]\gt ; ' Accad .
Lincei Atti , ' vol. 15 , pp. 63-74 .
' Nature , ' vol. 62 , p. 413 .
3 .
'Wied .
Ann vol. 26 , p. 427 .
S 'Soc .
Int. Elect. Bull 1911 , Ser. 3 , vol. 1 , pp. 49-57 .
'Ann . .
Physik , ' vol. 27 ( 1 ) , pp. 64-82 .
'Deutsch .
Phys. Gesell .
Verh vol. , pp. 185-204 ; 'Phi ] .
Mag. ' ) , vol. 3 , pp. 500-512 .
'Comptes Rendus , ' vol. 132 , pp. 917-920 .
'Ann . .
Physik , ' vol. 12 ( 4 ) , pp. SS 'Roy .
Soc. Proc vol. 70 , p. 398 .
'Comptes Rendus , ' June , 1894 .
lr 'Recent Researches in Electricity and Magnetism , ' p. 302 .
Phif .
Mag. ' 5 ) , vol. 38 , p. 426 .
'Wien .
Akad .
Sitz , p. 330 .
'Ann . .
Physik , ' vol. 12 ( 4 ) , pp. 869-874 .
SSS 'Comptes Rendus , ' June 11 , 1894 .
, etc. , under .
255 Method of Observation .
2 .
The method adopted in this investigation was to photograph the image .
of the spark produced by the condenser discharge reflected from a mirror ( see fig. 1 ) .
This method was similar to that originally adopted by : : : FIG. 1.\mdash ; Diagram to show arrangement of circuit .
Feddersen and Schiller , but in this case the surface of the mirror was flat and the light from the spark was focussed by a lens on to the photographic plate .
The mirror used in most of the experiments was formed with four silvered glass plates supported in a square aluminium frame , each milTor being 1 inch by 4 inches .
This could be run with safety up to 100 revolutions per second .
For higher speeds a small silver mirror was used with a reflecting face inch by 2 inches , which was run up to 230 .
revolutions per second .
The speed of the mirror was determined by a contact wheel driven by a worm cut in the shaft of the mirror , the contacts being arranged in one of the circuits of a double chronograph , the second circuit was connected to a clock giving seconds , and the speed could be measured to within* .
per cent. The spark gap was made with brass balls and its length was measured by a micrometer .
The circuit of the spark was closed through a contact-maker formed of a brush of fine wires attached to the frame of the mirror , which rubbed against a brass plate .
This contact-maker gave excellent results , chiefly because of the high speed at which the brush was revolved .
3 .
The condensers used consisted , in the first experiments , of a number of Leyden jars each having approximately microfarad capacity .
A second small condenser , sometimes used , consisted of nine plates of glass , 2 feet square , on which were pasted sheets of tinfoil 18 inches square , the glass being carefully shellacked and the whole well insulated .
The third condenser ( the one chiefly used ) consisted of 10 sets of 10 glass plates 3 feet square , coated with tinfoil 2 feet square , the glass being FIO .
2 .
Iron , etc. , under the Osciltatory Discharge .
257 different layers .
Each side of the square coil contained 90 turns , the diameter of each turn being approximately cm .
( see fig. 2 ) .
The object of using a square coil was to obtain , as far as possible , a closed magnetic circuit of the substances which were introduced , generally iron wire cores , as by this means calculation would be simplified .
With the assistance of Mr. Duddell , to whom the author is much indebted , $ very .
sensitive measuring apparatus was arranged , with the permission of 1 the late Prof. Ayrton , at the Central Technical College , for the measurement of this self-induction .
With this apparatus the value henrys was obtained correct to per cent. for the single-layer coil .
The resistance of this coil at C. was ohm .
Another square coil was used to obtain lower frequencies ; in this there were seven layers of wire on each side of the square , the whole containing 2670 turris of wire .
The self-induction of this coil was also carefully measured with a secohmmeter , found to be millihenrys .
The resistance of coil at C. was ohms .
A coil , used chietly preliminary experiments , consisted of a single bobbin of the same wire wound in seven layers ; this contained 750 turns of wire , and had a self-induction of app 1 millihenry .
Its self-induction was not measured by the secohmmeter , but was obtained by comparison with the square seven-layer coil by means of the air condenser .
