TRIUMF: Canada's national laboratory for particle and nuclear physics

Proceedings of the KAON PDS Magnet Design Workshop, Vancouver, October 3-5, 1988 Otter, A. J.; Strathdee, A. Mar 31, 1989

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


51833-TRI-1989-01.pdf [ 89.98MB ]
JSON: 51833-1.0107822.json
JSON-LD: 51833-1.0107822-ld.json
RDF/XML (Pretty): 51833-1.0107822-rdf.xml
RDF/JSON: 51833-1.0107822-rdf.json
Turtle: 51833-1.0107822-turtle.txt
N-Triples: 51833-1.0107822-rdf-ntriples.txt
Original Record: 51833-1.0107822-source.json
Full Text

Full Text

TRIUM FP R O C E E D IN G S  OF TH E K A O N  P D S  M A G N E T  D E S IG N  W O R K S H O P  VANCOUVER O C TO B ER 3-5, 1988 E d itors: A .J . O tter  and A . S tra th d eeCANADA’S NATIONAL MESON FACILITY OPERATED  AS A JO IN T  V ENTURE BY:UNIVERSITY OF ALBERTA SIMON FRASER UNIVERSITY UNIVERSITY OF VICTORIA UNIVERSITY OF BRITISH COLUMBIAUNDER A CON TRIBUTION FROM THE NATIONAL RESEARCH COUNCIL OF CANADA T R I-89-1T R I —89-1P R O C E E D IN G SOF TH EK A O N  P D S  M A G N E T  D E S IG N  W O R K S H O PVANCOUVER O C TO B ER 3-5, 1988E d itors: A .J . O tter  and A . S tra th d eeP o sta l address:T R IU M F4 0 0 4  W esb ro o k  M all V a n co u v er , B .C .C an ad a  V 6 T  2 A 3  M arch  1989PREFACEThese proceedings bring together the papers given at the M agnet Design Workshop (October 3-5) which was held to  kick off the KAON Factory PDS which was officially started  on O ctober 1, 1988.The workshop included sessions on power supplies and m easurem ents as well as syn­chrotron and kicker m agnet design. The aim of the m eetings was to  bring together experts who could advise us on m agnet and power supply techniques which, prior to  the KAON era, have not been required at TR IU M F. These include fast-cycling cyclotron m agnets and their power supplies, and the kickers needed to  switch the beam  from one ring to  another or to the experim ental areas. We also invited participation from industrial companies who will be potential m agnet suppliers when the KAON Factory is funded. It was a pleasure to  have representatives from six industrial companies amongst the participants.A to ta l of 51 people a ttended  the meetings. They came from a variety of places, as listed below:Scientists and engineers from other laboratories 23Scientists and engineers from TR IU M F 18Scientists and engineers from industry  9Government representatives 1The workshop was arranged in a little  over seven weeks so the good attendance was encour­aging and indicative of the world-wide enthusiasm  for our project.Several things were accomplished in the three days of presentations and discussions. We received lots of advice on how to build ac m agnets and how not to  build kickers. As a result of discussions and offers of help from other laboratories, our proposed kicker program  has been completely revised. We also met and forged links with “the experts” from around the world, m any of whom will visit us again as consultants during the PDS year.These proceedings include a m ajority of the presentations given at the technical sessions plus a sum m ary of the discussions which followed. We have also included as an appendix several w ritten contributions sent to  us as a result of ideas generated at the workshop and which are very pertinent to  our KAON study.The workshop was arranged by a com m ittee of Ew art Blackmore, Mike Craddock, A1 O tter and Lorraine King. Thanks are expressed to  all the participants and particularly  to  our invited speakers, who came at short notice and assisted us by chairing the discussion sessions:B. Berkes Paul Scherrer In stitu te  SwitzerlandD. F iander CERN SwitzerlandN. M arks ESRF FranceM. Harold RAL EnglandJ. Riimm ler DESY W est Germ anyV. Rodel CERN SwitzerlandH. Sasaki KEK JapanW . Praeg Argonne U.S.A.We would also like acknowledge the able secretarial assistance of Krish T hiruchittam - palam , M argaret Lear and M aureen W hite during the workshop and of Jana  Thom son in the preparation of these proceedings.CONTENTSPreface .............   iiiOpening rem arks, A . Astbury  ...............................................................................................................  1T R IU M F’s KAON Factory accelerators, M. Craddock .................................................................  2Fast-Cycling M agnetsAc m agnets for the KAON Factory accelerators, U. W ienands andR .V . S e rvra n c k x ............................................................................................................................ 11The design, m anufacture and testing of the ISIS fast-cycling m agnets,M .R. H a ro ld ....................................................................................................................................15Design and m easurem ent of ac m agnets a t DESY II, G. H e m m ie ........................................27A dc biased rapid-cycling m agnet system operating in a dual frequency mode,H. Sasaki, T. Adachi, H. Someya and I. Sakai ................................................................ 33ESRF booster synchrotron m agnet and power supply design, N. M a r k s .......................... 41Power supply considerations for m agnet design, K. R e in ig e r ................................................50Fermilab m agnet construction, J.C . Humbert ............................................................................51The sta tus of the booster dipole design and plans for the Project DefinitionStudy year, A . O t te r ....................................................................................................................56KAON Factory m agnets, V. V e rm a ...............................................................................................62KickersKicker requirem ents for the KAON Factory, U. W ie n a n d s ....................................................68Kickers and septa a t the PS complex, CERN, D. Fiander, K -D .M etzm acherand P. Pearce ................................................................................................................................ 71New stripekicker in the injection chain of HERA, J. R iim m le r ............................................80Overview of kicker m agnet systems at the SPS and LEP, V. Rodeland G.H. Schroder .......................................................................................................................87A kicker upgrade for Los Alamos proton storage ring, H.A. T h ie s s e n ...............................96A prelim inary design of the Los Alamos fast kicker m agnet pulser andpower supply, R .A . W inje ....................................................................................................... 101M easurem ents and Radiation-H ard M agnetsM agnet requirem ents for experim ental areas, E.W . Blackmore and A .J . O t t e r  106Radiation-hardening of m agnet coils, A . Harvey .................................................................... 