It will be of interest now to compare the results of the tests, especially of the sparking distance in air, with those found by other experiments. In Table XV. I have collected all the tests which were available to me, and represented a part of them in Fig. 9; with a very discouraging result however, for the values found by different observers disagree seriously.

Conceding even a large margin of uncertainty for the older tests, made by electrostatic machines, and leaving out of consideration those tests, where not sufficient explanation was given to permit critical discussion of the method, the discrepancies are still too large to permit of explanation by errors of observation. The different curves differ considerably in shape; the best and most reliable values however point fairly well to a formula :

ò = a V+b V2.

The tests made by Warren de la Rue, which were made by the chloride-of-silver battery, and hence are the only continuous potential tests free from the objection due to the electrostatic machine, agree completely with this formula, over the whole range up to 11,330 volts. Bourne's tests by means of alternating potentials, reaching up to 110,000 volts (?) agree fairly well also over the whole range. Other tests again show agreement with a quadratic formula only within a limited range.

Most noticeable, however, is the wide disagreement between the values of different observers, which seems to show that still other factors besides the difference of potential have a decisive influence on the sparking distance. Especially I suspect the frequency, for the values found with continuous potentials show the smallest, those with alternating potentials the largest distances, so that I almost believe that the sparking distance in air is larger with alternating than with continuous potentials, and is the larger, the higher the frequency is. This whole phenomenon of disruptive discharge, however, has been so little studied, in spite of the great importance it begins to assume by reason of the problem of long distance power transmission, that a closer investigation of the manifold phenomena connected therewith is very much to be desired.

Eickemeyer Laboratory, December 1892. *

DISCUSSION.

DR. J. B. WILLIAMS-I am satisfied that the discrepancies referred to by the author are largely due to leakage in the apparatus used by different experimenters, and that the figures obtained by calculation cannot always be accepted as correct. Referring to Mr. Steinmetz's apparatus, in which paraffined wood is used, I would say that this material cannot be relied upon to insulate currents of high voltage unless the wood be absolutely clean and dry before it is saturated with paraffin, and even then it must be used right after the saturation. If wood or any other porous material is saturated with paraffin, the wax rarely, if ever, solidifies in a solid form, but is in itself porous, and in a few hours absorbs moisture. I have demonstrated this fact by scores of experiments wherein dust and other bodies which would cause surface leakage were excluded. In 1886, when I succeeded in making an electrometer the quadrants of which would remain charged at potentials of 3000 volts and upwards for hours, with practically no leakage, I was astonished to find what an amount of leakage there was in the materials ordinarily used for insulating the various parts of testing instruments. Take, for example, hard rubber (ebonite). When this substance is pure, especially made for insulation, and fresh, its resistance to leakage, both through the mass and over the surface, is very great. But in order that it may retain its originally high insulating properties it must be kept scrupulously clean and dry, and also be kept coated with a material having at least as high a specific resistance as itself, which will prevent the deposition of moisture on the surface; and this coating must also be kept clean. Paraffin wax is an example of such a coating. I have seen electrostatic measuring instruments, made by European makers, in which hard rubber was used for insulation, and yet, upon examination, I saw that, owing to the condition, shape, and dimensions of the hard rubber, it could not possibly fulfill the conditions intended by the designers.

The perfect insulation of any electrical apparatus is a very important matter, and a difficult one as well, as any one may ascertain for himself if he has instruments which are sufficiently delicate to detect changes in the resistance of the dielectric on, say, six inches of a highly insulated wire, which result from changes of temperature, hygroscopic conditions of the atmosphere, etc.; and I am fully persuaded that if those who give us figures representing, and formula for determining, the voltage necessary to produce sparks of given lengths were provided with absolutely insulated apparatus-by "absolutely insulated" I always mean, unless otherwise specified, insulated to such perfection that no leakage can be detected during the time required for any one test, whether this time be five seconds or one hour, and at the highest potential used during the test-including

machinery and testing appliances, these discrepancies would not occur. Almost all kinds of glass, if exposed to the air, are practically worthless for insulation when high potential currents or charges are used for testing purposes.