The results of these experiments are iven later .
Measurement of the Pfates .
: 5 .
In order to obtain accurate measurement of the spark photographs , alarge mea microme'ter with a low-power microscope , with a oross wire in the eye-piece .
In most cases it was found possible to obtain measure- ments of not very quickly damped sparks accurately to within 2 per cent. , but with sparks whose period varied the acCuracy was very much lower , as black spots were often spread out and asymmetrical .
Theory of nents .
6 .
From the fundamental equations for the ischarge of a oondenser of farads capacity through a coil of self-induction henrys , and with a total resistance of ohms in the circuit , the-frequency of is given by .
The current at any instant pt amperes , where frequency and damping of the oscillations ; while even with fine iron wire cores the number of vibrations in many cases is reduced from 10 or 15 to 3 or 4 .
According to the investigation given above the damping factor is ; and accordingly , as increases , remaining constant ( the change caned by introducing the iron wire is minished , and hence the damping should be less rapid with the larger self-induction .
It is clear , therefore , that Phil. Mag January , 1903 , .
lffi .
'Recent Researches in Magnetism and Electricity , ' p. $04 .
Iron , etc. , umder the Oscillatory the cause of the very rapid damping must be the absorption of by the cores introduced , either by hysteresis or by eddy currents , or both .
Using the same notation as above , in .
which the magnetic force is represented by the real part of , the magnetic induction being assumed uniform along the axis of the wire , the rate of heat production per unit length of wire due to eddy currents is equal to the mean value of\mdash ; Sa real part of ) pt , where magnetic permeability , frequency , specific resistance of iron , radius of iron wire .
When is small the heat produced per unit length of core per second , being the number of wires and the maximum magnetising force .
The experiments of Sir J. J. Thomson .
( loc. cit. ) appeared to show that the hysteresis of iron accounted for a considerable part of the absorption of energy .
In these experiments it will be shown that hysteresis loss is almost negligible as compared with eddy-current loss .
Results of Experiments .
Part I. 9 .
The first series of was made in order to find how nearly the frequency determined by experiment agreed with that calculated from the observed self-induction and capacities used in the tests .
The tests made with the air condenser may first be considered .
The capacity of the air condenser was obtained both by calculation and by comparison with the standard 12-microfarad condenser .
The capacity found in this way microfarads .
On account of the bending of the plates , lt was found impossible to obtain a spark more than mm. long , and some difficulty was found in photographing this , as well as in maintaining the discharge by the large Wimshurst machine .
A series of determinations was first made of the self-induction of the circuit apart from the coil .
Using the small coil , the time for a complete 'Recent Researches in Electricity and Magnetism , ' J. J. Thomson , p. $20 .
These results are not given at great length as the paper by Battelli and Magri ( ' Phil. Mag June , 1903 , p. 620 ) covers the subject quite fully .
0.654 8.18 I-VII-V 0 s.5584 0.96 4.2 4.46 1.22 4.68 1.35 In all these experiments the spark was mm. and the values of are accurate to within 2 per cent. 11 .
In order to see whether the resistance .
of the circuit can have any effect on the frequenoy of the oscitlations ; the ce or the Iron , etc. , under the Oscillatory wire has been calculated from the expressions originally given by Lord Rayleigh .
For the highest frequencies used\mdash ; ( resistance to steady current ) .
Since ohm , ohm .
For this frequency , and The effect of the resistance of the circuit in this case is therefore negligible .
For the lower frequencies , when using the single-layer square coil , the total resistance neglecting the spark may be taken ohm , and so that in this case also the correction is less than 1 per cent. When the coil is used with an iron core , the value of is increased , and the co1Tection will therefore be smaller still .
It will be shown later that when the maximum current during an oscillation of the spark falls to less than 200 amperes , the resistance of the spark increases to about 2 ohms , and may be expected to become larger still as the current through the spark diminishes still further .
Under certain conditions , therefore , the spark resistance will have to be taken into account .
Part II .
12 .
The results of the experiments made to determine the effect of currents at high frequencies on iron may now be considered .
With freuencies .
from 100,000 to 5000 per second , the effect of an iron or nickel wire core is to produce a spark discharge which has no constant frequency , but in which the period for a half-oscillation rapidly increases as the discharge becomes damped by energy absorption .
