Chapter IV: Part 4

But for many purposes, especially where the voltage is high and the
current small, it is advantageous to add together the inductive effect
of the several poles of the multipolar machine by throwing the E.M.F's
of half the total number of elements into series, the number of
parallel circuits being conversely again reduced to two. This is
effected by the second method of winding the closed-coil continuous
current drum, which is known as "wave-winding." The front pitch is now
in the same direction round the armature as the back pitch (fig. 22),
so that the beginning of the second loop, i.e. element No. 15, lies
outside the first loop. After p loops have been formed and as many
elements have been traversed as there are poles, the distance covered
either falls short of or exceeds a complete tour of the armature by
two winding-spaces, or the width of two elements. A second and third
tour are then made, and so on, until finally the winding again closes
upon itself. When the completed winding is developed as in fig. 23, it
is seen to work continuously forwards round the armature in zigzag
waves, one of which is marked in heavy lines, and the number of
complete tours is equal to the average of the back and front pitches.
Since the number of parallel circuits from brush to brush is q = 2,
the _E.M.F. equation of the wave-wound drum_ is E_a = pZ_a (N/60)[tau]
x 10^(-8) volts. Only two sets of brushes are necessary, but in order
to shorten the length of the commutator, other sets may also be added
at the point of highest and lowest potential up to as many in number
as there are poles. Thus the advantage of the wave-wound armature is
that for a given voltage and number of poles the number of active
wires is only 1/p of that in the lap-wound drum, each being of larger
cross-section in order to carry p times as much current; hence the
ratio of the room occupied by the insulation to the copper area is
less, and the available space is better utilized. A further advantage
is that the two circuits from brush to brush consist of elements
influenced by all the poles, so that if for any reason, such as
eccentricity of the armature within the bore of the pole-pieces, or
want of uniformity in the magnetic qualities of the poles, the flux of
each field is not equal to that of every other, the equality of the
voltage produced by the two halves of the winding is not affected
thereby.

In appearance the two classes of armatures, lap and wave, may be
distinguished in the barrel type of winding by the slope of the upper
layer of back end-connexions, and that of the front connexions at the
commutator end being parallel to one another in the latter, and
oppositely directed in the former.

After completion of the winding, the end-connexions are firmly bound down by bands of steel or phosphor bronze binding wire, so as to resist the stress of centrifugal force. In the case of smooth-surface armatures, such bands are also placed at intervals along the length of the armature core, but in toothed armatures, although the coils are often in small machines secured in the slots by similar bands of a non-magnetic high-resistance wire, the use of hard-wood wedges driven into notches at the sides of the slots becomes preferable, and in very large machines indispensable. The external appearance of a typical armature with lap-winding is shown in fig. 24.

The commutator.

A sound mechanical construction of the commutator is of vital importance to the good working of the continuous-current dynamo. The narrow, wedge-shaped sectors of hard-drawn copper, with their insulating strips of thin mica, are built up into a cylinder, tightly clamped together, and turned in the lathe; at each end a V-shaped groove is turned, and into these are fitted rings of micanite of corresponding section (fig. 19); the whole is then slipped over a cast iron sleeve, and at either end strong rings are forced into the V-shaped grooves under great pressure and fixed by a number of closely-pitched tightening bolts. In dynamos driven by steam-turbines in which the peripheral speed of the commutator is very high, rings of steel are frequently shrunk on the surface of the commutator at either end and at its centre. But in every case the copper must be entirely insulated from the supporting body of metal by the interposition of mica or micanite and the prevention of any movement of the sectors under frequent and long-continued heating and cooling calls for the greatest care in both the design and the manufacture.

Forms of field-magnet.

On passing to the second fundamental part of the dynamo, namely, the field-magnet, its functions may be briefly recalled as follows:--It has to supply the magnetic flux; to provide for it an iron path as nearly closed as possible upon the armature, save for the air-gaps which must exist between the pole-system and the armature core, the one stationary and the other rotating; and, lastly, it has to give the lines such direction and intensity within the air-gaps that they may be cut by the armature wires to the best advantage. Roughly corresponding to the three functions above summarized are the three portions which are more or less differentiated in the complete structure. These are: (1) the magnet "cores" or "_limbs_," carrying the exciting coils whereby the inert iron is converted into an electro-magnet; (2) the _yoke_, which joins the limbs together and conducts the flux between them; and (3) the _pole-pieces_, which face the armature and transmit the lines from the limbs through the air-gap to the armature core, or vice versa.

Of the countless shapes which the field-magnet may take, it may be
said, without much exaggeration, that almost all have been tried; yet
those which have proved economical and successful, and hence have met
with general adoption, may be classed under a comparatively small
number of types. For bipolar machines the _single horse-shoe_ (fig.
25), which is the lineal successor of the permanent magnet employed in
the first magneto-electric machines, was formerly very largely used.
It takes two principal forms, according as the pole-pieces and
armature are above or beneath the magnet limbs and yoke. The
"over-type" form is best suited to small belt-driven dynamos, while
the "under-type" is admirably adapted to be directly driven by the
steam-engine, the armature shaft being immediately coupled to the
crank-shaft of the engine. In the latter case the magnet must be
mounted on non-magnetic supports of gun-metal or zinc, so as to hold
it at some distance away from the iron bedplate which carries both
engine and dynamo; otherwise a large proportion of the flux which
passes through the magnet limbs would leak through the bedplate across
from pole to pole without passing through the armature core, and so
would not be cut by the armature wires.

