On-Chip Spectral Analysis Using Recursive Discrete Fourier Transforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional spectral-domain analyses for mixed-signal designs require increasingly large sample sizes for greater frequency resolution, leading to increased power consumption, data rates, and manufacturing costs when performed on-chip using built-in self-testing techniques.
Innovation Solution
A digital or signal processor on a substrate performs a first-level and subsequent recursive discrete Fourier transforms on windowed test data to identify frequency bins with the greatest signal power, allowing for on-chip spectral characterization with reduced calculations and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional FFT/DFT algorithms are used for spectral analysis on-chip, then frequency resolution can be improved by increasing sample size, but power consumption and data rates increase
Solution Approach 1:
The patent segments the spectral analysis process into multiple stages: initial coarse spectral analysis to identify dominant frequency components, followed by targeted fine analysis only on those specific frequency bins. This segmentation allows achieving high frequency resolution for critical frequencies without performing full-spectrum high-resolution analysis, thereby reducing overall power consumption and computational burden.
Solution Approach 2:
The patent applies local quality by concentrating computational resources and power on specific frequency bins that contain dominant signal components, rather than uniformly processing the entire spectrum. The system identifies frequency bins with significant power and performs enhanced analysis only on those local regions, optimizing the trade-off between measurement precision and power consumption.
2Measurement precision
If larger number of samples are used for spectral analysis, then frequency resolution is improved, but circuit layout area and manufacturing costs increase
Solution Approach 1:
The patent divides the spectral analysis into coarse and fine stages, where the fine high-resolution analysis is performed only on selected frequency bins rather than the entire spectrum. This segmentation reduces the number of samples and computational operations required, thereby reducing the circuit layout area needed for implementing the spectral analysis function on-chip.
Solution Approach 2:
The patent applies partial action by performing high-resolution spectral analysis only on the necessary subset of frequency bins that contain dominant signal components, rather than analyzing the entire frequency spectrum at high resolution. This partial processing approach achieves the required measurement precision with reduced computational resources and smaller circuit area.
3Measurement precision
If full spectral analysis is performed on-chip, then measurement capability is improved, but I/O pin count and manufacturing costs increase
Solution Approach 1:
The patent extracts and processes only the critical frequency components that contain dominant signal power, rather than performing complete spectral analysis on all frequency bins. By extracting and focusing on the essential frequency information, the system achieves effective spectral characterization with reduced data output requirements, thereby reducing I/O pin count and device complexity.
Solution Approach 2:
The patent performs partial spectral analysis by concentrating computational and measurement resources on the essential frequency bins that dominate the signal spectrum. This partial analysis approach provides sufficient spectral characterization capability for mixed-signal design validation while significantly reducing the data volume and I/O requirements compared to full-spectrum analysis.
Data Source
AI summary
A computing device uses a recursive discrete Fourier transform (RDFT) engine to reduce time required by a frequency transform module, memory required to hold intermediate products, and/or computing resources used for the testing. In an embodiment the windowing function is integrated and processed simultaneously with the recursive DFT funcions. A frequency-bin power module is configured to determine the frequency bin within the set of frequency bins that has a greatest signal power at various levels of recursion.