]\gt ; On the Determination of thoe Absolute Unit of Resistance by Current Jlethods .
By ALBERT CAMPBELL , B.A. ( Communicated by R. T. Glazebrook , F.R.S. Received July 2 , 1912 .
) ( From the Natioual Physical Laboratory .
) CONTENTS .
PAGE 1 .
Introductory 391 2 .
Standard Mutual Inductance .391 3 .
Secondary Standard : Mutual Inductometer 396 4 .
Comparison of Resistance and Mutual Inductance by Two-phase Method 398 5 .
Method of Intermediary Condenser 402 1 .
Introductory.\mdash ; Recently at the Physical Laboratory we have constructed a standard of mutual inductance of novel type , whose value has been accurately calculated from the dimensions .
This inductance has formed the basis for the determination of the unit of resistance in absolute measure by two different methods , in both of which alternating current is employed .
Although there is no doubt that the accuracy attainable by these methods could be increased by greater elaboration of the apparatus used , the results already obtained seem to be of sufficient interest to warrant publication .
It should be mentioned that the accuracy here aimed at was of a considerably lower order than that contemplated in the determination of the ohm by the Lorenz apparatus which is at present out in the laboratory .
For the experiments here described , no apparatus was specially constructed , but use was made of instruments which had already been and set up for the measurement of inductance and capacity .
I shall first a brief description of the standard inductance and then pass on to the methods and results .
2 .
Standard Mutual Inductance.\mdash ; The design of the mutual inductance has already been described .
* The electrical circuits have the form and arrangement shown in section in fig. 1 .
The primary circuit consists of two single-layer coils connected in series , while the secondary is a coil of many layers wound in a com- paratively narrow channel .
The dimensions and positions of the coils are such that the magnetic field due to current in and is practically zero over the space occupied by the 'Roy .
Soc. Proc May , 1907 , , vol. 79 , p. 428 .
Mr. A. Campbell .
Absolute Unit of [ July 2 , windings of H. Thus ( a ) the calculation of the mutual inductance can be made with accuracy without an accurate of the dimensions of the many-layered coil , provided the dimensions of and can be accurately FIG. 1 .
measured ; ( b ) slight eccentric or axial displacements of the coil from the symmetrical position have very mall effect on the value of , and hence the of into its proper position can be done without difficulty .
The actual construction is shown in fig. 2 .
The coils and consisted of bare hard copper wire ( diameter mm. ) wound in right and left handed screw threads cut on a single white marble cylinder , the pitch of the screw being 1 mm. The treatment of the marble and the process of winding of these coils were very similar to those used by Mr. F. E. Smith in the construction of the standard ampere balance , need not be further described here .
The same remark applies to the measurement of the dimensions of the ooils and , which was carried out by Mr. L. F. Richardson , of the Department .
As the distance from wire to wire of the was about mm. it was not considered necessary to employ bifilar winding .
After winding , the coils were dipped in melted paraffin wax .
The leads were led out parallel to the axis and almost touching the outer surface of the winding .
The coil was of nearly square section and was wound of double silk covered copper wire ( diameter mm. ) in a channel accurately turned in a large white marble ring , supported by a ring of built mahogany having three solid ebonite feet and provided with screws by which the level and centrality could be independently adjusted .
The main part of coil was in four nearly equal sections brought 'Phil .
Trans , vol. 207 .
1912 .
] by Alternating Current Methods .
393 out to separate leads and having 97 , 145 , 145 , and 98 turns respectively .
thin strip of paper was interposed between each layer and the whole was well soaked in melted paraffin after winding .
An auxiliary coil ( for 1 millihenry ) was wound over the main coil ; it was not , however , used in the present iments .
Outside of this was wound a single turn which was connected in series with the main coil so as to make the total value of the mutual inductance almost exactly millihenries .
During winding , the number of turns was recorded by a counter , and when the coil was complete , the of the counting and the sufficiency of the insulation were checked by separately the mutual inductance of each section , using and as primary .
Duling the winding the irth of each layer was tested by a carefully calibrated steel tape .
The dimensions at C. were as follows:\mdash ; Coils and D.\mdash ; 75 turns each .
Mean diameter , cm .
1)istance between inner ends of coils , cm .
; between outer ends , Main Coil H.\mdash ; 485 turns .
Mean diameter , cm .
Axial depth , cm .
Radial depth , cm .
Mr. A. Campbell .