Next may be placed the "Manchester" field (fig. 26)--the type of a
divided magnetic circuit in which the flux forming one field or pole
is divided between two magnets. An exciting coil is placed on each
half of the double horse-shoe magnet, the pair being so wound that
consequent poles are formed above and below the armature. Each magnet
thus carries one-half of the total flux, the lines of the two halves
uniting to form a common field where they issue forth into or leave
the air-gaps. The pole-pieces may be lighter than in the single
horse-shoe type, and the field is much more symmetrical, whence it is
well suited to ring armatures of large diameter. Yet these advantages
are greatly discounted by the excessive magnetic leakage, and by the
increased weight of copper in the exciting coils. Even if the greater
percentage which the leakage lines bear to the useful flux is
neglected, and the cross sectional area of each magnet core is but
half that of the equivalent single horse-shoe, the weight of wire in
the double magnet for the same rise of temperature in the coils must
be some 40% more than in the single horse-shoe, and the rate at which
energy is expended in heating the coils will exceed that of the single
horse-shoe in the same proportion.

Thirdly comes the two-pole _ironclad_ type, so called from the
exciting coil being more or less encased by the iron yoke; this latter
is divided into two halves, which pass on either side of the armature.
Unless the yoke be kept well away from the polar edges and armature,
the leakage across the air into the yoke becomes considerable,
especially if only one exciting coil is used, as in fig. 27 A; it is
better, therefore, to divide the excitation between two coils, as in
fig. 27 B, when the field also becomes symmetrical.

From this form is easily derived the _multipolar_ type of fig. 28 or
fig. 29, which is by far the most usual for any number of poles from
four upwards; its leakage coefficient is but small, and it is
economical in weight both of iron and copper.

Materials of magnets.

As regards the materials of which magnets are made, generally speaking
there is little difference in the permeability of "wrought iron" or
"mild steel forgings" and good "cast steel"; typical (B, H) curves
connecting the magnetizing force required with different
flux-densities for these materials are given under ELECTROMAGNETISM.
On the other hand there is a marked inferiority in the case of "cast
iron," which for a flux-density of B = 8000 C.G.S. lines per sq. cm.
requires practically the same number of ampere-turns per centimetre
length as steel requires for B = 16,000. Whatever the material, if the
flux-density be pressed to a high value the ampere-turns are very
largely increased owing to its approaching saturation, and this
implies either a large amount of copper in the field coils or an undue
expenditure of electrical energy in their excitation. Hence there is a
limit imposed by practical considerations to the density at which the
magnet should be worked, and this limit may be placed at about B =
16,000 for wrought iron or steel, and at half this value for cast
iron. For a given flux, therefore, the cast iron magnet must have
twice the sectional area and be twice as heavy, although this
disadvantage is partly compensated by its greater cheapness. If,
however, cast iron be used for the portion of the magnetic circuit
which is covered with the exciting coils, the further disadvantage
must be added that the weight of copper on the field-magnet is much
increased, so that it is usual to employ forgings or cast steel for
the magnet cores on which the coils are wound. If weight is not a
disadvantage, a cast iron yoke may be combined with the wrought iron
or cast steel magnet cores. An absence of joints in the magnetic
circuit is only desirable from the point of view of economy of expense
in machining the component parts during manufacture; when the surfaces
which abut against each other are drawn firmly together by screws, the
want of homogeneity at the joint, which virtually amounts to the
presence of a very thin film of air, produces little or no effect on
the total reluctance by comparison with the very much longer air-gaps
surrounding the armature. In order to reduce the eddy-currents in the
pole-pieces, due to the use of toothed armatures with relatively wide
slots, the poles themselves must be laminated, or must have fixed to
them laminated pole-shoes, built up of thin strips of mild steel
riveted together (as shown in fig. 29).

However it be built up, the mechanical strength of the magnet system
must be carefully considered. Any two surfaces between which there
exists a field of density B_g experience a force tending to draw them
together proportional to the square of the density, and having a value
of B_g squared/(1.735 x 10^6) lb. per sq. in. of surface, over which the
density may be regarded as having the uniform value B_g. Hence, quite
apart from the torque with which the stationary part of the dynamo
tends to turn with the rotating part as soon as current is taken out
of the armature, there exists a force tending to make the pole-pieces
close on the armature as soon as the field is excited. Since both
armature and magnet must be capable of resisting this force, they
require to be rigidly held; although the one or the other must be
capable of rotation, there should otherwise be no possibility of one
part of the magnetic circuit shifting relatively to any other part. An
important conclusion may be drawn from this circumstance. If the
armature be placed exactly concentric within the bore of the poles,
and the two or more magnetic fields be symmetrical about a line
joining their centres, there is no tendency for the armature core to
be drawn in one direction more than in another; but if there is any
difference between the densities of the several fields, it will cause
an unbalanced stress on the armature and its shaft, under which it
will bend, and as this bending is continually reversed relatively to
the fibres of the shaft, they will eventually become weakened and give
way. Especially is this likely to take place in dynamos with short
air-gaps, wherein any difference in the lengths of the air-gaps
produces a much greater percentage difference in the flux-density than
in dynamos with long air-gaps. In toothed armatures with short
air-gaps the shaft must on this account be sufficiently strong to
withstand the stress without appreciable bending.

