[paper] An Open-Source AMS Circuit Optimization

Z. Li and A. Chan Carusone

An Open-Source AMS Circuit Optimization Framework Based on Reinforcement Learning: 

From Specifications to Layouts

in IEEE Access, vol. 12, pp. 150032-150045 (2024) 

DOI: 10.1109/ACCESS.2024.3478832

Abstract: This paper presents a fully open-sourced AMS integrated circuit optimization framework based on reinforcement learning (RL). Specifically, given a certain circuit topology and target specifications, this framework optimizes the circuit in both schematic and post-layout phases. We propose using the heterogeneous graph neural network as the function approximator for RL. Optimization results suggest that it can achieve higher reward values with fewer iterations than the homogeneous graph neural networks. We demonstrate the applications of transfer learning (TL) in optimizing circuits in a different technology node. Furthermore, we show that by transferring the knowledge of schematic-level optimization, the trained RL agent can optimize the post-layout performance more efficiently than optimizing post-layout performance from scratch. To showcase the workflow of our approach, we extended our prior work to optimize latched comparators in the SKY130 and GF180MCU processes. Simulation results demonstrate that our framework can satisfy various target specifications and generate LVS/DRC clean circuit layouts.

FIG: Proposed AMS IC optimizer overview. 

The picture is adapted from [Z. Li and A. C. Carusone; 2023]

Acknowledgment: The authors would like to thank Dr. Hossein Shakiba from Huawei Technologies
for his valuable discussions throughout this project.

[REF] Z. Li and A. C. Carusone, "Design and optimization of low-dropout voltage regulator using relational graph neural network and reinforcement learning in open-source SKY130 process," in Proc. IEEE/ACM Int. Conf. Comput. Aided Design (ICCAD), Oct. 2023, pp. 1–9.

[paper] Parameter Extraction Approaches for Memristor Models

Dmitry Alexeevich Zhevnenko^1,2, Fedor Pavlovich Meshchaninov^1,2, Vladislav Sergeevich Kozhevnikov^1,2, Evgeniy Sergeevich Shamin^1,2, Oleg Alexandrovich Telminov^1,2, and Evgeniy Sergeevich Gornev^1,2

Research and Development of Parameter Extraction Approaches for Memristor Models

Micromachines 2021, 12, 1220. 

DOI: 10.3390/mi12101220

   
1 Moscow Institute of Physics and Technology, Moscow, Russia;
2 JSС MERI, Zelenograd, Russia

Abstract: Memristors are among the most promising devices for building neural processors and non-volatile memory. One circuit design stage involves modeling, which includes the option of memristor models. The most common approach is the use of compact models, the accuracy of which is often determined by the accuracy of their parameter extraction from experiment results. In this paper, a review of existing extraction methods was performed and new parameter extraction algorithms for an adaptive compact model were proposed. The effectiveness of the developed methods was confirmed for the volt-ampere characteristic of a memristor with a vertical structure: TiN/HfxAl1-xOy/HfO2/TiN.

Fig: Model VACs with different numbers of inhomogeneities: 

(a) four inhomogeneities; (b) no inhomogeneities.

Acknowledgments: This research was funded by the Ministry of Science and Higher Education of the Russian  Federation, grant number 075-15-2020-791. Authors thank the Institute of Microelectronics Technology and High-Purity Materials RAS for access to experimental data on the study of graphene oxide memristor switching cycles.

[paper] Novel Design and Optimization and the gm/ID Ratio

A Novel Design and Optimization Approach for Low Noise Amplifiers (LNA) Based on MOST Scattering Parameters and the gm/ID Ratio

Juan L. Castagnola¹*, Fortunato C. Dualibe² Agustín M. Laprovitta¹ and Hugo García-Vázquez³

¹Facultad de Ingeniería, Universidad Católica de Córdoba, Córdoba 5017 (AN)

²Service d’Électronique et Microélectronique, Université de Mons (UMONS), 7000 Mons (BE)

³Departamento de Electrónica, Instituto de Astrofísica de Canarias (IAC), 38200 La Laguna (SP)

* Author to whom correspondence should be addressed.

Electronics 2020, 9(5), 785; https://doi.org/10.3390/electronics9050785

Received: 31 March 2020 / Revised: 30 April 2020 

Accepted: 9 May 2020 / Published: 11 May 2020

Abstract - This work presents a new design methodology for radio frequency (RF) integrated circuits based on a unified analysis of the scattering parameters of the circuit and the gm/ID ratio of the involved transistors. Since the scattering parameters of the circuits are parameterized by means of the physical characteristics of transistors, designers can optimize transistor size and biasing to comply with the circuit specifications given in terms of S-parameters. A complete design of a cascode low noise amplifier (LNA) in 65nm CMOS technology is taken as a case study in order to validate the approach. In addition, this methodology permits the identification of the best trade-off between the minimum noise figure and the maximum gain for the LNA in a very simple way.

Figure: gm/ID versus iD

Acknowledgement - This research was funded by Universidad Católica de Córdoba (Argentina), the Walloon Region DGO6 BEWARE Fellowships Academia Programme (1410164-POHAR, cofunded by the European Marie Curie Actions), the Belgian FNRS (Fond National pour la Recherche Scientifique) and the Argentinean MINCyT (Ministerio de Ciencia y Tecnología).

levmar : Levenberg-Marquardt nonlinear least squares algorithms in C/C++

Introduction page: <http://users.ics.forth.gr/~lourakis/levmar/> 

The site provides GPL native ANSI C implementations of the Levenberg-Marquardt optimization algorithm, usable also from C++, Matlab, Perl, Python, Haskell and Tcl and explains their use. Both unconstrained and constrained (under linear equations, inequality and box constraints) Levenberg-Marquardt variants are included. The Levenberg-Marquardt (LM) algorithm is an iterative technique that finds a local minimum of a function that is expressed as the sum of squares of nonlinear functions. It has become a standard technique for nonlinear least-squares problems and can be thought of as a combination of steepest descent and the Gauss-Newton method. When the current solution is far from the correct one, the algorithm behaves like a steepest descent method: slow, but guaranteed to converge. When the current solution is close to the correct solution, it becomes a Gauss-Newton method.

Interfaces for using levmar from high-level programming environments & languages such as Matlab, Perl Python, Haskell and Tcl are also available; please refer to the FAQ for more details.