Vector FIR Filtering for Complex Data With Real Coefficients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital signal processing systems face challenges in efficiently handling symmetric FIR filters with real coefficients and complex data, particularly in terms of computational efficiency and resource utilization, especially when supporting multiple filter types and arbitrary size inputs.
Innovation Solution
The proposed system and method utilize vector addition and multiplication to process multiple components of the filtered signal simultaneously, leveraging symmetry properties to improve computational efficiency while handling different FIR filter types and sizes with minimal additional hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional scalar processing is used for FIR filters, then implementation is simple, but computational efficiency and data throughput are low
Solution Approach 1:
The patent transitions from scalar processing to vector processing by organizing filter coefficients and input data into vectors. Multiple scalar operations are combined into vector operations, processing multiple data elements simultaneously. This dimensional change from 0-dimensional scalars to 1-dimensional vectors enables parallel processing without proportionally increasing hardware complexity.
Solution Approach 2:
The patent merges multiple scalar multiplications and additions into single vector operations. By combining several scalar operations that would normally be executed sequentially into unified vector operations, the system achieves higher throughput while using the same computational resources more efficiently.
2Productivity
If vector processing is used for FIR filters, then computational efficiency and data throughput are improved, but hardware complexity increases
Solution Approach 1:
The patent creates a universal vector processing framework that can handle multiple FIR filter types (symmetric, asymmetric, different lengths) using the same hardware infrastructure. The vector operations are designed to be type-agnostic, allowing a single hardware implementation to serve multiple filtering scenarios without requiring specialized circuits for each filter type.
Solution Approach 2:
The patent changes the parameters of processing by utilizing vector registers with configurable lengths and operations with variable parameters. Instead of fixed scalar operations, the system uses vector operations where the number of elements processed can be adjusted, enabling efficient handling of different filter sizes and types without hardware reconfiguration.
3Productivity
If symmetric FIR filter optimization is applied, then computational efficiency improves for symmetric filters, but flexibility to handle different filter types decreases
Solution Approach 1:
The patent introduces dynamic adaptability where the processing system can adjust its behavior based on the filter type being applied. The vector processing framework includes conditional logic that detects whether a filter is symmetric or asymmetric and automatically applies the appropriate optimization strategy, maintaining high efficiency across different filter types without sacrificing flexibility.
Solution Approach 2:
The patent handles both symmetric and asymmetric filter cases within a unified framework. For symmetric filters, the system exploits the symmetry property by processing only half the coefficients and mirroring results. For asymmetric filters, the full vector processing capability is utilized. This asymmetric handling approach allows optimization where applicable while maintaining general functionality.
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Systems and methods for applying a Finite Impulse Response (FIR) filter. First and second sub-vectors, and succeeding and preceding third and fourth sub-vectors, are iteratively extracted from received signal elements. First sub-vector is shifted left by at least one element and elements at right-most locations thereof replaced by elements of third sub-vector, second sub-vector is shifted right by at least one element and elements at left-most locations thereof replaced by elements of fourth sub-vector, thereby obtaining first and second dual left-shift vectors and first and second dual right-shift vectors. First dual left-shift vector and first dual right-shift vector are summed and multiplied by a multiplier from the FIR filter, second dual left-shift vector and second dual right-shift vector are summed and multiplied by another multiplier from the FIR filter, which outcomes are added to an accumulated vector comprising portions of the filtered data.