Calculation of

Frequency-Dependent
Impedance for
Conductors of
Rectangular
Cross Section
Alexander W. Barr
AMP Incorporated

ABSTRACT
Knowledge of the resistance and self-inductance of a
contact as well as the mutual inductance between
contacts is often necessary in determining the highspeed performance of connectors. Many of these connectors contain pins with rectangular cross sections.
Such geometries cannot be easily handled through
analytical techniques. This paper presents a numerical technique where the conductor cross sections are
broken into subconductors and the frequency-dependent current amplitude in each subconductor is found
by solving a system of equations relating the resistances, self-inductances, and mutual inductances of
all the subconductors. Once the currents in the sub
conductors are known, they are used to compute the
frequency-dependent resistance and self-inductance
of an individual conductor and the effective resistance and inductance of a conductor pair. Also the
currents of the subconductors are graphed to give a
discretized representation of the current density in
the conductors.

INTRODUCTION
Spiraling advances in electronic technology are placing an
ever increasing burden on connector manufacturers. With
computer clock frequencies doubling every few years and

semiconductor manufacturers fabricating chips requiring

hundreds of interconnects, systems packaging becomes critical. Customers want to produce densely packaged systems
to carry high-speed signals at low cost.
Several electrical challenges are presented by this desire.
First, high-speed signals usually contain pulses with fast rising and falling edges. Critical to the production of these fast
transitions is the high-frequency content of the pulse. Unfortunately, risetime degradation often occurs because these
high-frequency components are easily attenuated.
Impedance-matched interconnections can be used to minimize the possibility of compromising these edge rates.
Second, the significant voltage gradients attendant with fast
risetimes couple effectively with neighboring lines in the
form of crosstalk. Crosstalk can cause distortion of the original signal and possibly lead to false triggering of the connected circuitry. Dense packaging only tends to exacerbate
this problem. System designs need to be reviewed in terms
of their prospective susceptibility to crosstalk
problems.
Third, high-speed signals travel short distances between
voltage transitions. If interconnections are long, or of dissimilar lengths, propagation delays can affect signal synchronization and system integrity. Propagation delays need

Copyright 2004 by Tyco Electronics Corporation. All rights reserved.

AMP Journal of Technology Vol. 1 November, 1991

A.W. Barr

91

to be minimized and signal path lengths should be .

consistent.
To assess the impact of interconnections on these signal
integrity issues, it is necessary to compute the frequencydependent impedance of the interconnects. In the case of
circular-shaped wires or pins, this task can be accomplished
through analytical techniques. However, many connectors
contain pins or contacts of rectangular cross section. The
behavior of such conductors can only be accurately examined through use of numerical methods.
This paper describes a numerical technique for computing
the frequency-dependent resistance and inductance of conductors with rectangular cross sections. The cases of both a
single isolated conductor and two conductors experiencing
mutual coupling are addressed: The conductors are divided
into subconductors and the frequency-dependent current
amplitude in each subconductor is found by solving a system of equations relating the resistances and self-inductances of, and mutual inductances between, the
subconductors. Once the currents in the subconductors are
known, they are used to compute the frequency-dependent
resistance and self-inductance of each conductor and the
effective resistance and inductance of a conductor pair. Furthermore, the current amplitudes of the subconductors can
be graphed to give a discretized visual representation of the
current density in the conductors.

