2670 IEEE TRANSACTIONS ON MAGNETICS. VOL. 2 5 . NO 3.

MAY 10x9

The Optimal Form for Coreless Inductors

PAUL N . MURGATROYD

Abstract-A fundamental problem of inductor design is to obtain the of the inductance, of whatever value in a small range, was
greatest possible inductance with a piece of round wire. The optimal more important than the value itself. “Forwards” for-
design is a solenoid with many layers of circular turns, and the maxi-
mulas meet just this requirement. The second factor is not
mum inductance scales with the five-thirds power of the total length of
wire used. For practical reasons other design types, including solenoids historical, but fundamental. The “backwards” problem
and toroids, also are commonly used. These may have one, few, or has no unique solution. It is possible to achieve a required
many layers, and all have optimal forms with associated scaling rules. inductance value, even with a given type of wire, in any
By describing the class of coreless inductors in a common notation, a number of different ways. For example, in simple sole-
significant unification of design is achieved.
noids with circular turns, a few large turns, or many small
turns, or any combination in-between, will deliver the
I. INTRODUCTION value. Thus the trial-and-error process using a “for-
wards” formula can end, if inductance value is the sole
I N THE published literature, formulas for calculating in-
ductance tend to be of two main types. In handbook
collections for the engineer or technician, the formulas
objective, with all manner of different solutions which
may be more or less acceptable for other subsidiary rea-
are usually accurate only within a limited range of design sons. One such reason is the most economic use of wire.
parameters. However, the mathematical expressions usu- In “A Treatise on Electricity and Magnetism” James
ally are relatively short. In academic works, the opposite Clerk Maxwell [4] examined the following problem. “To
combination is found, and the formulas are more accurate find the form of a coil for which the coefficient of self-
in a wider range, but much longer. The largest collection, induction is a maximum, the total length and thickness of
mainly of the second type, was published in 1946 by the wire being given.” The solution appears in a chapter
Grover [ 11. With the increasing availability of computer headed Circular Currents, and perhaps it was obvious to
power to evaluate the complicated expressions, the classic Maxwell that a solenoid with many circular turns closely
geometries still attract detailed attention [2]. A common packed into many layers gives the best solution to the
feature of both main types is that they are “forwards” problem as posed. The conclusion of his analysis was that
formulas, meaning that the shape, size, and number of “if the channel in which the coil is wound has a square
turns are given, so the formula delivers the inductance transverse section, the mean diameter of the coil should
value. For the practical engineer or designer, these for- be 3.7 times the side of the square-section of the chan-
mulas are the wrong way round, because the given param- nel.”
eters will typically be a wire diameter, usually set by the From the point of view of a coil designer or manufac-
current rating, and a specified inductance value. There turer, Maxwell’s problem may appear somewhat aca-
may also be constraints on shape, cost, or external mag- demic. Such people have to make inductors with specified
netic field. For the practical person, shape, size, and num- values, using wire of sufficient sectional area to avoid
ber of turns are the end of the design, not the beginning, overheating at the specified current loading: making the
so the “forwards” type of formula can only be used in a maximum possible inductance from a fixed length of wire
sequential process of trial-and-error. Much more useful is not the standard problem. Consequently, few such
would be a “backwards” formula, that starts from the practical people have heard of the Brooks Inductor [5]
wire diameter and specified inductance, and delivers di- which gives a more accurate and simply described em-
rectly the required coil design. bodiment of the solution, and may be designed to deliver
It has been suggested [3] that the dominance of “for- a required inductance value in a direct and straightforward
wards” formulas in the literature is due to two main fac- way. This paucity of general awareness of the Brooks In-
tors. Firstly, the people who derived the accurate formu- ductor is unfortunate for several reasons, among which
las were, like Grover, mainly workers in Standards the ability to make the most economic use of wire is only
Laboratories. Their prime objective was to make calcu- one.
lable devices, fit to establish reliably the linkage between For various good technical reasons, many coreless in-
electrical standards and length standards. So the accuracy ductors deliberately are not made as many-layered sole-
noids. Coils with one layer, or just a few layers, are often
Manuscript received August 15, 1988; revised January 18, 1989. simpler to make and more easily cooled. If induced elec-
The author is with the Department of Electronic and Electrical Engi- tromagnetic interference is a problem, toroidal inductors
neering, Loughborough University of Technology, Loughborough, LEI 1
3TU, England. may be preferred to solenoids (even though they typically
IEEE Log Number 8926998. require twice as much wire for the same inductance value),

