Predictive-Filter DDS for Spurious Tone Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital frequency synthesis systems face challenges in achieving high spectral clarity while minimizing overhead, as increasing the address bus of look-up table memory and digital-to-analog converter capabilities lead to significant increases in memory and computational resources.
Innovation Solution
A frequency synthesizer architecture that incorporates a predictive filter to noise-shape the phase profile signal, separating it into quantized and error portions, and using a digital-to-analog converter with reduced precision to generate an analog sinusoidal signal, thereby suppressing spurious tones without increasing memory or DAC capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the address bus of the LUT memory is increased to suppress spurious tones, then the spectral clarity is improved, but the memory size and system overhead increase significantly
Solution Approach 1:
The patent segments the phase profile signal into multiple components: a quantized version that addresses the LUT and an error portion that feeds the predictive filter. This segmentation allows the system to process the signal in parts, using the error portion to generate corrections that suppress spurious tones without requiring a larger address bus or increased memory capacity.
Solution Approach 2:
The predictive filter performs preliminary action by generating predicted error values in advance based on the error portion of the phase profile signal. These predicted errors are subtracted from the quantized phase profile before it reaches the LUT, pre-compensating for quantization effects and suppressing spurious tones before they can degrade spectral clarity, thereby maintaining performance with reduced memory requirements.
2Measurement precision
If the DAC capability is increased to handle larger sample sizes for enhanced quality, then the signal quality is improved, but the device complexity and overhead increase
Solution Approach 1:
The patent segments the phase profile signal processing into distinct paths: one handling the quantized version for LUT addressing and another handling the error portion for predictive filtering. This segmentation allows the system to manage quantization errors separately, enabling the use of a lower-resolution DAC while maintaining overall signal quality through the predictive error correction mechanism.
Solution Approach 2:
The predictive filter acts as an intermediary between the quantization process and the DAC. It processes the error portion and generates correction signals that compensate for quantization effects, mediating the relationship between reduced precision components and the final output quality, thereby enabling the use of a simpler DAC with reduced capability.
3Ease of manufacture
If conventional DDS architecture is used with LUT, DAC, and anti-imaging filter, then the system is simple to implement, but the spectral clarity and SFDR are insufficient without increasing overhead
Solution Approach 1:
The patent merges the predictive filter functionality into the existing DDS architecture, combining it with the accumulator, quantizer, and LUT. This integration adds spectral clarity enhancement capabilities to the conventional simple architecture without requiring completely new components, maintaining ease of manufacture while improving SFDR performance through the combined action of the predictive filter and existing DDS elements.
Data Source
AI summary
Method and apparatus for digital frequency synthesis are described. A frequency synthesizer has an accumulator, an adder, and a predictive filter. The adder is configured to subtract a predicted error from a phase profile signal. A quantized version of the phase profile signal is separated from an error portion thereof. The predictive filter, set for a fraction of a sample frequency bandwidth, is coupled to receive the error portion for generation of a next predicted error. A storage device has digital representations of sinusoidal signals accessible responsive to the quantized version of the phase profile signal. A digital-to-analog converter is coupled to receive a digital representation of a sinusoidal signal obtained from the storage device to provide an analog sinusoidal signal. An anti-imaging filter is coupled to receive the analog sinusoidal signal and configured to filter out noise.