Absolute Unit of [ July 2 , The calculation of the mutual inductance was made in the following manner .
To with , an arrangement of coils was chosen , having the dimensions of and round numbers ; for convenience I shall refer to as " " the model it was not actually constructed .
The dimensions as shown in fig. 1 were , and cm .
respectively .
The mean radius of the main coil was chosen so as to give maximum mutual inductance between its central filament and the coils and .
This was done as already described* by calculating the mutual inductances between and connected in series and a series of circles midway between them of radii and cm .
respectively .
The results of the more accurate recalculation are give1l in able I the product of primary and secondary turns , being taken as 100,000 .
In the last column are given the corresponding values of Table I. cm .
, to 10 cm .
, If , then from the numbers in the table we find Hence is a maximum for variation of when ; also , when , we have For , the maximum value Integration over Cross-section of Main Coil.\mdash ; The mean radius of the actual main coil is cm .
When this is reduced by dividing by to corre- spond with the model ws have .
As the variation of with is very small , this value may be taken as a close enough approximation to , the value giving maximum M. This maximum refers to a single circle , the central filament of the rectangular cross-section of the 1912 .
] Resistance by Alternating Current Methods .
main coil .
In order to derive from ) due to the whole coil I make use of a theorem kindly communicated to me by Mr. G. F. C. Searle ; it gives in the present case , ( 1 ) where radial depth , axial depth of cross-section , and The second term in ( 1 ) vanishes , since , and the terms beyond that in will be found to be quite negligible .
For the model , 2 and 2 cm .
Thus we obtain millihenries .
to Dimensions.\mdash ; The actual dimensions of the coils are\mdash ; cm .
, cm .
, cm .
Dividing by , we have cm .
cm .
cm .
The differences of the last numbers from 5 and 10 give respectively and Thus is equivalent to a model with primary turns , and the annuluses ( equivalent to circles ) at distances 5 and 10 cm .
from the secondary coil , their turns being and respectively .
The fraction 5 The mutual inductances for the circles were calculated by the formula where and and are complete elliptic integrals ; for the inductances were and millihenry ( in a total of ) , a factor of Hence for the actual coils , in which millihenries .
Mr. A. Campbell .
Absolute Unit of [ July 2 , Prof. B. Ll .
Jones , of the Presidency College , Madras , very kindly carlied out the calculation by an entirely different method , using the series formula developed by Rosa*extended to nine terms to ensure accuracy of 1 in 100,000 , and applying the Purkiss formula to integrate over the section of the secondary coil .
He obtained as result millihenries .
The agreement between the two methods is very sa , tisfactory .
It thus appears that the calculation is accurate to within 1 in 100,000 .
The probable error in due to the uncertainties in measuring the dimensions may be taken as follows:\mdash ; Parts in 100,000 .
Diameter of single-layer coils channel coil Axial lengths Accordingly the total probable error in may be taken as in 100,000 .
3 .
: Mutual Inductorneter .
\mdash ; The primary standard already described is not of the most suitable form for direct comparison with resistance coils or other inductances ; the resistance of the secondary coil is rather high , being 140 ohms , and , besides , it possesses quite appreciable distributed capacity .
For this reason and for other purposes a secondary standard was constructed in the form of a mutual inductometer ( variable mutual inductance ) of very wide range .
It was of the type already described by me in which the larger subdivisions are obtained by multiple-stranded fixed secondary coils , while the smaller ones are given by a movable secondary coil which can be turned in a plane parallel to two primary coils and midway between them .
Its higher is from 11,000 microhenries down to microhenry , and the scale can be read to an accuracy of at least microhenry at any point .
For permanence the main working parts , including the bobbins the coils , were made of white marble .
To avoid eddy currents the coils were wound of highly stranded wire ( to mm. diameter ) and all metal was avoided as far as possible near the coils .
The proportionality of the ions throughout the whole range was checked both by Maxwell 's method of comparing mutual inductances and by other methods similar to those used in accurate resistance measurement where a building-up system is employed .
The errors from exact proportionality were very small , and were taken account of in all the experiments .
The inductometer was tested against the primary standard close to the value 10,000 microhenries by the method shown in fig. 3 , but before this test 'Bulletin of Bureau of Standards , ' No. 3 .
Phys. Soc. Proc vol. 21 , p. 69 ; and ' Phil. Mag January , 1908 , p. 155 .
1912 .
