v. · the quadrant, a coil was affixed to the core by glueing and clamping. These semi ... the turns of the coil, some to the use of metal tools which scratched the insulation ...

DESIGN OF NAL EXTERNAL BEAM LINE QUADRUPOLES

F. Ascolese, R. Billinget, H. Edwards, W. Hanson,R. Juhala, A. Maschke,*J. Michelassi, P. Reardon, S. Snowdon

National Accelerator Laboratory*Batavia, Illinois 60510

Abstract

The 3-inch aperture quadrupole mag- ­nets used in the extracted proton beamlines and in some secondary beam lines atthe National Accelerator Laboratory pro­duce a gradient of 5 kG per inch with 100ampere excitation and a voltage of 225 v.The 10-foot long laminated core consistsof four separate quadrants punched frommain ring magnet steel welded together andcompletely vacuum impregnated with coilsand vacuum chamber in place. The fourcoils each contain 118 turns of #3 U.S.gauge solid square copper insulated with apolyimide coating. Conduction cooling isprovided by water-cooled fins located be­tween conductor layers. The designedgradient is uniform to 1% over the 3 inchaperture.

I. Introduction

These three inch quadrupole magnetsare the focussing elements of the beamtransport system for proton beam from theaccelerator and also for some secondarybeam transport lines in the three externalexperimental areas. Because of the exten­ded beam lines, it is desirable to have amagnet which can be powered with a low­current, high-voltage supply to minimizethe cost of power transmission. The deci­sion to use this type of magnet was madeafter comparing the 10 year operating costfor magnets of several designs, includingtheir power consumption and cooling re­quirements, to the initial capital invest­ment for the magnets and power supplies.The coils in this magnet are designed with118 turns to provide -12000 AT requiredfor a field gradient of 5000 G/in. with avoltage of 225 V across the four coils.The design value for field gradient vari­ation is one part in 10 3 over 2/3 of theaperture area. The magnet core, the coils,and the vacuum chamber are simultaneouslyvacuum impregnated into a monolithic unit.

t Present address CERN, Geneva, Switzer­land* Present address BNL, Upton, New York

* Operated by Universities ResearchAssociation, Inc. under contract withthe U.S. Atomic Energy Commission

407

The length of the magnet, plus thefact that it is not only laminated butconsists of essentially six componentparts making up the iron path, posed somedifficulties in meeting the fairly tighttolerances on the aperture. In spite ofthe length, however, the magnet, whenvacuum impregnated, forms an extremelyrigid structure which shows essentially nodeflection regardless of how it is support­ed. The tight packing of the turns of thecoil led to some electrical insulationdifficulties which were solved by addition­al insulation in critical areas of thecoils. The magnets were built at threefacilities: NAL, Westinghouse of Emery­ville, California, and Magnetic Corporationof America (formerly Airco-Temescal) atBerkeley, California. Prior to, and dur­ing initial production, by the mentionedvendors, NAL built several magnets for tworeasons; i.e., (a) to develop tooling andprocedures to achieve the desired qualityof magnet, and (b) to establish manufac­turing tolerances consistent with the mag­netic specifications. The vendors werethen supplied punched laminations, spec­ially contoured end packs, pre-insulatedwire, and the expertise developed at NAL.This NAL experience was then supplementedby the experience of each vendor andresulted in the preparation of tooling andprocedures that would yield acceptablemagnets. A total of 170 magnets have beenbuilt with the Magnetic Corporation ofAmerica providing over 100. The magnetscost approximately $12,000 each and thepower supplies (provided by ACME and Ling)approximately $4,500 each.

II. Construction

Stacking of the Core

The magnet consists of four quadrantswhich are held together by weldments onfour keys. The laminations are punchedfrom the same type of steel used for themain ring magnets, namely 1/16 in. thickArmco super soft, specially modified todevelop large grain size, and control im­purities which adversely affect magneticcharacteristics. It is insulated with alight phosphate coating commonly known ascore-plate.

After a quality control check, thelaminations are vapor degreased and thenstacked together in special stacking

fixtures. There were two differentapproaches taken in the stacking of thecore. MCA chose to stack a single quad­rant at a time using the straight outsidesurfaces of the lamination as the guidingedge. The inner surfaces of a right anglewere used to support the laminations withthe pole facing upward. After compressingthe quadrant, a coil was affixed to thecore by glueing and clamping. These semi­rigid quadrants were then assembled intohalves with the wide key and the halvestack welded together with the narrow key.

