned model estimates S-parameters at non-zero frequencies based on both first and second information, improves accuracy by leveraging machine learning on specific training data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single learned model is used for S-parameter estimation across all frequencies, then the device complexity is reduced, but the measurement precision deteriorates especially at frequency transitions
Solution Approach 1:
The patent divides the frequency range into two segments: 0 Hz and non-zero frequencies. Two separate learned models are trained independently on training data corresponding to each segment. This segmentation allows each model to specialize in its frequency range, improving estimation accuracy without requiring a single complex universal model.
Solution Approach 2:
The patent applies different model architectures or training approaches tailored to specific frequency characteristics. The first learned model is optimized for DC conditions (0 Hz) while the second learned model is optimized for AC conditions (non-zero frequencies), allowing each model to have local quality optimized for its specific operating condition.
2Adaptability or versatility
If machine learning is performed on comprehensive training data covering all frequencies, then the adaptability improves, but the loss of time increases due to larger data processing requirements
Solution Approach 1:
The training data is segmented by frequency range, with separate datasets prepared for 0 Hz and non-zero frequencies. This reduces the computational burden on each individual training process while maintaining comprehensive frequency coverage through the combination of multiple specialized models.
Solution Approach 2:
The patent performs preliminary classification of operating conditions (0 Hz vs. non-zero frequency) before selecting which model to use. This preliminary action avoids the need for a single model to learn all frequency characteristics simultaneously, reducing overall training time while maintaining adaptability.
Data Source
AI summary
A non-transitory computer-readable recording medium having stored therein a program causes a computer to execute a process. The process includes acquiring first information related to a linear circuit having a plurality of ports for receiving or outputting a high frequency signal, and second information related to a frequency of the high frequency signal, estimating, when the second information indicates that the frequency is 0 Hz, an S-parameter for two ports of the plurality of ports when the frequency is 0 Hz, from the first information based on a first learned model, and estimating, when the second information indicates that the frequency is other than 0 Hz, an S-parameter at the frequency from the first information and the second information, based on a second learned model. The first and the second learned models are generated by performing machine learning on plural pieces of first training data and second training data, respectively.