001 8-9464/89/0500-2670$01 .OO 0 1989 IEEE

MURGATROYD: OPTIMAL FORM FOR CORELESS INDUCTORS 261 I

because toroids effectively enclose their own magnetic This answer is larger than is obtainable with any single-
fields. To disperse heat, or to reduce interturn capaci- loop shapes, essentially because all the mutual couplings
tance, spaced-out windings may be preferred. All these between points of the wire are positive. This result finds
features give coil designs of lower efficiency, meaning application, for short lengths, in computations of loop in-
more wire is required to achieve the specified inductance. ductances [6] by the Method of Sections. However, it is
Nevertheless, as this paper will show, the “given piece not of practical interest because the inductance cannot be
of wire” or GPW criterion can be applied to all the well- measured or used in isolation from other conductors which
known classes of coreless inductor, and the resulting have inductances of the same order of magnitude.
forms provide the best possible starting point for designs The inductance of a single closed turn of wire depends
incorporating other desired features beyond efficient wire on the shape. When carrying a current, the mutual mag-
utilization. netic forces between all parts of the turn would alter the
For the designer, GPW forms of inductor have simple shape, if the material was sufficiently flexible, into a plane
scaling properties. For example, in single-layer devices, circle. For a single turn this is the GPW maximum shape,
the GPW maximum inductance scales with the three- with inductance given [7] for large k by
halves power of the GPW length, and the optimum num-
ber of turns scales with the square root of the GPW length. L
- = k {In k - 0 . 8 1 5 3 ) . (2)
A further most desirable feature is inductance stability, or Lo
relative immunity to manufacturing error. Provided the
The inductance of single-turn regular polygons is obtain-
total wire length is correct, it is possible to have shapes
able from similar formulas [ 11, for example
and numbers of turns with significant errors, yet the effect
on the GPW maximum inductance is small. These fea- L
tures again commend GPW forms as the best starting point - = k{ln k - 0.9997) (hexagon) (3)
for inductance design, for if later stages of the design pro- Lo
cess have to be iterative, the designer can do no better L
than start from a solution that is economic, simple, and - = k {In k - 1.2177) (square). (4)
stable. Lo
In general, the inductance of a single turn decreases as
11. NATURAL UNITSAND WIREINDUCTANCES the area A decreases, according to the Bashenoff formula
In this paper all the inductor designs are given in “nat- [I1
ural” or dimensionless units based on the diameter of the
“given piece of wire.” Let the GPW have length W and
external diameter d. Then the dimensionless wire length
is k = W/d, and k typically ranges from a few hundreds where 4 is a term dependent on shape. Some of the indi-
up to some thousands. vidual formulas given above are plotted, for comparison,
The choice of d as the external diameter, measured over in Fig. 1.
any insulation layer, may possibly seem inappropriate be-
cause the metal conductor diameter governs both the dc 111. SOLENOID
INDUCTORS
resistance and proximity losses. However, all GPW de- A general feature of inductance formulas, covering most
signs at some point require the closest possible packing designs with well-coupled turns, is that inductance value
together of the wires, so it is the external diameter that scales approximately with the mean turn diameter and the
sets the spacing. square of the number of turns. As a benchmark suppose
The natural unit of inductance is chosen to be Lo = then that the GPW is made into a single circular turn,
p0d/27r henries. This is typically a small unit, for ex- providing a simple inductor. Suppose next that the GPW
ample if d = 5 mm, Lo is 1 nH. The smallness presents is rewound into two turns, closely coupled together. This
no conceptual or practical difficulty. The objective of the change increases the inductance by four through the num-
design optimization becomes the ratio L I L O ,which may ber of turns, and decreases it by two through the mean
be typically many thousands. turn diameter, giving an overall gain of about two. Evi-
The “internal” self-inductance of the GPW, due to the dently if GPW is rewound into N turns, there is a gain of
magnetic field energy stored entirely within the wire if it about N over the single-turn inductance, except that as N
remains straight, is po W/8a henries. This is 0.25k in units becomes yet larger the quality of mutual coupling be-
of Lo, and such a term appears in the formulas for the tween turns must degrade, because a large number cannot
complete inductance in many different geometries. The all be in close proximity together. This process explains
“external” self-inductance of the GPW, assumed straight, the physical origin of the optimum number of turns in
is given by [6] GPW inductor designs.
The simple helix, with one layer of wire on a cylindri-
L
- = k{lnk + 0.393). (1)
cal former, is cheap and easy to make. For small values
Lo and high current ratings, the former may be withdrawn
2672 IEEE TRANSACTIONS ON MAGNETICS, VOL. 25. NO. 3, MAY 1989

