tests of this type. This difficulty we Surge-Voltage Breakdown Characteristics overcame by making a test barrel in which the electrodes were mounted, and its lead-in bushing of insulating material for Electrical Gaps in Oil having about the same dielectric constant as insulating oil. Fortunately, we had available for making this device two ROYAL W. SORENSEN cylinders made of paper and varnish, FELLOW AIEE known by the trade names Micarta and Herkolite, respectively. The larger cyl­ Synopsis: This paper, emanating from a 50-kv 0.5-microfarad General Electric inder, 40 inches in diameter and 72 inches request for information regarding surge capacitors and a suitable d-c charging long, forms the test chamber, and the voltages required to break down gaps in oil, 3 circuit. The test gaps in oil were formed smaller one, 23 /4 inches in diameter and gives surge breakdown voltage values with 48 inches long, serves as a lead-in bushing. different gap spacings for y4-inch, i/2-inch, by using specially constructed electrodes and 1-inch round rod electrodes with hemi­ assembled in a test barrel devised by the These cylinders were open at both ends spherical ends immersed in transformer oil. author and shown in figures 1 and 2. when obtained. An oil-tight bottom for As expected, these values are a function of All tests were made with a standard the test barrel was obtained by placing "time to breakdown" but for each condi­ the insulating cylinder upon a steel plate tion curves show constant breakdown 1V2 x 40 negative wave. The voltage for values for breakdown times of the order of each discharge was measured by means of on the surface of which a circular gasket about 16 microseconds and longer with the a cath©de-ray oscillograph of the cold- seat had been machined. A cork gasket 1V2X40 negative wave, used on all the cathode continuous-beam type designed which could be clamped tightly between tests reported. Particular attention is and built by Doctor Howard Griest,1 the end of the cylinder and the gasket called to the relatively small amount of oil seat was used. A steel clamping band, required for tests of this type, made possible and a resistance potentiometer con­ by the special type of test barrel used. structed according to the required stand­ as shown in figure 2, drawn tightly around ards for such work. Figure 3 shows a the Micarta cylinder by a bolt and drawn typical oscillogram. The voltages thus down hard upon the cork gasket by bolts through studs welded to the band and NSULATION co-ordination and oil- measured were at intervals checked as to engaging threaded holes in the steel circuit-breaker designs make desirable peak values by comparison with standard I bottom, provided means for keeping this a knowledge of surge voltages required to sphere gaps. cork-gasket joint tight. Any oil leakage break down insulating oil. The lack of All gap test electrodes used were round which might occur, due to accident or data pertaining to this subject was called rods terminating as hemispheres having other cause, was amply provided for by to our attention by E. K. Sadler and J. L. the same diameter as the rods. This placing the test barrel in a shallow oil- Thompson, engineers with the Kelman type electrode was chosen in preference to tight steel receptacle, large enough to hold Electric and Manufacturing Company. spheres mounted on rods or shanks in all the oil used. With this arrangement, The test results published in this paper order to avoid under surge conditions, requiring only eight barrels of oil, the were made possible by the co-operation of uncertainties regarding the influence of test electrodes are freer from field dis­ J. N. Kelman, who provided some of the shank diameter upon the field uniformity tortion due to adjacent materials than equipment required for making the tests, around the gap, and, hence, upon the would be possible with the use of con­ and by a great deal of arduous and careful results obtained by our tests. tainers of metal or other conducting or work done by graduate students, particu­ Electrodes ranging from one-fourth even semiconducting material, unless larly G. D. McCann and A. E. Harrison, inch to six inches in diameter were pre­ such containers were made very large. in the high-voltage laboratory at the pared. This report contains data for That is, the use of oil in a barrel of in­ California Institute of Technology. electrodes having diameters of one-fourth, sulating material having about the same Voltage for making the tests was ob­ one-half, one inch, and two inches, re­ dielectric constant as oil, makes possible tained by means of a 1,500,000-volt Marx spectively. The data for the two-inch with a small amount of oil, electric gaps circuit surge generator comprising 30 electrodes are less complete than for the immersed in what, in effect, is a column other rods, because the voltage available, of oil insulation surrounded by the at­ 1,500 kv, would not permit tests with the Paper 39-115, recommended by the AIEE com­ mosphere of the test room. mittee on instruments and measurements, and two-inch electrodes at the larger spacings. presented at the AIEE combined summer and Pa­ cific Coast convention, San Francisco, Calif., An increase in surge-generator voltage is Calculations before construction, and June 26-30, 1939. Manuscript submitted Novem­ being made and further tests will be re­ tests after construction, show, for all ber 30, 1938; made available for preprinting April 24, 1939. ported in a later paper. practical purposes, that this device pro­ ROYAL W. SORENSEN is professor of electrical One reason for the absence of data of vides a means of having test gaps in oil engineering at California Institute of Technology, the type presented in this paper has been completely free from field distortion due Pasadena. the difficulty encountered in providing a to surrounding objects. 1. For all numbered references, see list at end of paper. convenient lead-in bushing for making The Herkolite cylinder serving as the lead-in bushing is closed at the lower end with a fiber and metal bottom designed to hold oil between it and a steel tubing periods of subnormal loads, such as Sun­ naturally to accommodate itself to the concentric with the insulating cylinder days and late night hours, it permits the increased burden. Under all conditions, and through which the upper and adjust­ highest degree of concentration of work. it handles the telegrams with greater able electrode extends into the test barrel. In times of abnormal loads, such as speed and accuracy than ever before, and This metal part of the bottom is con­ Christmas, New Year's, Easter, or special thus has brought about significant prog­ structed in such a way as to provide for a events of national interest, it expands ress in the telegraphic art. well-distributed electric field around the