* Having observed the effect , the explanation is obvious .
It is clear that at these frequencies iron retains very much the same properties as it does under a steady magnetic field , i.e. the permeability of the iron increases as the flux density in the iron diminishes .
13 .
Experiments have been made with a variety of soft iron wire cores , and it will be convenient first to consider the effects produced with discharges from the set of ten large condensers through the single-layer coil , when the latter is arranged with an iron wire core .
* This phenomenon has now been Battelli and Magri To obtain the " " effective permeability\ldquo ; of the iron the lines of force may be assumed to be uniform density ; on the introduction of an iron *The reason for this is clear from the curves drawn graphically in fig. 5 .
Phil. Mug Jsnmry , Iron , etc. , urbder the Oscillatory Dischvxrge .
263 oore a certain number of lines go through the iron .
We may put , there fore , , where is the area of one turn of the coil ( calculated from the mean value of the self-induction ) and is a constant .
When a core is introduced , where is the of the iron core , and is a quantity which may be designated the 1 " " effective\ldquo ; permeability .
Hence The quantity may be defined as the permeability of the iron which * would give the same period of oscillation for the spark as is obtained 4 experimentaly , the " " effective permeability\ldquo ; being constant and independent .
of the value of the magnetising force .
With the single-layer square coil sq .
cm .
sq .
cm .
( 1 ) Area of section of the No. 28 S.W.G. iron wire core ( 2 ) ( 3 ) ( 4 ) , , , 28 , , nickel , , 14 .
In the table are shown the results obtained when the discharge from No. 1 set of the large condenser flows through the singlelayer square eoil into which a core [ ( 1 ) above ] made up of 660 No. 28 S.W.G. iron wires has been inserted , the spark gap is mm. , corresponding to a P.D. before discharge of 9400 volts .
Large condenser , No. I set .
sec. , , volts sec. T. Lx 18.4 164.0 .
In the first place , the value of the maximum current during the first half-oscillation was estimated , and was found to be 210 amperes .
The corresponding value of the magnetising force due to this current C.G.S. units approximately .
The value of corresponding to this , obtained from Ewing 's results for Low Moor cvht iron , giving a value of approximately , a result which is of the same order of magnitude as the effective permeability calculated above .
M. Klemencio* and others that the hysteresis loss is greater at high ' Wien .
Akad .
Sitzberg , ' 1898 , vol. 107 , pp. 330-360 .
M. Klemencic seems to have determined the hysteresis loss by a of the decrement of in a Leyden jar discharge .
No allusion is made to which to be ced for eddy current losses .
also Warburg and Honig , -Wied .
, p. ; ' Phil. Mag Sspt9 1889 ; Battelh and Magri , [ Aecad .
} Atti , ' vol. 15 , pp. 485-492 ; Corbino , 'Accad .
Lincei Atti , ' vol. 16 , pp. ; 'Atti dell ' ASSOG Elett vol. 7 , p. 606 ; F. Piola , ' Elett .
Romi , ' vol. 6 , pp. 4-6 .
Iron , etc. , under the Oscilla tory frequencies than low .
Now , the total volume of the iron in the cores .
approx. Hence the total loss of energy due to hysteresis for the half-oscillation .
This loss is quite small as compared with the eddy-current losses .
The loss due to eddy currents during the first half-oscillation calculated from the formulae given above ergs .
Adding to this the hysteresis loss as calculated above , the total iron loss The ohmic loss due to the resistance of the coil is compalatively small .
In this case , assuming the root mean square value of the current to be of the maximum value , it amounts to ergs .
for the first half-oscillation .
The total energy stored ergs .
The value given above shows that the eddy-current and hysteresis losses amount to more than one-third of the total energy of the discharge during the first half-oscillation , whereas the loss due to ohmic resistance is only of the total energy .
The damping of the oscillations with iron cores might be expected to be much more rapid than when the cores are not present , and this is precisely what is observed .
The total number of half-oscillations on the plate with the iron cores is three , whereas without them ten half-oscillations , and even more , have been photographed .
This result nfirms the observation that the permeability of the iron is not very different at this frequency from that observed under similar conditions with a steady magnetising force , for , if the magnetic induction were very different from the value that has been assumed , the total calculated eddy-current loss would not have agreed so nearly with experiment .
