Multi-Rate FIR Filtering Without High-Pass Filter Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital signal filtering methods require high processing requirements to achieve high resolution for low frequencies, especially in audio systems, and multi-rate filters introduce computational intensity or signal distortion.
Innovation Solution
A method that approximates a target Finite Impulse Response (FIR) filter by eliminating the need for a separate high pass filter, using parallel full-rate and reduced-rate paths with linear phase FIR filters, and implicit low pass filtering through down-sampling, which reduces computational requirements and maintains constant group delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single FIR filter is used to filter digital signals at high sample rate, then high frequency signal components are preserved, but processing requirements become excessively high and filter length must be very long to achieve high resolution for low frequencies
Solution Approach 1:
The patent divides the filtering task into two separate parallel paths: a full-rate path that processes high frequency components at the original high sample rate, and a reduced-rate path that processes low frequency components at a lower sample rate after down-sampling. This segmentation allows each path to use appropriately optimized filters, reducing overall computational requirements while maintaining high resolution for both frequency ranges
Solution Approach 2:
The patent changes the sampling rate parameter for different frequency bands by implementing down-sampling for the reduced-rate path. This parameter change allows the reduced-rate path to process low frequency components with fewer computations per second, while the full-rate path maintains the original high sampling rate for high frequency components that require it
2Productivity
If traditional multi-rate filters with separate high pass and low pass filters are used, then processing requirements are reduced, but computational intensity increases due to the need for both high pass and low pass filtering
Solution Approach 1:
The patent extracts and eliminates the high pass filter from the traditional multi-rate filter architecture. Instead of implementing both high pass and low pass filters separately, the invention uses the full-rate path to handle high frequency components directly and only implements a low pass filter in the reduced-rate path, thereby reducing computational intensity while maintaining processing efficiency
Solution Approach 2:
The patent merges the high frequency processing function into the full-rate path by applying the filtered reduced-rate signal directly to the high frequency band without requiring a separate high pass filter. This merging eliminates redundant filtering operations and reduces overall computational complexity
3Productivity
If IIR high pass and low pass filters are used in multi-rate filters, then processing requirements are reduced, but signal distortion occurs due to differing group delay above and below transition frequency
Solution Approach 1:
The patent changes the filter type parameter from IIR to FIR for the low pass filter in the reduced-rate path. This parameter change ensures constant group delay characteristics are maintained, preventing signal distortion while still achieving reduced processing requirements through the lower sampling rate
4Reliability
If linear phase FIR filters are used for high and low pass filtering in multi-rate filters, then signal distortion is minimized, but computational intensity increases
Solution Approach 1:
The patent segments the filtering operations so that linear phase FIR filtering is applied only to the low pass filter in the reduced-rate path, while the full-rate path processes high frequency components without requiring additional high pass filtering. This segmentation reduces the total computational intensity required for linear phase filtering while maintaining signal fidelity
Data Source
AI summary
A filtering method approximates a target Finite Impulse Response (FIR) (or transversal) filter and reduces computational requirements by eliminating high pass filtering required by known multi-rate filters. An input signal is copied into two identical signals and processed in parallel by a full-rate path, and by a reduced-rate path. Parallel filters are computed and applied in each path, the reduced-rate signal is up-sampled, and the two signals summed. The high pass filter required by known multi-rate filters is eliminated and the low pass filter in the prior art is implicit in a down sampling. Linear phase FIR filters are used for down and up sampling, resulting in constant group delay. Added benefits include the option of zero added latency through the filtering and the constant group delay added to the target FIR. The user may choose criteria such as minimum resolution in each band.