78 TRANSACTIONS Sorensen—Breakdown in Oil ELECTRICAL ENGINEERING Figure 1. Test bar­ reason for some of the variable results ob­ rel and lead-in bush­ tained, it may be necessary to determine ing showing bushing oil puncture strength with a precision raised For examina­ which will show small variations in di­ tion and polishing electric strength in the range above 30 electrodes kv, as determined with the standard test cup. Early in our test program we found that, for consistent breakdown values, only a few discharges could be made be­ tween electrode polishings. This was more noticeable for the smaller electrodes than for the larger ones. On rare oc­ casions only one or two discharges which, perchance, occurred from the same spot on one of the electrodes, would make necessary a repolishing of electrodes be­ fore consistent breakdown values could be obtained. In all tests it was necessary to observe a time interval of not less than 5 minutes between successive discharges, and for the longer gaps time intervals of 15 to 20 minutes were required for the oil to return to what might be called a normal state. These influence factors, plus the all- prevailing time-lag influence upon surge- breakdown values of voltage for gaps in oil, added to those well-known difficulties encountered in obtaining similar data at power frequencies, soon made evident the fact that breakdown values for gaps in lower end of the bushing. The lower The tests reported in this paper show oil would have to be considered as statis­ electrode, supported by and attached to only the surge voltages required to break tical values obtained from a large number the bottom of the barrel, was not made down gaps in oil formed between spheri­ of tests, some of which would show values adjustable, but constructed so as easily cally ended cylindrical electrodes of of breakdown voltage varying a con­ to be changed or removed for cleaning and different diameter. All oil used was siderable amount from any curve drawn polishing. During tests the part of the carefully selected, filtered, and tested to show the probable reasonable break­ lead-in bushing between outer and inner according to standard oil testing practice down values. In fact, surge-voltage- cylinders is kept filled with oil and flash- before starting the tests and as often breakdown values in oil seem to be even over from lead-in conductor to the barrel during any test as was necessary to keep is avoided by having its lower end project the dielectric strength at top value, this into the test barrel a few inches below the top value being determined by a standard surface of the oil in the barrel. This oil gap test set capable of applying 30 arrangement was proved entirely satis­ kv to the one-tenth inch test gap formed factory as a means of limiting all arc-overs between one-inch-diameter flat electrodes. to the test gap. The oil was considered in proper condi­ tion when it would not break down for a Figure 2. Cross- two-minute application of this voltage. section sketch of To keep the oil up to this standard during test barrel and a test program was not easy, filtering lead-in bushing often being found necessary after only a few oil punctures had been made; and IRON STRAP , GAP ADJUSTMENT i-V· METAL TUBE frequently after the puncture value of the Figure 3. Typical oscillogram, showing oil H·OIL LEVEL oil had been decreased, it was necessary to puncture occurring at 860 lev crest and 9l/2 M AIR l-HERKOLITE CYLINDER put the oil through the filter several times microseconds time lag for eight-inch gap, one- to restore its dielectric strength to 30 kv. half-inch-diameter electrodes Whenever the dielectric strength of the

OIL LEVEL oil dropped to a value of 22 kv or less, the more sensitive to the influences which METAL CAP surge voltage required to break down any make testing such gaps at power fre­ METAL TUBE MICARTA CYLINDER given gap was reduced an amount which quency a major operation, than is the indicates that perhaps as further tests of case for the power-frequency tests; and ELECTRODES this type are made, the dielectric strength it may almost be said that on some days

•SAFETY RESERVOIR of the oil during a test run should be sport breakdown values occurred, for

\, CLAMPING BAND determined after practically every oil which there just seems to be no way of CORK GASKET IL· BOTTOM PLATE breakdown. Also, if we wish to know the accounting.