The other method of stacking employedboth at NAL and Westinghouse consisted ofstacking a half magnet with the coilsloosely in place and then joining the twoquadrants with the wide key. In this case,the guiding surfaces were two points onthe pole face and the mating surface ofthe back leg.

The stacking fixture shown in Figure2, the one used at NAL, consists of twoouter 2 in. diameter steel cylinderswhich were case hardened and a V-blockholding a 1.5 in. diameter steel cylinderin the center. The fixture was opticallyleveled and aligned to a tolerance of± .001 in. By pressing on the outer cor­ner in such a manner as to force the lam­inations against the guiding surfaces, thequadrants were held in place for joiningtogether with the wide key. It was foundthat under this compressing force, thelaminations tended to skew such that theywould no longer be at right angles withthe guiding rails. This was corrected byprestacking 9 in. hard packs on a veryprecise fixture. The laminations in thehard pack were TIG welded together.Usually four of these hard packs were usedin each quadrant, spaced at equal inter­vals with loose laminations stackedbetween them. Before welding the wide keyinto place, it was found necessary to firstTIG weld the laminations in the threelocations shown in Figure 1. If the weldto the wide key is made first, the lamin­ations pull, causing a wavy surface dueto uneven shrinking of the welded joints.The wide key was stress relieved to reducewarping during welding, and machined to awidth of 7.750 ± .0005 in. The quadrantsare terminated at the ends with 2 in. widehard end packs that have been contoured toallow a smooth transition to the zerofield outside.

The magnet was then assembled fromthe completed half cores using a granitesurface table. The halves were compressedtogether by clamping to the surface table.A round key is used to align the back legsof the laminations so they are not shift­ed with respect to each other. Obviously,the halves must be of the exact same

408

width or the magnet will not fit togetherproperly. Originally, the design calledfor a .250 in. dowel to be used. However,it was found necessary to use a somewhatsmaller dowel (.247 in.) to allow forsmall irregularities in the stack. A 10ft long, 4~ in. wide adjustable parallelwas placed between the halves in the aper­ture to maintain the proper distancebetween the vertical tips. At this pointthe vertical tip to tip spacing was set.005 in. smaller than the assembled spac­ing. It was in this manner than an allow­ance was made for weld shrinkage occurringat the back leg. The most critical pointof the assembly procedure (aside fromcareful stacking) is in joining the twohalves together. They are best joined byTIG welding the back leg gap togethe~ asnecessar~ to pull the mating surfaces flatto each other. Excess weld here wouldcause the vertical tip to tip dimension toopen up. The actual amount of weld usedhere was usually a 1/2 in. bead every fourto six in. With this weld complete themagnet could be moved about and would re­main intact. The next step was to weldthe small key in place. On the firstmagnets produced, a full weld on bothsides of the key was used. The stressesproduced as this weld shrunk increasedthe spacing between the vertical tips byas much as 1/32 in. Therefore, tack weldswere used to hold this key in place spaced9 in. apart and staggered so that no twowelds were directly opposite each other(see Figure 1).

Coil

The coils are wound from #3 square OFHCsolid copper with 118 turns per coil.There are approximately 2500 ft of wire ineach coil (no splices were allowed). Theinsulation is a .0015 in. polyimide filmcoating which provides good resistance toradiation damage and mechanical abrasion.Cooling is provided by two edge cooledaluminum fins which are inserted betweenthe coil layers. Because of the thinnessof the insulation and the close packingof the turns, special precautions werenecessary to avoid trapping foreign mater­ial in the coil, which could lead to fail­ure. In fact, during the early productionof magnets, a considerable loss in coilswas experienced, some of which wasapparently due to poor coating of the wire,some to foreign matter getting in betweenthe turns of the coil, some to the use ofmetal tools which scratched the insulationand some due to abrasion of the coatingduring winding.

In addition, the handling of the copperand the final sizing process at the manu­facturer were such as to produce nicks andslivers periodically in the bare copper.