At the November meeting of the Institute it was suggested that the results of the tests of the insulated wires which were sent to the Institute be recorded for the use of the members. As the sizes of the wires and thicknesses of the dielectrics were not uniform, letters were sent by the Secretary to manufacturers of first-class wires requesting them to send short samples of their best wires of a specified size and thickness of dielectric. All but four complied. Of the four kinds that were lacking I found one on sale and obtained it. These wires are to be tested at a potential of over 3,000 volts; and I invite any and all who would like to assist me or to witness the tests-which will probably occupy the afternoons of 3 or 4 days-to do so at my office at No. 44 Broadway, Room 518. Notice of the dates of the tests will be given by mail. As the wires will be tested in short lengths I will make a large absolutely insulated air condenser which will not only have a much greater capacity than that of the wires, but will also serve to nullify any effects which might be due to differences in the specific inductive capacity of dielectrics of such short lengths.

I would add that I expect to soon commence a long series of tests upon all kinds of insulating material including insulated wires. The currents to be used in these tests will be alternating, and the periods per second, voltage, etc., can be varied as desired. Every portion of the apparatus used will have absolute insulation whenever necessary. Some of these tests I propose to exhibit before the Institute.

PROF. FRANCIS B. CROCKER :-This paper of Mr. Steinmetz is especially interesting to me, as it relates to a subject that I have already done considerable work upon and intend to follow still further. At the December meeting of the Institute when this subject was discussed in a general way, I then stated that it was important to have insulation tests made with a proper source of current. Mr. Steinmetz agreed with me entirely at that time. The insulation we require in electrical engineering is not for an electrostatic charge from a Leyden jar or anything of that kind. What we need is insulation which will stand direct or alternating currents of high potential and consequently to be reliable, the tests should be made with some precisely similar source of current. If we can get reliable data, the calculation of the thickness of insulation required in any given case will be as definite as that of the size of the conductor, and it would be as important an element in the designing of a dynamo as any other. It would be a very desirable state of affairs if we could be reasonably sure of accurate results in the calculation of the disruptive strength of insulation. In testing insulating materials, the alternating current is, of course, better than the electrostatic charge because of

the power behind it, but differences will be found in testing with alternating currents due to different frequencies. The form of the alternating current wave would probably also affect the results. A perfectly smooth current wave might be a less severe test than a sharp one or vice versa. The alternating is a peculiar current, it is a special case. The current which I have been working with is a steady, direct current of high potential. It seems to me that this gives the most perfect conditions. If, however, insulation is to be used for alternating current apparatus then of course the test should be made with alternating currents. I have been using for some time a direct current dynamo of 5,000 volts. M. F., and have just constructed one of 10,000 volts In the 5,000-volt machine the remarkable fact is that there are but 32 commutator bars in all. I have run this machine up to 5,500 volts for an hour or more at a time without any trouble. I have measured 500 volts between two adjacent commutator bars at the point of maximum P. D. (half way between the two brushes.) The discrepancies between different experimenters referred to by Mr. Steinmetz, can be explained as Dr. Williams suggests, as being due to leakage, the tests having been made with electrostatic apparatus at a time when the art of insulation was very crude.

DR. WILLIAMS:-If a dielectric will leak at a certain potential when subjected to an electrostatic stress, it will also leak when subjected to the action of a steady current having the same potential. This is an unscientific way of stating the case, but what I mean is this:-If we find that any dielectric will allow excessive leakage through itself of the small charge which is on the quadrants and a few inches of wire, why subject the same to the action of a steady current having the same potential as that of the charge? We know that it will break down if the current is continued long enough, because work is done on the material of the dielectric by the current. Electrostatic tests are simple and quickly performed, and a few inches of an insulated wire or a thin sheet of insulating material having a small area will suffice for the tests. If we test any dielectric by the electrostatic method and find that its specific resistance is low, we know that any wire insulated with the same material and conveying a direct current of the same voltage as that used during the tests would be dangerous. If the wire should suddenly become ruptured at a point distant from the generator, and the same amount of dielectric surface that was used during the tests could come in contact with a grounded body, the thickness of the dielectric on the wire being substantially that of the specimen tested. Conversely, we would be justified in using any suitable material for insulating wires if electrostatic tests showed that the material possessed a high specific resistance. Furthermore this method of testing will enable us to quickly and accurately compare the specific resistances of any two or more kinds of in