FEBRUARY 1940, VOL. 59 Sorensen—Breakdown in Oil TRANSACTIONS 79 1200 1200

1000 1000 \ en CO \ !J 8 800 s 8" >6AP v §800 s 3 V \ 6" fiM 4" • 600 ^600 V 2" S* 4" CO ù UJ •Ã a: 400 ?" tu °400 o 1" Ul ce o a: 200 200

0 4% (3 12 16 0 24 28 32 36 0 4 8 12 16 20 24 28 32 36 ôIM E LA(3 - I rilCF 0SEC0NDS TIME LAG- MICROSECONDS Figure 4. Time-las curves for one-fourth-inch ternately raised and lowered above and Figure 5. Time-lag curves for one-half-inch electrodes with hemispherical ends, 1 Vix40- below the probable breakdown values. electrodes with hemispherical ends, 1%x40- microsecond wave Often a one-step charfge on the surge microsecond wave generator (about three per cent) in the crest voltage value either way would In making the volt-time tests, the change breakdown performance so that time to breakdown for higher voltage operating procedure was as follows: The with a three per cent lowering of the crest values. Figure 7 shows the same curve surge generator was adjusted to give a voltage there would be no breakdowns at for two-inch spacing, as shown on figure 1 6, but shows also the exact location of the l /2x40-microsecond negative wave hav­ all; whereas with a three per cent in­ ing a crest value well below the mini­ crease in crest voltage practically every points used in plotting the curve. mum flashover voltage required to strike surge application would cause a break­ The probable breakdown curves shown across the gap under test, and for this down. in the paper are considered correct as to setting a surge was applied to the gap. Having in this way established the values given for prevailing conditions The voltage was then increased by small probable minimum breakdown voltage within a value of plus or minus three per steps until a voltage was reached which for a given gap setting and electrode size, cent. Occasionally, however, voltages would puncture the oil in the gap. When the balance of the volt-time curve was more than ten per cent above the prob­ 1 able breakdown values shown in the breakdown occurred, the surge voltage obtained by applying the l /2x40-micro- which first caused breakdown was applied second wave at increasing voltages and curves would be required to cause break­ a number of times. If breakdown oc­ taking cathode-ray oscillograms of the down. We think such discrepancies curred for every surge application, the voltage across the oil gap under test. were due to undetermined variable con­ voltage setting on the generator was re­ For each surge-voltage application an ditions of oil and electrode surfaces, but duced until breakdown stopped, after oscillogram was made to show the value up to the present time our analysis does which the voltage was again raised until of the wave from the time of tripping the not permit a statement as to what these it reached the minimum value which circuit to breakdown. Figure 3 is a conditions were. would give breakdown for a majority of typical oscillogram. In most cases break­ When one of these extremely high first- the applications. down occurred on the tail of the wave. breakdown voltage conditions was en­ As a check on this breakdown value, the The crest values of the waves thus re­ countered, the test procedure outlined surge-generator voltage would be al- corded were plotted as ordinates on the curve sheets shown in figures 4, 5, 6, and Figure 7. Time-lag curves for one-inch elec­ 7, with the time to breakdown shown as trodes with hemispherical ends, showing Figure 6. Time-lag curves for one-inch elec­ corresponding abscissas. The curves spread of actual plotted values from which

trodes with hemispherical ends, iy2x40- thus obtained show minimum breakdown curve was obtained; 1 y2x40-micro$econd microsecond wave voltages and breakdown voltage versus wave

ItiVV 1200,

1000 I00C+

co Î3 4" GAP >O800 h 3 \ s 2" o * 1" 1 600 200 "cOO

0 A ii 12 28 3 2 3 6 12 16 20 2~4~ ""ST "36 TIK> i - Ml CRO°SE C,0N 1 ?S TIME LAG - MICROSECONDS

80 TRANSACTIONS Sorensen—Breakdown in Oil ELECTRICAL ENGINEERING Figure 8. Curve wave form as compared to 60-cycle break­ showing relation of down voltages in oil. gap length to mini­ To obtain these further data, we invite mum surge crest co-operation from others interested in voltage required for high-voltage phenomena and recommend breakdown between cylindrical elec­ the 'use of test barrels and lead-in bush­ trodes with hemi­ ings made of insulation material having spherical ends im­ about the same dielectric constant as the mersed in oil, insulating liquids to be tested and built 1 V2x40-micro- according to the general plan of the ap­ second wave paratus described in this paper.