FREQUENCY-DEPENDENT IMPEDANCE FOR

A SINGLE CONDUCTOR
Before computing the impedance of an individual conductor
of rectangular cross section, it is instructive to understand
the motivation behind pursuing the announced technique. It
is well known that at dc, the current in a conductor is evenly
distributed across the cross section. As frequency increases
the current tends to crowd toward the outside of the conductor, until at high frequencies it is almost completely confined to a thin skin near the surface. If the conductors cross
section is subdivided into much smaller sections, the current
density in each subconductor will be almost uniform across
the subconductors cross section. By assuming that these
subconductor current densities are indeed uniform, the subconductors can be treated as circuit components. Once a circuit model representation of the original conductor has been
constructed from these components, the subconductor currents can be computed and the impedance of the conductor
can be found.
Figure 1 shows a rectangular post of width W, thickness T,
and length [, being driven by a sinusoidal generator operating at angular frequency co= 2rcf. If this post is divided into
Nw subconductors across the width and N T subconductors
across the thickness, there will be N = NwNT total subconductors. Each of these subconductors will have a width, w =
W/N w, and a thickness, t = T/NT.
To find the resistance and self-inductance for this post at the
frequency of interest, the resistance and inductance for the

92

A.W. Barr

Figure 1. Single conducting post driven by a sinusoidal

generator.
individual subconductors must first be computed. If the conductivity s for the post is known, then the resistance of a
subconductor will be

If (! is large compared to w and t (at least five times greater),

Keiser1 has shown that the self-inductance, L, for the subconductors will be

if [], w, and t are in meters. Similarly, Grover2 has shown

that the mutual inductance between two posts is ostensibly
the same as between two filaments located at the posts centers. Therefore, the mutual inductance between the subconductor in the ith row and jth column and the subconductor in
the mth row and nth column will be

where dij,mn is the distance between the center axes of the

subconductors, [ is the length of both subconductors, and
dij,mn and L are in meters. Note that this formula exhibits
some inaccuracy when the subconductors are close together
(distances on the order of the cross-sectional dimensions).
However, it has been found that the minor errors this equation sometimes introduces have an insignificant impact on
the frequency-dependent impedance computation.

AMP Journal of Technology Vol. 1 November, 1991

Once calculations have been completed to find the resistance and self-inductance of the subconductors as well as
the mutual inductances between all the subconductors, a
system of equations can be constructed to determine the
subconductor currents.
The laws of electromagnetic dictate that the voltage drop
across the length of a conductor is the same regardless of
where, in the conductor cross section, the current travels.

Therefore, if the voltage drop across the post is V, the drop

across each of the subconductors is also V. Ohms law says
that the voltage drop will be equal to the impedance multiplied by the current. In this case, the impedance is a combination of the resistance and self-inductance of the given
subconductor as well as the mutual inductance between the
chosen subconductor and each of the other subconductors.
For the subconductor in the ith row and jth column, this voltage will be

where v = 2nf. A similar expression can be developed for each of the subconductors, creating a system of N equations in N
unknowns where the unknowns are the subconductor currents:

This equation is just a matrix expression of Ohms law.

Since the voltage vector V and the elements of the
impedance matrix are already known, the system can be
solved for the current vector.

Consequently, the impedance of the whole post is

Solving this matrix involves finding the inverse of the

impedance matrix and premultiplying it by the voltage vector. Numerous ways exist to find the inverse of square
matrices. However, matrix size, conditioning of the elements, desired accuracy, and required speed of solution all
influence the choice of an algorithm. Because of its efficiency and resilience to roundoff errors,3 LU Decomposition
was used to invert the impedance matrix. This method
involves decomposing the impedance matrix into the product of a lower triangular matrix and an upper triangular
matrix. Substitution can be easily used to solve the resulting
two systems of equations and find the inverse of the original
matrix. More discussion of this method can be found in the
references.

where Rpost and Lpost are the resistance and self-inductance

of the post at the frequency f.

Once the current vector is known, the currents in each of the

subconductors have been determined. Since these sub-currents all flow in parallel, the total current through the post is

AMP Journal of Technology Vol. 1 November, 199 I

To illustrate the development of the skin effect with rising

frequency, the subconductor currents can be plotted relative
to their position in the post to give a discrete approximation
to the cross-sectional current density. Such plots are shown
later in this paper.