To guard against this, special handlingwas required and additional inspectionsperformed. It was found necessary to dothe coil winding in a clean area shut offfrom other activities. It was also notedthat attempts to clean the copper withacetone or similar cleaners tended tosoften the insulation which apparently wasnot fUlly cured at the time it left themanufacturer. Thus, only a dry wipe wasused and a procedure of post-curing at3750 F for 15 minutes was employed afterwinding but before compressing into thecore. Also, in the end areas and in thecross-over areas, the conductors werewrapped with a few layers of either glassOr 5 mil Kapton tape. The eight turnlayers of the coils were separated bystaggered strips of 5 mil Kapton tape toprevent abrasion during winding and handl­ing of the coils. Mica insulation wasused in the ground wall.

Two slightly different procedureswere used by the suppliers to wind thecoils and hold the turns in place. West­inghouse wound half coils and brush coatedthem with epoxy after winding, pressedthem to size, pre-cured the half coil unit,and then assembled the rigid half coilswith cooling fins and ground wrap. MCApreferred to wind the whole coil, insertthe cooling fins, and ground wrap with B­stage mica-glass tape, then pre-cure thecoil without attempting to bring it tofinal size. In this case the turns wereloose, being held together by the outerground wrap only.

Both procedures had advantages anddisadvantages. The brush coating of thehalf coils provided an incomplete impreg­nation which could contain trapped voidsthat would not be filled during the finalvacuum impregnation process, and in anycase would hinder the epoxy flow. Thecured B-stage, on the other hand, tendedto form a closed outer shell which alsohindered penetration by the epoxy duringfinal impregnation. (This was latersolved by perforating this outer shell.)Also, the loose turns rubbed against eachother during handling, thus creating anopportunity for slivers and chips to digthrough the insulation. (As mentionedabove, this was later solved by separatingthe layers.)

Although the original subcontractrequired only a final test of coil integ­rity, it was desirable to subject them toinduced voltage testing after each separ­ate phase of construction. They weresubjected to a maximum of ten volts perturn. In cases where a failure occurred,usually the coil could not be repaired.The additional inspections at the coppermanufacturer, the imposition of tight

409

standards of cleanliness at the vendor,the improved insulation system, and theimproved handling techniques resulted insuch a dramatic decrease in coil and mag­net losses that no failure of either coilsor magnets occurred during the productionof the last hundred magnets.

Assembly and Impregnation

After welding the quadrants of themagnet together, it remained, of course,a very flexible unit. At this stage verycareful handling was required so as toavoid overstressing any point. Because ofthis flexibility it was necessary to designa rigid carrier for constraining the magnetduring impregnation. Such a unit was con­structed from two I-beams welded togetherand stress relieved. Here again, cylind­rical rails, optically levelled, were usedto support the magnet as shown in Figure3.

Since the magnets were laminated,they had to be wrapped tightly around theoutside with a covering, which was essen­tially vacuum tight, to prevent the epoxyfrom leaking through the surface. Therewere three different techniques used.Westinghouse chose to build a rectangularbox out of 1 in. thick steel to enclosethe magnet. This, unfortunately, did notallow for the possibility of straighteningthe magnet, once inside, to the requiredtolerances. MCA used thin steel sheetingthat was sealed with end caps and a-ringmaterial. This eventually worked outquite well after some redesign of thesupporting base. The technique employedat NAL was to use a 0.010 in. mylar wrapglued at the overlap and sealed at theends with steel covers contoured to fitclosely around the coil ends. To keep therails from cutting into the mylar, 1/4 in.x 1 in. aluminum straps were placed on topof the rails before setting the magnet onthe g~rder. Side plates and a bottomplate were then banded around the magnetto minimize the bulging of the mylar wrapduring impregnation.

After the magnet was placed on thegirder and before insertion of the vacuumchamber (a 2.950 in. diameter x 1/16 in.wall stainless steel tUbe), the spacingbetween tips and the pole to pole distancewere measured at 6 in. intervals over theentire length of the magnet. The tip totip dimensions are to be equal within atolerance of ± 0.005 in. The averages ofthe two bore dimensions on a given magnetare to be equal to within 0.003 in. Byconstraining the magnet along the approp­riate diagonal, it was possible to staywell within the 0.003 in. tolerance on thebore dimension. The method of clampingused to adjust these dimensions is shown

in Figure 3.

