The impact of parasitic elements on the overall circuit performance keeps increasing from one technology generation to the next. In advanced process nodes, the parasitic effects dominate the overall circuit performance. As a result, the accuracy requirements of parasitic extraction processes significantly increased, especially for parasitic capacitance extraction. Existing parasitic capacitance extraction tools face many challenges to cope with such new accuracy requirements that are set by semiconductor foundries (< 5% error). Although field-solver methods can meet such requirements, they are very slow and have a limited capacity. The other alternative is the rule-based parasitic capacitance extraction methods, which are faster and have a high capacity; however, they cannot consistently provide good accuracy as they use a pre-characterized library of capacitance formulas that cover a limited number of layout patterns. On the other hand, the new parasitic extraction accuracy requirements also added more challenges on existing parasitic-aware routing optimization methods, where simplified parasitic models are used to optimize layouts.
This dissertation provides new solutions for interconnect parasitic capacitance extraction and parasitic-aware routing optimization methodologies in order to cope with the new accuracy requirements of advanced process nodes as follows.
First, machine learning compact models are developed in rule-based extractors to predict parasitic capacitances of cross-section layout patterns efficiently. The developed models mitigate the problems of the pre-characterized library approach, where each compact model is designed to extract parasitic capacitances of cross-sections of arbitrary distributed metal polygons that belong to a specific set of metal layers (i.e., layer combination) efficiently. Therefore, the number of covered layout patterns significantly increased.
Second, machine learning compact models are developed to predict parasitic capacitances of middle-end-of-line (MEOL) layers around FINFETs and MOSFETs. Each compact model extracts parasitic capacitances of 3D MEOL patterns of a specific device type regardless of its metal polygons distribution. Therefore, the developed MEOL models can replace field-solvers in extracting MEOL patterns.
Third, a novel accuracy-based hybrid parasitic capacitance extraction method is developed. The proposed hybrid flow divides a layout into windows and extracts the parasitic capacitances of each window using one of three parasitic capacitance extraction methods that include: 1) rule-based; 2) novel deep-neural-networks-based; and 3) field-solver methods. This hybrid methodology uses neural-networks classifiers to determine an appropriate extraction method for each window. Moreover, as an intermediate parasitic capacitance extraction method between rule-based and field-solver methods, a novel deep-neural-networks-based extraction method is developed. This intermediate level of accuracy and speed is needed since using only rule-based and field-solver methods (for hybrid extraction) results in using field-solver most of the time for any required high accuracy extraction.
Eventually, a parasitic-aware layout routing optimization and analysis methodology is implemented based on an incremental parasitic extraction and a fast optimization methodology. Unlike existing flows that do not provide a mechanism to analyze the impact of modifying layout geometries on a circuit performance, the proposed methodology provides novel sensitivity circuit models to analyze the integrity of signals in layout routes. Such circuit models are based on an accurate matrix circuit representation, a cost function, and an accurate parasitic sensitivity extraction. The circuit models identify critical parasitic elements along with the corresponding layout geometries in a certain route, where they measure the sensitivity of a route’s performance to corresponding layout geometries very fast. Moreover, the proposed methodology uses a nonlinear programming technique to optimize problematic routes with pre-determined degrees of freedom using the proposed circuit models. Furthermore, it uses a novel incremental parasitic extraction method to extract parasitic elements of modified geometries efficiently, where the incremental extraction is used as a part of the routing optimization process to improve the optimization runtime and increase the optimization accuracy.
School of Sciences and Engineering
Electronics & Communications Engineering Department
PhD in Engineering
Committee Member 1
Committee Member 2
Committee Member 3
Institutional Review Board (IRB) Approval
Not necessary for this item
Saleh, M. S.
(2022).Integrated Circuits Parasitic Capacitance Extraction Using Machine Learning and its Application to Layout Optimization [Doctoral Dissertation, the American University in Cairo]. AUC Knowledge Fountain.
Saleh, Mohamed Saleh Abouelyazid. Integrated Circuits Parasitic Capacitance Extraction Using Machine Learning and its Application to Layout Optimization. 2022. American University in Cairo, Doctoral Dissertation. AUC Knowledge Fountain.
Available for download on Wednesday, May 24, 2023