ISSN 10637397, Russian Microelectronics, 2011, Vol. 40, No. 7, pp. 473–475. © Pleiades Publishing, Ltd., 2011.
Original Russian Text © V.N. Perminov, S.V. Makarov, S.A. Kokin, A.E. Veselov, 2010, published in Izvestiya Vysshikh Uchebnykh Zavedenii. Elektronika, 2010, No. 2, pp. 42–47.
The very largescale integration circuits (VLSI cir
cuits) are divided into custom and semicustom circuits
by the style of design and implementation. Semicus
tom VLSI can be classed into cellbased VLSI circuits
and arraybased VLSI circuits. The libraries of presep
arated library cells or custom generators of such cells,
including memory cards, which form the tracing of a
cell by its functional description, are used in cellbased
Development based on library cells implies the use
of standard cells or cells’ generators for the implemen
tation of primitives. It is customary to use generators
for the synthesis of storage cells, programmable logic
arrays, and complex deliverers and processing units of
data streams, such as multipliers and multiplexors.
The cell generators are fully parameterized and can
provide different capacity, memory, and other param
The development of library cells is the prime prob
lem for the design of custom digital VLSI. The
decreased design periods of the library can be attained
using customized CAD. The analysis of the market of
library cells’ CAD points to the fact that these prod
ucts are basically absent in the market. Such systems
are generally constituents of “in house CAD” and are
not for sale.
The following types of library cells are used in the
design of VLSI: highperformance, highload, and
lowpower; with low supply voltage; space saving with
minimum area; high noiseprotected; and high radia
As the process flow of VLSI production changes, all
library elements need to be designed. Such correction
of process flow typically occurs annually or biannually.
Demand has been created for a new set of libraries
in factory’s learning a new process flow, e.g., on pass
ing from 130 to 110 or 90 nm. The problem of main
taining the libraries of VLSI digital cells for a factory is
currently a central problem.
The process of the design of libraries of VLSI digital
elements is managed in the following manner. The
parametric optimization from the given electric cir
cuits of each library cell is carried out to obtain the
optimum in terms of the key criterion for a given type
of library. Then, the library cell layout is formed. The
process of designing a library ends in the characteriza
tion of all library cells, which makes it possible to use
the library on the stage of the synthesis of VLSI electric
circuit. For error recovery, the return to earlier stages
of design is possible as the need arises.
A peculiarity of the process of VLSI digital cells’
design is the great number of library cells. Their num
ber typically ranges up to several hundreds of cells,
which leads to the large volume of calculations at the
stage of parametric optimization.
The design of current nanoelecrtonic circuits raises
the problem of the design of custom interfaces of
nanolelectronic circuits and devices. The point is that
the joint functioning of the library cells operated in
terms of new physical principles and standard CMOS
circuits needs to be provided.
At the present time, a lot of algorithms of multidi
mensional and onedimensional optimization have
been developed [1, 2]; however, the following algo
rithms have become a defacto standard for the major
ity of spicesimulators and simulation programs: the
bisection method for the onedimensional case and
the Levenberg–Marquardt method for multidimen
sional optimization . The larger software compa
nies use their unique algorithms; however, the princi
ple is identical for all.
The algorithm solves the problem of nonlinear
optimization in terms of the leastsquares method.
The Use of Parameter Optimization in the Design of VLSI Library
Cells and Custom Interfaces of Nanoelectronic Devices
V. N. Perminov
, S. V. Makarov
, S. A. Kokin
, and A. E. Veselov
UniqueICs LLC, Moscow, Russia
Lukin State Research Institute for Problems in Physics, Moscow, Russia
Moscow State Institute of Electronic Technology, Moscow, Russia
—The modified Levenberg–Marquardt algorithm applied to the parametric optimization of VLSI
library cells and custom interfaces of nanoelectronic devices is presented.
CIRCUIT TECHNOLOGY AND DESIGN