DAC Gradient Error Correction for INL and DNL Linearity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital to analog converters (DACs) face systematic errors due to gradient inconsistencies in silicon wafers, leading to non-linear outputs and challenges in compensating for Integral Non-Linearity (INL) and Differential Non-Linearity (DNL), with existing methods requiring additional resources or a priori knowledge of gradients.
Innovation Solution
A hybrid DAC with a gradient correction module that models and applies a second-order polynomial correction term to address vertical and horizontal gradients, improving linearity by adjusting the signal path or output in the digital or analog domain, and incorporating an oversampled DAC to enhance resolution while slowing conversion rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a separate calibration DAC with look-up table is used to compensate for gradient errors, then INL and DNL are improved, but device complexity and current consumption increase
Solution Approach 1:
The patent extracts only the essential gradient compensation function from a full calibration DAC, implementing a simplified correction mechanism that applies gradient error compensation without requiring a complete separate calibration DAC. This reduces device complexity while maintaining INL and DNL improvement.
Solution Approach 2:
Instead of using a full calibration DAC that would require characterizing all DACs across process and temperature corners, the patent uses a simplified model that copies the gradient error characteristics and applies correction through a reduced mechanism, eliminating the need for extensive LUT entries while maintaining compensation effectiveness.
2Manufacturing precision
If a separate calibration DAC with look-up table is used to compensate for gradient errors, then INL and DNL are improved, but space and current consumption increase
Solution Approach 1:
The patent extracts only the essential gradient compensation function from a full calibration DAC, implementing a simplified correction mechanism that applies gradient error compensation without requiring a complete separate calibration DAC. This reduces current consumption while maintaining INL and DNL improvement.
3Manufacturing precision
If switching sequences are optimized to compensate for gradient errors, then INL is improved, but the solution is hard-wired and cannot adapt to different fabrication conditions
Solution Approach 1:
The patent implements a dynamic gradient error correction mechanism that can adapt to different fabrication conditions. The correction terms are calculated based on actual gradient measurements and can be adjusted for different process corners, temperature conditions, and fabrication variations, making the system flexible rather than hard-wired.
Solution Approach 2:
The patent changes the parameters used for gradient correction based on fabrication conditions. By measuring actual gradient errors and adjusting correction terms accordingly, the system adapts to different process corners, temperature ranges, and fabrication plants, maintaining effectiveness across varying conditions.
4Measurement precision
If oversampled DAC is used to achieve higher resolution, then resolution is improved, but conversion rate decreases
Solution Approach 1:
The patent applies partial oversampling by using gradient correction terms that are calculated at a lower rate and applied to compensate for gradient errors in the main DAC operation. This provides the benefits of higher resolution through correction without requiring the entire system to operate at the slower oversampled rate, thus maintaining higher conversion rates.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A digital-to-analog converter is disclosed. The converter includes a gradient correction module that generates a correction term based on a model of gradient error. The correction term is then applied to the signal path in the digital domain or applied to the output of the digital-to-analog converter in the analog domain. The model used to generate the correction term is based on a vertical gradient error in the array of current source elements, which may be modelled and calibrated using a second-order polynomial. Further, a digital-to-analog converter having a Nyquist DAC and an oversampled DAC is disclosed. When the oversampled DAC is enabled, the resolution of the Nyquist DAC may be increased while slowing the conversion rate.