19 .
The other results obtained with the set of large condensers may now be dealt with .
Ihese will correspond with lower frequencies than those considered in the preceding paragraphs .
The results are tabulated according to the method described in S13 .
Large condensers , Nos. , volts . .
llaximum ourrent during first oscillation amperes .
Naximum magnetising force , , , , C.G.S. units .
Estimated maximum magnetising force first oscillation - C.G.S. units .
Considering , in the first instance , the first half-oscillation of the spark after the discharge begins , it will be noticed that when iron cores are intro duced the change in frequency produced by using different capacities is comparatively slight , varying from 28 , per second with microfarad to 18,800\mdash ; per second with microfarad .
It will also be noticed that the decrease of effective permeability with increase of capacity is not quite uniform , but this is amply accounted for by possible errors in observation . !
: Iron , etc. , under the .
267 curve has been plotted between the values of the effective permeability maximum magnetising force , and is shown in fig. 4 .
20 .
Considering next the half-oscillations of the spark other than the first it will be seen that the same law holds .
Increase in capacity diminishes the apparent effective permeability .
Thus the second half-oscillations with microfarad corresponds with a greater value of the effeotive permeability .
It is noticeable from these figures that the damping of the discharge is of sine wave shape .
6 Fromtheresultsobtainedwiththelargestcondenseranattempthasbeenpercent.fromthatobtainedbyassumingbhecalculated\ldquo ; current t made to determine the magnitude of the resistance of the spark itself .
From the values calculated from the\ldquo ; effective\ldquo ; permeability the corresponding value of\ldquo ; \ldquo ; for each half-oscillation of the spark can be found , and hence the maximum value of the current , assuming a sine wave shape for the curve as a first approximation .
From this value for the current , the maximum voltage at the condenser at the beginning of each half-oscillation has been determined , and hence the loss of energy during each half-oscillation * The values of spark reeistance referred to later have been taken account of in drAwing this perceptible difference in the frequency of the oscillation .
* Large condenser , \mdash ; Capacity N.B.\mdash ; If the value of the spark tance for the 4th half-oscillation be assumed equal to 2 ohms , the total energy loss during the 4th half-oscillation becomes ergs .
Similar calculations have been made on other typical discharges , in particular , with the discharge of a condenser of 1 microfarad at an initial potentia ] difference of 7800 volts ; the average resistance for the spark- which it is necessary to assume to bring the energy losses into greement is 048 ohnl at 770 amperes , ohm at 440 amperes , and ohms at 230 amperes .
The accuracy of these figures is not great , and they are of value as showing the order of magnitude of the quantity involved , and in this respect they confirm the results which have been ooiven above .
21 .
Experiments were made in order to determine how the spark and , consequently , the potential difference at the spark gap just before the discharges , affected the observed effect .
It is clear that a shorter spark will give a smaller maximum current during the first half-oscillation , and hence , to the above results , the apparent self-induction of the coil should be increased .
Below given the results obtained with spark gaps 2 mm. and .
in length .
This result is in accordance with that obtained by Battelli and Magri , 'Phil .
Mag June , 1903 , vol. 5 , pp. 620-643 .
VOL. X lron , etc. , under the Discharge . .
Large condenser , No. sec. , volts T. Lx 0.048 6.5 5463 4443 0 .
lIaximum ourrent during first half-oscillation amperes .
Maximum magnetising force , , , , C.G.S. units .
Large condenser , .
Nos. sec. , volts sec. T. 6 .
54.1 45.0 Maximum current duriug half-ost illation amperes .
Maximum magnetising folce , , , , C.G.S. units .
aec .
Large condensers , Nos. sec. , volts T. Lx 0 10$x6..8x 0.158 Maximum magnetising fMaximum current during , , ation Large condenser , Nos. sec. , volts k- sec. T. Lx 7 .
' 16.4 0.12 0.139 8 .
21.0 16.6 ' Maximnm current during first hation resMaximum magnetising force , units .
Large condenser , Nos. VIII .
sec. , volts sec. T. Lx 0.06 12.6 8.5 0.150 0.17 9 .
16.0 0.20 0.24 ' 51.0 41.0 0.36 20 .