References

5 5~ 1. A LOW-VOLTAGE HIGH-SPEED CATHODE-RAY GAP LENGTH- INCHES OSCILLOGRAPH, R. Howard Griest. Thesis for doctor of philosophy degree, California Institute of Technology, 1937.

2. OIL BREAKDOWN AT LARGE SPACINGS, Douglas was reversed and, subsequent to the first failed to puncture the oil. Puncture F. Miner. AIEE TRANSACTIONS, 1927, page 248. breakdown, surges of lower crest voltage finally occurred at a crest voltage 17 per 3. DIELECTRIC PHENOMENA IN HIGH-VOLTAGE ENGINEERING (a book), Peek. McGraw-Hill Book than the one causing this first break­ cent higher than the probable minimum. Company, 1929, page 227 down, but higher than the recognized July 14, 1938, seven surges which failed minimum breakdown, were applied. The to puncture the oil had crest values more voltages for the successive tests were than 10 per cent above the probable gradually reduced until the probable minimum value applied to one-inch elec­ Discussion minimum breakdown voltage for any trodes spaced for a two-inch gap. The given gap was reached. The minimum breakdown finally occurred at a crest D. C. Prince (General Electric Company, Philadelphia, Pa.): This paper adds valu­ values of voltages found in this manner voltage 21 per cent higher than the prob­ able information to our store of knowledge were in agreement with those found for able minimum value for this arrangement. on the behavior of insulating oils. One the same gaps by the more commonly Figure 8 shows data rearranged to point of particular value to circuit-breaker used increasing-voltage procedure. For enable one to obtain readily the relation designers is brought out very clearly. Un­ der all the conditions investigated the oil the one-fourth-inch rods no excess- between gap lengths and minimum punc­ presents an impulse ratio of from 2 to 3. voltage first breakdowns were in evidence. ture voltages for the gaps included in this Ordinary air clearances, rod gaps, bushings, For the one-half-inch rods this effect was report. and the like, associated with circuit break­ indicated only, and then was limited to It is well to note that the limited ers, have impulse ratios of the order of 1.2. the longer gaps. The one-inch rods amount of data available for design cal­ It follows that failure of the oil under im­ pulse is not limiting, a fact of obvious value showed a very pronounced tendency to be culations as gleaned from the 1,500 surge to designers. subject to this phenomena at small spac­ tests made to obtain the data presented ings and for the longer gaps high break­ in this report point strongly to the need of down points definitely occurred. The continuing this work in a broader test V. M. Montsinger (General Electric Com­ pany, Pittsfield, Mass.): Professor Soren- two-inch rods showed occasional high program extended to a number of our sen has given us some very interesting im­ breakdown voltages for short gaps. If high-voltage laboratories. A search for pulse data on the breakdown of oil. Ap­ continued tests show any appreciable data pertaining to similar tests for a-c parently, by exercising great care, he has increase in the percentage of high-break­ 60-cycle voltage which might be used in been able to eliminate the wide variations in down-voltage values, it may be necessary making comparison between 60-cycle and oil-breakdown values that usually occur in 2 ordinary methods of testing. for the larger electrodes to indicate arc- surge dielectric strengths of oil gaps dis­ There are, however, two or three points over voltage by a broad curve showing the closed the fact that published 60-cycle I wish to comment on. 4 spread, rather than by a single-line prob­ data in agreement as to values for gaps The author apparently used hemispheri­ able-breakdown curve. formed between different types of elec­ cal-shaped electrodes to simulate conditions in circuit breakers. The variation of the To show more explicitly what this trodes in oil are by no means adequate for breakdown voltage with spacing does not phenomenon of high breakdown signifies, our present requirements. Although the seem to follow any law that I have ever the following specific instances of such ratio of surge crest lV2x40-microsecond observed. For example, his figure 8 shows occurrence are noted. wave voltages to 60-cycle crest voltages that for the one-fourth-inch and one-half- required to puncture oil when such volt­ inch-diameter electrodes the breakdown July 11, 1938, five successive surges, voltage varies roughly as the spacing raised all with crest voltage values more than ages are applied to immersed sphere gaps to the 0.37 power. For one-inch-diameter 10 per cent above the probable minimum varies with the size of spheres and gap electrodes the exponential value of the spac­ breakdown value applied to the one-inch spacing, the present available data indi­ ing is even less than 0.37 and bends over, gap between one-inch electrodes, failed to cate that no great trouble will be en­ indicating that considerably greater spac­ ings than four inches would give very little puncture the oil. The sixth, or punctur­ countered in designing apparatus if this increase in breakdown values. ing surge, had a crest value 24 per cent ratio of crest surge voltage to power- Generally when spheres are used with above what appears to be the probable frequency root-mean-square voltage is spacings several times their diameter they minimum puncture value for the gap. taken as three to one until sufficient data act like sharp-cornered electrodes, so far as breakdown versus spacing is concerned. July 13, 1938, with the same electrodes for better conclusions are available. 3 This is especially true when air is the insu­ and same setting, five surges with crest Tests made by F. W. Peek when he in­ lating medium. voltages more than 10 per cent above the troduced the idea of time lag show this We have a large amount of data on break­ probable minimum breakdown voltage ratio of three to one for surges of unknown down of oil for square-edged and round