FREQUENCY DEPENDENT IMPEDANCE

FOR TWO CONDUCTORS CARRYING EQUAL
CURRENTS
Two posts, each carrying current I, and connected in series
to a generator oscillating at frequency f, are illustrated in
Figure 2. Post 1 is T1 thick, W1 wide, and ~ long. Post 2 is
T2 thick, W2 wide, and 8 long. If the posts are divided into
across their widths and N 1T and
N 1w,.,and N2w subconductors
,
N2T subconductors across their thicknesses, the total number
of subconductors in each will be N1 = N1TN 1w and N2 =
N2TN2W, for post 1 and post 2, respectively. The widths and
thicknesses for the subconductors in each post will be

A.W. Barr

93

inductance between subconductors. However, care must be

taken to consider the directions of the currents in the two
subconductors. If the currents are in the same direction, the
mutual inductance will be positive; if the currents are in
opposite directions, the mutual inductance will be negative.
Consequently, the mutual inductances between subconductors in the same post will always be positive; mutual inductances between subconductors in different posts will be
positive or negative depending on the direction of current.

Figure 2. Two conducting posts in series driven by a

sinusoidal generator.

As in the calculation for the single post, equations(1) and

(2) are used to compute the resistance and self-inductance of
each subconductor. Equation (3) also yields the mutual

Once the resistance and inductance values were determined

for the various subconductors in the single post, equation (4)
was constructed expressing the voltage drop across a given
subconductor. A somewhat different approach is taken this
time. It was initially assumed that the posts were connected
in series so that the currents would be the same in each.
Because the voltage V of the generator is dropped across the
two posts together, any voltage drop equations that are constructed must include the drop across a subconductor in post
1 and a subconductor in post 2. It is only by choosing these
pairs of subconductors-one from each postthat equations
relating to the generator voltage V can be constructed. Consequently, a subscripting system is adopted using three integers. The first subscript designates the post, the second
designates the row of the subconductor, and the third designates the column of the subconductor. Using this system, the
voltage drop across the subconductor in the i th row and jth
column of post 1 and the subconductor in the mth row and
nth column of post 2 has the form:

where R1 and R2 are the resistances for the subconductors in posts 1 and 2, and L1 and L2 are the self-inductances of subconductors in post 1 and post 2.

94

A.W. Barr

AMP Journal of Technology Vol. 1 November, 1991

Combining the impedance terms for the same subconductor

currents from equations (9) and (10) generates a single
impedance coefficient for each subconductor current. By
choosing various pairs of subconductors-one from post 1
and one from post 2a system of equations can be constructed. Unlike the previous case with a single post, a total
voltage drop equal to V is not the only boundary condition
to the problem. To force a solution for the subconductor currents which insures the total current in post 1 is the same as
in post 2, one of the equations in the system must specify
that the difference between the sum of the subconductor currents in post 1 and the sum of the subconductor currents in
post 2 is zero:

LU Decomposition is again an advantageous method for

inverting the matrix to find the subconductor currents. Once
these currents have been determined, a method similar to the
single-post problem can be used to determine the impedance
of each post.

If the total number of subconductors in the system is given

as N = N1 + N2, then a system of equations containing N 1
voltage drop equations (each using a unique pair of subconductors) and the current boundary condition will determine
the unique solution for the subconductor currents.
As in the single-post problem, this system of equations can
again be expressed in matrix form. The first N 2 rows express
the drop across post 1, subconductorl, and post 2, subconductormn where m = 1,2, . . . ,N 2T and n = 1,2, . . . ,N 2W. Next
the drop across post 1, subconductorl2 and each of the subconductors in post 2 is used to generate as many as the next
N2 rows. In this manner the first N 1 rows of the matrix
are generated. The N th row contains equation (12) to enforce
the subconductor current boundary condition. Equation ( 13)
shows the matrix equation for the two-post problem.

across post 2, where V = V1 + V2. Next, the total current in

each post can be found in the same manner as for the single
post:

