Macromodel of High Speed Interconnect using Vector Fitting Algorithm

IJSRD - International Journal for Scientific Research & Development| Vol. 1, Issue 2, 2013 | ISSN (online): 2321-0613
All rights reserved by www.ijsrd.com 70
Macromodel of High Speed Interconnect using Vector Fitting Algorithm
Naresh Punani1
Ravi Mehta2
Ashish Sarvaiya3
1
PG Scholars, Shantilal Shah Engg. College, India
2
Assistant Professor, Shantilal Shah Engg. College, India
3
Assistant Professor, Govt. Engg. College Bhavnagar, India
Abstract— At high frequency efficient macromodeling of
high speed interconnects is all time challenging task.
We have presented systematic methodologies to
generate rational function approximations of high-speed
interconnects using vector fitting technique for any type of
termination conditions and construct efficient multiport
model, which is easily and directly compatible with circuit
simulators.
Keywords— Macromodel, Vector Fitting, Frequency
Response
I. INTRODUCTION
Transient simulation and effective macromodeling of high
speed interconnect is all time challenging task. As high
frequency transmission line is best suited or modelled by
frequency domain data and most of the other non linear
devices and systems are represented in time domain
effectively, hence it is the problem of mixed domain and not
easily solved. Mixed domain problem is also CPU expensive
[1].
we propose methodologies for rational function
approximation of high- speed interconnect network
responses. The method relies on the use of frequency-
domain data, obtained either through measurement or
numerical simulation obtained through 3D electromagnetic
simulator.
Interconnect can be modelled either using lump
RLC component or using distributed RLC component or
Macromodel [1].
Lump Model: This model is used to represent interconnect
when the frequency of operation is low. In this model
interconnect is modeled by connecting lump RLC
component and it will make RLC network. Disadvantage of
this model is that when the frequency of operation is high
we can’t represent RLC component in lump form.
Distributed Model: This model is used to represent
interconnect when the frequency of operation is high.
Interconnect is model using RLC component but define in
per unit length. In this model due to large number of internal
nodes this model is become CPU expensive during
simulation.
Macromodel from tabulated data: The macromodeling
process for transient simulation of Tabulated data is
illustrated in Figure1.In the recent years, the
macromodeling-based transient simulation approaches have
gained popularity due to their computational advantages.
Figure.1 Macromodel process from tabulated data
In section II, vector fitting method is described
theoretically. This method relies mainly on selection of
starting poles this signification explained in this section.
II. THEORY OF VECTOR FITTING
A. Vector fitting by pole relocation [2]:
Consider the rational function
( ) ∑ (1)
The residues cn, and poles an, are either real quantities or
come in complex conjugate pairs, while d and h are real. We
have to estimate all coefficients in (1) so that a least squares
approximation of f(s) is obtained over a given frequency
interval. We note that (1) is a nonlinear problem in terms of
the unknowns, because the unknowns an, appear in the
denominator. Vector fitting solves the problem (1)
sequentially as a linear Problem in two stages, both times
with known poles.
1) Stage-1: Pole identification
Specify a set of starting poles ̅̅̅ , in (1), and multiply f(s)
with an unknown function σ(s). In addition we introduce a
rational approximation for σ(s). This gives the augmented
problem:
Tabulated Data
Vector Fitting
Passivity Checking
RLC circuit
Transient Simulation
using circuit simulators

Macromodel of high speed interconnect using Vector Fitting Algorithm
(IJSRD/Vol. 1/Issue 2/2013/0006)
All rights reserved by www.ijsrd.com
71
[
( ) ( )
( )
] [
∑
̅̅̅̅
∑
̃
̅̅̅̅
] (2)
Multiplying the second row in (2) with f(s) yields the
following relation
(∑ ̅̅̅̅
) (∑
̃
̅̅̅̅
) ( ) (3)
Or (σf)fit(s) ( ) ( ) .
A rational function approximation is obtain from equation
(3). This become evident if each sum of partial fraction in
(3) is written as:
( ) ( )
∏ ( )
∏ ( ̅̅̅̅)
, ( )
∏ ( )̃
∏ ( ̃ )
(4)
( )
( ) ( )
( )
∏ ( )
∏ ( )̃ (5)
equation (5) shows that poles of f(s) becomes equal to zeros
of σfit(s). So by calculating the zeros of σfit(s) we get good
set of poles for fitting original function f(s).
2) Stage-2: residue identification:
In principle we could now calculate the residues for f(s)
directly from(8). However, a more accurate result is in
general obtained by solving the original problem with the
zeros of σ(s) as new poles an for f(s). This again gives an
over determined linear problem of form Ax = b where the
solution vector x contains the unknowns Cn, d and h.
B. Starting Poles
1) Significance of starting poles:
Successful application of vector fitting requires that the
linear problem can be solved with sufficient accuracy. In our
experience, difficulties may arise due to poor starting poles
in the following ways:
1) The linear problem (6) becomes ill-conditioned
if the starting Poles are real. This may result in an inaccurate
solution.
2) A large difference between the starting poles and
the correct poles may result in large variations in f(s) and
σ(s)f(s). Because a least squares approach is used when
solving linear problem, a poor fitting may result where these
functions are small.
The first problem is overcome by usage of complex
starting poles. The second problem is overcome by sensible
location of the starting poles, and by using the new poles as
starting poles in an iterative procedure.
2) Recommended procedure for starting poles:
The starting poles must be complex conjugate pair with
imaginary part β linearly distributed our the frequency of
interest .Each pair is chosen as follows:
an= -α + jβ and an+1= -α – jβ
Where
α = β/100.
In section III same method’s computational algorithm/code
has been explained with necessary subroutines and at last
validating Macromodel of single microstrip via obtain
using vector fitting .
III. COMPUTER CODE FOR RATIONAL FUNCTION
APPROXIMATION
Based on a previously developed technique (vector fitting),
a set of callable routines have been written in the Matlab
language. These routines makes rational approximation with
a common set of stable poles, automatic selection of initial
poles, passivity enforcement, and creation of an equivalent
electrical network that can be imported into circuit
simulator.
Figure. 2 Matlab subroutines
A rational function and its approximation and also
equivalent electrical network of interconnect from tabulated
data is obtain using this routines:
1. spavectfit.m
2. mtrxvectfit.m
3. passive.m
4. netgen.m
5. mainfit.m
For the purpose of matrix fitting, the routines have been
integrated within the routine mainfit.m. Fig.2 shows the
hierarchy of routine for matrix fitting. Several other
capabilities are built in, including plotting of results. The
following gives a brief description of the respective routines.
matlab script file
Y(s),s,N,weight(s)
mtrxvectfit.m
h(s),s,N,weight(s)
spavectfit.m
A,B,C,D,E
passive.m
A',B',C',D',E'
netgen.m
Electrical Network
RLC.out
circuit simulator

