Orthogonal-Function Compensation for Mixed-Signal Nonlinearity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-linearity in mixed-signal ICs, particularly in ΔΣ DACs and ΔΣ FN-PLLs, limits their performance by causing noise and distortion, which cannot be effectively removed by simple filters, and existing compensation techniques are inadequate for high-performance applications.
Innovation Solution
An adaptive non-linearity identification and compensation scheme using orthogonal kernels and the Least Mean Square (LMS) method to accurately represent and compensate for non-linearity in circuit blocks, allowing for robust performance across Process, Supply, and Temperature (PVT) variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If non-linearity compensation is not applied, then device complexity is low, but measurement precision and signal quality deteriorate due to noise and distortion
Solution Approach 1:
The non-linearity compensation is segmented into identification and compensation stages. The identification stage uses orthogonal kernels to decompose the non-linear distortion into separable components, while the compensation stage applies inverse operations to each component separately, reducing overall system complexity
Solution Approach 2:
Orthogonal kernels serve as intermediaries between the non-linear circuit block and the compensation mechanism. These kernels decompose complex non-linear distortions into manageable spectral components that can be independently compensated, bridging the gap between simple filtering and complex non-linearity correction
2Measurement precision
If simple filtering is used to remove noise and distortion, then device complexity is low, but measurement precision deteriorates because shaped out-of-band noise folds intermodulation products into the baseband
Solution Approach 1:
The invention converts the harmful folded intermodulation products into identifiable spectral components through orthogonal kernel decomposition. By characterizing the distortion spectrum using orthogonal functions, the compensation mechanism can selectively target and remove specific intermodulation products while preserving the desired signal and shaped noise
Solution Approach 2:
The compensation scheme changes the spectral parameters of the distorted signal by applying inverse orthogonal kernel transformations. This selectively modifies the amplitude and phase of specific frequency components (intermodulation products) while leaving the signal band intact, achieving high SFDR improvement without complex multi-stage filtering
3Measurement precision
If existing compensation techniques are used, then device complexity is moderate, but measurement precision is insufficient for high-performance applications due to inadequate compensation
Solution Approach 1:
The compensation scheme dynamically adapts to varying operating conditions through continuous non-linearity identification. The orthogonal kernel decomposition is performed in real-time, allowing the compensation parameters to adjust automatically to changes in circuit characteristics due to PVT variations, achieving high performance across different operating points
Solution Approach 2:
The invention implements a feedback mechanism where the output of the non-linear block is continuously monitored, decomposed using orthogonal kernels, and fed back to generate compensating signals. This closed-loop approach ensures that compensation remains accurate even as circuit characteristics drift due to temperature, process, or voltage variations
4Measurement precision
If adaptive compensation with orthogonal kernels is applied, then measurement precision and signal quality improve, but device complexity and computational requirements increase
Solution Approach 1:
The computational load is segmented across multiple orthogonal kernel channels, each processing a specific frequency band or distortion component independently. This parallel decomposition allows complex non-linearity compensation to be distributed across simpler sub-processing units, reducing the computational complexity of each individual processing stage while maintaining overall high SNR and SFDR performance
Data Source
AI summary
A feedback divider in a mixed-signal circuit is modulated by a frequency control word controlling a delta-sigma modulator. An accumulated quantization error from the delta-sigma modulator is compared to a residual error in the circuit by a Least-Mean Square (LMS) correlator for gain calibration to adjust for linear errors. Upper bits of the accumulated quantization error access a lookup table to find two outputs of the compensation function that are interpolated between using lower bits of the accumulated quantization error. The interpolated result is an adjustment subtracted from the loop to compensate for non-linear errors. A set of orthogonal kernels is generated from the accumulated quantization error and calibrated using another LMS correlator and inverse transformed to generate updates to the non-linear compensation function in the lookup table. The kernels can be Walsh Hadamard (WH) and the inverse transformer an inverse WH transformer.