112Search coil m easurem ents of particle accelerator m agnets, K .N . Henrichsen  ................119Ac m agnetic m easurem ents of the ALS booster synchrotron dipole m agnetengineering model, M .I. Green, E. Hoyer, R. Keller and D.H. N e ls o n .....................124Loss m easurem ent program s at TR IU M F, A . Otter and W. Neves ................................. 132M ethods of estim ating iron losses and field errors in ac m agnets,P .A . Reeve .................................................................................................................................... 136Reports on W orking DiscussionsKickers, D. F ia n d e r ...........................................................................................................................139Fast-cycling m agnets, M .R. Harold ..............................................................................................142M agnet power supply system , N. Marks ....................................................................................144Industrial participation, A. O t t e r ...................   148A ppendixA shielded energy-storage choke for rapid-cycling synchrotrons, W.F. Praeg ...................150Dual versus single frequency ring m agnet power supplies for rapid-cyclingsynchrotrons, W .F. P ra e g .........................................................................................................157Calculating the frequency response of m agnets w ith lam inated cores,W .F. P ra e g ....................................................................................................................................164A prototype kicker m agnet for the kaon factory, V. R o d e l ................................................... 172List of participants ...................................................................................................................................177viOPENING REMARKSA. AstburyTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V6T  2A3Let me begin by welcoming you to  this M agnet W orkshop which is a  part of what is known as the KAON Factory Project Definition Study. In particular I would like to  thank our experts, m any of whom have crossed several tim e zones in coming to  Vancouver to  help us. We believe we have some challenging problems, we know you have the expertise to  solve them . It is perhaps interesting to  outline the steps -  but not all of them! -  which have led to  the present stage of the TRIU M F KAON Factory Project.Our proposal for a 30 GeV 100 /iA facility was worked on during 1984 and 1985, and finally tabled in Septem ber 1985. The two Canadian Research Councils (NRC and NSERC) struck two committees. The first of these -  a Joint International Technical Review Committee -  m et in February 1986 and gave the proposal an extremely high rating  for science and potential realisation w ithin the proposed budget and timescales. The second com m ittee -  a Jo in t “C anadian” Review Com m ittee -  was charged with examining the proposal in the very broad context of Canadian Science. They endorsed the excellence of the science, bu t could not agree th a t the funding of such a m ajor initiative in C anada would not grossly d istort the support to  basic science in the country. The councils of NRC and NSERC accepted this view, but of course the m anagem ent of TRIU M F could not. In February of 1987 the KAON Factory Proposal was presented to  the Economic Development Com m ittee of the Cabinet of the British Columbia Provincial Government where it found im m ediate support to  the extent of a full com m itm ent of B.C. to  provide $98M of civil engineering.During 1987 the idea emerged of a jointly funded project which would define better the design and costs of the KAON Factory, explore fully the possibility of international contributions in a HERA style model, and study the Canadian industrial capabilities for building, and potential future enhancem ents which might be gained from the experience of building. This is now our Project Definition Study, and the jo in t Federal and Provincial agreement was signed by the M inisters Mr. Frank Oberle and Mr. Stan Hagen in Vancouver on July 21, 1988, m aking available $11M for the study.We have subsequently presented our budgets to  the Steering Com m ittee and in fact these were put in place -  effective October 1 . Here we are now on October 3, the first official working day of the project, kicking off with our M agnet W orkshiop. We believe we are on the way towards building our KAON Factory, we know we have m any problems to  solve, we can’t th ink of a  more pleasant beginning then sharing some of these problems with you.2TH E TR IU M F KAON FACTORY ACCELERATORS M.K. Craddock*TR IU M F, 4004 Wesbrook Mall, Vancouver, B.C. C anada V 6T  2A3ABSTRACTThe TR IU M F KAON Factory proposal has m ade considerable progress on both  tech­nical and political fronts over the last year. For the m ain rings a racetrack-shaped lattice has now been adopted in conjunction with a three-elem ent slow extraction system in an effort to  reduce losses to  the 0.1% level. Hardware studies have continued on both  m agnet power supplies and on rf cavities -  the la tte r  work gaining an ex tra  dimension from a recently in­s titu ted  form al collaboration w ith LAM PF. The H " extraction system  for the cyclotron has been tested  successfully w ith 66 /xA pulsed and 10 /xA average beams. On the  political side, British Columbia has agreed to fund the buildings and tunnels (Cdn $87M) and is con tribu t­ing jointly  with the Canadian federal government to  an $11 million pre-construction R & D study over the next 15 months. This will allow construction of prototypes of m agnets, power supplies, kickers, rf  cavities, ceramic beam  pipes, targets and controls. It will also perm it economic and environm ental im pact studies and formal consultations abroad. These will follow up on the exploratory talks last year, when it was found th a t a num ber of countries would consider m aking significant contributions to  the cost. These steps would pave the way for project approval in mid-1990.INTRODUCTIONThis talk  will outline the design of the KAON Factory accelerators and review the technical studies now under way. The TR IU M F K aon-A ntiproton-O therhadron-N eutrino Factory is fully described in the original proposal1 and outlined in various papers .2’3 The prim ary aim is to  provide a 100 /xA beam  of protons a t 30 GeV — about a hundred times more than  available a t present. The TRIU M F H " cyclotron, which routinely delivers 150 /iA beam s a t 500 MeV, would provide a ready-m ade and reliable injector. It would be followed by two fast-cycling synchrotrons, interleaved with 3 storage rings, as follows:A Accum ulator: accum ulates cw 450 MeV beam  from the cyclotron over 20 