黃埔學報第七十一期民國一○五年 131

WHAMPOA - An Interdisciplinary Journal 71(2016) 131-140

Analytical GMD Calculation for Inductions of Rectangular

Conductors

Che-Wei Su 1, Yung-Tao Liu 2

1
Department of Computer and Communication Engineering, Chienkuo Technology
University, Changhua 50094, Taiwan
2
Department of Electrical Engineering, R.O.C. Military Academy
Feng-Shan 83059, Taiwan

Abstract

In this paper, the theoretical basis and development of geometrical mean distance (GMD)
to determine the inductance formula for straight and parallel conducts of rectangular cross
section are reviewed. Sample calculations for the self and mutual inductance of dual- and
triple-geometric configurations are given by using the proposed formula of GMD theorem.

Keywords: geometrical mean distance, GMD, inductance, mutual inductance.

1. Introduction given to the skin effects. In either case,

expeditious solution of the complex
The improvements of semiconductor equations is possible only with the aid of
process technologies have enabled the computers.
design of on chip inductors with multilevel The purpose of this report is to review
interconnects [1-3]. The pre-computed the theoretical basis of the inductance
tables reported in Refs. [4 and 5] can be formula for straight, parallel conductors of
employed to calculate the geometrical mean rectangular section, illustrating the role of
distance (GMD) of two rectangular cross Maxwell’s GMD theorem. Detailed
sections with finite thickness. Self- and examples are presented for the single
mutual-inductance effects are importance conductor and for arrangements of two and
considerations in the design of electrical three conductors. The general inductance
circuits because energy storage in magnetic formula and applicable GMD equations
fields can produce voltage transients have been used with excellent results for
resulting in noise, feedback, and other applications involving printed circuit cables.
undesirable phenomenon. The problem of The inductance formula is accurate for the
calculating low-frequency current is solely range of frequencies where skin depth
one of geometry. For conductors carrying exceeds conductor thickness.
high-frequency current, the additional
complication of non-uniform current of
distribution requires that consideration be
132 黃埔學報 第七十一期 民國一○五年

2. Development of Inductance Equations t dI  I1

W2 = ∫  L12  I 2' dt = ∫ L12 I 2' dI =L12 I 2' I1 (2)
for Straight and Parallel Conductors 0
 dt  0

The energy of the system of circuit 1 and 2

Consider a conductor element in a is calculated by allowing I 1 in circuit 1 to be
circuit path, supplied by a source, in which it established first. The current in circuit 2 is
is desired to establish a current I 1 . As the then allowed to rise to I 2 while I 1 is constant.
current rises from zero it induces an The rise of current in circuit 1 from zero to
electromotive force opposing the current rise I 1 involves the storage of energy (L 1 I 1 2)/2 in
and this requires the source to supply energy the magnetic field. As I 2 is established,
if the current is to be maintained against the energy (L 2 I 2 2)/2 is supplied by the source 2
induced electromotive force (emf). The while source 1 supplies energy (L 12 I 2 I 1 )/2 to
power expanded in forcing the current maintain a constant current I 1 . The total
against the induce emf e = -L(dI/dt) is P = energy of the system circuit 1 and circuit 2
LI(dI/dt). The total energy supplied in can be expressed as
raising the current to the final value I 1 in the 1 1
WT = L1 I12 − L12 I 2 I1 + L2 I 22 . (3)
2 2
time interval t is
The sign of the second term is negative
t dI I1 1
W1 = ∫ LI dt = ∫ LIdI = LI12 . (1) if the induced emf is in such a direction that
0 dt 0 2
it aids the flow of current I 2 and
This energy is stored in the magnetic field
supplements the energy of source 2. The
and is available to the circuit when the parameter of L12 is termed the mutual
inductance M and defined as the additional
source is disconnected.
energy available, or required, from the
Consider next a conductor element in vector addition of the two magnetic fields
each of two circuits supplied by separate [6]. A mutual inductance of one henry
gives rise to an induced emf of one voltage
sources. If, while current I 1 is being
when the rate of change of the inducing
established in circuit 1, a current I’ 2 is current 1 A/s. If the emf induced in circuit 1
maintained in circuit 2, which is linked by by a current changing at rate of 1 A/s in
circuit 2 is equal to e, the same emf e is
magnetic field interaction with circuit 1,
induced in circuit 2 when current in circuit 1
then during the rise of I 1 an emf e = changes at the rate 1 A/s. The mutual
-L 12 (dI/dt) is induced in circuit 2. The inductance can also be regarded as the
number of flux linkages with circuit 1 due to
energy required to force the current I’ 2
unit current in circuit 2. The reverse is also
against this emf is equal to true. Self-inductance is simply a special case
 Analytical GMD Calculation for Inductions of Rectangular Conductors 133