105

10‘
0
2
\
J

“
W
c
e
U

103 Fig. 2 . The “Most Economical Coil Shape” among single-layer sole-
U noids.
-c
W

-
c
U

e
0
101

10 -T---

I I I
10 100 1000
Dimensionless Wire Length k=w/d

Fig. 1. Comparison of inductance formulas for straight wires and single

turns.

after winding, for the rigidity of the wire maintains the

shape. To maximize inductance in a GPW design, suc-
cessive turns must be in continuous contact, and such
windings can be analyzed [8] as cylindrical current sheets.
These designs are reviewed by Grover [ 11, who considers
a range of practical examples including the “Most Eco-
nomical Coil Shape” [ l , p. 1611 which is the GPW op-
timum. The solution, illustrated in Fig. 2, has the coil
length 0.408 times the turn diameter. This particular op-
timum design is typical of all GPW designs in that the
inductance maximum, when the number of turns is var-
ied, is very flat. If the length is made 0.39 times the di-
ameter, the inductance is only 17 parts in lo5 below the ‘“a-..._ (e)
maximum, and for 0.42 times the diameter it is only 7
parts in lo5 below the maximum. The optimum number Fig. 3 . Alternative forms of single-layer coils. (a) Clockspring, in rect-
angular section strip. (b) Flat, for printed circuit. (c) Pancake coil, in
of turns is 0.360k1/2and the maximum inductance is given solid round wire (plan). (d) Pancake coil, in solid round wire (section).
by L / L o = 0.661k3/2. The numerical coefficients here (e) Round wires on conical former.
provide a basis of comparison with all the other single-
layer inductor geometries, which all obey the same scal-
ing rules. used, with thick self-supporting copper strip, to make
Alternative forms of single-layer geometry are used for “pancake coils” in power electronics. Formulas are given
special purposes. “Clockspring” coils using rectangular by Grover [l]. Flat spirals usually with square turns,
section strip conductor, with section shown in Fig. 3(a), shown in plan in Fig. 3(b), are used in printed circuits.
have been used in radio-frequency work, sometimes in In the foregoing examples the turns are often well sepa-
closely coupled coaxial pairs so that tuning is possible by rated, either to reduce self-capacitance or to improve
changing the separation. Similar arrangements have been cooling. A GPW design requires compact spiral winding,
MURGATROY D: OPTIMAL FORM FOR CORELESS INDUCTORS 2613

as shown in plan and section in Fig. 3(c) and (d). Cal-

culation of the inductance of these coils [9] uses an equiv-
alent circuit approach. Each turn, with self-inductance Li,
has mutual inductance M?, with all the others, and all the
terms in the summation
N N N

