University of Minnesota
Characterization of Distance-Dependent Variations and Methodologies for Analog Layout Automation in FinFETs
Abstract
dc:description.abstractThe demand for analog and mixed-signal integrated circuits (ICs) is increasing due to their essential role in telecommunications, automotive, consumer electronics, healthcare, and industrial systems. The global market for mixed-signal ICs reached $121.9 billion in 2023 and is projected to grow to $184.0 billion by 2032. To meet this demand, enhancing analog electronic design automation (EDA) methodologies and tools is necessary, enabling designers to innovate quickly, achieve high performance, and reduce time-to-market and risk. However, unlike digital circuits, analog circuits have varied requirements and performance metrics, making it difficult to standardize EDA methodologies. Another major challenge is the lack of models to quantify the impact of phenomena that affect circuits post-fabrication but are not observed during the design stage, such as process variations. Addressing this gap is crucial for making optimal trade-offs between performance, area, and effort in analog design. The first part of this thesis focuses on the characterization and modeling of distance-dependent variations and their impact on analog circuit design. This research aims to address the gap in modeling local process variations, which can be random or distance-dependent. While random mismatches are well understood and decrease with larger device sizes, distance-dependent variations, caused by factors like process gradients, are not typically included in SPICE models provided in recent process design kits (PDKs). The latter can lead to mismatches between devices as the distance between them increases. Developing models for distance-dependent variations would help designers balance trade-offs between area, parasitics, and mismatch by allowing them to quantitatively select appropriate layout patterns. This thesis details our work on the measurement and modeling of distance-dependent variations. More than 30 dies fabricated using a commercial 12 nm FinFET technology node have been measured, and the chip design, floorplan, and measurement setup are presented. Device parameters such as threshold voltage (Vₜₕ) and current factor (𝛽) were measured to understand the distance-dependent variation surface and its impact on different device sizes and layout patterns. Furthermore, the impact of distance-dependent variations, layout-dependent effects (LDEs), layout parasitics, and layout area was examined on the design of amplifiers and current steering digital-to-analog converters (DACs). A methodology was developed to compare the performance and area of various layout patterns, allowing designers to make optimal choices by estimating the routing and layout area before starting the layout design process. The second part of the thesis focuses on analog layout design automation to address the time-consuming and tedious nature of layout design tasks such as floorplanning, placement, and routing. Previous efforts to automate analog layout synthesis have not been widely adopted due to their limitations in handling various design classes and their inability to match the expertise of human designers. A crucial component of automated analog layout methodologies is the constraint generation framework, which identifies the requirements that must be met during layout generation. These requirements vary for each circuit topology and can be time-consuming to evaluate during optimization. Therefore, there is a need to quickly infer and optimize these constraints. This thesis proposes a novel methodology to optimize a circuit layout, such that circuit requirements are inferred and satisfied, by defining and optimizing performance metrics for primitive cells or analog functional blocks. The primitive layouts are optimized while considering the impact of layout parasitics and LDEs within a primitive, and the routing between primitives to generate good-quality layouts. Optimization performed at the primitive level involves quick simulations to evaluate performance metrics and hence is faster compared to circuit-level optimizations. The resulting automatically generated layouts are shown to be competitive with manual layouts for multiple FinFET-based designs. This thesis also presents an overview of ALIGN (Analog Layouts Intelligently Generated from Netlists), which is a hierarchical analog layout tool that generates a layout from a sized schematic netlist. In this work, ALIGN was applied to layout several classes of analog circuits to identify limitations and to suggest areas of improvement within ALIGN as it was being developed. In the first part, ALIGN is improved and adapted to laying out ten high-frequency ring-oscillator-based VCO circuits in a commercial 12 nm PDK in an LVS and DRC clean manner, ready for tapeout. The application of ALIGN and learnings from other circuits such as an analog-to-digital converter (ADC) and switched-capacitor filter circuits are also captured.
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Madhusudan, Meghna
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Citation
- Madhusudan, Meghna. (2025). Characterization of Distance-Dependent Variations and Methodologies for Analog Layout Automation in FinFETs. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/280358.