ms periodsB Booster: 50 Hz synchrotron; accelerates beam to  3 GeV; circumference 214 mC Collector: collects 5 Booster pulses and m anipulates longitudinal em ittanceD Driver: main 10 Hz synchrotron; accelerates beam  to 30 GeV; circumference1072 mE Extender: 30 GeV stretcher ring for slow extraction for coincidence experimentsIt can be seen from the energy-time plot (Fig. 1) th a t this arrangem ent allows the cyclotron ou tpu t to  be accepted w ithout a break, and the B and D rings to  run continuous acceleration cycles w ithout wasting tim e on flat bottom s or flat tops; as a result the full 100 /xA from the cyclotron can be accelerated to 30 GeV for either fast or slow extraction.The use of fast-cycling synchrotrons lowers the proton charge per pulse Ne to  2 /xC (Booster) and 10 /xC (D river), levels a t which the space-charge tune shift a t injection is tol­erable (A Q < 0.2). Since AQ ex Ne*(/?72)inj the use of a Booster perm its a smaller normalized*On leave from Physics D ept., University of British Columbia, Vancouver, B.C ., V61 2A6.30 100 200Fig. 1. E nergy -tim e p lo t show ing th e  progress o f th e  b eam  th ro u g h  th e  five T R IU M F  K A ON  F actory  rings.em ittance e* and hence reduces the aperture  and cost of the Driver m agnets. The Booster energy is chosen to  minimize the to ta l cost of the project. Since this depends mainly on m agnet costs, the minimum is found to  occur when the em ittances set by the space charge tune shift form ula are the same for both machines.3,4The use of a Booster also simplifies the rf design by separating the requirem ents for large frequency swing and high voltage (33% and 600 kV respectively for the Booster, and 3% and 2400 kV for the Driver). These high rf voltages are associated with the high cycling rates; the use of an asym m etric m agnet cycle with a rise tim e 3 tim es greater than  the fall (Fig. 1) reduces the voltage required by one-third, and the num ber of cavities in proportion.Figure 2 shows a proposed site layout together w ith cross-sections through the tunnels, w ith the Accum ulator above the Booster in the small tunnel, and the Collector and Extender rings above and below the Driver in the m ain tunnel. Identical lattices and tunes are used for the rings in each tunnel. This is a natural choice providing structu ra l simplicity, similar m agnet apertures and straightforw ard m atching for beam  transfer.F ig. 2. Possib le s ite  layou t w ith  a  ra c e tra c k  la ttic e .4Separated-function m agnet lattices are used w ith a FODO quadrupole arrangem ent. In the  A and B rings missing dipoles are arranged to  give superperiodicity S = 6 . This au­tom atically provides space for rf  and beam  transfer equipm ent. It also m odulates the dis­persion function r/x and drives its mean value < tjx > towards zero, enabling transition to be driven above top energy. In the absence of zero-dispersion stra ights, nearby synchro- be ta tron  resonances are suppressed by placing the rf cavities sym m etrically w ith the m agnet superperiodicity.For the  C,D and E rings a racetrack lattice  has now been adopted in place of the originally-proposed S=12 lattice. As explained below, this provides the long straights required for a  super-efficient slow extraction and collimation system in the E ring, while keeping 7 * above top energy.Injection into the Accum ulator is achieved by stripping the H-  beam  from the cyclotron (see below) enabling m any turns to  be injected into the same area of phase space. The small em ittance beam  from the injector is in fact painted over the much larger three-dimensional acceptance of the A ccum ulator to  lim it the space charge tune shift. Painting also enables the optim um  density profile to  be obtained and the number of passages through the stripping foil to  be lim ited.Work on the design has continued since the proposal was issued in 1985. Up to  now lim ited funds have restricted  hardw are development to  rf cavities, m agnet power supplies, targets and H~ extraction from the cyclotron. But w ith the onset of the $11M Project Definition (pre-construction) Study this m onth prototype construction will be extended to m agnets, kickers, ceramic beam  pipes and controls. The following section describes individual developments. A nother m ajor development was the institu tion  in August 1987 of a formal collaboration w ith LA M PF on accelerator studies. The first areas for collaboration have been rf studies and beam  commissioning work on the Los Alamos Pro ton  Storage Ring. The m utual benefits of this arrangem ent are already apparent.TECHNICAL STUDIESM agnet Lattices and Slow ExtractionConsiderable thought has been given to  the possibility of reducing the losses a t slow ex­traction  to  0.1% rather than  the 1% typically obtainable w ith existing system s. U. W ienands5 has suggested the use of a short additional pre-septum  to  dilute the beam  density a t the main septum , dem onstrating, in a sim ulation, a  loss of only 0.2%. A th ird  septum  could not easily be accom m odated in the superperiodicity-12 lattice  of our m ain ring reference design. In­stead, R.V. Servranckx6 has proposed a racetrack lattice  (Fig. 2) with two dispersionless 167 m long straight sections. Horizontal /3 values of about 100 m are obtained near the focusing quadrupoles, providing low density locations for the septa. The achrom atic 180° arcs contain 24 cells, and are tuned to  5 x2tt (~  75°) per cell. The tune of the  straight sections may be adjusted to  give a to ta l tune variation for the ring of ± 1  in each plane independently. A half-integer resonance may be used for extraction, to  simplify the collimation process. Such a racetrack la ttice  is convenient for the Driver synchrotron as well as for the Extender, providing more flexibility either for the insertion of Siberian snakes7, or for tuning for low depolarization w ithout snakes, using high-periodicity arcs and sp in-transparent straight sec­tions. Investigation of the properties of the lattice in detail show th a t it is no more sensitive to  beta tron  resonances than  the old circular design, and hence it has been adopted as the new reference design.5Racetrack lattices are also being investigated for the Booster and Accum ulator rings, where they would provide dispersion-free regions for rf cavities and beam  transfer. A FODO lattice, similar to  th a t proposed for the main rings, bu t with 4 superperiods and a tune of 3 per arc, has been considered, but requires ra ther short dipoles and too m any ceramic-steel vacuum joints. A lternatives under investigation are doublet and trip le t lattices, and a hybrid using combined-function dipole magnets. U. W ienands8 gives more details of the lattices and m agnet specifications.M agnet DevelopmentDesign studies are underway on both  separated and combined function dipole magnets for the Booster ring and