µ0 I 2 2
of mutual inductance, which will shortly
Aiz =
4πa 2
(a − r ) (8)
become apparent.
The magnetic flux φ, needed in Eq.(1) and
to evaluate the inductance coefficients, can
µ0 I r
be obtained from the magnetic vector Aoz = − ln , (9)
2π a
potential A where φ = ∫ A ⋅ dr and r is the For the case of the field of uniformly
vector of a spherical radius . The equation B distributed current in a straight conductor of
= curl A relates the vector potential A and a
any section of area S, the section may be
magnetic field B. In the case of a straight
rod of circular cross section (radius = a) divided into elements ds, each carrying
carrying uniformly distributed current (J of current Ids/S. Equation (5) may be written
magnitude I /(πa2)) and with permeability of µ0  Ids 
free space µ 0 , the use of the equation curl B
A=C−
2π
∑  S 
 ln r (10)

=µ 0 J results in the circumferential magnetic where r is the distance between each

flux density (external to the rod)
element ds and any point P where A is to be
µI
Boφ = 0 , r > a. (4)
2πr calculated. Equation (10) may also be
For this geometry, the vector potential has written as
only a component parallel to the current
µ0 I
2πS ∫
flow. The equation B = curl A reduce to B φ A=C− ln rds. (11)
= -dA z /dr. Integration of this yields Computation is greatly facilitated if the
µI
Aoz = C − 0 ln r, . (5) section S is divided into small equal
2π
elements (ds = S/n), n in number, and if the
where C is constant of integration, inside the distance from each element to the point P is
measured as r 1 , r 2 …r n . Therefore
rod
µI n  S
µ0 Ir A = C − 0 ∑  ln ri 
Biφ = , r<a (6) 2πS i =1  n
2πa 2

and
µ0 I  ln r1 + ln r2 + ... + ln rn 
=C−  . (12)
µ Ir 2 2π  n 
Aiz = C − 0 2 . (7)
4πa
Let R equal the geometric mean of r 1 , r 2 ,…
By requiring the vector potential be
r n . Then r 1 r 2 …r n = Rn and
continuous at the periphery of the rod, and ln r1 + ln r2 + ... + ln rn = ln (r1r2 ...rn )
by arbitrarily assigning a value of zero to the ( )
= ln R n = n ln R (13)

magnitude of vector potentials inside (A iz ) Equation (12) may now be rewritten as

µ0 I
and outside (A oz ) become A=C− ln R. (14)
2π
134 黃埔學報 第七十一期 民國一○五年

If the n subdivisions are infinitely small, R Since R 12 , R 11 , and R 22 are the geometric

can be defined by the equation mean distances, R 12 of area S 1 from area S 2 ,

1
S∫
ln R = ln rdS . (15) R 11 of area S 1 from itself, and R 22 of area S 2

from itself. The logarithm of each of these

Since R is the geometric mean of all possible
r distance, it is called the GMD between the distances is termed the logarithmic mean
point P and the conductor area S. Utilizing distance. From Eq. (3),the bracket portion of
Eq. (1) and these forms for the vector
Eq. (16) can be set equal to L/2 and the
potential, Gray [7] writes Eq. (3) in the form
general inductance formula of Snow [8]
 2 
WT = I  ∫B2 ∫B1 ln r12dS1dS2 
2

 S1S 2 results for conductors of length l, assuming

uniform current distribution in both

 1 
+ I  − 2 ∫B1 ∫B1 ln r11′ dS1dS1′ 
2

 S1 conductors comprising the return circuit,

 1  L 1
+ I 2  − 2 ∫ ∫ ln r22′ dS 2 dS 2′ . (16) = − 2 ∫ ∫ ln r11′ dS1dS1′
 S2 B2 B2
 l S1 B1 B1
1
where B 1 and B 2 are two long straight −
S 22 ∫ ∫
B2 B2
ln r22′ dS 2 dS 2′

parallel wires of any form of cross section 4

carrying current in opposing directions, r 12
+
S1S 2 ∫ ∫
B2 B1
ln r12 dS1dS 2 . (18)

is the distance between a filament in each of The first two terms are referred to as L 1 and

wires 1 and 2, r11′ is the distance between L 2 , and called the self inductance of

two filaments in wire 1 (and conductors 1 and 2. The last term is called
correspondingly for r22′ in wire 2), and dS1′ the mutual inductance of conductors 1 and 2.