Substituting the subconductor currents into equations ( 10)

and (11 ) will yield the voltage drop across chosen subconductors in post 1 and post 2. Since the voltage drop across
any subconductor in the post is the same as the voltage drop
across the entire post, finding the voltage across one subconductor in each post supplies V1 across post I and V2

AMP Journal of Technology Vol. 1 November, 1991

A. W. Barr

95

Since the voltage and current across each of the posts is now
known, the impedance for each member of the coupled pair
can be found:

These impedances are the effective result of the resistance

and self-inductance of each post as well as the mutual
inductance between posts.
Comparing the impedance results for the two-post solution
with the self-impedance found for a single post allows one
to quantify the effective difference, at a particular frequency, that the mutual inductance has. Furthermore, graphing
the currents for the subconductors allows one to observe
how the mutual coupling between the posts, in conjunction
with the skin effect, alters their current densities. This task
will be undertaken in a later section.

VERIFICATION OF THE IMPEDANCE OF

A SINGLE POST
Attempting to measure the impedance of a single conductor
is difficult. The fixturing necessary to mount a device under
test (DUT) generally contributes some impedance of its own
to the measurement. Coupling, in the form of capacitances
and mutual inductances between the DUT and the fixturing,
is usually difficult to quantify. Consequently, it is often a
formidable task to extract the self-impedance of the conductor from the measured data. For these reasons a different
approach was taken toward verifying the post impedance.
Well-known formulas exist for computing the frequencydependent impedance of a round wire. If such a piece of
wire is compared to a square post with the same cross-sectional area, similar impedances can be expected at each frequency.
Since both conductors have the same cross-sectional area,
their resistance will be the same at dc and should remain
similar until the skin effect is well developed. As frequency
is increased and the current crowds toward the conductor
surface, it will distribute itself equally around the circumference of the wire. However, because the comers of the
square post are the farthest points from the center of the
conductor, the current in the post will tend to be disproportionately concentrated at the comers. For this reason one
would expect to see the resistance of the post increase more
quickly than that of the wire.
As the current crowds toward the surface of a conductor, the
mean distance between current elements increases. With

96

A.W. Barr

this increase in mean distance comes a decrease in magnetic

flux inside the conductor. As the magnetic flux decreases,
the self-inductance of the conductor will also decrease. The
same phenomenon responsible for increasing the conductors resistivity serves to decrease its self-inductance. If the
inductance is separated into two parts-internal inductance
and external inductancea qualitative prediction can be
developed for the inductance of the wire versus the post.
The external inductance is only related to the magnetic flux
present outside of the conductor. Because this flux is independent of the changing current density inside the conductor, it is also independent of frequency. Flux is determined
by the magnetic field distribution. As soon as one moves a
short distance from the surface of the post, the field lines
begin to assume the same circular orientation as is present
around the wire. Consequently, the external inductance of
the wire and post will be ostensibly the same and constant
over frequency.
The internal inductance, however, is strongly dependent on
the internal current distribution. Because the comers of the
post are geometrically farther from the interior of the post
than the circumference of the wire is from the interior of the
wire, the internal inductance of the post will be slightly
lower than the internal inductance of the wire. As the frequency increases, this tendency will become increasingly
noticeable. Eventually, the frequency will be high enough
that the current is so crowded into a thin skin around the
perimeter of the conductor, that the internal inductance is
virtually zero. As the frequency is increased from this point,
the self-inductance of the conductors will remain nearly
constant while maintaining a difference attributable to the
slight difference in their external inductances.
To test the algorithm and verify these expectations, a piece
of 20 AWG copper wire (diameter = 0.032 inch) was compared to a square post with the same cross-sectional area
(length of side= 0.0283 inch). Resistance and self-inductance values were computed for a 1-inch length from 1 kHz
to 100 MHz.
A three-part calculation was performed to find the
impedance of the wire. First, the frequency-dependent resistance of a wire has been shown by Ramo and Whinnery4 to
be

and r is the radius of the wire, L is the length of the wire, f is

the frequency of interest, s is the conductivity of copper,
and is the permeability of copper (the same as air).

