Segmented DAC Subtractive Dither for Inter-Segment Linearity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Segmented digital-to-analog converters (DACs) suffer from inter-segment errors that cause non-linear distortion, particularly in large narrowband signals and deep digital back-off conditions, which are not effectively addressed by existing calibration methods that increase complexity and power consumption.
Innovation Solution
Implementing subtractive dithering by introducing overrange DAC cells and a dither control circuit to randomize inter-segment errors, which is independent of DAC segment architecture and can be used with or without additional calibration methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If segmented DAC architecture is used to achieve high resolution, then conversion precision is improved, but inter-segment errors cause non-linear distortion and increase noise floor
Solution Approach 1:
The patent applies preliminary action by adding dither to the digital input data before it reaches the segmented DAC. This dither is added in advance to randomize the inter-segment errors that will occur during conversion, preventing them from causing systematic non-linear distortion. The dither preparation happens in the digital domain before the analog conversion process begins.
Solution Approach 2:
The patent converts the harmful inter-segment errors into beneficial randomized noise through dithering. Instead of attempting to eliminate the inherent mismatches between DAC segments, the invention accepts these errors and uses dither to transform them from correlated distortion sources into uncorrelated noise that can be filtered more effectively, thereby improving overall linearity and spectral purity.
2Manufacturing precision
If calibration is performed to reduce inter-segment errors, then linearity is improved, but calibration complexity and power dissipation increase
Solution Approach 1:
The patent replaces the mechanical/calibration-based approach with a digital signal processing approach. Instead of physically adjusting or calibrating the DAC segments to match each other, the invention uses digital dithering and spectral filtering to achieve the same linearity improvement. This substitution eliminates complex calibration circuits and procedures while maintaining or improving linearity performance.
Solution Approach 2:
The patent changes the parameter domain from physical calibration adjustments to digital signal processing parameters. By modifying the digital input data with dither and applying spectral filtering in the digital domain, the system achieves linearity improvement without altering physical DAC segment characteristics or requiring complex calibration procedures.
3Manufacturing precision
If dither is added to randomize inter-segment errors, then linearity is improved, but noise floor increases
Solution Approach 1:
The patent moves the noise problem from the time domain to the frequency domain. By applying spectral filtering after dithering, the invention separates the dither-induced noise from the desired signal in the frequency domain. The filter removes out-of-band noise components while preserving the in-band signal, effectively managing the noise floor through dimensional transformation from time to frequency analysis.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A segmented digital-to-analog converter (DAC) includes DAC segments, an overrange DAC, and a dither control circuit. Each DAC segment includes a plurality of DAC cells for generating an analog output signal based on input data to each DAC segment. The overrange DAC generates an analog output signal based on a control signal. The dither control circuit adds a dither to first input data supplied to a higher-order DAC segment, subtract a portion of the dither from second input data supplied to a lower-order DAC segment, and generate the control signal for subtracting a remaining portion of the dither from an output of the segmented DAC in an analog domain. The dither added to the first input data may be one of +1, 0, and -1 and the portion of the dither subtracted from the second input data may be a half of the dither added to the first input data.