The magnetic circuit.

Reference has already been made to the importance in dynamo design of the _predetermination of the flux_ due to a given number of ampere-turns wound on the field-magnet, or, conversely, of the number of ampere-turns which must be furnished by the exciting coils in order that a certain flux corresponding to one field may flow through the armature core from each pole. An equally important problem is the correct proportioning of the field-magnet, so that the useful flux Z_a may be obtained with the greatest economy in materials and exciting energy. The key to the two problems is to be found in the concept of a magnetic circuit as originated by H.A. Rowland and R.H.M. Bosanquet;[16] and the full solution of both may be especially connected with the name of Dr J. Hopkinson, from his practical application of the concept in his design of the Edison-Hopkinson machine, and in his paper on "Dynamo-Electric Machinery."[17] The publication of this paper in 1886 begins the second era in the history of the dynamo; it at once raised its design from the level of empirical rules-of-thumb to a science, and is thus worthy to be ranked as the necessary supplement of the original discoveries of Faraday. The process of predetermining the necessary ampere-turns is described in a simple case under ELECTROMAGNETISM. In its extension to the complete dynamo, it consists merely in the division of the magnetic circuit into such portions as have the same sectional area and permeability and carry approximately the same total flux; the difference of magnetic potential that must exist between the ends of each section of the magnet in order that the flux may pass through it is then calculated _seriatim_ for the several portions into which the magnetic circuit is divided, and the separate items are summed up into one magnetomotive force that must be furnished by the exciting coils.

The chief sections of the magnetic circuit are (1) the air-gaps, (2)
the armature core, and (3) the iron magnet.

The _air-gap_ of a dynamo with smooth-core armature is partly filled
with copper and partly with the cotton, mica, or other materials used
to insulate the core and wires; all these substances are, however,
sensibly non-magnetic, so that the whole interferric gap between the
iron of the pole-pieces and the iron of the armature may be treated as
an air-space, of which the permeability is constant for all values of
the flux density, and in the C.G.S. system is unity. Hence if l_g and
A_g be the length and area of the single air-gap in cm. and sq. cm.,
the reluctance of the double air-gap is 2l_g/A_g, and the difference
of magnetic potential required to pass Z_a lines over this reluctance
is Z_a.2l_g/A_g = B_g.2l_g; or, since one ampere-turn gives 1.257
C.G.S. units of magnetomotive force, the exciting power in
ampere-turns required over the two air-gaps is X_g = B_g.2l_g/1.257 =
0.8 B_g.2l_g. In the determination of the area A_g small allowance
must be made for the fringe of lines which extend beyond the actual
polar face. In the toothed armature with open slots, the lines are no
longer uniformly distributed over the air-gap area, but are graduated
into alternate bands of dense and weak induction corresponding to the
teeth and slots. Further, the lines curve round into the sides of the
teeth, so that their average length of path in the air and the air-gap
reluctance is not so easily calculated. Allowance must be made for
this by taking an increased length of air-gap = ml_g, where m is the
ratio _maximum density/mean density_, of which the value is chiefly
determined by the ratios of the width of tooth to width of slot and of
the width of slot to the air-gap between pole-face and surface of the
armature core.

The _armature core_ must be divided into the teeth and the core proper
below the teeth. Owing to the tapering section of the teeth, the
density rises towards their root, and when this reaches a high value,
such as 18,000 or more lines per sq. cm., the saturation of the iron
again forces an increasing proportion of the lines outwards into the
slot. A distinction must then be drawn between the "apparent"
induction which would hold if all the lines were concentrated in the
teeth, and the "real" induction. The area of the iron is obtained by
multiplying the number of teeth under the pole-face by their width and
by the net length of the iron core parallel to the axis of rotation.
The latter is the gross length of the armature less the space lost
through the insulating varnish or paper between the disks or through
the presence of ventilating ducts, which are introduced at intervals
along the length of the core. The former deduction averages about 7 to
10% of the gross length, while the latter, especially in large
multipolar machines, is an even more important item. Alter calculating
the density at different sections of the teeth, reference has now to
be made to a (B, H) or flux-density curve, from which may be found the
number of ampere-turns required per cm. length of path. This number
may be expressed as a function of the density in the teeth, and f(B_t)
be its average value over the length of a tooth, the ampere-turns of
excitation required over the teeth on either side of the core as the
lines of one field enter or leave the armature is X_t = f(B_t).2l_t,
where l_t is the length of a single tooth in cm.