and dS2′ are differential areas in wires 1 Snow also writes total inductance as
L L1 L2 M
and 2. The permeability in cgs units of the = + − 2 12 , (19)
l l l l
wires and intervening space are taken as
and
unity.
L
= 4 ln R12 − ln R11 − ln R22 . (20)
The following expressions can be l

written: Consequently, the determination of circuit

1
ln R12 =
S1S 2 ∫ ∫
B2 B1
ln r12 dS1dS 2 inductance for straight, parallel conductors

1 of any section carrying uniformly distributed

ln R11 = 2
S1 ∫ ∫
B1 B1
ln r11′ dS1dS1′ (17)
current in a closed path depends only on the
1
ln R22 =
S22 ∫ ∫
B2 B2
ln r22′ dS2 dS2′ . GMD of the cross-sectional areas from
 Analytical GMD Calculation for Inductions of Rectangular Conductors 135

  2l  R 
themselves and one another. M = 2 × 102 l ln  + − 1 . (25)
X=0 dx X=L
+I
 R l 

r R When the medium between the filaments has

X1 = 0 dx1 X1 = L
-I
a relative permeability of one (H will be in

Fig. 1 Current-Carrying filaments.

nH).

When the separation of two conductors

Neumann’s integral [9] can be used to
of arbitrary cross section is large relative to
calculate the mutual inductance between the
the size of their relative cross-sectional
two parallel wire filaments shown in Fig. 1.
dimensions, the mutual inductance of the
The integral is developed from Biot and
combination will be essentially the same as
Savart’s law for the magnetic field intensity
that of two filaments along their axes.
produced at an external point by a current I
Typically, however, the cross sections will
in an element ds of a circuit. The integral is
ds1ds2 cos ε be too large to justify filament substitution.
M = ∫∫ . (21)
r
Each conductor must be divided into an
After the substitution of ds 1 = dx 1 , ds 2 = dx, infinite number of filaments and integration
cos ε = 1 r = R 2 + (x − x1 )
2
and , the will accomplish an average of all possible
integral becomes pair combinations. The change to the basic
l l dx
M = ∫ dx1 ∫ . (22) formula is limited to R, which becomes a
R + ( x − x1 )
0 0 2 2

GMD function as opposed to a simple

Curtis [9] shows the details of integration, centerline distance. Eq. (25) still applies, but

which results in the mutual inductance R will be replaced by a formidable

formula expression best solved independently. The

  l 2 + R2  
 − l + R + R .
2 2 mutual inductance of the two conductions
M = 2 l ln
  R 
 l l
 will be equal to that of two filaments

(23) separated by a distance corresponding to the

When l >> R, this equation is frequently GMD of the two cross sections.

simplified as As is evident in Eq. (18), the self

  2l  R 
M ≈ 2 l ln  + − 1 . (24) inductance of a conductor of any cross
 R l 
section is equal to the sum of the mutual
This is an expression in cgs units. The MKS
inductances of all filament pairs of the
equivalent is
136 黃埔學報 第七十一期 民國一○五年

section. As before, the self inductance of a mean distance formula for two rectangles
conductor of any section is equal to the arbitrarily located in a quadrant, both by the
mutual inductance of two filaments multiple integration method and by the use
separated by a distance corresponding to the of complex variables. The multiple-integral
GMD of the cross section from itself. For equation solved by the two techniques is
T + T1 + T2 W1 + D +W2
(W1T1W2T2 ) ln R12 = ∫0 ∫0 ∫T +T
T1 W1
this reason, self inductance is often
1
∫
W1 + D

described as a special case of mutual

inductance.
ln (x2 − x1 )2 + ( y2 − y1 )2 dx2dy2dx1dy1, (29)

The calculation of inductance, then, is where dx 1 dy 1 and dx 2 dy 2 are differential

indeed a problem of geometry. Equation (25) areas in the respective conductors. The

can be expressed as follows: conductor arrangement is shown in Fig. 2.