L = C L ; + 2 i =Cl
i= 1
CM,,
j>i (6)
1000
are different. Using series approximations for the M,,the
GPW maximum single-layer solenoid inductance was cal-
culated to be L/Lo = 0.635k3/’ (k I 1000) while for
the flat spiral coil, as in Fig. 3(d), it was L/Lo =
0.631k3I2(k 1 500) with optimum turns N = 0.347k’12. z
YI
The similarity of these results, for two quite different ge- c
L
c
3

ometries, was not expected, and it prompted inductance c

0

calculations for conical coils, illustrated by example in 100 ti

E
Fig. 3(e). The solenoid and the spiral are the limiting 1
cases, with angle I9 = 0 and I9 = a / 2 , respectively. Cal- E

culations [9] showed there is no preferred intermediate

a
.-
c
0”
angle, and the GPW inductance falls smoothly through
the range with the k3/’ scaling maintained in all cases.
The single-layer GPW solenoid can be used, approxi-
mately, to indicate the possible gains of inductance if the
same wire is wound into two or more layers. Useful ad- I 00
5
10‘
vantage is expected, because the general level of mutual
Dimensionless Wire Length k:w/d
coupling must increase, though other properties of the
system, notably capacitance, may also change signifi- Fig. 4. Maximum inductance and optimal number of turns for air-cored
solenoids.
cantly. Starting from a GPW solenoid with N turns, let
the wire be rewound into N & turns in two layers, with
N / h turns in each layer. Both the average turn radius
and the axial length are thereby reduced by h,the coil where a is the mean radius of the turns and b is the self-
maintains the correct ratio of length to (average) turn ra- geometric-mean-distance of the transverse section, which
dius, N 2 is increased by 2, so the inductance is increased Maxwell called the “channel.” For a square channel the
by h.The argument obviously extends, and the limiting inductance was maximized [4] with the mean turn diam-
inductance ideally increases as the square root of the num- eter 3.7 times the side of the square. With a more elabo-
ber of layers, with the proviso that the buildup, or differ- rate starting formula due to Rosa [ 111 of the National Bu-
ence of radii between outermost and innermost layers, re- reau of Standards, Shawcross and Wells [12] corrected
mains small compared with the axial length. Detailed the ratio to about 3.0, but their graph suggests that ratios
calculations, using the turn-by-turn summation method of 2.5 and 3.5 give inductance only about 1 percent below
mentioned above, confirm these indications, provided the the peak. The last word, in calculations of this type which
dimensionless wire length k is large. The results, together do not treat the turns and layers individually, was by
with the optimum number of turns, are plotted in Fig. 4. Brooks [SI who found the ratio 2.967, noting that for all
It is noticeable, however, that as k is made smaller the practical purposes 3.0 was good enough, as it provided
GPW inductance scales more steeply than k3/* and this inductance only twenty parts in a million below the true
trend increases with the number of layers. There is no peak. The Brooks inductance formula is
simple relationship describing this change. The GPW
maximum inductance for a given k approaches a limit, the
top curve in Fig. 4, which is found by letting the coil axial
length equal the buildup. where c is the side of the square channel. In the dimen-
The first calculations of the GPW solenoid with sionless notation of the present paper LB/Lo = 0.318k5l3
“many” layers, by Gauss [lo] and Maxwell [4] started with optimum number of turns 0.235k2/3, and these ap-
from an approximate formula equivalent to pear as limiting lines labeled B in Fig. 4.
An alternative approach to the multilayer solenoid,
using the method of individual turns and (6), allows other
L = poaN2[ln 8a - 2 1
(7) channel shapes than squares or rectangles to be studied.
The results [3] suggest that hexagons, which can be made
2614 IEEE TRANSACTIONS ON MAGNETICS. VOL. 2 5 . NO. 3. MAY 1989

with regular packing of round wire, give more inductance

than the Brooks square by about 0.9 percent, and that cir-
cles, if such coils could be made, would give an improve-
ment of about 1 . 1 percent. These gains are essentially due
to the removal of wires near comers, where the mutual
coupling to the whole section is weakest. Conversely, it
was found that equilateral triangle sections are inferior to
the Brooks square by about 3 percent. For all these cases
it was found [3] once again that for fixed wire length the
inductance value is remarkably insensitive to the number
of turns, near the GPW optimum designs.