In the core proper below the teeth the length of path continually
shortens as we pass from the middle of the pole towards the centre
line of symmetry. On the other hand, as the lines gradually accumulate
in the core, their density increases from zero midway under the poles
until it reaches a maximum on the line of symmetry. The two effects
partially counteract one another, and tend to equalize the difference
of magnetic potential required over the paths of varying lengths; but
since the reluctivity of the iron increases more rapidly than the
density of the lines, we may approximately take for the length of path
(l_a) the minimum peripheral distance between the edges of adjacent
pole-faces, and then assume the maximum value of the density of the
lines as holding throughout this entire path. In ring and drum
machines the flux issuing from one pole divides into two halves in the
armature core, so that the maximum density of lines in the armature is
B_a = Z_a/2ab, where a = the radial depth of the disks in centimetres
and b = the net length of iron core. The total exciting power required
between the pole-pieces is therefore, at no load, X_p = X_g + X_t +
X_a, where X_a = f(B_a).l_a; in order, however, to allow for the
effect of the armature current, which increases with the load, a
further term X_b, must be added.

In the continuous-current dynamo it may be, and usually is, necessary
to move the brushes forward from the interpolar line of symmetry
through a small angle in the direction of rotation, in order to avoid
sparking between the brushes and the commutator (_vide infra_). When
the dynamo is giving current, the wires on either side of the diameter
of commutation form a current-sheet flowing along the surface of the
armature from end to end, and whatever the actual end-connexions of
the wires, the wires may be imagined to be joined together into a
system of loops such that the two sides of each loop are carrying
current in opposite directions. Thus a number of armature ampere-turns
are formed, and their effect on the entire system of magnet and
armature must be taken into account. So long as the diameter of
commutation coincides with the line of symmetry, the armature may be
regarded as a cylindrical electromagnet producing a flux of lines, as
shown in fig. 30. The direction of the self-induced flux in the
air-gaps is the same as that of the lines of the external field in one
quadrant on one side of DC, but opposed to it in the other quadrant on
the same side of DC; hence in the resultant field due to the combined
action of the field-magnet and armature ampere-turns, the flux is as
much strengthened over the one half of each polar face as it is
weakened over the other, and the total number of lines is unaffected,
although their distribution is altered. The armature ampere-turns are
then called _cross-turns_, since they produce a cross-field, which,
when combined with the symmetrical field, causes the leading
pole-corners ll to be weakened and the trailing pole-corners tt to be
strengthened, the neutral line of zero field being thus twisted
forwards in the direction of rotation. But when the brushes and
diameter of commutation are shifted forward, as shown in fig. 31, it
will be seen that a number of ampere-turns, forming a zone between the
lines Dn and mC, are in effect wound immediately on the magnetic
circuit proper, and this belt of ampere-turns is in direct opposition
to the ampere-turns of the field, as shown by the dotted and crossed
wires on the pole-pieces. The armature ampere-turns are then divisible
into the two bands, the _back-turns_, included within twice the angle
of lead [lambda], weakening the field, and the cross-turns, bounded by
the lines Dm, nC, again producing distortion of the weakened
symmetrical field. If, therefore, a certain flux is to be passed
through the armature core in opposition to the demagnetizing turns,
the difference of magnetic potential between the pole-faces must
include not only X_a, X_t, and X_g, but also an item X_b, in order to
balance the "back" ampere-turns of the armature. The amount by which
the brushes must be shifted forward increases with the armature
current, and in corresponding proportion the back ampere-turns are
also increased, their value being c[tau]2[lambda]/360 deg., where c = the
current carried by each of the [tau] active wires. Thus the term X_b,
takes into account the effect of the armature reaction on the total
flux; it varies as the armature current and angle of lead required to
avoid sparking are increased; and the reason for its introduction in
the fourth place (X_p = X_g + X_t + X_a + X_b), is that it increases
the magnetic difference of potential which must exist between the
poles of the dynamo, and to which the greater part of the leakage is
due. The leakage paths which are in parallel with the armature across
the poles must now be estimated, and so a new value be derived for the
flux at the commencement of the _iron-magnet_ path. If P = their joint
permeance, the leakage flux due to the difference of potential at the
poles is z_l = 1.257X_p x P, and this must be added to the useful flux
Z_a, or Z_p = Z_a + Z_l. There are also certain leakage paths in
parallel with the magnet cores, and upon the permeance of these a
varying number of ampere-turns is acting as we proceed along the
magnet coils; the magnet flux therefore increases by the addition of
leakage along the length of the limbs, and finally reaches a maximum
near the yoke. Either, then, the density in the magnet B_m = Z_m/A_m
will vary if the same sectional area be retained throughout, or the
sectional area of the magnet must itself be progressively increased.
In general, sufficient accuracy will be obtained by assuming a certain
number of additional leakage lines z_n as traversing the entire length
of magnet limbs and yoke (= l_m), so that the density in the magnet
has the uniform value B_m = (Z_p + z_n)/A_m. The leakage flux added on
actually within the length of the magnet core or z_n will be
approximately equal to half the total M.M.F. of the coils multiplied
by the permeance of the leakage paths around one coil. The
corresponding value of H can then be obtained from the (B, H) curve of
the material of which the magnet is composed, and the ampere-turns
thus determined must be added to X_p, or X = X_p + X_m, where X_m =
f(B_m)l_m. The final equation for the exciting power required on a
magnetic circuit as a whole will therefore take the form

X = A[Tau] = 0.8B_g.2l_g + f(B_t)2l_t + f(B_a)l_a + X_b + f(B_m)l_m. (3)

If the magnet cores are of wrought iron or cast steel, and the yoke is
of cast iron, the last term must be divided into two portions
corresponding to the different materials, i.e. into f(B_m)l_m +
f(B_y)l_y. In the ordinary multipolar machine with as many
magnet-coils as there are poles, each coil must furnish half the above
number of ampere-turns.