are fully described by A .J. O tte r .9 One of these designs will be selected for prototyping, while design studies continue on the various o ther m agnets needed in both  accelerators and beam lines.M agnet Power SuppliesAs explained above, dual-frequency magnet excitation is planned for the synchrotrons, w ith a rise tim e three times longer than  the fall. To test the performance of such a system a high-power test stand has been set up (Fig. 3). Four m agnets from the decommissioned NINA synchrotron are used, one as the load and three in series as the resonant 81 mH choke. A 1000 /rF capacitor bank may be switched in parallel with a 125 /rF bank to  change the resonant frequency from 100 Hz to  33 Hz. This stand has been operating for a short while now and successful tests have been carried out a t fixed frequencies (Reiniger10).Fig. 3. Iligh-pow er te s t s ta n d  for dual-frequency  m ag n e t ex c ita ­tio n  stud ies.6KickersW ith  the increased funding available from the Project Definition Study work has re­cently begun on the design of the extraction kicker for the Booster ring -  probably the kicker w ith the m ost challenging specifications -  about 60 kV across an 8 cm gap over a length of 2 m and with a  rise tim e <80 ns. Our aim  is to  have a prototype kicker operating by the end of 1989.Radio Frequency SystemsThe reference design for the Booster cavities is based on those used in the Fermilab booster. A full scale prototype cavity is alm ost complete and should be ready for tests with an air tuner soon (Poirier et a/.11). The collaboration with LAM PF has also enabled us to  study the possibility of using a version of the Los Alamos cavity which employs perpendicularly- biased microwave ferrite. Under dc bias conditions this has produced relatively high voltages (140 kV), potentially  reducing the num ber of cavities required and, more im portantly , the im pedance presented to  the beam  and the likelihood of inducing coupled-bunch instabilities. In Septem ber 1987 the TRIU M F group was able to  make m easurem ents on the Los Alamos cavity and dem onstrate its operation w ith good Q-values down to  and below the lowest frequencies (46 MHz) required at TR IU M F (Fig. 4).F R E Q U E N C Y  ( M H z .  )Fig. 4. P erm eab ility  m easu rem en ts  on th e  Los A lam os booster cav ity  show ing good beh av io u r over th e  en tire  frequency  range requ ired  a t  T R IU M F .Enegren and Poirier12 have calculated the transmission-line cavity modes for the Los Alamos and Fermilab-style cavities. The effect of dam ping was also investigated. These studies indicate a further advantage of the Los Alamos cavity, whose shortness reduces the num ber of modes in a given frequency interval.W ith  the Los Alamos group transferring their activities to  the development of a main ring cavity, they have generously m ade their booster cavity available on loan to  TR IU M F, where it will be tested  under ac bias conditions -  the crucial rem aining test of its viability.7R. Burge13 and W . R oberts14 have studied control of the rf system s under high beam loading. Burge presents designs of feedback circuits for phase and am plitude control. S. Koscielniak15 has m ade an analytical study of radial and phase control of the rf taking explicit account of tim e delays. Local pickups appear to  be preferable to  shared ones and the radial loop control signal should drive the m aster oscillator ra ther th an  a  phase adjuster upstream . As part of its collaboration with LAM PF TR IU M F will build the low-level control system and also a solid s ta te  driver for the main ring cavity. Kwiatkowski et al.16 have described the design of the power amplifier for the Booster cavities.Beam Pipe fc VacuumThe vacuum requirem ents for all five rings are being carefully reviewed (Oram  & B aartm an17). The high circulating beam current makes ion desorption from the walls the m ost critical process. This requires a hydrocarbon-free system w ith all m etal elements pre­baked to  300°C, and pum ps spaced no more than  5 m apart. This will autom atically result in vacua be tte r than  10 8 Torr. An additional concern in the Extender ring, where the beam may be debunched, is the possibility of electron-proton oscillations; electrostatic collector plates will be needed to  suppress these.Com puter Control SystemA six-m onth study of the KAON Factory control system has been completed with the help of two visitors from CERN. A comprehensive review was carried out of both  hardw are and software options and a full report is now available (Dawson et al.)18. A test platform  is being assembled based on a VAX3200 workstation w ith a bridged E thernet connexion to 2 VM E crates.H~ E xtraction from the cyclotronTo ex tract H_ ions (instead of stripping them  to protons as in norm al operation) a conventional extraction system is being developed. Laxdal et al.19 report th a t with 18 kV onPROBE RRDIUS ( INCHES)Fig. 5. In teg ra l an d  d ifferen tia l p ro b e  scans in th e  H~ e x tra c tio n  region on  th e  T R IU M F  cyclo tron .the  rf deflector and 50 kV on the electrostatic deflector 90% of the beam  (66 fiA  macropulses a t 1% duty  factory) is transm itted  through the la tte r (Fig. 5). The 10% not transm itted  is stripped by a narrow foil shadowing the septum  and protecting it from irradiation; the resulting protons may be dum ped or steered in to  an experim ental beam  line. The differential scan in Fig. 3 illustrates the intensity m odulation and improved tu rn  separation produced by the rf deflector in conjunction with the Qr = 3 /2  resonance. In recent tests the average beam current was successfully raised to  10 //A. Design of the 4-segment m agnetic channel which will steer the H_ beam  out of the cyclotron is now alm ost complete. Detailed design of the front end of the  external beam  line is under way.PROGRESS TOWARDS FUNDINGGood progress has been made over the last year, a crucial factor being the British Columbia provincial governm ents’s strong support for the project as its top priority among federal projects and the centrepiece of its high-tech development strategy. In February 1987 the B.C. cabinet gave the project formal approval in principle; i.e. agreement to  fund the civil works ($87M Canadian) provided the federal government funds the technical equip­m ent. A tangible symbol of the B.C. governm ent’s com m itm ent has been its commissioning of specially-labelled “KAON P ro jec t” wine, which has already emerged the winner of an in­ternational tasting  contest and which you will have the opportunity  of sampling during this workshop!On the federal side the M inister of S tate  for Science and Technology, Mr. Frank Oberle, last year agreed to  in stitu te  jo in t federal-provincial studies of additional university involve­m ent and international contributions to  the funding. On the first issue the four