FEBRUARY 1940, VOL. 59 Sorensen—Breakdown in Oil TRANSACTIONS 81 1000 • Figure 1 material which has a dielectric constant • 5 6 A 8 I about the same as oil. It is true there will > s * ,: 5i§7 o · Ã ■ 13 # · ■—·— o« be little field distortion at the oil-Herkolite * o èï boundary, but very close to it is the Herko- «/> lite-air boundary, where there is a change of dielectric constant and a distortion of field. u JULY 18, «938 Ui ao 5 A. £. Harrison (California Institute of 12 3 4 5 Technology, Pasadena): A diagram show­ <) TIME IN HOURS ing the time of application of surges to the one-inch electrodes at four-inch spacing lOOOf ^STIRRED OIL ..? •° /INSULATOR has been prepared to illustrate the phenome­ 8 ¥ .^ 12 ° ^-FLASHOVER non of application of crest voltages above —· ·· the probable minimum value without break­ down. Figure 1 of this discussion shows two runs made July 18 and July 19, 1938. JULY 19. 1939 The ordinates of the points represent crest voltages, while the time in hours gives the intervals of time between application of surges. The figures above certain points TIME IN HOURS give the breakdown time in microseconds. No breakdown occurred for surges indicated by open circles. rods (both parallel and at right angles) and Professor Sorensen, except to say that the I interpret these data as follows: Appli­ in all cases the breakdown is roughly pro­ oil was periodically filtered. cation of surge voltage less than that re­ portional to the spacing raised to the two- Oil in contact with air undoubtedly con­ quired for breakdown changes the condi­ third power. It appears from Professor tains water in two forms: water in solution tion of the electrode surface, and/or the Sorensen's data that for limited spacings— and water suspended as small droplets which condition of the oil so that the gap will not up to six or eight inches—and with small gradually settle out and collect as free mois­ break down when the probable minimum diameter electrodes, the breakdown voltage ture at the bottom of the tank. The bal­ value of crest voltage is exceeded. The in­ increases quite slowly with spacing. It ance between the two types is controlled by fluence of operating procedure has not been would be interesting to know if circuit- the temperature. As the oil cools less water established, although recent evidence points breaker engineers find that this is true for is held in solution, and more in suspension. to the conclusion that the stirring of the the kind of electrodes used in circuit break­ I do not believe that moisture in one form oil after breakdown has occurred may ac­ ers, or do they find that the breakdown will have the same effect as moisture in the count for this behavior. Stirring the oil strength varies roughly as the spacing raised other, so I feel that before consistent results when breakdown had not occurred, and to the two-third power. can be obtained the total amount of mois­ waiting 15 minutes (the next to last point on ture the oil contains and the temperature July 19 and other similar tests not illus­ must be controlled and recorded. trated) would not cause breakdown to oc­ Eric A. Walker (Tufts College, Medford, If oil contains moisture in suspension it cur, however. Once breakdown has oc­ Mass.): The paper by Professor Sorensen appears difficult to reproduce results because curred at a high voltage, breakdown will then on the "Surge-Voltage Breakdown Char­ each successive shot draws more moisture occur at the probable minimum voltage. acteristics for Electrical Gaps in Oil" is of into the gap. The reason is self-evident. (Note point at time two hours on July 19.) considerable interest to me because it is Water has a higher dielectric constant than This unusual behavior was quite consist­ very similar in nature to a study which has oil and so is impelled into the region where ent on days when it occurred, but there been started at Tufts College. the electrical stress is greatest. were times when the phenomenon failed to In making surge breakdown tests I have In our tests a number of sludged samples occur for no apparent reason. Also, this felt that one of the most important factors of oil were found, which had relatively high phenomenon did not seem to occur when the in determining the breakdown strength of breakdown strengths when tested on a short oil around the gap was not stirred after a the oil is the amount of moisture the oil con­ chopped wave; and relatively low strengths breakdown, although failure to stir the oil tains; a factor which is not mentioned by when tested at 60 cycles in the standard gap. seemed to reduce the strength of the oil in The power factor of these sludged samples was high. Figure 2. Time-las curves for one-inch elec­ I wonder if Professor Sorensen does not Figure 3. Time-lag curves for two-inch elec­ trodes with hemispherical ends, iy2x40- emphasize too much the advantages of con­ trodes with hemispherical ends, 1^x40- microsecond negative impulse wave structing a tank of Herkolite or some other microsecond negative impulse wave V \ \ leoo I200 \ s \ 8" GAP -V- I000 k > \ \ S \ 2" GAF * J \ — _. 4' I 800 J8X)0 \ > \ s _ 2" J SAP *"- -^ _ I" 600 vN \ \ v *v  -^ I/2" I/2' GAF ' > > ) ^£00 200