] Resistance by Alternating Current Methods .
the secondary coil of the primary standard ( I ) was adjusted to be ( a ) exactly midway between the two primary coils , and ( b ) coaxial with them .
Adjustment was effected by connecting the two primary coils of ( I ) in series but opposing one another magnetically , while the secondary was connected to a telephone .
Alternating current at ) second was passed through the primary circuit and the position of the secondary coil was adjusted until silence was reached in the telephone .
Adjustment ( b ) was made the comparison with the inductometer , the relative positions of the coil-axes being altered until the minimum reading on the inductometer was obtained .
FIG. ,3 .
In fig. 3 the inductometer ( m ) and the standard ( I ) have their primary coils connected in series to , a source of alternating current of frequency their secondary coils being connected in series ( and opposition ) to a tuned vibration galvanometer G. By varying the value of an approximate balance can be obtained ; the want of exact balance is due to the difference in distributed capacity of the two secondaries ( and of the primaries also ) .
The balance can be made exact by adjusting a variable condenser connected across the seQondary of ( in our case ) .
If the resultant distributed capacity effects be represented by a capacity across , and if , and be the self-inductances and resistances as in the figure , and , it is easy to show that .
( 2 ) When is nearly equal to ?
, we have as a sufficient approximation .
( 3 ) The actual values were heury , ohms , Hence millihenries .
heury .
oh1ns .
Mr. A. Campbell .
A bsolute Unit of [ July 2 , The comparisons were made at three frequencies .
For the alternating current was replaced by direct current with a single reversal , and the vibration galvanometer by a ballistic one .
In Table II are given the results .
Table II .
ApparentError ofPar .
The corrected values show satisfactory yreement , and thus when the inductometer reads 10000 , the true value is microhenries .
It was found that when a current of 1 ampere was passed through the primary circuit of the inductometer for several minutes , the slight of the coils caused a gradual change ( of the order of 1 to 2 parts in 100,000 ) in the error .
A small correction for this was applied when necessary in the experiments described below .
4 .
of esistance and Mutual by Two-Phase JIethod .
The most direct comparison of resistance with mutual inductance was made by the method already briefly described by m which two-phase alternating currents are used .
In fig. 4 , is the inductometer , a standatd oil-cooled resistance , and FIG. 4.\mdash ; Two-phase Method .
'Roy .
Soc. Proc 1908 , vol. 81 , p. 450 .
1912 .
] Resistance b.y Alternating Current Methods .
a vibration galvanometer tuned to the frequency of the alternator .
Let A and be the instantaneous values of currents in quadrature ( as shown ) ; then if alvanometer shows no deflection .
( 4 ) In practice A is very nearly equal to , but is not perfectly non-inductive .
Let its self-inductance be , causing circuit A to be slightly out of quadrature with by a small .
Then for a balance ( when ) we have ; or , since is small compared with Thus , as before , and If is even as as 1/ 1000 , will not differ from by more than 1 in 2,000,000 .
The small phase displace1oent is balanced in the adjustment of .
The actual value of for the coil used was microhenry , while was microhenries ; thus the correction is quite negligible .
Determination of \mdash ; The ratio of A to ( which is taken very nearly equal to unity ) is found by observing the effective values of the two currents by the respective potential differencesproduced in the ( very nearly ) equal resistances and , tb es voltages being read alternately on a very sensitive electrostatic voltmeter .
If the wave forms are not exact sine curves , the instantaneous values of the currents may be written and respectively , where , 3 , In this case the ratio of the effective values will be .
( 5 Now in the alternator used , the wave forms had been carefully traced and it was known that and were neither of them of an order greater than 1/ 100 ; hence expression ( 5 ) will not differ from by more than 1 part in 20,000 .
The error , if any , can , howeyer , be ehminated by repeating the experiment with the coils of the alternator interchanged .
of Distributed Capacity Inductometer.\mdash ; The coils of the inductometer have small distributed capacities , the total effect of which may represented by a small capacity across the ends of the secondal'y coil of self inductance L. It is easy to show that for equation ( 4 ) we must write .
( C ) Mr. A. Campbell .
Absolute Unit of [ July 2 , The correction is proportional to the square of the frequency .
As will be shown later , it was evaltlaCed by testing an air condenser against the inductometer at various up to second .