IV. TOROIDAL
INDUCTORS
The principal reason for choosing a toroidal inductor is
the very small external magnetic field. Leakage can be a
problem with solenoids, especially in power electronics
circuits where it can cause interference in nearby equip-
ment. Whereas the external field of a solenoid drops away
approximately as the inverse cube of the distance, that of
a toroid falls much more steeply, and can be made vir-
tually negligible if a reverse loop is used to counter the
one-turn of the winding progression. If economic use of
wire is the first objective of design, then air-cored toroids
need be considered no further. Solenoids have a signifi- Fig. 5 . Two alternative designs of single-layer toroid using the same length
cant advantage, mainly resulting from their use (or per- of wire.
haps it should be called pollution) of the surrounding
space. However, if the desirable properties of toroids are
needed, only the best possible toroids should be used, and
a study of GPW designs is required.
Textbook formulas for the inductance of toroids nor-
mally assume a single layer of fine wires, justifying a cur-
rent-sheet approximation. “Forwards” formulas are eas-
ily derived for any shape of ring section 1131, though
circular and rectangular sections are most widely known
161. A particular shape called the Shafranov D or Prince-
ton D has the property 1141, [15] of maximizing the in-
ductance for a given turn perimeter.
Two alternative GPW designs are illustrated in Fig. 5
where the turns are made to touch along the inner circle Fig. 6 . The optimal window shape for single-layer toroidal inductors
of the structure. If the coil has many turns the inductance
increases as N 2 but is reduced as the turn area is dimin-
ished, and conversely. The optimum GPW windings have of improvement in devices having two layers or more. A
been obtained 1161 with a circular turn window, for which simple argument, similar to that given earlier for sole-
+
L / L o = 0.272k3l2 0.25k and N = 0.817k1/2,and a noids, suggests the limiting inductance increases as the
square turn window, for which L / L o = 0.252k3/2 + square root of the number of layers, together with signif-
0.25k and N = 0.633k’I2.Approximately 15 percent more icant reductions of component volume. Cross sections are
inductance than the circle is theoretically available by indicated in Fig. 7 . Such devices are not easy to wind
using a D-shaped window with ratio of outer radius to neatly, for the layer in contact with the (nonmagnetic) core
inner radius optimized at 5.3. The section is shown in must be evenly spaced if the succeeding layers are to reg-
Fig. 6 . For this, the best of all single-layer GPW toroids, ister correctly. Detailed calculations [ 171 showed that
+
L I L O = 0.314k3/2 0.25k and N = 0.565k1/’. It is in- there are geometrical limits to this particular model, which
teresting to note that all these toroidal designs scale (if restrict the available range of dimensionless wire-length
the small contribution of the “internal” inductance is dis- k. These limits are due to the impossibility of regular
regarded) similarly to the single-layer solenoid. packing in the center, with three or more layers and k
A possible disadvantage of the single-layer toroid is the small, and insufficient space at the outer extremity to fit
large volume of space occupied, and there are prospects all the turns into one layer. In summary, the limits mean
MURGATROY D : OPTIMAL FORM FOR CORELESS INDUCTORS 2675

1ooog
3

IA
0
K
w
gz
z
Fig. 7. Cross sections of two-layer and three-layer toroids. 100 3
5
n

y,
0
that no six-layer GPW design is possible, in this defined
model, and only a small range of five-layer designs is pos-
sible. The inductance of these devices is calculated [17]
from the self-inductance of each layer, and from the mu-
tual inductances between layers. The dimensions of the
lo4
1o3
, I ,,,,,
1o4
,