3 2 I6 .0 >4 2 8 52 6 0 A 3 I 2 I o 2 0 2 4 2 8 3 2 i * ( T! ME LAG MIC**0S JEC O NDS \ TIME LAG - MICROSECONDS

82 TRANSACTIONS Sorensen—Breakdown in Oil ELECTRICAL ENGINEERING I400 I200i \ 4" D AMETEfc ELEC mopES 1200 V 1000 / 2" CAP / é *- O J^. 1000 \ >800 / / o \ // ,-· ^ —- - — "T ""t J Y y '"* ^. I/2J -- -- daoo \ 600 1 L. - > 1" ut -- 1/4: — - | -~ T """ o_l ^^.- ·- -^ir^ ^00 V 1 (Ë 1 \ ^ \ U"T £J400 1/2 ·' / o 200l· o a:

FEBRUARY 1940, VOL. 59 Sorensen—Breakdown in Oil TRANSACTIONS 83 is perhaps found in D. C. Prince's discussion of our paper, wherein he says "under all the conditions investigated (and reported in the Sensitive Ground Protection for paper) the oil presents an impulse ratio of from 2 to 3, whereas the parts of the breaker which are in air have impulse ratios of the Radial Distribution Feeders order of 1.2." In other words, what we have done is verify the fact that oil under impulse is not a limiting factor in design be­ cause oil strength under normal and re­ LLOYD F. HUNT J. H. VIVIAN covery voltage has taken care of the impulse FELLOW AIEE ASSOCIATE AIEE factor. E. A. Walker's discussion calls attention to the importance of giving careful attention HERE are three reasons for isolating sand soil, where there had been no rain in making these studies to the amount of Tor de-energizing grounded feeders: for several months, show that practically moisture in the oil under test and the form first, danger to life; second, danger to no current flows. A series of tests were in which the moisture is present. Very property; and third, to permit joint use made on a 7-kv impedance-grounded early in our experiments, as was stated in the paper, we found that breakdown values of pole lines with the telephone com­ neutral system by dropping a 2/0 bare were very sensitive to the oil moisture con­ panies. The first two are very difficult conductor on a green lawn. In this case, tent. So far we have simply avoided most to evaluate, but in certain territories a with about 20 feet in contact, the current of the difficulties of variability due to this means of isolation is very important. was in. the order of 2 to 3 amperes during moisture content by using only oil of high The third reason can be given a definite the few minutes of each test. In another dielectric test; for example, oil which would stand 30,000 to 40,000 volts when tested value, and the economies resulting from test, with a ground connected to a five- with a standard 0.1-inch oil testing gap. joint use of pole lines can more than gallon can in mud, 35 amperes flowed in In fact, as Mr. Walker suggests, there is no justify this proposed scheme of sensitive an 11-kv solidly grounded neutral circuit. doubt a fertile field for investigation to de­ ground protection for radial feeders. Tests were made on 16.5-kv solidly termine the exact quantitative influence of water content and also of temperature on This paper will not discuss the merits grounded neutral circuit by dropping a surge-voltage-breakdown values for oil. of leaving a ground-faulted feeder in partial span in a dry stubble field. It I do not think, as Mr. Walker has stated, service for the continuity of service, as was necessary for the conductor to touch that we have at all overemphasized the ad­ may be the practice by some utilities, a green weed or come in contact with a vantages of using a test tank of material but discusses the methods of clearing conducting portion of earth before current which has a dielectric constant approxi­ mately the same as that of oil, because after ground faults that may normally occur would flow. Some of the times the con­ all the prime reason for the use of a tank on radial feeders. ductor would lie in this