founding uni­versities of TR IU M F have already been joined by the University of M anitoba and L’Universite de M ontreal as associate members, while the University of Toronto is to  join shortly.On the second item , Mr. Oberle himself, in a speech to  the OECD nations in Paris, sta ted  “we are anxious to  seek and develop o ther jo in t ventures” , giving us an example “ partnership in the construction of the Kaon Factory” . A Canadian delegation was appointed to explore the potential for such partnership  and in November and December 1987 visited W est Germany, Italy, Japan  and the U.S.A. Each country agreed to consider financial involvement in construction, and indeed the possibility of support is being explicitly allowed for in the planning scenarios of both  Germany and Italy. If negotiations are successful the external contributions will am ount to  considerably more than  the $75M recommended by the Kaon Factory Review Com m ittee. Besides the countries m entioned above, Belgium, B ritain , Israel and the People’s Republic of China have all expressed interest in participating in experim ents and in some cases in accelerator design and construction.These discussions will now continue more formally under the aegis of the $11-million Project Definition Study which began officially on 1 October, funded jointly  by the govern­m ents of C anada and British Columbia. The projects being funded are as follows:Accelerator Design Project M anagem entR F Prototypes Building DesignM agnet Prototypes Tunnel DesignM agnet Power Supplies Service and Power D istributionBeam PipeKickersCyclotron Beam Extraction9Shielding and SafetyTargetsControlsSystems Integration Experim ental Areas Science WorkshopsIndustry  Development In ternational Consultations Economic Assessment Legal & Environm ental StudiesThis pre-construction study is expected to  be complete by the end of 1989, leaving the way clear for final approval of the project in early 1990.ACKNOW LEDGEM ENTSIt is a pleasure to  acknowledge the efforts of all those who have worked to  improvethe KAON Factory design over the past year. The au thor is particularly  grateful to  JanaThom son for the accuracy of the typing.REFERENCES1. KAON Factory Proposal, TR IU M F, Septem ber, 1985.2. M.K. Craddock, R. B aartm an et al. IEEE Trans. Nucl. Sci. NS-32, 1707 (1985).3. M.K. Craddock, Proc. In t. Workshop on Hadron Facility Technology, Santa Fe, Febru­ary 1987, ed. II.A. Thiessen, Los Alamos Report LA-11130-C, pp 8-31 (1987).4. U. W ienands and M.K. Craddock, TRI-DN-86-7 (1986).5. U. W ienands and R.V. Servranckx, “Towards a Slow E xtraction  System for the T R I­UM F KAON Factory Extender Ring with 0.1% Losses” , Proc. 1st European Particle Accelerator Conference, Rome, June 1988 (to  be published).6 . R.V. Servranckx, TR IU M F Report TRI-DN-88-3, January  1988.7. U. W ienands, “Discrete Helical Spin R otato rs” , Proc. 1st European Particle Accelerator Conference, Rome, June 1988 (to  be published).8 . U. W ienands, “Ac M agnets for the KAON Factory Accelerators” , these proceedings.9. A .J. O tter, “The S tatus of the Booster Dipole Design and Plans for the Project Defi­nition Study Year” , these proceedings.10. K. Reiniger, “Power Supply Considerations for M agnet Design” , these proceedings.11. R.L. Poirier et al., “Parallel Bias versus Perpendicular Bias of a Ferrite-Tuned Cavity for the TR IU M F KAON Factory Booster Ring” , Proc. 1st European Particle Accelerator Conference, Rome, June 1988 (to be published).12. T.A . Enegren and R.L. Poirier, Proc. of the AHF Accelerator Design W orkshop, Los Alamos, February 1988, ed. H.A. Thiessen, Los Alamos Report LA-11432-C, p 308, (1989).13. R. Burge, ibid., p 298.14. W. R oberts, ibid., p 353.15. S. Koscielniak, ibid., p 262.16. S. Kwiatkowski et al., “Design of the 150 kW  46-62 MHz Power Amplifier for the T R I­UM F KAON Factory Booster Ring” , Proc. 1st European Particle Accelerator Confer­ence, Rome, June 1988 (to  be published).17. C .J. Oram  and R. B aartm an, “Vacuum requirem ents for the KAON Factory” , TRI- DN-88-K 6 (1988).1018. W .K. Dawson, R.W . Dobinson, D. Gurd and Ch. Serre, “A Conceptual Design for the TR IU M F KAON Factory Control System ” , TRIU M F Report TRI-87-1, July 1987.19. R.E. Laxdal et al., “Progress Towards H~ Extraction at T R IU M F” , Proc. 1st European Particle A ccelerator Conference, Rome, June 1988 (to  be published).11AC M AGNETS FO R TH E KAON FACTORY ACCELERATORSU. W ienandsTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V 6T 2A3R.V. ServranckxSaskatchewan Accelerator Laboratory, Saskatoon, Sask., C anada S7N 0W0and TRIU M FA BSTRACTThe requirem ents for the m agnets for the KAON Factory accelerators are given. Besides a tabulation  of the m agnet da ta , field uniformity is discussed and lim its are derived from tracking studies. Dynamic tolerances are discussed.INTRODUCTIONA ltogether, about 800 m agnets will be needed for the 5 rings of the KAON Factory accelerator complex. Of these m agnets 500 will be dc m agnets, 200 ac m agnets cycling at 10 Hz, and 100 ac m agnets cycling at 50 Hz. In this paper the requirem ents for the ac dipole and quadrupole m agnets for the KAON Factory accelerators are discussed. The m agnets will be looked at from the point of view of the beam physicist, i.e., w ith regard to  field quality, aperture, and effective lengths, ra ther than  from the m agnet builder’s point of view, who may be concerned with other problems such as lam ination thickness, pole profiles, etc. For specifying the m agnets of the Driver ring we will take the newly developed racetrack lattice1 as the reference design, while Booster m agnet d a ta  for two lattices are given, for the circular lattice  as given in the KAON Factory proposal and for an alternative racetrack lattice currently under investigation. The la tte r  uses defocusing combined-function dipole m agnets and separate focusing quadrupoles, hence its being referred to  as hybrid lattice. In Figs. 1 and 2 beam envelopes for the racetrack lattices are given for the nom inal em ittance of the beam  at injection, epsx =  140 w m m -m rad, epsy = 62 7r m m -m rad for the Booster, and epsx = 3 6 .1 7r m m -m rad, epsy =  16.8 7r m m -m rad for the Driver.M AGNET DATABasic d a ta  for the dipoles are given in Table I and for quadrupoles in Table II. The beam sizes contain a safety factor of two in the em ittance and allowance for closed-orbit distortions and for dispersion. The beam  em ittances are determ ined by space-charge tune shift th a t is to  be less than  0.15 at injection and contain some allowance for em ittance blowup in the ring and during transfer from ring to  ring. The gap height contains 1 cm allowance on each side for vacuum chamber thickness and clearance to  the pole faces. The maximum  fields drop with repetition ra te , hence the lower peak field of 1.05 T for the Booster, while the Driver bending m agnets peak at 1.35 T. The different rise and fall tim es of the (sinusoidal) magnet cycle reflect the proposed use of a dual frequency m agnet excitation in order to  reduce the maximum rf accelerating voltage needed. Of the two Booster scenarios, the hybrid racetrack Booster needs much less aperture than  the circular lattice; however, the dipoles have a field index of about 50 and will be more difficult to  design and build and also may have higher stored energy.12T able  I. D ipole m agnets .R ing D river B ooster circ. B ooster (n = 50) U nitshor. vert. hor. vert. hor. vert.V W e 4.85 2.92 5.74 3.82 3.10 2.86 cmdpr) 1.70 0 0.78 0 0.23 0 cm2 c.o .d . 