DIMENSIONLESS WIRE-LENGTH
, , / , , , 1
10’
10

layers are calculated assuming perfect wire-packing ge- k=w/d

ometry, which sets all the spacings and angles. The re- Fig. 8. Maximum inductance and optimum number of turns for air-cored
toroids.
sults for GPW designs are shown in Fig. 8. For large k
the GPW inductance scales as k3/2and as the square root
of the number of layers. Measurements [17] on devices
with up to three layers gave good agreement with the cal-
culations, for inductors wound on solid nonmagnetic cores
and also for an open structure, shown in Fig. 9, of un-
supported wires threaded through holes in an insulating
base.
It is apparent, in Fig. 8, that as the number of layers is
increased the inductance scales more steeply than k3l2.
Since the detailed model with regular packing breaks down
with five or more layers, a completely different approach
is required to investigate a “many-layer’’ model. Irreg-
ular or wild winding can be represented by a filling factor
(fraction of the normal cross section occupied by conduc-
tors), and the winding is regarded as a conductor with
uniform area though varying shape, and uniform current
density. Calculations which concentrate attention on the Fig. 9 . Toroidal inductors with one, two, and three layers: wires are
continuously distributed magnetic field energy, rather than threaded through holes in the insulating plane.
the individual layers of wire [17], suggest a theoretical
upper limiting design, having the section sketched in Fig.
10, which may be regarded as the toroidal counterpart of
the Brooks solenoid. With the optimum ratio of outer to
inner radii, of the air space, at 2.35, the GPW inductance
is L / L o = 0.122k5/3with N = 0.49k2/3.In several re-
spects this result is approximate. Even if the smallest and
first-wound turns were circular, the turns in subsequent
layers would not be so. The model shape is, therefore,
somewhat arbitrary, the best possible shape is not yet
known, and the practicalities of winding would make the
assumed perfect hexagonal packing of wires impossible
it 235
i
to achieve. Notwithstanding these problems, the result is --L
~

conceptually important, both because it has the same scal- i- 2554 U

ing properties as the Brooks Inductor, and because it in- Fig. 10. Longitudinal section of the theoretical optimum GPW toroid with
dicates a fundamental factor of about 2.6 in favor of the many layers of circular turns
2676 IEEE TRANSACTIONS ON MAGNETICS. VOL. 75. NO. 3. MAY 1989

TABLE I tice that imperfections of turn shape, or uneven wire com-

THEORETICAL GPW M A X I M UI hM DUCTANCES
(Z,,l = p , , d / 2 ~k, = ~ v / d )
pacting, will result in coils slightly below these theoreti-
cal paradigms, but the shortfalls are often quite small. The
Geometrical Form Inductance/l, Turns results are also expressed in Fig. 11 which, together with
the more detailed graphs for solenoids and toroids, sug-
Straight wire (internal) 0 251 -
gests a simple design method. For any specified coil, the
Straight wire (external) h {Ink + 0 393) -
One circular turn k { I n k - 0 815) I inductance and wire diameter together set the ratio L I L O ,
Single-layer solenoid 0 661k’ ’ 0 360k’ ’ so a horizontal line ruled across Fig. 11, or Fig. 4 or Fig.
Brooks solenoid 0 318k’ ’ 0 235k’ ’ 8, gives the necessary k value and hence the minimum
Single-layer toroids
(circle window) 0 272k ’ ’ + 0 25h 0 8171’ ’ length of wire required, to achieve the specification in
(square window) 0 252k’ ’ + 0 25k 0 633h ’’ each of the various geometries. The optimum number of
(D-shape) 0 324k”’ + 0 25k 0 565k‘ ’
turns is also obtainable graphically, from Fig. 4 or Fig.
Many-layer toroid
(circle window) 0 122k5/’ 049k’ ’ 8.

VI. CONCLUSION
At first sight, the GPW principle may seem inappro-
priate to practical inductor design. However, it has been
shown to provide a common basis for comparison of all
the best available shapes and winding schemes for the im-
portant topologies of coreless inductor. Furthermore, es-
pecially in short-run and prototype manufacture, the GPW
basis gives simple formulas for the minimum required
wire length, and designs that are usefully robust against
manufacturing error. If the wire length is correctly mea-
sured in advance of winding, it is quite difficult to get the
inductance value wrong! The scaling properties of single-
layer and multilayer windings are the basis of straightfor-
ward design methods, graphical or by formula, to solve
the “backwards” design problem directly. The original
objective in the GPW analysis was economic use of wire,
and it is pleasing to note that methods following from that
principle make it easier to get the most economic design
than it is to get any inferior design by trial and error.
ACKNOWLEDGMENT
Some of the illustrations in this paper are reproduced
by permission of the Institution of Electrical Engineers,
London, UK.

