Spectrogram-Based Sample Rate Conversion for Flexible Decimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for decimating sample rates in signal processing are inefficient and inflexible, requiring large filters or complex calculations, and often struggle with high relative bandwidths and parallel processing, limiting their applicability in applications like measurement technology and radiomonitoring.
Innovation Solution
A device and method that involves a two-stage process: first, a forward transformer reduces the sampling rate efficiently using a transformation with a specific length, followed by an inverse transformation and time domain interpolation to achieve a flexible and efficient change in sampling rate, allowing for adjustment of the second sampling rate using a time domain interpolator and frequency windows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a polyphase filter with heavily oversampled impulse response is used for decimation, then decimation with variable sampling rate is achieved, but computation time and memory requirements disproportionately increase
Solution Approach 1:
The patent divides the decimation process into two separate stages: a first decimation stage that reduces the sampling rate by a first factor, and a second decimation stage that reduces it by a second factor. This segmentation allows each stage to use smaller, more efficient filters rather than one large oversampled filter, thereby reducing overall computational complexity and memory requirements while maintaining variable sampling rate capability.
Solution Approach 2:
The patent changes the decimation factors from a single large factor to two smaller factors applied sequentially. By parameterizing the decimation process into multiple stages with adjustable factors, the system achieves variable sampling rate conversion with reduced computational burden compared to using a single heavily oversampled filter.
2Quantity of substance
If linear interpolation between coefficients is used, then memory requirements are reduced, but calculation complexity increases
Solution Approach 1:
The patent applies linear interpolation separately in each decimation stage rather than using a single complex interpolation scheme. This segmented approach reduces memory requirements by using smaller filter coefficients at each stage, while the interpolation complexity is managed through the staged architecture rather than concentrated in one complex calculation.
3Productivity
If Farrow interpolator with short FIR filter and polynomial interpolation is used, then computation time is reduced, but SFDR deteriorates to approximately 50 dB
Solution Approach 1:
The patent uses two sequential decimation stages instead of a single Farrow interpolator stage. This segmentation allows each stage to achieve moderate decimation with acceptable SFDR, and the cumulative effect of two stages provides the required overall decimation ratio while maintaining better SFDR performance than a single Farrow interpolator would achieve alone.
4Adaptability or versatility
If decimation in frequency domain using fast convolution with different FFT lengths is used, then decimation is achieved, but implementation efficiency is very low due to integer constraints on transformation lengths
Solution Approach 1:
Instead of using frequency domain fast convolution with FFT as the primary decimation method, the patent inverts the approach by using time domain filtering and decimation. This inversion avoids the integer constraints and efficiency problems of frequency domain methods, as time domain implementations can more flexibly handle variable decimation factors without requiring specific integer relationships between transformation lengths.
Data Source
AI summary
The invention relates to a device for changing a sampling rate, comprising a forward transformer for forming a first version of a spectogram by means of a transformation with a first transformation length from an information signal, which has a first sampling rate. The device comprises a processor for forming a second version of the spectogram, which has a smaller bandwidth in relation to the first version. The device comprises a reverse transformer for forming a roughly pre-changed information signal having a first sampling rate that is reduced in relation to the first sampling rate, by means of a reverse transformation of the second version of the spectogram with a second transformation length that is reduced in relation to the first transformation length. The device comprises a time domain interpolator for obtaining an information signal having a third sampling rate, which is changed in relation to the second sampling rate, by means of interpolation of the pre-changed information signal.