86.0 0.48 ?
225.0 193.0 Maximum current during first hation esMaximum magnetising force , nits .
Large condeIlser , Nos. sec. , volts sec. T. 0.14 8 .
11.7 0.18 16.6 0 .
42.0 34.0 52.0 Maximum current during first oscillation 570 amperes .
Maximum magnetising force .
, , , C.G.8 .
units .
Large condenser , Nos. sec. , volts sec. . .
0 .
0.0 8.6 4.4 18.2 11.4 0.24 ll Maximum magnetising force , unitsMaximum current during first , tion 24 .
These results appear , at first sight , somewhat remarkable , as it will be seen , on comparing them with the results obtained with a core composed of much finer wire , No. 28 S.W.G. , that there is very little difference in the value of in the tables .
On comparing the areas of the surface of the wires per centimetre length of the core , it is seen that , whereas for the No. 28 core the surface is sq .
cm .
, the surface for the No. 18 core is sq .
cm .
; Iron , etc. , under the Discharge .
snd since it is the surface of iron in the core which would appear to be most important , it might be expected that the change in frequency produced by the core would be much less noticeable .
The explanation , however , is obvious .
With the larger diameter wire core the loss due to eddy currents is greatly increased .
According to the approximate formula given above ( S8 ) it depends on the fourth power of the diameter , hence the loss of energy during the first half-oscillation would be more than ten times as great for a core having the same amount of iron in it , and made of the No. 18 wire , as it would be for one of No. 28 S.W.G. wire .
The actual magnetising current rapidly falls off , therefore , on account of this loss in energy .
There is another tending to diminish the magnetising force inside the iron , what is called the self-demagnetising force due to the magnetisation on the surface of the wire , at the place where the netic lines leak .
This netising force is increased with increase in diameter of the wire .
The diminution in the magnetising force owing to these causes will correspond with a greater permeability in the iron , which ( although the increase in permeability of the iron diminishes the depth to which the magnetisation penetrates ) is sufficient to counterbalance the much greater surface of iron exposed , with ] the finer core .
As in this case the permeability must clearly be higher than in the previous cases , we may assume taking into account the ratio of the surfaces ) , as a first approximation , [ As the results obtained with the large condenser , No. I , seem to be in very fair accordance with those given for the other sparks , the result for the first half-oscillation has been used as a basis of calculation .
] Proceeding exactly as in S16 , Hence cm .
The total induction through the core since the area of the iron sq .
cm .
This gives a value for the effective permeability , instead of as found by experiment .
It is of interest to observe the comparatively great lag in the maximum magnetic induction behind the maximum magnetising force .
With these The radius of the finer No. 28 is } cm .
, and of the No. 18 wire is cm .
The energy loss for the same mass of iron will be proportional to the squares of the diameters of the wires .
larger and smaller diameter .
The results have been arranged below the same way as the previous ones .
The notation of S13 has been adopted .
Large condenser , No. ' sec. , volts seo .
' .
3 .
0.11 0.12 Maximum magnetising force , lfaximumcurrent during ftion Large condenser , Nos. sec. , volts sec. T. 0.110 4 .
14.6 4 .
15.4 Maximum mgnetising force , ximum current during first hlation Large condenser , Nos , sec. , volts sec. T. 0.143 5 .
12.0 21.6 16.4 24.1 19.0 Maximum current during first half-oscillation amperes .
Maximum magnetising for , , , , C.G.S. units .
Large condenser , Nos. sec. , volts sec. Lx 12.8 0.20 0.216 8 .
20.0 16.4 0.228 8 .
22.4 18.0 0.34 21.0 Maxi netiMaximum oduring first half-oscillation The value of , the area of iron in this core , is equal to sq .
cm .
TIME IN OF A SECOND FIG. 5.\mdash ; Curves showing increase in permeability during successive oscillations of discharge ( from condenser ) of iron wire ( No. 28 B.W.G. ) core placed in coil through which discharge passes .
Numbers opposite are maximum magnetising forces during first half-oscillation .
Part III .
lVickpl Wire Cores .
29 .
Besides the experiments made on iron as described in Part II , a series !
of through coils having cores of nickel wire was photographed , the results from which are tabulated below .