At second , the used , the correction was approximately 6 in 1,000,000 .
of \mdash ; By a worm and wheel connected to the nator spindle , at every revolutions an electr.ic contact was made and this caused a record to be made by the pen of a raph to which at the same time the standard clock was sending signals each second .
A comparison of their respective marks over a sufficiently long interval gave the avel a frequency .
of \mdash ; The alternator was run at as steady a speed as possible , the vibration galvanometer was tuned to resonance with the frequency , and at every half minute a balance was obtained by slight alteration of the , and also of the phase of the circuit , the latter adjustment being made by a small variable self-inductance ( fig. 4 ) .
At the a second observer switched the voltmeter allernately from to , obtaining a at each half minute .
The ratio was usually about and thus the readings were all taken at very nearly the same part of the scale of the eCer .
The scale was very open , giving about 2 mm. for a difference of 1 part in 10,000 .
The measured part of each run lasted 15 minutes , and the mean frequency was alwnys detel.mined over the whole interval ; the maximum variation from the mean to 3 parts in 1000 .
The vibration ) alvanometel .
was of t , he -coil type such as I have already described As it gave a deflection of 10 mm. at 1 metre for a current of 1 microampere at 100 ] ) second , it was ) sensitive for the purpose and usually had a resistance placed in series with it .
The resistances and consisted of two large frames of wire ( wound non-inductively ) ; they were nominally 100 ohms each , the ratio being .
The resistance was a )-ohm coil of silk- covered manganin wire wound on a mica frame and immersed in oil .
Its value at C. was found , by comparison with the laboratory standards , to be international ohms , with a temperature coefficient of per degree C. In each experiment the mean temperature of the oil was observed and the result reduced to C. by the above coefficient .
Corrections.\mdash ; In addition to the resistance temperature corrections , the following corrections were applied to the value of deduced from each experiment by formula ( 4 ) 'Phil .
October , 1907 , p. 498 .
1912 .
] by Alternating Iethods .
Correction .
Parts in 100,000 .
For clock rate For heating of inductometer For capacity of , , Results.\mdash ; In Table are shown the final results of expe1ineuCs , the second column the values of deduced from , while the third column gives the ratio of this absolute value of to the value measured in international ohms .
The experiments are arranged in pairs , as systenlatic interchanges were made ( for each ) in the connections of the two pairs of leads to the voltmeters and those to , in order to eliminate as far as possible the slight inequality in capacity and leakage of the leads and any effect of stray field from the coil .
In experiments and the connections to the alternator coils were altered from the arrangement in the other experiments so as to interchange the positions of coils relative to and M. The mean of D6 and D8 is so near the eneral meam that the whole set have been together .
Table It will be noticed in the third column , the reatest variation from the mean is 8 in 100,000 .
This is probably only partly due to errors of experiments , as the of the leads , etc. , was expected to cause small variations in the conditions .
The general mean gives the value of the ratio of the international ohm to the true ohm as The probable error is discussed at the end of the next ) ( S5 ) .
With regard to the above method it may be remarked that the resistance evaluated is much higher than that given by methods like that of Lorenz ; also that the method affords a means of inductances in terms of known resistances .
VOL. LXXXVII.\mdash ; A. Mr. A. Campbell .
Absolute Unit of [ July 2 , 5 .
of \mdash ; In the second method the capacity of a condenser is evaluated in terms of a resistance ( B ) and a frequency by Maxwell 's comlnutator method , and in terms of two resistances and P ) and a calculated mutual inductance by Carey Foster 's method .
From a comparison of the two values we obtain in absolute measure the value of the unit in which the resistance been measured .
Thus in fig. 5 , if the frequency of ( and discharge ) by the commutator is per second , then where is the small correction depending on the resistances of the galvanometer , battery , etc. As will be shown below , in the Carey Foster method , corrections , PKR M. Hence , ( 8 ) which gives in terms of a mutund inductance and a frequency .
JTethod.\mdash ; Maxwell 's method is now so familiar that it is unnecessary to do more than mention some of the details of the setting ) .
The rotating commutator was very similar in design to that used by Thomson and Searle in 1889 , but special care was taken with the insulation .
The frequency was held extremely steady , and was determined by a chronograph just as in the two-phase method .
The galvanometer had a resistance of about 700 ohms , and the arms and were usually nominally 10 and 2,000 ohms respectively , their exact ratio being very carefully checked from time to time .
The SenSitiyity was of the order of 4 to 8 mm. deflection at 2 meters ' scale-distance for a of 1 in 100,000 .
