New Progress in the Field of Nonlinear Exponential Integrating Circuit Simulation Techniques Made by Chen Quan Research Group at School of Microelectronics

2020-08-11 Research News views：12014

Recently, Chen Quan, an assistant professor at School of Microelectronics, Southern University of Science and Technology (SUSTech), developed and innovated the nonlinear exponential integrating circuit simulation techniques, which significantly improve the performance of ultra-large-scale time-domain circuit simulation. Relevant results have been published in the IEEE Design Automation Conference (DAC) 2020, a top conference in the field of EDA, under the title of "A Robust Exponential Integrator Method for Generic Nonlinear Circuit Simulation".

Circuit simulation techniques are the most commonly used tool in electronic design automation (EDA). No matter in digital, analogy or RF IC design, a large number of circuit simulations are required to optimize and verify the design. With the development of semiconductor technology and the increasing complexity of circuit design, designers often need to face the ultra-large-scale circuits with tens or even hundreds of millions of nodes. However, at present, the mainstream circuit simulation tool is still the SPICE framework developed by the University of California, Berkeley in the 1970s, which has not changed significantly for decades, whose performance lags behind the needs of the industry seriously, and which often takes several days for large-scale circuit simulations. The exponential integration method is a new time-domain integration method which is most promising to break through the SPICE framework in recent years and greatly improve the efficiency of circuit simulation without reducing accuracy. Currently, it has already demonstrated 1-2 orders of magnitude speed improvement in linear circuit applications. However, there are still many challenges in nonlinear circuit simulation, such as low numerical stability and weak nonlinear processing ability, which makes exponential integration method unusable in many nonlinear circuits.

To address these problems, researchers analyse the origin of numerical instability of the exponential integration method in this paper, and innovatively propose a differential-algebraic equation normalization method that can maintain sparsity of the matrix, which initially solves the problem of numerical instability of the exponential integration method in the nonlinear system. In addition, the paper also proposes a new nonlinear solution technique based on Newton iteration method, which can significantly improve the processing ability of exponential integration method for strong nonlinear systems, with a 5-fold speedup over SPICE at the same accuracy. These theoretical and technological breakthroughs make the exponential integration method promising to become the next generation of general high-performance circuit simulation techniques. 

Chen Quan is the sole author of this paper, SUSTech is the first unit involved, and this is the first time SUSTech has published a paper in IEEE DAC. IEEE DAC is the top conference in the field of EDA, with over 5,000 attendees per year and only 23% of the articles accepted.

Paper link：http://www2.dac.com/events/eventdetails.aspx?id=295-57

Brief introduction to Chen Quan Research Group:

The main research directions of Chen Quan Research Group of the School of Microelectronics are ultra-large-scale circuit simulation in the field of electronic design automation (EDA), quantum physics simulation of emerging nano-electronic devices and multi-physics modelling simulation. And the Research Group has achieved a series of internationally influential results. Such research results have been published in top EDA journals and conferences, including TCAD, IEEE Circuits Syst. Mag., DAC, ICCAD, DATE, ASP-DAC, etc.

This is also the first time that SUSTech has published a paper