International Conference «Mathematical and Informational Technologies, MIT-2013»
(X Conference «Computational and Informational Technologies for Science,
Engineering and Education»)

Vrnjacka Banja, Serbia, September, 5–8, 2013

Budva, Montenegro, September, 9-14, 2013

Rusakov S.G.   Gourary M.M.   Ulyanov S.L.   Zharov M.M.  

The Concept of Redundancy for Techniques of Model Order Reduction in Circuit Simulation

Reporter: Rusakov S.G.

Model order reduction (MOR) techniques are widely used in practical design of integrated circuits.  The projection methods of MOR are based as a rule on Krylov subspace methods. But the trends of electronic industry advance new requirements for MOR techniques. In particular the actual important problem of MOR is the problem of reduction of the networks with many inputs and outputs. As known “Krylov subspace projection methods will be extremely inefficient“ for reducing the models with many ports. For instance the number of basic vectors added to the subspace in the case of multiport applications for block Krylov subspace methods is often proportional to the number of inputs. But in practical MOR problems the result order of reduced model obtained by this way is often much larger than it is required to provide desired accuracy.
This is reason of introducing the concept of redundancy into MOR problems in circuit simulation. The decreasing the redundancy means the decreasing the order of reduced models in comparison with MOR results obtained by Krylov subspace based methods. To avoid the redundancy of reduced models is the one of the main goal of improvement of the MOR techniques.
In frames of this challenging research topic a new approach to construct multipoint projection based reducing algorithms is proposed. The process of basis generation is performed under error control with worst-case analysis to determine frequency point and input excitation corresponding to the maximal user defined error norm. Some experimental results are presented.

To reports list

© 1996-2019, Institute of computational technologies of SB RAS, Novosibirsk