Magnetic leakage.

Since no substance is impermeable to the passage of magnetic flux, the
only form of magnetic circuit free from leakage is one uniformly wound
with ampere-turns over its whole length. The reduction of the
_magnetic leakage_ to a minimum in any given type is therefore
primarily a question of distributing the winding as far as possible
uniformly upon the circuit, and as the winding must be more or less
concentrated into coils, it resolves itself into the necessity of
introducing as long air-paths as possible between any surfaces which
are at different magnetic potentials. No iron should be brought near
the machine which does not form part of the magnetic circuit proper,
and especially no iron should be brought near the poles, between which
the difference of magnetic potential practically reaches its maximum
value. In default of a machine of the same size or similar type on
which to experiment, the probable direction of the leakage flux must
be assumed from the drawing, and the air surrounding the machine must
be mapped out into areas, between which the permeances are calculated
as closely as possible by means of such approximate formulae as those
devised by Professor G. Forbes.

Excitation of field-magnet.

In the earliest "magneto-electric" machines permanent steel magnets,
either simple or compound, were employed, and for many years these
were retained in certain alternators, some of which are still in use
for arc lighting in lighthouses. But since the field they furnish is
very weak, a great advance was made when they were replaced by soft
iron electromagnets, which could be made to yield a much more intense
flux. As early as 1831 Faraday[18] experimented with electromagnets,
and after 1850 they gradually superseded the permanent magnet. When
the total ampere-turns required to excite the electromagnet have been
determined, it remains to decide how the excitation shall be obtained;
and, according to the method adopted, continuous-current machines may
be divided into four well-defined classes.

The simplest method, and that which was first used, is _separate
excitation_ from some other source of direct current, which may be
either a primary or a secondary battery or another dynamo (fig. 32).
But since the armature yields a continuous current, it was early
suggested (by J. Brett in 1848 and F. Sinsteden in 1851) that this
current might be utilized to increase the flux; combinations of
permanent and electromagnets were therefore next employed, acting
either on the main armature or on separate armatures, until in 1867 Dr
Werner von Siemens and Sir C. Wheatstone almost simultaneously
discovered that the dynamo could be made _self-exciting_ through the
residual magnetism retained in the soft iron cores of the
electromagnet. The former proposed to take the whole of the current
round the magnet coils which were in series with the armature and
external circuit, while the latter proposed to utilize only a portion
derived by a shunt from the main circuit; we thus arrive at the second
and third classes, namely, _series_ and _shunt_ machines. The starting
of the process of excitation in either case is the same; when the
brushes are touching the commutator and the armature is rotated, the
small amount of flux left in the magnet is cut by the wires, and a
very small current begins to flow round the closed circuit; this
increases the flux, which in turn further increases the E.M.F. and
current, until, finally, the cumulative effect stops through the
increasing saturation of the iron cores. Fig. 33, illustrating the
_series_ machine, shows the winding of the exciting coils to be
composed of a few turns of thick wire. Since the current is undivided
throughout the whole circuit, the resistance of both the armature and
field-magnet winding must be low as compared with that of the external
circuit, if the useful power available at the terminals of the machine
is to form a large percentage of the total electrical power--in other
words, if the efficiency is to be high. Fig. 34 shows the third
method, in which the winding of the field-magnets is a _shunt_ or
fine-wire circuit of many turns applied to the terminals of the
machine; in this ease the resistance of the shunt must be high as
compared with that of the external circuit, in order that only a small
proportion of the total energy may be absorbed in the field.

Since the whole of the armature current passes round the field-magnet
of the series machine, any alteration in the resistance of the
external circuit will affect the excitation and also the voltage. A
curve connecting together corresponding values of external current and
terminal voltage for a given speed of rotation is known as the
_external-characteristic_ of the machine; in its main features it has
the same appearance as a curve of magnetic flux, but when the current
exceeds a certain amount it begins to bend downwards and the voltage
decreases. The reason for this will be found in the armature reaction
at large loads, which gradually produces a more and more powerful
demagnetizing effect, as the brushes are shifted forwards to avoid
sparking; eventually the back ampere-turns overpower any addition to
the field that would otherwise be due to the increased current flowing
round the magnet. The "external characteristic" for a shunt machine
has an entirely different shape. The field-magnet circuit being
connected in parallel with the external circuit, the exciting current,
if the applied voltage remains the same, is in no way affected by
alterations in the resistance of the latter. As, however, an increase
in the external current causes a greater loss of volts in the armature
and a greater armature reaction, the terminal voltage, which is also
the exciting voltage, is highest at no load and then diminishes. The
fall is at first gradual, but after a certain critical value of the
armature current is reached, the machine is rapidly demagnetized and
loses its voltage entirely.

