Modeling Substrate Noise Coupling in Mixed-Signal ICs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Designing mixed-signal ICs is challenging due to noise coupling through the substrate, which affects analog circuitry performance unpredictably, and existing modeling techniques are either too computationally expensive or inadequate for complex geometric arrangements.
Innovation Solution
The method involves modeling electrical impedance using horizontal and L-shaped impedance models to simulate substrate noise coupling, combining elements into fewer groups, and representing these in circuit simulations to predict noise behavior before fabrication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If numerical analysis methods (Pisces, finite-element, finite difference) are used to model substrate noise coupling, then measurement precision is improved, but productivity deteriorates due to high computation cost
Solution Approach 1:
The substrate is divided into discrete regions with different conductivity types (highly doped regions, lightly doped regions, intrinsic regions). Each region is modeled with appropriate simplified impedance characteristics, allowing the complex substrate to be broken down into manageable segments that can be processed faster while maintaining accuracy.
Solution Approach 2:
The patent transforms the continuous substrate conductivity distribution into discrete parameter categories (highly doped, lightly doped, intrinsic). This parameter discretization enables the use of simplified impedance models for each category, dramatically reducing computation time while preserving the essential noise coupling behavior.
2Measurement precision
If full domain discretization is used to handle large and dense designs, then measurement precision is improved, but productivity deteriorates due to excessive computation cost
Solution Approach 1:
Instead of discretizing the entire substrate domain, the patent segments only the regions containing active elements and their immediate surroundings. These segmented regions are then connected through simplified impedance models representing the substrate regions between them, avoiding the need for full domain discretization while maintaining accuracy for critical areas.
Solution Approach 2:
The patent extracts and models only the critical substrate regions that contain noise sources and noise receivers, along with the connecting paths between them. Non-critical substrate areas are represented by simplified impedance parameters rather than being fully discretized, reducing the overall computational burden.
3Productivity
If formula-based macromodeling techniques are used, then productivity is improved due to faster evaluation, but measurement precision deteriorates for complex geometric arrangements
Solution Approach 1:
The patent extends formula-based macromodeling by introducing discrete conductivity parameters (highly doped, lightly doped, intrinsic) that can represent complex substrate geometries. These parameterized models maintain the computational speed of formula-based approaches while improving accuracy for complex arrangements through the segmented regional approach.
Data Source
AI summary
Aspects of the disclosed techniques relate to techniques for modeling substrate noise coupling. Electrical impedance between two contacts in the presence of one or more other contacts is modeled based on a horizontal impedance model and an L-shaped impedance model. The one or more other contacts may be clustered together in four regions first and then are represented by the horizontal impedance model and/or the L-shaped impedance model. The electrical impedance is inserted into netlist for circuit simulation.