0.32 0.68 0.7 0.5 0.7 0.5 cmT o ta l 6.87 3.6 7.3 4.4 4.1 3.4 cmAll values :are 1 /2  values: to ta l rounded  up  to  n ex t m m .gap 9.2 0° 00 6 .8 cmleng th 5.2 3 .18 2 .4 mBmin 0.16 0 .277 0 .277 T^max 1.35 1.05 1.05 T/rep 10 50 50 HzI rise 75 15 15 m sI fall 25 5 5 m snu m b er 96 24 32T able II. Q u ad ru p o le  m agnets .R ing D river B ooster circ. B ooster (n = 5 0 ) U nitsQF QD QF QD QFy J W t 5.1 2.21 6.56 4.19 4.42 cmdpi) 1.81 0.82 2.14 0.9 0.34 cm2 c.o .d . 0.32 0.68 0.7 0.5 0.7 cmT o ta l 7.3 3.8 9.4 5.6 5.5 cmAll values a re  1 /2  values: to ta l rounded  up  to  nex t m m .rad iu s 8.3 4.8 10.4 6.6 6.5 cmlen g th 1.4 0.82 0.65 0.43 1.1 THtip,min 0.12 0.12 0.17 0.17 0.17 TFhip,max 1.01 1.01 0.63 0.63 0.63 T/rep 10 50 50 IIz ^rise 75 15 15 m sIfall 25 5 5 m snum ber 48 48 24 24 3213i 1 i ■ 1—i 1 i 1I  ■ I 1 I I I I I I  1 1 1___ I___ I----- 1----- 1----- 1----- 1----- 1----- 1----- 1----- 1----- I----- 1----- L.E jr =  140.00 n  m m - m r  A p /p  =  0 .00342XE-Q  0  -\ \  ' V v ^ V v V v V v V V v V V " - 'AAAA A /V 's  s < *3 ICQ 8 -  , x* - 5  -V ' A V v V M V . V A  / \  / \ x /eT =  62.00 7r m m - m r1 I 1 1 ’ 1 I 20 40r I 1 1“60 80 100 120D IS T A N C E  ( m )Fig. 1. B eam  envelopes for th e  hyb rid  B ooster la ttice .i 1 i 1 i ' i V i'a  xO 04'—NX  E- Q 02 2  < *W |m i -  5_1---- 1---- 1---- 1-----1-----1---- 1-----1---- 1-----1-----L_e x =  36.10 7T m m - m rA p /p  =  0.00273, i ,  ;  /' i  A , a / \ ( i / i  , ' i ,  A.*, a > f. / . t ,  | V  V V  V  ' \ A / V v  i / ' W . A f v V'/ V  v ti v  v V v v 11 1/ v ^AA/si iwV ^ /XA/V'A'VvVVVvV^Vv^/VyV^JV \/\/\/\/VVeT =  16.80 7T m m - m r~i— i— i— i— |— i— i— i— i— r—i— i— i— i— |— i0 100 200 300 400 500 600D I S T A N C E  ( m )Fig. 2. B eam  envelopes for th e  race track  D river la ttice .FIELD QUALITYIn order to  examine the effects of nonuniformity of the field, the horizontal field profile of a prelim inary design for a Driver dipole2 was param etrized in form of field harmonics. These field harm onics were sim ulated by multipole elements in a com puter model of the Driver ring using the code DIM AD .3 Figure 3 shows the field shape used for the Driver dipole m agnet. The field uniformity is about ± 4 x l0 ~4 over the apertu re  of 7.3 cm. For the quadrupoles m easured field harm onics of a standard  TR IU M F 20 cm quadrupole were used, having typical field harm onics of a few 10" 3 of the pole-tip field a t 8 cm radius. Tracking runs showed the acceptance of the machine with these m agnets to be about 1.3 tim es the design em ittance of the beam. This is acceptable although in view of finite alignment tolerances14X ( c m )Fig. 3. H orizon ta l field un ifo rm ity  of th e  D river dipole; b2, b4 an d  b6 are  th e  field expansion  coefficients for sex tupo le , decapo le an d  14-pole com ponen ts , respectively .of the m achine and resulting orbit excursions, the requirem ents m ay eventually tu rn  out to  be somewhat more stringent. A similar study done for the circular Booster indicates an acceptance of 1.5 times the beam em ittance for ±7.5 cm good-field region. Acceptance was lim ited by the field uniformity of the dipoles ra ther than  of the quadrupoles.A nother issue with ac m agnets is the dynamic behaviour, i.e., errors in m agnet field setting , especially a t in jection/extraction, tracking errors, etc. In general we found variations on the order of 10-4 to  be tolerable w ithout special corrections, while errors on the order of 10-3  need corrective action by trim  elements; this is true  for field setting  and field track ­ing errors. Reproducibility from cycle to cycle is alm ost impossible to  correct; therefore, a variation of less than  10~4 from cycle to cycle is needed. For the field index of the combined function m agnets of the hybrid Booster a tolerance of 10~3 appears to  be tolerable.REFEREN CES1. R.V. Servranckx, TR IU M F design note TRI-DN-88-3 (1988).2. E.M . Gibson, “The Booster M agnet for the TR IU M F KAON Factory” (A. O tter, private com m unication).3. C. M anz, TR IU M F design note TRI-DN-87-31 (1987);G. W ellman, Horizontal Studies for the KAON Factory Driver Ring” (U. W ienands, private com m unication).15The design, manufacture and testing of the ISIS fast-cycling magnets M R HaroldRutherford Appleton Laboratory September 1988IntroductionPreliminary design work on a spallation neutron source [l] was begun in 1975. The SNS (the name was later changed to ISIS) was approved in 1977, and beam was first accelerated and extracted in 1984.ISIS uses a synchrotron designed to provide 180 pA (2.5 E13 protons per pulse) at 800 MeV with a pulse repetition frequency of 50 Hz. Injection at 70 MeV is achieved by the stripping of H“ ions, and acceleration in two bunches is provided by six RF stations. The beam is extracted in the vertical plane and taken to the U 238 target by means of a beamline 150 m long (fig 1). ISIS presently operates at a current of about 100 ytA at 750 MeV.The synchrotron has a mean radius of 26 m and superperiodicity of10. Much of the circumference consists of straight-section space, due to the requirements of injection, extraction, diagnostics and the RF. The design emittances are 540 and 430 mm-mrad in the horizontal and vertical planes respectively, figures which lead to large aperture magnets.Each superperiod (fig. 2) consists of a 36 deg dipole, a quadrupole doublet and a singlet defocusing quadrupole, all of which are electrically in series. The dipole has a weak transverse gradient to provide horizontal focusing. Associated with each of the doublet quads is an independently-powered trim quadrupole, and there are eleven orbit- correcting dipoles [2]. Spaces have been reserved for sextupoles and octupoles; these have been designed but not yet manufactured. The magnet parameters are given below.Magnet Dipole Doublet F(D) Singlet D TrimNumber of magnets 10 10(10) 10 20Field at injection (T) 0.176Field at 800 Mev (T) 0.697Normalized gradient -0.06785 0.6377 -0.7262 0.103Aperture, inscribed dia. (mm) 160 274 212 274Good field region H (mm) 190 252 104 220Good field region V (mm) 140 186 148 176Core length (mm) 4400 609(592) 303 203Magnetic length (mm) 4400 730(715) 402 314Turns/pole 42 22 15 15Inductance/magnet (mH) 143 