  2l  R 
L or M = 2 × 102 l ln  + − 1 . (26) Higgins’ solution for the logarithmic mean
 R l 
distance, which is the logarithm of the GMD
The expression for the total inductance of
R 12 , is
two parallel nonmagnetic conductors (1 and
(W1T1W2T2 ) ln R12 = − 25 (W1T1W2T2 )
2) carrying uniformly distributed current in 12

opposing directions is −
1 4 4
∑∑ (− 1)i + j k (Ai , B j ), (30)
24 i =1 j =1
LT = L1 + L2 − M 12 − M 21 . (27)

When the conductors are the same size and where

shape, Eq. (27) simplified to ( )

k (Ai , B j ) = Ai4 − 6 Ai2 B 2j + B 4j ln Ai2 + B 2j
LT = 2 L1 −2 M 12 . (28)
B  B 
− Ai3 B1j tan −1  j  + Ai1B tan −1  j , (31)
Y Y’  Ai   Ai 
W1 D W2
Y2 T1
Con. 2
A1 = W1 + D + W2 ; A2 = D + W2 ; A1 = D ; A2 = W1 + D
R12

X’
Y1 T (32)
X
x1 x2
Con. 1
and
Figure 2. General arrangement of rectangular B1 = T1 + T + T2 ; B2 = T + T2 ; B1 = T ; B2 = T1 + T .
conductors.
(33)

3. GMD of Rectangular Conductors The purpose of the second coordinate

system (x’, y’) in figure 2 is to indicate that

Higgins [10,11] has derived the logarithmic D and T are both positive only when
 Analytical GMD Calculation for Inductions of Rectangular Conductors 137

Case GMD (mm) Cal. (nH) Measured (nH)

conductors 1 and 2 are offset as shown. The (a) 1.693 16.4 16.8
(b) 2.442 30.9 29.7
two values locate only the left and bottom (c) 4.697 56.8 54.8
(d) 7.244 73.0 72.5
edges of conductor 2 with respect to the

right and top edges of conductor 1. For the 4. Inductance Calculations for Two
characteristic over/under orientation of Parallel Rectangular Conductors

equal-size rectangular conductors, D = -W 1

Figure 3 shows in cross section an
and T is some positive value. For the
arrangement of four simple return circuits,
edge-to-edge orientation of equal-size
each 100 mm long, which were fabricated
conductors, T = T 1 and D is some positive
for the purpose of comparing measures and
value.
calculated for the purpose of comparing
The logarithmic mean distance of a
measures and calculated inductances. All
single rectangle relative to itself can be
conductors were 0.034 mm (0.0014 inch)-
easily found from the above equations by
thick copper and were printed on 0.770 mm
letting W 1 = W 2 , T 1 = T 2 , D = -W 1 , and T =
(0.032 inch)-thick Fiberglass-reinforced
-T 1 . Accordingly, A 1 = A 3 = W1 , A 2 = A 4 = 0,
epoxy. Table 1 lists the GMD for each
B 1 = B 3 = T 1 , and B 2 = B 4 = 0. The result is
conductor pair, calculated for Eq. (30), the
log R11 = −
25 1
2 2
12 6W1 T1
[(
W14 − 6W12T12 + T14 )
total inductance, calculated from Eqs. (26)
T1
ln W12 + T12 − 4W13T1 tan −1 and (27), and the total inductance measured
W1
W1  on a 100 kHz Impedance meter. Agreement
− 4W1T13 tan −1 − 4W14 ln W1 − 4T14 ln T1  . (34)
T1  between calculated and measured inductance
A simple but accurate approximation for the
is within 4%. When Eq. (19) is used to
GMD of a single rectangle is R 11 =
0.2235(W 1 + T 1 ). calculate the inductance, the values for cases

(a) through (d) are 16.6, 31.4, 58.2 and 75.4

4.97 4.95
0.77 mm 0.054 nH, respectively. While agreement with the
2.44
(a) (b) measured values is still satisfactory, more
4.80 2.0 2.64
4.39
accurate results can generally be obtained
4.94
(c) (d) with Eq. (25). Experience has shown that