The epoxy system 1 used in the magnetalong with the curing cycle, has been usedin several other magnets at NAL. Oneaddition was made to this formulation;namely, tabular alumina filler. This mix­ture contains equal parts by weight ofalumina and resin. The filled epoxy hasbetter resistance to radiation, shrinkage,and has better mechanical properties withregard to rigidity, and most important forthis magnet, it has a thermal heat conduc­tivity five to ten times greater than theunfilled epoxy. Since the proper coolingof this magnet hinges on a very thoroughimpregnation with no voids, special pre­cautions were taken to degas the resin andeliminate trapped air pockets by elevatingone end of the magnet during curing.

Figure 4 shows a photograph of amagnet complete with electrical connectionsand water manifold. It is supported on astand used in the secondary beam lines ofthe Meson Facility.

III. Testing and Performance

The magnets are subjected to thefollowing measurements and tests: borediameter and tip to tip measurements beforepotting, straightness and flatness bothbefore and after potting, induced voltagetest, inductance and resistance measure­ments. In addition, a few magnets wereselected for a thermal cycling test, othersfor power tests to determine the effective­ness of water cooling and finally, ofcourse, magnetic measurements. The datapresented in the table below (with theexception of the power test data) repre­sent the averages of measurements made onseveral magnets.

The quantities labelled Twist, Flat­ness, and Straightness, were determinedby optical measurements at 13 sectionsalong the length of the magnet 2

• Thepoints on the surfaces at which themeasurements were taken are indicated inFigure 1 by a circled X. The data pre­sented here are taken from 11 magnetschosen at random and in each case thepoint on the magnet at which the maximumdeviation occurred is the one chosen foruse in determining the average. The"Twist" then is the maximum angle betweenany two sections of the magnet. The tol­erance on the "Twist" is 1 mrad. The"Flatness" refers to the deviation ofpoints measured on the wide face of themagnet from a plane passing through themeasured points 24" in from the ends. Thetolerance for the "Flatness" is 0.030 in.The "Straightness" refers to deviationsof points from a line in the narrow faceof the magnet which passes through the

410

two points 24 in. in from the ends. Thetolerance of the "Straightness" is 0.030in.

The diameter of the bore and the tipto tip spacing were measured at 6 in.intervals. These measurements were madewith the magnet under constraint justbefore impregnation. At this point adjust­ments could be made in these dimensions byaltering the constraints. Thus, for eachmagnet the averages consist of 42 pointsfor the bore diameter and 84 points forthe tip to tip spacing. The average ofthese data for eleven magnets are presen­ted in the table with their correspondingaverage deviations. The "Bore Diameter"is 2.997 ± .0015 in. The tip to tipspacing is 1.033 ± 005 in. Finally, theaverage dc resistance and equivalentseries inductances for the eleven magnetsare given.

TABLE I. Averages of Mechanical andElectrical Measurements onthe 3 in. EPB Quadrupoles

Twist 0.6 ± 0.3 mradFlatness 0.0072 ± 0.0016 inchStraightness 0.0052 ± 0.0011 inchBore Diameter 2.9985 ± 0.0014 inchTip to Tip Spacing 1.0323 ± 0.0006 inchD.C. Resistance 1.620 ± 0.022 nInductance 1.34 ± 0.034 mHy

(at 1 kHz)

Power tests were performed under con­ditions likely to occur in actual use ofthe magnets. In certain areas, for exampleit is intended to use air-cooled heat ex­changers to cool the LCW water used in themagnet cooling system. In this case thewater could be exgected uo enter the mag­net at around 113 F during the hottestsummer days. A magnet was powered at 100Ausing water at a supply pressure and tem­perature of 115 psi and 1050 F, respec­tively. At steady state the voltage dropacross the magnet was 225 V. The temper­ature of the water leaving the magnet was1500 F. Temperatures at points on theiron and coil surfaces ranged from 2050 Fto 2120 F. Since the flow was 2.5 gpm,the amount of power absorbed by the waterwas 16500 W with 6000 W going into theexterior surroundings. In another test,this time with no water cooling, the mag­net when powered at 40 A with 85.2 Vreached a temperature of 2120 F inside thecoil and approximately 2000 F on thesurface.