72
1) spavectfit.m
This routine is identical to the previous vectfit.m developed
in [2], except that the matrix problem in is now solved
using sparse arithmetic. This greatly reduces the memory
requirements and increases the computational efficiency for
matrices of big dimensions. All poles produced by the
routine are stable.
2) mtrxvectfit.m
This routine essentially stacks the columns of matrix into a
single column, and calls spavectfit.m. The symmetry of
matrix is taken advantage of by fitting only the upper
triangle of, which leads to a shorter column to be fitted by
spavectfit.m.
3) passive.m
This routine enforces passivity by the Simplistic Approach
in [4] at frequency samples provided in the input. These
frequency samples should in general be chosen to cover a
wider frequency range than the samples used in the fitting
because nonpassivity may occur outside the fitting range.
Upon output, modified matrices are produced if
nonpassivity was detected.
4. netgen.m
This routine takes as input the rational approximation of as
given by matrices A,B ,C ,D ,E and calculates an equivalent
network. The equivalent network is written to file (RLC.out)
as branch cards for circuit simulator and then this file is
imported directly in circuit simulator.
5. mainfit.m
This routine integrates the above routines into a complete
procedure for Matrix Fitting. It takes as input the frequency
domain data, does the rational function approximation,
enforces passivity, and generates an electrical network
which can be included in circuit simulator.
IV. MACROMODEL OF SINGLE MICROSTRIP VIA FROM
TABULATED DATA USING VECTOR FITTING.
Single microstrip via design in CST is shown in fig.3. Single
microstrip via is interconnect used to connect discrete
component on multilayer PCB. Y parameter of via is obtain
from CST by 3 D simulation used as tabulated data.
Figure.3 : Single microstrip via
After getting tabulated data vector fitting algorithm is
applied on this data. Vector fitting algorithm generate state
space model of via from this tabulated data.
 Comparison of Y parameter of state space model
obtain using vector fitting and Y parameter obtain through 3
D simulation in CST
Figure.4 : Y parameter of via obtain through 3 D simulation
Figure.5: Y parameter of state space model develop using
vector fitting from tabulated data.
RMS error obtain in this fitted model obtain using vector
fitting is 1.059e-05.
 Comparison of Transient simulation
Figure.6 : transient simulation result of via from CST

73
Figure.7 : transient simulation result of develop macromodel
in HSPICE
V. CONCLUSION
From the comparison of transient simulation of develop
macromodel in HSPICE and in CST we concluded that
the macromodel of single via develop using vector is
completely characterized the via behavior.
REFERENCES
[1] IEEE microwave magazine ‘High Speed Interconnect
‘ by Ramachandra Achar Augest,2011
[2] IEEE Transactions on Power Delivery, Vol. 14, NO. 3,
July 1999,’ rational approximation of frequency domain
responses by vector fitting’, by Bjorn Gustavsen and
Adam Semlyen.
[3] IEEE transactions on power delivery, vol. 17, no. 4,
october 2002, ‘Computer Code for Rational
Approximation of Frequency Dependent Admittance
Matrices’ Bjorn Gustavsen, Member, IEEE
[4] IEEE transactions on very large scale integration (vlsi)
systems, vol. 13, no. 7, july 2005 ‘Global Passivity
Enforcement Algorithm for Macromodels of
Interconnect Subnetworks Characterized by Tabulated
Data’ Dharmendra Saraswat, Student Member, IEEE,
Ramachandra Achar, Senior Member, IEEE, and
Michel S. Nakhla, Fellow, IEEE
[5] IEEE transactions on electromagnetic compatibility,
vol. 45, no. 3, august 2003 ‘SPICE Equivalent Circuits
of Frequency-Domain Responses’ Giulio Antonini,
Member, IEEE

Macromodel of High Speed Interconnect using Vector Fitting Algorithm

More Related Content

What's hot(20)

Viewers also liked(16)

Similar to Macromodel of High Speed Interconnect using Vector Fitting Algorithm(20)

More from ijsrd.com(20)

Recently uploaded(20)

Macromodel of High Speed Interconnect using Vector Fitting Algorithm