It will be seen that the phenomena are similar to those obtained with iron cores , though the effects : : are less marked with the nickel wires .
The core consisted of 500 No. 28 wires , sq .
cm .
Maximum magnetising force , nits .
Large condenser , Nos. sec. , volts * Maximum current during first half-oscillation amperes .
Maximum magnetising force , , , , C.G.S. units .
Large condenser , Nos. , volts sec. 1 . . .
0 .
4 .
1.7 ' ll Maximum current dul:ing first half-oscillation S80 amperes .
Maximum magnotising force , , , , ) C.G.S. units .
30 .
We may proceed in exactly the same way as in the case of the No. 28 iron wire cores to determine more exactly the actual permeability necessary to produce the observed results , making allowances for the " " skin effect First , calc.ulating the value of may be assumed as a first approximation , C.G.S. units , whence cm .
Hence the magnetic induction through a single wire oos , where , giving an effective permeability which does not differ appreciably from 7 .
The lag in maximum netic induction behind maximum magnetising force is 1leglecting ysteresis .
mental error .
: When a similar core of iron wires was inserted , all the characteristic effects both of change of frequency and of damping were observed .
When a core of solid brass was introduced , the time for a complete oscillation seconds , thus showing an increase in frequency as found by Hemsalech .
With solid soft iron core the number of half-oscillations visible on the plate is reduced to two , thus showing the very great eddy-current loss ; while * Results published by Hemsalech confirm some of these observations .
'Journ .
Phys 1908 , pp. 76-90 ; 'Comptes Rendus , ' vol. 140 , pp. 1322-1326 .
'Comptes Rendus , ' vol. 140 , pp. 1322-1325 .
indicate a slight decrease in frequency , due to the permeability of the iron .
With the solid iron core , the permeability , therefore , is sufficiently great to overcome the effect of eddy currents , and tends to.increase the selfinduction , and to produce a slight decrease in the frequency .
32 .
It may be of interest here to notice some experiments in which a coil similar to the above was wound on a brass tube , about 1/ 32 inch thick .
In this case the brass seemed to screen off the action of the iron entirely , and no difference whatever could be observed when the iron cores were introduced , either in damping or in change of frequency , this result is also in agreement with that obtained by Hemsalech using a tube of zinc ( loc. cit Genered .
1 .
With an air condenser , the capacity of which was measured ballistically , and an air-core self-induction coil of known value , the frequency of the oscmations of the condenser discharge agrees ( within limits of experimental error ) with the values calculated by Kelvin 's formula .
2 .
The resistance of a spark between spheres of 1 inch diameter , mm. apart , has been estimated ; the resistance of such a spark when the maximum current is greater than 200 amperes does not exceed 2 ohms .
If the maximum current through the spark is greater than 500 amperes , the resistance does not exceed ohm .
Effecf , Produced by an lron Wire Core in the Self-induction .
3 .
When an iron wire is inserted in the self-induction coil , the time for each consecutive half-oscillation increases with the duration of the discharge .
4 .
With a series of discharges from a given condenser in which the maximum value of the magnetising current varies , the time for the first oscillation decreases with in the strength of the current .
5 .
The increase in the time of a half-oscillation with the duration of the discharge is due to the increase in the permeability of the iron , as the current , and consequently the magnetisin force , dies away .
6 .
The permeability of the iron wires calculated from the observed increase in the self-induction of the coil , decreases with increase of the magnetising force .
The curve connecting the ' effective\ldquo ; permeability , , the netising force , agrees , within limits of experimental error , with that obtained on the assumption that the permeability of the iron is the same as it is under a steady magnetising force .
oduced by of other Jfaterials . .
: 10 .
Nickel wire cores produce similar effects to those observed with iron , but of diminished intensity , the permeability of the nickel wire with rapidly oscillating magnetic force similar to that found with steady magnetising forces .
11 .
With cores made up of insulated copper wires , no , either in damping or self-induction , are apparent .
In conclusion , the author 's best thanks are due to the late Lord Blythswood for the loan of the apparatus described above , and to the late Lord Kelvin for the kind encouragement given by him during the early stages of the research , to Mr. Duddell for the measurements of the self-induction of the various coils used , and for his criticisms , and to Dr. Silvanus Ihompson for his criticisms and suggestions .