At higher temperatures and corres­ponding increased stress on the coil, onemight expect the insulation to fail atareas where slight flaws existed. To testthis possibility, "life tests" were con-

ducted with a few of the first magnetsproduced. These tests consisted of power­ing the magnet for four hours at 100 Awith water cooling and then turning off thepower and allowing the magnet to cool forfour hours. These tests were carried onfor 8-10 weeks. Of the magnets so tested,none were found to fail; failure wouldhave been indicated by shorts to ground orturn to turn shorts.

The pole contour was determined bythe method of conformal transformation ofa complex variable from which a profile isgenerated based on the desired fieldcharacteristics. 3 The requirement on thefield gradient is that it shall be uniformto within 1% out to a radius of 1.5 in.along the horizontal and vertical axes.The profile based on the above criteriawould theoretically produce a gradientuniform to within 0.1% at a radius of.925 in. A plot of the ideal relativegradient is shown in Figure 5.

Based on the results of the methoddescribed abov~, the die for punching thelaminations was constructed. As a checkon the design, profile measurements weremade on the first few laminations. Theseprofile measurements were used as input toa magnet design-aiding program (LINDA)which calculated the corresponding field.The relative gradient ~Gx/Gx determinedfrom these calculations is also shown inFigure 5. The term relative gradient asused here means:

dBy/dx -1 d ~Gy = dBx/dy -1dBy/dxlx=o an Gy dBX/dyly=o •

Magnetic measurements have been madeon several quadrupoles. Measurements ofthe relative gradient along the horizontaland vertical axes have been made at NAL.A measurement to determine the contributionof higher multipoles to the gradient hasbeen completed at the Stanford LinearAccelerator by the Magnetic MeasurementsFacility 4. Also, an excitation curve ofgradient versus current has been measured.

The relative gradients were measuredusing a bucked pair of 2 in. long rectang­ular gradient coils. Because of the coilwidth, measurements could only be made outto x = ± 0.8 in. The average deviation inthe measured values of ~G/G is ± 0.0001.These data are averaged from measurementsmade at 3 points inside the magnet spaced6 in. apart starting at 12 in. from theend. The magnet used is one manufacturedby MCA (#87) which based on the mechanicalmeasurements, is representative of the typeof magnet produced after the productiontechniques were well established. Theresults in Figure 5 show that the

411

measured gradients along both axes andthe theoretical gradient provide a widergood field region than that which wasdetermined from the measured profile.

For the harmonic analysis a magnetproduced at MCA was sent to SLAC. Measure­ments were made at 40 A and 100 A. The40A data were taken without water cooling;therefore, approximately the same coil andcore temperatures existed during the twomeasurements. The measuring techniqueSemployed a uniformly rotating 1 in. longcoil which sensed the field out to a radiusof 1.355 in. The instantaneous voltagewaveform produced by the coil is frequencyanalyzed to determine the harmonic coef­ficients. Approximately 15 points weremeasured at each end. The maximum distanceinto the magnet that could be reached bythe probe was 29 in. from the ends. Basedon these measurements, the average relativegradients at both levels of excitationwere calculated. The results are shownin Figure 6. Based on successive measure­ments, the average deviations are approxi­mately ± (~G/G) x (0.15). This magnet(MCA #5) has a somewhat narrower good fieldwidth over the region measured than thatmeasured for MCA #87 shown in Figure 5.This could be due in part to the fact thatit is among the first produced.

The gradient has been measured as afunction of current. For this measurementa rectangular coil 2 in. by 3/8 in. widewas moved from x = -.5 to x = +.5 in.across the aperture at several points wellinside the magnet. The search coil andintegrator are calibrated using an NMRfield sensor and a uniform field dipolemagnet. The largest uncertainty is inthe shunt calibration which is ± 0.25%.All other uncertainties are less than± 0.05%. A straight line is a very goodapproximation to the curve up to 80 A.The magnet (MCA #87) was cycled severaltimes to 110 A before the data were taken.

IV. Conclusions

As the result of the experiencegained during the construction of thesemagnets, it might be appropriate to pointout some of the areas where a slightmodification of the lamination would haveperhaps facilitated the constructionsomewhat. To do this we refer back toFigure 1, which shows the lamination. Asit turned out, the small key really didnot serve any essential part in the con­struction of the magnet; therefore, byeliminating the notch for this key, onegains by having a wider back leg which,although not necessary from the standpointof flux density, would help in stabilizingthe tip to tip dimension between halves.Then the two halves could have been joined

simply by TIG welding as necessary toachieve a minimum back leg gap. Anotherpoint of difficulty turned out to be the 1.large radius on the portion of the lamin­ation which is welded to the wide key.The large amount of heat and rod requiredto fill this gap caused stresses which 2.tended to distort the magnet half. Itwould have been better to have a very smallradius here so that a TIG weld requiring 3.no additional material and much less heatwould have sufficed.