AMP Journal of Technology Vol. 1 November, 1991

Second, Ramo and Whinnery4 show the internal inductance

of the wire at f to be

Both equations (18) and (19) are in terms of the Ber and Bei
Bessel functions where

ness (0.01415 inch) and twice the width (0.0566 inch) of the
square post. This rectangular post possesses the same crosssectional area as the wire and square post. The data for the
rectangular post are consistent with expectations. Both the
resistance and self-inductance should be slightly lower than
the square post. In the graphs the subscripts mean the following: w = wire, s = square post, and r = rectangular post.

Table 1

and Ber' and Bei' are their derivatives.

The third part of the calculation was to find the constant
value for the external inductance of the wire. Grover2 gives
the equation for the self-inductance of nonmagnetic round
wire at dc as

where all dimensions are in meters. At dc the internal inductance for a round wire is 50 nH/m or 1.27 nH/inch. Subtracting this value from equation (20) yields the external
inductance for the wire. Adding the external inductance
value to the result from equation (19) furnishes the selfinductance for the wire sample at the chosen frequency.
To compute the impedance of the post, a C program called
Post_L was written. Given T, W, ~, NT, Nw, and f, Post_L
computes the frequency-dependent impedance of a post
using the previously described method. As with any finite
approximation method, the greater the number of subconductors in the model, the better the approximation. For this
study, Post_L was written and executed on an MS-DOSbased machine. Memory limitations allowed the square post
to be divided into an array of subconductors no larger than
14 by 14. It will be seen that further subdivision is necessary
to maintain model accuracy at high frequencies.
Table 1 shows the impedance values for the 20 AWG wire
and 0.0283-inch-square post.
As expected the resistance and self-inductance values are
nearly identical below 50 kHz. This observation is anticipated because below 50 kHz, the skin effect has not yet developed significantly. Consequently, the resistances should be
the same while a-minor difference in self-inductance is
found due to the difference in cross-sectional geometry. In
the range from 50 kHz to 1 MHz the resistance of the post
rises above that for the wire. Across this same frequency
range, the self-inductance of the post decreases slightly
faster than the wire. Both results were predicted.
The graphs in Figures 3 and 4 illustrate the resistance and
self-inductance data for the wire and square post. Also
included are data for a rectangular post with half the thick-

AMP Journal of Technology Vol. 1 November, 1991

Figure 3. Graph of resistance versus frequency for 20

AWG wire, 0.0283-inch-square post, and 0.0142 x
0.0566-inch-rectangular post.

Above 1 MHz, the graphs show a disturbing trend in the

impedance data, particularly the resistance. The resistance
calculated for both posts drops off sharply from the well-

A.W. Barr

97

known df dependence displayed by the wire. Likewise,

below 1 MHz the self-inductance of the square post decreases faster than the wire. At 10 MHz the self-inductances for
the two are about equal. At about 100 MHz the self-inductance of the wire has dropped below that of the post. It
would have been expected that the post would maintain a
lower inductance than the wire through and beyond the
point that the skin effect becomes completely developed.

larly at the corners. This graph was made by representing

each subconductors current at its position in the cross section of the square post.

Figure 5. Surface plot of current density in 0.0283inch-square post.