The last method of excitation, namely, _compound-winding_ (fig. 35),
is a combination of the two preceding, and was first used by S.A.
Varley and by C.F. Brush. If a machine is in the first instance
shunt-wound, and a certain number of series-turns are added, the
latter, since they carry the external current, can be made to
counteract the effect which the increased external current would have
in lowering the voltage of the simple shunt machine. The ampere-turns
of the series winding must be such that they not only balance the
increase of the demagnetizing back ampere-turns on the armature, but
further increase the useful flux, and compensate for the loss of volts
over their own resistance and that of the armature. The machine will
then give for a constant speed a nearly constant voltage at its
terminals, and the curve of the external characteristic becomes a
straight line for all loads within its capacity. Since with most prime
movers an increase of the load is accompanied by a drop in speed, this
effect may also be counteracted; while, lastly, if the series-turns
are still further increased, the voltage may be made to rise with an
increasing load, and the machine is "over-compounded."

Commutation and sparking at the brushes.

At the initial moment when an armature coil is first short-circuited by the passage of the two sectors forming its ends under the contact surface of a brush, a certain amount of electromagnetic energy is stored up in its magnetic field as linked with the ampere-turns of the coil when carrying its full share of the total armature current. During the period of short-circuit this quantity of energy has to be dissipated as the current falls to zero, and has again to be re-stored as the current is reversed and raised to the same value, but in the opposite direction. The period of short-circuit as fixed by the widths of the brush and of the mica insulation between the sectors, and by the peripheral speed of the commutator is extremely brief, and only lasts on an average from (1/200)th to (1/1000)th of a second. The problem of sparkless commutation is therefore primarily a question of our ability to dissipate and to re-store the required amount of energy with sufficient rapidity.

An important aid towards the solution of this problem is found in the effect of the varying contact-resistance between the brush and the surfaces of the leading and trailing sectors which it covers. As the commutator moves under the brush, the area of contact which the brush makes with the leading sector diminishes, and the resistance between the two rises; conversely, the area of contact between the brush and the trailing sector increases and the resistance falls. This action tends automatically to bring the current through each sector into strict proportionality to the amount of its surface which is covered by the brush, and so to keep the current-density and the loss of volts over the contacts uniform and constant. As soon as the current-density in the two portions of the brush becomes unequal, a greater amount of heat is developed at the commutator surface, and this in the first place affords an additional outlet for the dissipation of the stored energy of the coil, while after reversal of the current it is the accompaniment of a re-storage of the required energy. This energy, as well as that which is spent in heating the coil, can in fact, in default of other sources, be derived through the action of the unequal current-density from the electrical output of the rest of the armature winding, and so only indirectly from the prime mover.

In practice, when the normal contact-resistance of the brushes is low relatively to the resistance of the coil, as is the case with metal brushes of copper or brass gauze, but little benefit can be obtained from the action of the varying contact-resistance. It exerts no appreciable effect until close towards the end of the period of short-circuit, and then only with such a high-current-density at the trailing edge of the leaving sector that at the moment of parting the brush-tip is fused, or its metal volatilized, and sparking has in fact set in. With such brushes, then, it becomes necessary to call in the aid of a reversing E.M.F. impressed upon the coil by the magnetic field through which it is moving. If such a reversing field comes into action while the current is still unreversed, its E.M.F. is opposed to the direction of the current, and the coil is therefore driving the armature forward as in a motor; it thus affords a ready means of rapidly dissipating part of the initial energy in the form of mechanical work instead of as heat. After the current has been reversed, the converse process sets in, and the prime mover directly expends mechanical energy not only in heating the coil, but also in storing up electromagnetic energy with a rapidity dependent upon the strength of the reversing field. The required direction of external field can be obtained in the dynamo by shifting the brushes forward, so that the short-circuited coil enters into the fringe of lines issuing from the leading pole-tip, i.e. by giving the brushes an "angle of lead." An objection to this process is that the main flux is thereby weakened owing to the belt of back ampere-turns which arises (_v. supra_). A still greater objection is that the amount of the angle of lead must be suited to the value of the load, the corrective power of copper brushes being very small if the reversing E.M.F. is not closely adjusted in proportion to the armature current.

On this account metal brushes have been almost entirely superseded by carbon moulded into hard blocks. With these, owing to their higher specific contact-resistance, a very considerable reversing effect can be obtained through the action of unequal current-density, and indeed in favourable cases complete sparklessness can be obtained throughout the entire range of load of the machine with a fixed position of the brushes. Yet if the work which they are called upon to perform exceeds certain limits, they tend to become overheated with consequent glowing or sparking at their tips, so that, wherever possible, it is advisable to reinforce their action by a certain amount of reversing field, the brushes being set so that its strength is roughly correct for, say, half load.

In the case of dynamos driven by steam-turbines, sparkless commutation is especially difficult to obtain owing to the high speed of rotation and the very short space of time in which the current has to be reversed. Special "reversing poles" then become necessary; these are wound with magnetizing coils in series with the main armature current, so that the strength of field which they yield is roughly proportional to the current which has to be reversed. These again may be combined with a "compensating winding" embedded in the pole-faces and carrying current in the opposite direction to the armature ampere-turns, so as to neutralize the cross effect of the latter and prevent distortion of the resultant field.

Heating effects.