The battery was reversed in the middle of each run . ?
Foster Method.\mdash ; The Carey Foster method of capacity is not so familiar , although in our experience it is one of the most direct and convenient .
When current is used , it is essential to employ Heydweiller ' modification in which an adjustable resistance is placed in series with the condenser .
For the accuracy here required we must consider the general case where none of the resistance coils employed can be assumed to be quite noninductive .
In fig. 6 let be the alternator , the vibration galvanometer or telephone , and the condenser ; and let A be the adjustable coils of the mutual inductometer , the fixed coils being in the arm BC with a 'Ann .
der Phys 1894 , vol. 53 , p. 499 .
312 .
] Resistance Current Methods .
suitable added resistance .
Let , and be the resistauces and self-inductances respectively of the three bran ches as shown including a part the absorption in the condenser ) .
When , by adjustment of the inductometel and , the current in the is reduced to FIG. 6 .
FIG. 6 .
zero , let be the of the inductometer , and let the instantaneous values of the currents be , and as .
Then the current in CD will also be .
Let ( being the frequency ) , and let Then , since at every moment the potentials of and are equal , or .
( 9 ) Also .
( 10 ) Hence .
Separating the real and inary parts , we have ( 11 ) and MS .
( 12 ) When the residual inductances and are yibly small , these equations reduce to the ordinary case and .
( 13 ) , ( 14 ) If be read in microhenries , while and are in ohms , will come out in microfarads .
By a suitable series of coils for and , capacities from a few microfarads vards can be directly measured .
If be obtained Mr. A. Campbell .
Absolute Unit of [ July 2 , by equation ( 12 ) while is the actual resistance added to the condenser branch , then enables the energy loss in to be evaluated .
If the condenser carries a current of effective value I of pure sine wave form at frequency then the power loss in the condenser is When the power factor is very small , as in good mica condensers , both and the residual inductance must be accurately determined .
In most of the present experiments the values of the quantities in equation ( 11 ) were approximately as follows:\mdash ; microhenries , .
It will be seen , therefore , that the correction involving in equation is less than one part in 1,000,000 and is igible for our purpose .
Effect of .\mdash ; The primary and secondary circuits of the mutual inductometer have distributed capacity , and this introduces a small error which may become quite appreciable at frequencies of 1000 or second .
With inductometers in which the subdivision is done by stranding the wires , the capacity effects are somewhat increased .
The following investigation deals the fairly ooeneral case in which both the primary and secondary circuits have each distributed capacity .
In fig. the distributed capacities be represented by capacities and in parallel with the coils of the inductometer , and being respectively FIG. 7 .
1912 .
] Resistance by Current Methods .
the resistances and self inductances of these coils and their mutual inductance .
Let the current in the galvanometer be zero , the instantaneous values of currents in the other branches being , , , as shown , and let instantaneous value of the potential difference between and F. Then we have and Also ; therefore .
( 15 ) Also But , and therefore Thus whence .
( 16 ) This expression for is applicable to other methods in which a mutual inductometer is used .
Since the galvanometer terminals are at the same potential we have or , ( 17 ) .
( 18 ) From these two equations and ( 16 ) , by eliminating , and only the real part , we obtain ( 19 ) In the actual inductometer used henry , henry , while and are each less than being about 18 ohms , and not greater than 6 ohms .
Thus up to second we may neglect and in equation ( 19 ) and we then have M KB .
( 20 ) Accordingly , the correction for distributed capacity is here practically proportional to the square of the frequency .
This was found experimentally to be true by testing air condensers ( presumably constant with frequency ) against the inductometer at various frequencies from 50 up to Mr. A. Campbell .
A bsolute Unit of [ July 2 , second .
From these and other experiments in which self-inductances with negligible self-capacities were tested , the error of the inductometer as used was found to be about parts in 1000 at 2000 per second , and hence about 1 in 100,000 at 100 per second , which was the order of the fi'equency used in the experiments .
This value of the correction was also corroborated by tests against another inductometer of later design in which the error was scarcely appreciable .
* Coefficicnt of Inductometer.\mdash ; Since the exlJeriments on the two-phase method were completed , the inductometer has been compared with the primary standard on several occasions and at ious temperatures .
In the results to standard temperature it has been assumed that the primary standard has a temperature coefficient equal to that of the white marble on which the coils are wound , for the value of the is almost entirely governed by the dimensions of the bare wire coils which are wound under iderable tension into screw threads on the marble cylinder .