Figure 4. Graph of self-inductance versus frequency for
20 AWG wire, 0.0283-inch-square post, and 0.0142 x
0.0566-inch-rectangular post.
The explanation for this behavior can be traced to the precision of the model. With the square post divided into 14 subconductors in each dimension, each subconductor was a post
about 0.0020-inch square. At 1 MHz, the skin depth in copper was calculated to be 0.0026 inch. At this frequency and
below, there is more than one subconductor spanning the
skin depth. However, as the frequency increases beyond this
point, the skin depth soon becomes more shallow than the
dimensions of one subconductor. This condition forces the
model to represent current as flowing deeper in the conductor than is actually true. Consequently, the 14 by 14 subconductor model becomes a poor approximation at higher
frequencies.
These inaccuracies can be alleviated by using more subconductors in the model. For instance, at 100 MHz the skin
depth in copper is 0.00026 inch, and a conductor 0.0283 inch square with one subconductor spanning the skin depth
would need a 109 by 109 array of subconductors. Computations of this magnitude can only be handled by large mainframe computers. Therefore, it can be seen that the
single-post algorithm performs quite well, but constructing
accurate high-frequency models requires extensive computational power.
A surface plot of the current density in the square post at
500 kHz is shown in Figure 5. It is readily seen how the current crowds toward the perimeter of the conductor, particu-

98

A.W. Barr

VERIFICATION OF THE IMPEDANCE OF

TWO POSTS EXPERIENCING MAGNETIC
COUPLING
Rather than attempting to measure the impedance of the
two-post structure shown in Figure 2, a comparison was
made with a known reliable calculation methodthe finite
element method. Finite element modeling has become an
accurate and accepted practice for numerically solving the
differential equations governing electromagnetic behavior.
The method attempts to approximate the geometry of interest with a mesh of triangles. To solve the problem, a system
of equations is constructed defining the appropriate differential equations behavior at the boundaries of each of the triangles in the mesh. Solving the matrix from this system
yields the solution to the problem.
To test the two-post impedance algorithm, the following
problem was analyzed. Two 0.025-inch-square posts, each
O. S-inch long, were spaced at an 0.050-inch centerline distance. The post material was assumed to be copper. The currents in the posts were identical but flowing in opposite
directions. The effective impedances of the posts (including
mutual inductances) were computed over the frequency
range from 1 kHz to 1 MHz, This frequency range was chosen because of the model inaccuracy problems experienced
in the single-post computation.
Ansoft Corporations MaxwellTM package was used for the
finite element modeling. A mesh containing more than 3000
triangles was generated for the geometry. Maxwells frequency-dependent magnetic solver EddyTM was used to
compute the frequency-dependent resistance and inductance

AMP Journal of Technology Vol. I November, 1991

of the two-post structure. Maxwell is a two-dimensional

field solver; it assumes that the length of the geometry is
infinite and supplies answers per unit length. To obtain the
actual resistance and inductance for the 0.5-inch posts,
Ansofts answers must be multiplied by the post length.
To implement the two-post algorithm, a program called 2
CONDuctor MUTual was written for an 80386-based
machine. Because of the memory limitations experienced
with the single-post solution, the program was written and
executed using MATLABTM by The Math Works Inc. MATLAB runs in protected mode on 386-class computers, allowing the user to circumvent DOS-imposed memory barriers.
Consequently, each of the two posts could be represented by
a 14 by 14 matrix of subconductors in 2CondMut.

This table indicates excellent agreement between Ansoft and

2CondMut. Differences between resistance values ranged
from virtually zero below 10 kHz to 6% at 1 MHz. Inductance values, on the other hand, showed a consistent difference of 2.8% to 3.4%. It is suspected that much of the
difference in the inductance values can be attributed to
Ansofts characteristic of assuming infinite lengths. Such an
approach does not account for end effects, i.e., the decrease
in inductance per unit length exhibited by the portions of the
conductors near the ends of the posts. Therefore, it is likely
that the Ansoft results are slightly inflated.

Table 2 displays the data collected using Ansoft and 2CondMut at each frequency, while Figures 6 and 7 display graphs
of the same values:
Table 2

Figure 7. Graph of inductance versus frequency for the

two-post problem as solved with Ansoft and 2CondMut.