10 3 2Peak current (A) 1062 1062 1062 250RMS current (A) 720 720 720 15016Pre-approval design work.The design of the magnets was dominated by the already existing power supply. This consists (fig. 3) of a resonant choke having 10 secondaries, with the make-up power being provided by an impulse circuit (this was replaced by a rotary machine giving continuous make-up). The power supply and capacitors were donated by Daresbury Laboratory, who were replacing NINA with a synchrotron light source. The choke was ratedat 14.32 kV, 722 A rms, but had never been run to its full capacity.Calculations showed that the stored energy of the ISIS magnets at 800 MeV would be very close to the power supply limit, and in order to both gain experience and test the calculations two dipoles were constructed in the laboratory.The dipoles were roughly 1/3 rd in cross-sectional scale, 0.46 m in length and made from 0.5 mm silicon steel laminations glued together. They were placed in a White circuit (fig. 4), so that DC could be injected to produce a biased field. Instrumentation was modest, but despite this a measureable increase in make-up power was registered when the bias current was applied. This increase amounted to an additional hysteresis loss of about 75% compared with the theoretical loss with no bias.Looking for eddy-current effects (sextupole) in the guide field, no out-of-phase field component could be detected with search coils at the edge of the good field region. We saw the expected temperature rise atthe pole ends; these had a Rowgowsky rolloff which was approximated insteps rather than continuously. A neat method of measuring change in effective length against radius, at any field level but at 50 Hz, was successfully tried [3]. The DC field quality was in reasonable agreement with the GFUN [4] predictions.For a variety of reasons it was decided at this time that all magnetic measurements would be done under DC or quasi-DC conditions. Much of the period was spent in planning the quadrupole measurement system, which was to use a rotating harmonic coil, and in determining the basic parameters of all the magnets, such as conductor composition, number of turns, yoke sizes, etc.The choice of conductor lay between the indirectly cooled type, as used at CEA, DESY and Daresbury, or one which was directly cooled. The former is very economical in power consumption and requires only simple manifolding, but is difficult to make, has a very low average current density and the coil overhangs use up valuable straight-section space. With the latter, a higher current density is possible at the expense of power consumption and more extensive manifolding. Greater reliability in service was also expected from this type of coil.The fields in the coil slots were determined using BIM2D [5], and the eddy-losses calculated by hand (the results were confirmed some years later using PE2D [6], which was not available at the time). For all the series-connected magnets it was found that a cable consisting of four water-cooled conductors, insulated from one another but wound in17parallel, would result in acceptable losses and average temperature rises of about 20 deg C. Because of the harsh environment in which the coils were operating (high radiation, high voltage, possible vibration) the temperature rise was to be kept as low as possible to reduce mechanical stresses.The dipole coil, because of its 4.4 m length, required manifolding (and transposition of the conductors) for each of its six layers, one layer consisting of seven turns. For the doublet quadrupoles, transposition and manifolding occurred at each pole, but for the singlet the water path could extend over two poles and this led to a much simpler and neater design.The steel to be used for all the series-connected magnets was selected; this was British Steel Corporation's TRANSIL 315-35, a silicon transformer steel, 0.35 mm thick, with good permeability and an inorganic insulation coating.Doublet quadrupolesThe F and D quadrupoles were so similar in strengths that they were made identical (fig. 5) apart from a difference of 17 mm in core length. Made in four quadrants, the laminations were glued together and straps welded down the outside. The 10 mm-thick endplates did not extend over the poles, and through-bolts were used to guard against de-lamination. The peak field at the pole-tips being only 0.43 T, there was no roll-off at the pole-ends: any saturation effects would be corrected with the trim quadrupoles. To reduce eddy-heating at the ends, slits were cut into the poles of the end laminations.The pole profile consisted of a hyperbola with tangent shims, and was calculated using BIM2D and GFUN. The tolerances on gradient uniformity were not very severe because of the short (10 msec) acceleration time, and since the measurements (fig. 6) agreed with GFUN to better than 0.5% in integrated gradient, no changes were necessary to the profile. Because of our inexperience, however, after the prototype had been tested many changes were necessary to the manifolding and the supports.In order to reduce future radiation exposure to personnel, magnet replacement was to require the minimum of alignment; a master base and alignment fixture were built and all the quadrupoles adjusted on their supports so as to be interchangeable. These supports therefore had to be particularly robust and free from 'stiction'. The doublet and trim quadrupoles were mounted on a single concrete base and then moved as a whole into the synchrotron hall. In the event of a major fault occurring, the set of four would be removed and replaced with a spare set.The manufacturer (no longer trading) initially had difficulty in producing coils which were properly impregnated, the first coil being completely dry inside. This was traced to the fact that the resin in the mould was not adequately covering the coil tails; when air was admitted to the vacuum chamber it, rather than the resin, entered between the18insulation and the copper at the tails. Because of the 5 mm of insulation ground wrap, the resin was very slow to penetrate during the soaking period.The quadrupoles were all measured with the harmonic coil system, and showed very good uniformity in field quality and magnetic length (del L/L = 7 E-4). They were then tested for a few hours under power at 50 Hz but without the DC component of current. Initially the manifolding vibrated with very large amplitudes, but improved coil clamps wereinstalled and the vibration reduced to such an extent that it was justdetectable, by hand, with an insulating rod. Some tests were done with anaccelerometer, but these were difficult to interpret and in the end it came down to personal judgement as to what was an acceptable level of vibration.The hysteresis loss in the cores was calculated to be about 1 kW,and at each pole end there would be an extra loss due to eddy currents. Core cooling had