A meall value of in 1,000,000 per degree C. was taken for this coefficient .
The resulting calibration values obtained for the inductometer when plotted yainst actual temperatures indicate that the inductometer has temperature coefficient of approximately in 1,000,000 per degree C. As might be expected this lies between the coefficients of marble and copper respectively ) since the coils are of stranded copper wire wound not too tightly on marble bobbins .
In Table are the results of the comparisons on the Table Calibration of Inductometer DecJune .
* In a mutual inductometer with a range from up to 10 millihenries , when care is taken to arrange the coils so as to minimise the capacity effects , good accuracy can be obtained even at frequencies of per sec. 1912 .
] Resistamce by Alternating Current Methods .
different dates , all reduced to a temperature of C. , the last column , differences from a slightly weighted mean value , 1lamely , 10 , which was that used finally .
In experiments , 8 , 9 , and 10 the temperatures were not so accurately known as in the others , by reason of quicker variation of room temperature and other causes .
When we take into account the fact that the were made over a period of two and a half years , at various temperatures , both by ballistic and alternating current methods , and with fresh adjustments of the primary standard on each occasion , the constancy of the inducton ) eter as shown by Table is satisfactory .
of \mdash ; The resistances used were of oil-cooled type and consisted of silk-covered manganin wire wound ( so as to very small in ductance ) on mica frames and shellacked .
Their values were determined in international ohms by Mr. : E. Smith by comparison with the laboratory standards .
Their temperature coefficients were determined , and hout the course of each experiment the temperatures were observed and the appropriate corrections made .
The olute value of the resistance was not required , as it is eliminated by the combination of the two sets of measurements .
In a few cases the two halves of the used in the Foster test were } ) in parallel to form the -arm in the commutator test , but this procedure still allowed the absolute value to be eliminated .
The resistance was made up of two parts\mdash ; an accurately known manganin coil , and the copper wire primary coil of the inductometer .
The coil being 18 ohms , formed a considerable fraction of the total , which was 200 ohms , and hence , before and after each reading of the illductometer , careful had to be taken of the resistance .
The current used was of the order of ampere through the primary coil , the source being a wire interrupter through a transformer with an earthed screen between the primary and secondaly coils .
To eliminate the effects of unbalanced capacities ( of bridge coils , etc. ) to earth , the point where the current entered the coil was put to earth , and , in addition , every reading was repeated with the leads from the source reversed , the mean of the two directions being taken .
The difference on reversal was very small , being only about or 4 parts in 100,000 .
In { cases ( as also in the comnrutator method ) the condenser was disconnected and measurements of the capacity of the leads , which was of the order of .
This was done without difficulty , for the scale of the inductometer is very open near the zero point .
In each test the current was on for only one or two minutes ; the temperature of the inductometer was usually deduced from the observed resistance of the primary coil .
Mr. A. Campbell .
Absolute Unit of [ July 2 , In a single set of tests the observations were generally made in the order:\mdash ; Temperatures ( of all coils ) .
Temperatures .
Resistance of primary coil .
Second Carey Foster tests .
First Carey Foster tests .
Resistance of coil Temperatures .
Temperatures .
Commutatol run ( 15 to 20 mins This procedure was necessitated by the fact that all the condensers tested had temperature the order of to parts in 10,000 per degree C. During the tests the condenser was kept in a well box , and the effect of change of temperature was largely elim inated by taking the mean of the first and second Carey Foster tests .
Variation of pacity due to Chamge of the purpose of comparisons such as are described above , air condensers would appear the most suitable , but when put to the test of experiment those available did not prove quite satisfactory , probably mainly for two reasons ; the total capacity .
was rather too small to give the required accuracy , and the whole volume of the condensers was so as to make the capacity of the outside to earth too prominent .
, all the comparisons were made with mica condensers .
The fact that all such condensers , in greater or less degree , show absorption , formed the crucial difficulty of the whole ation .
To elucidate the matter a variety of experiments were made , but I shall confine myself to describing only those that are important for the present purpose .
In both the Carey Foster and the commutator methods the effective capacity diminished in all cases as the frequency was raised .
For example in fig. 8 is shown the behaviour of one of the condensers ( A1 ) with Carey Foster 's method over a wide range of frequency .
Before examining and comparing actual curves of change with frequency , I would first draw attention to one of the fundamental differences between the two methods .