From the moment that a dynamo begins to run with excited field, heat
is continuously generated by the passage of the current through the
windings of the field-magnet coils and the armature, as well as by the
action of hysteresis and eddy currents in the armature and
pole-pieces. Whether the source of the heat be in the field-magnet or
in the armature, the mass in which it originates will continue to rise
in temperature until such a difference of temperature is established
between itself and the surrounding air that the rate at which the heat
is carried off by radiation, convection and conduction is equal to the
rate at which it is being generated. Evidently, then, the temperature
which any part of the machine attains after a prolonged run must
depend on the extent and effectiveness of the cooling surface from
which radiation takes place, upon the presence or absence of any
currents of air set up by the rotation of itself or surrounding parts,
and upon the presence of neighbouring masses of metal to carry away
the heat by conduction. In the field-magnet coils the rate at which
heat is being generated is easily determined, since it is equal to the
square of the current passing through them multiplied by their
resistance. Further, the magnet is usually stationary, and only
indirectly affected by draughts of air due to the rotating armature.
Hence for machines of a given type and of similar proportions, it is
not difficult to decide upon some method of reckoning the cooling
surface of the magnet coils S_c, such that the rise of temperature
above that of the surrounding air may be predicted from an equation of
the form t deg. = kW/S_c, where W = the rate in watts at which heat is
generated in the coils, and k is some constant depending upon the
exact method of reckoning their cooling surface. As a general rule the
cooling surface of a field-coil is reckoned as equal to the exposed
outer surface of its wire, the influence of the end flanges being
neglected, or only taken into account in the case of very short
bobbins wound with a considerable depth of wire. In the case of the
rotating armature a similar formula must be constructed, but with the
addition of a factor to allow for the increase in the effectiveness of
any given cooling surface due to the rotation causing convection
currents in the surrounding air. Only experiment can determine the
exact effect of this, and even with a given type of armature it is
dependent on the number of poles, each of which helps to break up the
air-currents, and so to dissipate the heat. For example, in two-pole
machines with drum bar-armatures, if the cooling surface be reckoned
as equal to the cylindrical exterior plus the area of the two ends,
the heating coefficient for a peripheral speed of 1500 ft. per minute
is less than half of that for the same armature when at rest. A
further difficulty still meets the designer in the correct
predetermination of the total loss of watts in an armature before the
machine has been tested. It is made up of three separate items,
namely, the copper loss in the armature winding, the loss by
hysteresis in the iron, and the loss by eddy currents, which again may
be divided into those in the armature bars and end-connexions, and
those in the core and its end-plates. The two latter items are both
dependent upon the speed of the machine; but whereas the hysteresis
loss is proportional to the speed for a given density of flux in the
armature, the eddy current loss is proportional to the square of the
speed, and owing to this difference, the one loss can be separated
from the other by testing an armature at varying speeds. Thus for a
given rise of temperature, the question of the amount of current which
can be taken out of an armature at different speeds depends upon the
proportion which the hysteresis and eddy watts bear to the copper
loss, and the ratio in which the effectiveness of the cooling surface
is altered by the alteration in speed. Experimental data, again, can
alone decide upon the amount of eddy currents that may be expected in
given armatures, and caution is required in applying the results of
one machine to another in which any of the conditions, such as the
number of poles, density in the teeth, proportions of slot depth to
width, &c., are radically altered.

It remains to add, that the rise of temperature which may be permitted
in any part of a dynamo after a prolonged run is very generally placed
at about 70 deg. Fahr. above the surrounding air. Such a limit in ordinary
conditions of working leads to a final temperature of about 170 deg.
Fahr., beyond which the durability of the insulation of the wires is
liable to be injuriously affected. Upon some such basis the output of
a dynamo in continuous working is rated, although for short periods
of, say, two hours the normal full-load current of a large machine may
be exceeded by some 25% without unduly heating the armature.

Uses of continuous current dynamos.

For the electro-deposition of metals or the electrolytic treatment of ores a continuous current is a necessity; but, apart from such use, the purposes from which the continuous-current dynamo is well adapted are so numerous that they cover nearly the whole field of electrical engineering, with one important exception. To meet these various uses, the pressures for which the machine is designed are of equally wide range; for the transmission of power over long distances they may be as high as 3000 volts, and for electrolytic work as low as five. Each electrolytic bath, with its leads, requires on an average only some four or five volts, so that even when several are worked in series the voltage of the dynamo seldom exceeds 60. On the other hand, the current is large and may amount to as much as from 1000 to 14,000 amperes, necessitating the use of two commutators, one at either end of the armature, in order to collect the current without excessive heating of the sectors and brushes. The field-magnets are invariably shunt-wound, in order to avoid reversal of the current through polarization at the electrodes of the bath. For incandescent lighting by glow lamps, the requirements of small isolated installations and of central stations for the distribution of electrical energy over large areas must be distinguished. For the lighting of a private house or small factory, the dynamo giving from 5 to 100 kilo-watts of output is commonly wound for a voltage of 100, and is driven by pulley and belt from a gas, oil or steam-engine; or, if approaching the higher limit above mentioned, it is often directly coupled to the crank-shaft of the steam-engine. If used in conjunction with an accumulator of secondary cells, it is shunt-wound, and must give the higher voltage necessary to charge the battery; otherwise it is compound-wound, in order to maintain the pressure on the lamps constant under all loads within its capacity. The compound-wound dynamo is likewise the most usual for the lighting of steamships, and is then directly coupled to its steam-engine; its output seldom exceeds 100 kilo-watts, at a voltage of 100 or 110. For larger installations a voltage of 250 is commonly used, while for central-station work, economy in the distributing mains dictates a higher voltage, especially in connexion with a three-wire system; the larger dynamos may then give 500 volts, and be connected directly across the two outer wires. A pair of smaller machines coupled together, and each capable of giving 250 volts, are often placed in series across the system, with their common junction connected to the middle wire; the one which at any time is on the side carrying the smaller current will act as a motor and drive the other as a dynamo, so as to balance the system. The directly-coupled steam dynamo may be said to have practically displaced the belt- or rope-driven sets which were formerly common in central stations. The generating units of the central station are arranged in progressive sizes, rising from, it may be, 250 or 500 horse-power up to 750 or 1000, or in large towns to as much as 5000 horse-power. If for lighting only, they are usually shunt-wound, the regulation of the voltage, to keep the pressure constant on the distributing system under the gradual changes of load, being effected by variable resistances in the shunt circuit of the field-magnets.