Figure 8. Current density in 0.025-inch-square posts

on 0.050-inch centers at 500 kHz.
Figure 6. Graph of resistance versus frequency for the
two-post problem as solved with Ansoft and 2CondMut.

AMP Journal of Technology Vol. 1 November, 1991

Typical current densities of the two posts are graphed in

Figure 8. This plot is based on computations at a frequency
of 500 kHz. The current density in each post is highest at the

A.W. Barr

99

inside edges. This distribution is a result of the strong negative mutual inductance occurring between the posts. If the
currents had been in the same direction, the greatest density
would have been located at the outer edges. The figure also
illustrates the influence of the skin effectthe current concentrates at the post perimeter, particularly at the comers.
Clearly, the observed current density is dictated by the combined influence of the mutual inductance between the posts
and the skin effect.

CONCLUSION
Two methods have been described for computing the
impedance of conducting posts with rectangular cross sections. The first method, intended for computing the resistance and self-inductance of an isolated post, divides the
post into subconductors and constructs a matrix equation
based on the voltage drop across the post. The second
method analyzes the combined impedance of two posts
experiencing mutual coupling. This later technique is not
only based on a uniform voltage drop across each post, but
must also satisfy the requirement of providing identical current magnitudes in each post.
Comparison of the results of these techniques with other
benchmarks revealed accurate correlation. The expectations
for marginally higher resistance and slightly lower selfinductance were realized for the single post when compared
to a round wire of identical cross-sectional area. Computational accuracy was found to be sensitive to the number of
subconductors spanning the skin depth. Because this condition affects the frequency range over which a given size
model can be accurately used, a reduced frequency range
was applied to the two-post problem. Comparison between
Ansofts finite element solution for this problem and the
previously described algorithm revealed agreement within a
few percent across the frequency range. It was postulated
that some of the observed discrepancy may be due to Ansoft
assuming infinite lengths for the conductors and not
accounting for end effects.
Because of the generality of these calculations, it is interesting to note that no constraint exists barring the application
of these techniques to conductors of arbitrary cross sections.
Furthermore, the technique applied to the coupled post prob-

100

A.W. Barr

lem could be expanded to assess the impedance of systems

containing more than two conductors. However, the computational horsepower needed to tackle such problems with
higher frequencies or multiple conductors dictates the use of
work stations or mainframe computers.

REFERENCES
1. B.E. Keiser, Principles of Electromagnetic Compatibili[y (Artech House Inc., Dedaham, Massachusetts,
1979), p. 102, equation (5.4).
2. F.W. Grover, Inductance Calculations, Working Formulas, and Tables (Dover Publications, New York,
1946).
3. W.N. Press, B.P. Flannery, S.A. Teukolsky, and W.T.
Vetterling, Numerical Recipes in C, The Art of Scientific Computing (Cambridge University Press, New York,
1989), pp. 39-45.
4. S. Ramo, J.R. Whinnery, and T. Van Duzer, Fields and
Waves in Communication Electronics (John Wiley &
Sons, New York, 1984), pp. 178-182.
5. P. Lorrain and D.R. Corson, Electromagnetic Fields
and Wave.s (W. H. Freeman and Company, San Francisco, I 970).
6. W.T. Weeks, L.L. Wu, M.F. McAllister, and A. Singh,
Resistive and Inductive Skin Effect in Rectangular
Conductors, IBM J. Res. Dev. 23(6), 652-660(1979).
Alexander W. Barr is a Development Engineer in the
Technology Group at AMP Incorporated in Harrisburg,
Pennsylvania.
Mr. Barr received a B.A. in physics from Franklin & Marshall College in Lancaster, Pennsylvania, and is currently
pursuing an M.S. in electrical engineering from the Pennsylvania State University. Since joining AMP in 1984, he
has been responsible for analyzing the electrical performance of connectors, contacts, and related devices through
computer modeling, time-domain reflectometry, and
microwave measurements.

AMP Journal of Technology Vol. 1 November, 1991