Figure 3. Dual-conductor arrangements. inductance values calculate from Eq. (6) for

various arrangements of two-conductor

Table 1 Calculated vs. measured inductance
138 黃埔學報 第七十一期 民國一○五年

return circuits of rectangular cross section nH

agree with measured values within 10%. The GMD 12 = GMD 23 = 2.430 mm

inductance and GMD equations have been The inductance of many different triple-

used principally for circuit less than 1 m conductor arrangements can be calculated

long, conductors less than 15 mm wide for by this method. As another example, if the

dielectric less than 1.57 mm thick. center conductor in Fig. 4 is half as wide,

but a symmetrical arrangement is retained,

4.97 I1 = -I/2
0.25 0.034 the following calculated values are obtained:

GMD 1, 3 = 2.239 mm
I2 = +I I3 = -I/2 GMD 2 = 1.124 mm L T = 16.9 nH
Conductor length = 100.0 mm
GMD 12 = 2095 mm
Figure 4. Three-conductor circuit. GMD 13 = 2095 mm
Eq. (36) and applicable GMD equations give
5. Inductance Calculations for Three consistently good results for shielded flat
Rectangular Conductors cables, and agreement with measured values
of inductance is usually within 10%. The
The instantaneous energy storage in a equations can also be used to calculate the
inductance of the more unusual case of a
network of n loops is given by Chen [12] as
nonsymmetrical arrangement of three
1 1 n n
WT = LT I 2 = ∑∑ Lkh ik ih . (35) conductors.
2 2 k =1 h

For the three-conductor circuit shown in Fig. 6. Conclusions

4, n = 3. When Eq. (35) is solved for total
Accurate analytical formulas for
inductance L T , the expression is obtaining the GMD of multilevel conductors
LT =
1
(L1 + L3 ) + L2 + 1 M 13 − M 12 − M 23 . for the Greenhouse method have been
4 2
presented. The mutual inductances of
(36) multilevel conductors calculated using the
Note that when subscripts k and h are proposed formulas. Proposed expressions
are scalable and e fficien
t for calculating
unequal, M kh (for mutual inductance) is used
mutual inductances of interconnects for 2D
in place of L kh . Eq. (26), (27), (30),and (36) inductors in VLSI circuits.
were used to calculate the following results:

GMD 1, 2, 3 = 2.239 mm

GMD 1, 3 = 2.630 mm L T = 9.1

 Analytical GMD Calculation for Inductions of Rectangular Conductors 139

of Complex logarithmic and Geometric

References Mean Distances’, AIEE Transactions,
[1] H. Y. D. Yang, “Design considerations of vol. 66, pp. 12-16, 1947.
differential inductors in COMS [11] W. H. Chen, Linear Network Design
technology for RFIC“, IEEE RFIC Symp. and Sythesis, McGraw-Hill Book Co.,
Dig., TX, pp. 449–452, June 2004. Inc., pp. 57, New York, 1964.
[2] A. Zolfaghari, A. Chan, and B. Razavi,
“Stacked inductors and trans- formers in
CMOS technology,” IEEE J. Solid-State
Circuits, vol. 36, no. 4, pp. 620–628,
April 2001.
[3] P. Park, C. S. Kim, M. Y. Park, S. D.
Kim, and H. K. Yu, “Variable inductance
multilayer inductor with MOSFET
switch control,” IEEE Electron Device
Lett., vol. 25, no. 3, pp. 144–146, March
2004.
[4] F. W. Grover, “The calculation of the
inductance of single-layer coils and
spirals wound with wire of large cross
section,,” Proc. the Institute of Radio
Engineers, vol. 17, no.11, pp.
2053–2063, Nov. 1929
[5] F. W. Grover, Inductance Calculations,
Van Nostrand, pp. 1-5, New York, 1946.
[6] J. M. Ham and G. R. Slemn, Scientific
Basic of Electric Engineer, Van
Nostrand, New York, 1946.
[7] A. GRAY, ‘Formulars for Computing
Capacitance and Inductance’, C. Snow,
Absolute Measurements in Electricity
and Magnetism, NBS Circular 544, pp.
40-41, 1954.
[8] H. L. Curtis, Electricity Measurements,
McGraw-Hill Book Co., Inc., pp. 138-
139, New York, 1937.
[9] T. J. Higgins, ‘Formulars for the
Geometric Mean Distance of
Rectangular Areas and of of line
Segments’, Journal of Physics, N.Y., vol.
14, pp. 188-195, 1943.
[10] T. J. Higgins, ‘Theory and Application
140 黃埔學報 第七十一期 民國一○五年

適用於長方形導體電感之解析 GMD 計算

蘇哲尉 1、劉永道 2
1
建國科技大學電腦與通訊工程系
2
陸軍軍官學校電機系

摘要

本文評述幾何平均距離 (GMD)之理論基礎和推演來確定直線和平行矩形截面導體
的電感公式。利用 GMD 定理公式提供自感和互感之二線和三線的導體幾何配置範例
計算。

關鍵詞：幾何平均距離, GMD、電感、互感.