Generators used for supplying current to electric tramways are commonly wound for 500 volts at no load and are over-compounded, so that the voltage rises to 550 volts at the maximum load, and thus compensates for the loss of volts over the transmitting lines. For arc lighting it was formerly usual to employ a class of dynamo which, from the nature of its construction, was called an "open-coil" machine, and which gave a unidirectional but pulsating current. Of such machines the Brush and Thomson-Houston types were very widely used; their E.M.F. ranged from 2000 to 3000 volts for working a large number of arcs in series, and by means of special regulators their current was maintained constant over a wide range of voltage. But as their efficiency was low and they could not be applied to any other purpose, they have been largely superseded in central stations by closed-coil dynamos or alternators, which can also be used for incandescent lighting. In cases where the central station is situated at some distance from the district to which the electric energy is to be supplied, voltages from 1000 to 2000 are employed, and these are transformed down at certain distributing centres by continuous-current transformers (see TRANSFORMERS and ELECTRICITY SUPPLY). These latter machines are in reality motor-driven dynamos, and hence are also called _motor-generators_; the armatures of the motor and dynamo are often wound on the same core, with a commutator at either end, the one to receive the high-pressure motor current, and the other to collect the low-pressure current furnished by the dynamo.

In all large central stations it is necessary that the dynamos should
be capable of being run _in parallel_, so that their outputs may be
combined on the same "omnibus bars" and thence distributed to the
network of feeders. With simple shunt-wound machines this is easily
effected by coupling together terminals of like sign when the voltage
of the two or more machines are closely equal. With compound-wound
dynamos not only must the external terminals of like sign be coupled
together, but the junctions of the brush leads with the series winding
must be connected by an "equalizing" lead of low resistance;
otherwise, should the E.M.F. of one machine for any reason fall below
the voltage of the omnibus bars, there is a danger of its polarity
being reversed by a back current from the others with which it is in
parallel.

Owing to the necessary presence in the continuous-current dynamo of
the commutator, with its attendant liability to sparking at the
brushes, and further, owing to the difficulty of insulating the
rotating armature wires, a pressure of 3000 volts has seldom been
exceeded in any one continuous-current machine, and has been given
above as the limiting voltage of the class. If therefore it is
required to work with higher pressures in order to secure economy in
the transmitting lines, two or more machines must be coupled _in
series_ by connecting together terminals which are of unlike sign.[19]
The stress of the total voltage may still fall on the insulation of
the winding from the body of the machine; hence for high-voltage
transmission of power over very long distances, the continuous-current
dynamo in certain points yields in convenience to the alternator. In
this there is no commutator, the armature coils may be stationary and
can be more thoroughly insulated, while further, if it be thought
undesirable to design the machine for the full transmitting voltage,
it is easy to wind the armature for a low pressure; this can be
subsequently transformed up to a high pressure by means of the
alternating-current transformer, which has stationary windings and so
high an efficiency that but little loss arises from its use. With
these remarks, the transition may be made to the fuller discussion of
the alternator.

_Alternators._

Frequency.

The frequency employed in alternating-current systems for distributing power and light varies between such wide limits as 25 and 133; yet in recent times the tendency has been towards standard frequencies of 25, 50 and 100 as a maximum. High frequencies involve more copper in the magnet coils, owing to the greater number of poles, and a greater loss of power in their excitation, but the alternator as a whole is somewhat lighter, and the transformers are cheaper. On the other hand, high frequency may cause prejudicial effects, due to the inductance and capacity of the distributing lines; and in asynchronous motors used on polyphase systems the increased number of poles necessary to obtain reasonable speeds reduces their efficiency, and is otherwise disadvantageous, especially for small horse-powers. A frequency lower than 40 is, however, not permissible where arc lighting is to form any considerable portion of the work and is to be effected by the alternating current without rectification, since below this value the eye can detect the periodic alteration in the light as the carbons alternately cool and become heated. Thus for combined lighting and power 50 or 60 are the most usual frequencies; but if the system is designed solely or chiefly for the distribution of power, a still lower frequency is preferable. On this account 25 was selected by the engineers for the Niagara Falls power transmission, after careful consideration of the problem, and this frequency has since been widely adopted in similar cases.

Alternator construction.