stubble field as made of insulation material rather than if dead ; then all of a sudden a breakdown metal is the avoidance of the difficult lead-in Ground Faults to earth would occur which was sufficient bushing problem and the use of the large to cause current to flow. From this amount of oil required when steel tanks are used. The fact that the dielectric constant The magnitude of ground currents on particular group of tests it seemed as of the insulation material used is so nearly lower-voltage feeders with solidly though at certain times, when the con­ the same value as the constant for trans­ ductor was dropped, it would break down, former oil and that the dielectric constants grounded neutral transformers varies for both the oil and the insulation material is more according to the contact resistance only two to three times the constant for air, of the fault than it does for the location is a happy coincidence which makes our of the fault. This effect is greater, the plan practical; whereas it would not be so lower the voltage of the circuit.1 The good if we had to use a barrel made of some material such as porcelain, which has a usual voltages used for radial feeders much higher dielectric constant than the range from 2.3 kv to 22 kv. material of which our test barrel is made. Many ground tests have been made on At present I have no explanation for the 2.3-, 4-, 7-, 11-, and 16.5-kv systems of phenomena described in Mr. Harrison's the Southern California Edison Company, discussion and can only confine my remarks to expressing appreciation of the large Ltd. These tests all have shown con­ amount of careful, conscientious work which clusively that the contact resistance has he has done in obtaining for me much of the a controlling influence on the magnitude data used in this paper. OIL [toi 1 <-|TRlP of ground current. SWITCH j y f Mr. Lovoff also has added to the paper J r COIL TRIPPING Tests on a 4-kv solidly grounded neu­ /TRANSFORMER data which he has obtained and which ex­ tend the curves used in the paper to values tral circuit show very small amounts of beyond those available at the time the paper current when bare wire is ground- was written. To him and to other stu­ faulted as follows: In dust 0.3 to 0.6 dents who have helped in this work, I am amperes, in grass 1 to 3 amperes, in deeply indebted, for without their help the |L_ large amount of testing necessary to produce muddy water 25 to 30 amperes. Other this report could not have been obtained. tests on 4 kv on dry pavement and dry T Y i vW]l· hi In fact, like most research work of today, the 1 1 If /V\\fl. ËËË 1 ' 1 1 * POT ENTIA results presented in this paper are available Paper 39-131, recommended by the AIEE com­ 1 1 TRAh4SFORK /"E R only because of much co-operation by many mittee on protective devices, and presented at the ! 1 11— persons and it would certainly be unfair to AIEE combined summer and Pacific Coast con­ H vention, San Francisco, Calif., June 26-30, 1939. POLE- ~7£h close our discussion without expressing ap­ TYPE 1 h preciation to the Kelman Electric and Manuscript submitted April 14, 1939; made avail­ CAPACI- I L able for preprinting May 26, 1939. ôçñò 1 1 Manufacturing Company, which provided LLOYD F. HUNT is protection engineer and J. H. II "1 VOLTAGE RELAY much of the equipment used in making the VIVIAN is technical assistant and superintendent of tests, and to Messrs. Kelman, Sadler, and relays, Southern California Edison Company, Ltd., Thompson, who contributed valuable sug­ Los Angeles. Figure 1. Sensitive ground protection for gestions. I,. For numbered reference, see end of paper. single-circuit delta system

84 TRANSACTIONS Hunt, Vivian—Ground Protection ELECTRICAL ENGINEERING