REFERENCES
7 I I
10‘ 105 [ I ] F. W. Grover, Inductance Calculations. New York, NY: Van Nos-
Dimensionless Wire Length k=w/d trand, 1947.
[ 2 ] D. Xu and K . S . Han, “Self-inductance of air-core circular coils with
Fig. 11. Collation of maximum inductance design limits for coreless so-
rectangular cross section,” IEEE Trans. Magn.. vol. MAG-23, pp.
lenoids and toroids.
3916-3921, 1987.
[ 3 ] P. N . Murgatroyd, “The Brooks inductor: A study of optimal sole-
noid cross-sections,” Proc. Inst. Elec. Eng. E.. vol. 133, pp. 309-
solenoid. Comparison of single-layer designs shows a 314. 1986.
corresponding factor of about 2.4 in favor of the solenoid. [ 4 ] J . C. Maxwell, A Treatise on Electricity and Magnerism, vol. 2 .
Oxford, 1892, Article 706.
V. COLLATION OF OPTIMAL FORMS [ 5 ] B. Brooks, “Design of standards of inductance and the proposed use
of model reactors in the design of air-core and iron-core reactors”
The inductances of the optimal forms of coreless in- (Res. Paper 342). J . Res. Nat. Bur. Stand. I , pp. 289-328, 1931.
ductor, all normalized to the common unit Lo = pod/27r [ 6 ] B. D. Popovic, Introductory Engineering Electromagnetics. Read-
ing, MA: Addison-Wesley, 1971.
based on the wire diameter, are collated in Table I. For [ 7 ] Lord Rayleigh and A. Schuster, “On the determination of the ohm in
the standard geometries studied, these inductances rep- absolute measure,” Proc. Roy. Soc., vol. 32, pp. 104-141, 1881.
resent the greatest achievable values with a “given piece 181 V. G. Welsby, The Theory and Design of Inductance Coils. Lon-
don, England: Macdonald, 1960.
of wire” and represent the natural upper limits achievable [ 9 ] S . Gupta, B.Sc. degree thesis, Loughborough University, Dept. of
with normal coil winding technology. It is likely in prac- Engineering Mathematics, 1986.
MURGATROYD: OPTIMAL FORM FOR CORELESS INDUCTORS 2677

C. F. Gauss, Werke, vol. 5 . Gottingen, Germany, 1867, pp. 621- [ 171 P. N. Murgatroyd, “Optimal designs for multilayer air-cored toroidal
622. inductors with circular windows,” Proc. Inst. Elec. Eng. E . , vol.
E. B. Rosa and F. W. Grover. “Formulas and tables for the calcu- 135, pp. 202-209, 1988.
lation of mutual and self inductances,” Bull. Nut. Bur. Srund., vol.
8 , pp. 33-98, 1912.
R. E. Shawcross and R . I. Wells, “On the form of coils to give max-
imum inductance for a given length and thickness of wire,” Elecrri-
cian (UK), vol. 75, p. 65, 1915.
P. N. Murgatroyd, “Some optimum shapes for toroidal inductors,” Paul N. Murgatroyd was born in Rugby. England, in 1942. He received
Proc. Insr. Elec. Eng. E . , vol. 129, pp. 168-176, 1982. the B.Sc. degree in 1964 and the Ph.D. degree in 1971, both from the
V . D. Shafranov, “Optimal shape of a toroidal solenoid,” Sov. University of Bristol, Bristol, England.
Phys.-Tech. P h j s . , vol. 17, pp. 1433-1437, 1973. Following service in government laboratories he joined the Department
J . File, R. G . Mills, and G. V . Sheffield, “Large superconducting of Electronic and Electrical Engineering at the Loughborough University
magnet designs for fusion reactors,” IEEE Trans. Nucl. Sri., vol. of Technology, Loughborough, England, where he is currently Reader and
NS-18, pp. 205-207, 1971. teaches electromagnetism. He has published over 50 papers on insulators,
P. N. Murgatroyd and D. Belahrache, “Economic designs for single- magnetic suspensions, statistical topics, and inductor design.
layer toroidal inductors,” Proc. Inst. Elec. Eng. E . , vol. 132, pp. Dr. Murgatroyd is a Fellow of the IEE and a Fellow of the Institute of
315-318, 1985. Physics.