Polyphase Filter for Non-Integer Sample Hold
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sigma-delta digital-to-analog converter systems face challenges in performing non-integer sample/hold operations, which require high computational complexity and are inefficiently implemented with traditional filter architectures, especially when dealing with non-integer ratios.
Innovation Solution
A method and apparatus utilizing a polyphase filter constructed in the virtual upsampled domain with a long zero-order hold and short FIR filter, reducing the need for complex computations by only calculating a few output samples immediately after an input transition, thus enabling efficient non-integer sample/hold operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional digital filter architectures are used for non-integer sample/hold operations, then the system can perform sample rate conversion, but the computational complexity becomes excessively high and implementation becomes inefficient
Solution Approach 1:
The polyphase filter is divided into multiple phases (e.g., 8 phases for a 125/8 ratio), where each phase processes a subset of the input samples. This segmentation allows the system to handle non-integer sample rate conversion by distributing the computational load across parallel phases, reducing the complexity compared to a single monolithic filter.
Solution Approach 2:
The system pre-calculates and stores filter coefficients for each phase of the polyphase filter. By preparing these coefficients in advance, the runtime computation is significantly reduced, as the filter only needs to select and apply pre-computed values rather than performing complex calculations for each output sample.
2Adaptability or versatility
If traditional filter architectures are used for non-integer ratios, then sample rate conversion can be achieved, but the implementation requires a large number of computations and coefficient storage
Solution Approach 1:
The filter coefficients are segmented into multiple polyphase components, with each phase storing only the coefficients it needs. For a 125/8 ratio, the filter is divided into 8 phases, each storing 1/8 of the total coefficient set. This segmentation reduces the memory requirements compared to storing all coefficients in a single buffer.
Solution Approach 2:
Each polyphase phase is designed with local coefficients optimized for its specific function in the conversion ratio. Rather than using identical coefficients throughout, each phase has tailored coefficients that are only valid for that phase's specific time slots, reducing overall storage requirements while maintaining conversion accuracy.
3Measurement precision
If a polyphase filter is implemented with full calculation of all output samples, then accuracy is maintained, but computational efficiency is reduced
Solution Approach 1:
The invention extracts only the essential output samples that need to be calculated for non-integer sample rate conversion. In a 125/8 ratio conversion, not all 125 output samples need full filtering computation - only specific samples at phase boundaries require complex calculation, while intermediate samples can be derived more simply from previous computations.
Solution Approach 2:
The system performs full filtering calculations only when necessary (at phase transitions and boundary conditions), while using simplified computations for intermediate samples. This partial action approach maintains accuracy where required while improving overall computational efficiency by avoiding redundant full calculations.
Data Source
AI summary
A method and apparatus for implementing non-integer sample hold operations in a sigma-delta digital-to-analog converter system includes an interpolation filter, a polyphase filter circuit, and a modulator. The polyphase filter circuit is used to virtual upsample a digital input signal by a predetermined non-integer upsample ratio of a relatively large number. The polyphase filter circuit is formed of a long zero-order hold and a short FIR filter so that only several branches associated with the polyphase filter circuit corresponding to output samples immediately after a transition of the digital input signal is required to be calculated, thereby reducing the need to store a large number of filter coefficients and eliminating complex computations.