In the Carey Foster method the behaviour of the given condenser with frequency is entirely specified by an effecDive capacity in series with a resistance .
If this be equivalent to a capacity with a parallel resistance , then it is well known*that ( 21 ) and .
( 22 ) In most of the experiments for a frequency of second the term was of the order of to , and hence to the * M. Wien , ' Wied .
Ann , p. 681 .
Resistance by Alternating Current Methods .
aimed at ; thus the method here gives the effective capacity independent of the actual leakage resistance ( absorptive and ohmic ) .
In the commutator method , on the other hand , while a series resistance of a few ohms will have no appreciable effect on the observed capacity , an ohmic leakage resistance makes the observed capacity , where ratio of period of charge to whole period of cycle .
Accordingly the commutator method does not separate the ohmic effect from that of pure capacity .
Hence a condenser in 1 , ooo 1,500 2,000 encg se .
8\mdash ; Change in Effective Capacity .
which the ohmic is appreciable will show different apparent capacities by the two methods .
Now if it is attempted to determine the ohm by ring the values of capacity given at the sanne frequenc , ( say 100 per second ) by the two methods , it is found that various mica condensers quality give results disagreeing with one another by as much as two or three in 10,000 .
The differences cannot be accounted for by the effect of the ohmic resistance in the commutator method , for such condensers have a direct current insulation of the order of 10,000 megohms ( for 1 mfd and would be less than one two-millionth part of .
The discrepancies are really due to the other fundamental difference between the two methods , namely , that the .
A. Campbell .
A bsolute Unit of [ July 2 , wave form of the applied voltage is entirely different in the two cases .
In the Carey Foster method it is a pure sine wave , while in the ordinary form of the commutator method it is very appl'oximately a pulsating wave of rectangular form never passing below the zero line .
Tbis wave is equivalent to a similar alternating wave of voltage superposed on a continuous voltage of half its maximum value .
With good condensers , o.t least , this continuous component has no appreciable effect , for practically the same results were obtained with a reyersing commutator ( as in Maxwell 's original description ) as with the charging and short-circuiting cornmutatol commonly employed .
* Witl ] an air condenser the sine wave and square topped wave will give same capacity , but when absorption is present the square topped wave is found to give , for the same frequency , higher value of the effective capacity than that iven by the sine wave ; and the difference is greater the more the condenser .
Also for an absorptive condenser the curve of change of capacity with frequency for the sine wave form is different from that for the square topped wave .
FIG. 10 .
per sec. FIG. 9 .
Fig. 9 illustrates the nature of the difference , the curve ABC showing the change of capacity with sine wave folYn , and AD the change with square topped form , taking100 per second as the standard frequency in each case .
The question now arises : For square topped form of frequency , is it possible to specify an equivalent frequency , which with a sine wave form will give the same capacity as does with the square form ?
In order to answer this question * See also .
Mattenklodt , ' Ann. der Phys 1908 ( 28 ) , p. 359 .
1912 .
] Resistance by Alternating several condensers of various absorptive powers were tested for variation of capacity with frequency both for square topped and sine wave forms , curves like those in fig. 9 being obtained .
It was found in each case that the curve practically fitted a part BC of the curve , when was plotted to equivalent frequencies , where ) Thus in fig. 10 the forms ( of same maximum ) five the same variation with frequency when area area Odo or The equivalence assumption here adopted may be put into the following form : The two methods will the same value of the capacity of a well insulated condenser , provided that in the commutator method the eriod of charge ( and of equal short circuit ) is equal to times the half period of the sine wave in the Carey Foster method .
In estimating the periods of charge and short circuit , account is here taken of the air on the mutator , which in most of the experiments formed about 10 .
cent. of the whole circumference .
The above assunlption is probably only an approximation to the truth .
It is quite arbitrary and only justiIiable by tlJe way it Frequenc9 sec. brings the various divergent results into reement .
For example , in fig. 11 are given the curves of change of capacity with frequency for two different condensers ( Al and E ) .
The points marked by circles were obtained with sine wave form ; Mr. A. Campbell .
Absolute Unit of [ July 2 , those marked by crosses with square topped form , the actual fiequency reduced to the equivalent sine wave frequency by the assumption stated above .
It will be seen that the two methods are thus into reement as far as the frequency variation curves are concerned .
We can now proceed to the comparison of the absolute values given by the two methods Sor a number of different condensers .