References

R. Juhala, "Procedure for VacuumImpregnation," NAL TM-3l6 (July 9,1971)

R. Juhala, "Optical Survey of EPBMagnets," NAL TM-31S (July 9, 1971)

S. Snowdon, IEEE Transactions onNuclear Science, Vol. NS-18, No.3,pp 848, (1971)

4. J. Cobb, Private Communication

5. J. Cobb and R. Cole, SLAC-PUB-133,September, 1965

PLAN VIEW

SECTION A-A

I

T.IG. WELD FULLLEHGTH-TYP 7

'aol'ol.o")--A----"v-----l

~~~~~tt=~====~~

From the results of the power testsit can be observed that powering the mag­nets at full excitation (5000 G/in. at100 A) does lead to some undesirable heat­ing. One method of combating this wouldbe to cool the iron since the coil con­tacts the iron over most of its perimeter.One possible way of doing this is to adda one-inch diameter hole in the uppercorner of the lamination into which a thinwall stainless steel tube could be insert­ed. The tube could then be expanded underhydraulic pressure to provide positivecontact with every lamination. In fact,four one-inch diameter tubes were added toone magnet in the space bounded by thecoils, pole tips, and vacuum chamber.This resulted in coil and surface tempera­ture reductions of 300 F when powered at100 A.

There are approximately 100 of thesequadrupoles now installed in the variousextracted proton beam lines throughout theLaboratory. In addition, several of theexperimental areas will be supplied withthese quadrupoles as components in thespectrometer lines.

The magnetic measurement data/thoughnot extensive at this poin~ do inaicatethat substantial differences can exist inthe good field widths of these magnets.The excitation curve indicates that 105 Aare needed to achieve a gradient of 5 kG/in.With the limited water-cooling capabilitynow present in the magnets, maintainingthis level of excitation is probably notpractical for long periods of time.

Acknowledgments

Thanks are due to the Meson Labora­tory technical support for their extraefforts during the building of prototypemagnets. Appreciation is expressed toP. Foster for his thorough inspection ofthe magnets before shipment.

Fig. 1. 3-in. EPB Quadrupole with CrossSection.

The fine work of M. Pastor and W.Heilbraun with MCA on the contract tobuild these magnets is acknowledged. Theassociation with them has been veryenjoyable.

412

//>7=~-:3- f---li""'7-~

<~'/(L::'" '

lJI

Fig. 2. Stacking Fixture for 3-in. EPBQuadrupole.

Fig. 4. EPB Quadrupole.

RELATIVE GRADIENT OF EPB QUADRUPOLE

(MEASURED DATA FROM MCA81)

--I::. - MEASURED PROFILE

i.I

+1.0 -+ 1.5+0.5

CALCULATED

o-0.5

--e--

-1.0

6

I

i

-1.5

-0_2

-0_4

-0.3

0.3

0.4

0.2

I­Z~ 0.1c(It:(!)

:!: 0

'"(!)

zc(

ri -0.1

,t

Fig. 3. Potting Fixture, 3-in. EPBQuadrupole.

DISTANCE FROM CENTER (INCH)

Figure 5.

413

120

CURRENT (A)

0.4 ---0-- 40A. JI 6AG

--e-- 100A.~XIOO

X

0.3

I S10.2 I

x-{ r-:-z

~ /1 ';..z x I C) 4 .I&l 0.1

\ ~~:II:

Q x '-.Jcr:

\)(

III:

r!~~ ZC!»

~ 0 ~

IIIlIJ <]C!»Z

~3

cr::I: Z0 -0.1

J.-A- 40A. x I&l

AG

~a

~ T X'OO cr:III:

--X-100A. Y C)

-0.2

II2 EPB QUADRUPOLE

EXCITATION CURVE

-0.3 RELATIVE GRADIENT OF EPB

QUA DRUPOLE( MCAS)

IAS MEASURED BY SLAC

-0.4

Figure 6. Figure 7.

414