In Table are given the nominal of these condensers , their power factors at second , and wave form corrections obtained from observed change of capacity with frequency for sine wave form in each case .
The values iven in the table indicate considerable variety in absorption effects .
The fact that the capacity temperature coefficients of these condensers were of the order of per degree C. necessitated great care in the experiments by which the changes of capacity with frequency were determined .
The values in Table refer to temperatures near C. The condenser marked A2 was the microfarad section of Al , which was subdivided into live sections .
Condenser has been cluded in the table as it has already been referred to ( fig. 11 ) .
Condensers and have power factors much higher than the others ; they will be discussed separately below .
tesults .
Table are yiven the final results obtained with the first five condensers of Table , the dates of the various experiments being also shown .
The numbers in the third column give the values of the ratio , where is the resistance in absolute ohms , while .
is its value in international ohms .
The last column gives the mean for each condenser .
For condenser the mean is a weighted one , as the raph record on January 23 was not so good as the others .
In the earlier experiments ( June and July , 1911 ) , the comparisons were not quite so direct as the equations and ( 7 ) indicate ; the 's were not in the two methods , but were each tested against a standard box .
able V 1912 .
] Resistance by Current Iethods .
Table The results in the last column of Table show that the wave-form corrections of Table have the various condensers into good agreement ; without these corrections the extreme would have been in 10,000 .
To illustrate the behaviour of inferior condensers , Table VfII boives the results obtained from the condensers and G. Table VIII .
Paraffined a It was found that the insulation of was not ) ; with direct and 10 seconds electrification it only gave 1250 megohms per microfarad , while condenser Al gave 11,000 megohms .
Thus the direct leakage was sufficient to invalidate to some extent the comparison between the commutator and the Carey Foster methods .
The value given by the pal.aHined paper condenser ( G ) shows that even a poor condenser ives a fair approximation to the true result .
Reverbing to the normal results ( Table ) , we may remark that and should not carry so much weight as the others , and hence the meaIl of may be taken as .
This number is alnlost identical with that already obtained by the two-phase method .
The closeness of the agreement is probably accidental ; it must be remembered that ) results are derived from the same mutual inductance standard and by the help of the same inducto1neter ; and any errors in either of these standards 414 Absolute Unit ofResistance by Alternating Current .
may affect both equally .
It is not easy to estimate the probable error in the final results .
For the primary inductftllce standard it has already been estimated in 100,000 ( see S2 ) .
] with this the uncertainties in the experimental part of the work , the probable error of the final would be about .
in 100,000 .
International ohm/ true ohm The results indicate that , in measurements of resistance , inductance , and capacity with alternating current , it is possible to obtain consistency to abouC 1 part in 10 , conchnsion which will , I am sure , be satisfactory to oGhel .
experimenters .
\mdash ; The wave-form difficulty inherent in the commutator method is avoided in 's metho this was not foumd to be capable of giving sufficient accuracy .
Another method already partly described by the was tried ; the results obtained showed it to be worth further ration .
In conclusion , my best thanks are due to a number of persons who helped in the research : in the first place to Mr. D. W. Dye , who assisted in most of the experiments ; to his reat skill in and his in the discussion of difficulties a large part of the accuracy is due . .
F. E. Smith gave most important help , not only in regard to the construction and winding of the nlarble cylinder , but also by the resistance coils .
The late Mr. and Mr. NIurfitt carried out the construction of the inductance standards with admirable accuracy , and all the easurements of the dinlensions made by Mr. L. F. Richardson .
For the two-phase method Mr. C. C. Paterson pnt at my the Siemens alternator of the Electrotechnical Department and one of the long- scale ostatic voltmeters with which he has succeeded in such extremely accuracy of reading .
Dr. G. F. .
Searle very kindly supplied the formula by which the integration over the of winding in the primary standard was eifected , and Prof. E. Ll .
Jones carried out an independent calculation for that standard to a very high of accuracy \mdash ; a piece of work involving skill and toil .
Condenser Al was presented to the laboratory by Dr. Alexander Muirhead , and several of the other condensers used were kindly lent by him and by .
H. Tiusley .
Lastly , I would express my thanks to our director , Dr. Glazebrook , for his kind interest valued help in the ] opment of the methods and by which the research has been carried out .
See Rayleigh , ' Phil. Mag 1886 .
Phys. Soc. Proc February , 1912 , p. 110 .