not been thought necessary, and this was borne out by the tests.In the four years of operation so far, no major problems have arisen. Maintainence consists of an occasional tightening of coil clamps and visual inspection. On a few of the quadrupoles a black substance has oozed out to form pools under the quads. This has been analysed and found to be ethylene propylane, which is a high-hysteresis rubber with a melting point which decreases with oxidation. It is assumed that the manufacturer used this material for coil packing instead of the specified polyurethane. No corrective action has been taken, since the coils still appear to be well clamped.DipolesThe dipoles (fig. 7) were by far the most difficult magnets to construct and to measure. Being 36 deg magnets, curving them was essential, and since at the design stage it was thought that they would have to be split in order to introduce the ceramic vacuum chamber, they had to be made in two halves. Fanning of the laminations was also necessary, since the beam was to enter normal to the magnetic field. Finally the pole ends required a roll-off which kept both the field and gradient integrals correct across the aperture.The method adopted was to build each half-yoke out of six modules, which in turn were constructed from eight mini-modules. Tolerances had to be specified and held at each stage. The manufacturer tried to achieve the fanning by applying a tapered coating of epoxy glue to the laminations, but this was not a success. The method finally adopted was to punch dimples of graded depth across every sixth or seventh lamination, and to vacuum impregnate the minimodule with epoxy. The composition of the epoxy was not disclosed, but it was tested for radiation hardness. It must have been very fluid, however, since the magnet reference faces did not require cleaning-off afterwards. Penetration and adhesion were satisfactory.19The mini-modules were built into modules about 0.7 m long, using cold-curing resin. 'Dog-bones' were inserted into dovetails which had been punched into the laminations, and the modules attached to one- another by means of these and more resin. Bolts from upper to lower dog- bones were used to clamp the two halves of the magnet together. Again the end-plates did not extend over the poles: de-lamination of the pole-ends is prevented only by the adhesive.The estimated core losses were 10 kW, and therefore surface cooling pipes were let into pre-punched slots in the laminations. The dipole was mounted on a fabricated steel frame, and again the principle of replaceability was kept, although with such a weight (about 34 tonnes) the design of the supports had to be modified slightly to overcome stiction.The coils presented no particular manufacturing difficulties, despite their bulk and the thick ground wrap. They are held in the coil slot by clamps as shown in figure 8. Pieces of epoxy-glass laminate span the coil slot and grub screws bear down on the coil. Radiation tests showed that throughout the magnets' lifetime creep or permanent set in the laminates should not be a problem, but this solution to an awkward problem is not felt to be completely satisfactory. In fact, additional clamping of the coil overhangs was found necessary after power tests.The prototype was measured completely using short search coils, and the field quality in the body of the magnet found to be very good. The end fields were not quite right and modification of the roll-off profiles was necessary. Because the contract was running late, the ten production dipoles were only checked dimensionally and compared with the prototype for the field integral along R = 0 by means of a long, curved search coil. The rms variation in this integral was about 3 E-4, which was considered to be very satisfactory.Power tests again consisted of some hours soaking at 50 Hz with no DC current. Because all the coil tails were accessible, it was possible to vary the way in which the transpositions were made from layer to layer, and in fact a small improvement in the losses was made by making alterations to those deepest in the coil slot. The inductance was measured to be about 14% greater than expected, which is a large and so far unexplained discrepancy. Unfortunately it swallows up all the voltage safety margin for the choke.The extra clamping of the coils has already been mentioned. With this in place the vibration levels are very low, with virtually nothing detectable by hand on the yoke itself (but see below). If one had the chance of designing these dipoles again, three changes would probably be made:a) because the vacuum section were able to produce a ceramic chamber which could be inserted without splitting the magnet, every effort would be made to have a one-piece lamination;b) the coil clamps would be made more robust;c) conductors would be removed from the region of highest fringe field near the pole edges. Although the average temperature rise in the coil20is modest, the inside turns get very much hotter than the rest.Singlet quadrupoleThe singlet quadrupole has an asymmetric yoke because the one in superperiod 1 lies directly under the extraction septum magnet, and so was reduced in height to provide clearance. The design was such that the magnet was made in two halves, with a coil just able to slip onto a pole without fouling its neighbour. The manifolding was much simpler here since only two water circuits were required.Because the aperture was 212 mm and the core length only 303 mm, end effects were very significant. The dimensions of one of the ISR Terwilliger quadrupoles [7] were scaled to our aperture, and a small octupole component (to provide some Landau damping to the circulating beam) included in the profile. This all worked very well, with the length variation being measured as 9 E-4, and the octupole value being as predicted (fig 9). The coil clamps consist of epoxy-glass laminate shaped so as to slip between the coil-end and the yoke. They are pulled up by means of bolts. Although neat and simple, these clamps always need some tightening and may possibly need replacing in the future.Trim quadrupolesThese quadrupoles have three purposes: to vary the tunes during theinjection period of 450 jisec so that while the field is falling the constant-energy incoming beam has the same Q-values; to compensate for any magnetic saturation effects; and to fine-tune the machine so that maximum intensity might be accelerated. Each has its own bipolar programmeable power supply, but to date all the F-quads and all the D- quads have been fed with identical functions.The quadrupole diameters (274 mm) are greater than the yoke lengths (203 mm) and again the profile has been adapted from one of the ISR correction quads. Although the magnetic pass-band required is considerably greater than 50 Hz, the standard 0.35 mm-thick laminations were certainly good enough. With a peak current of only 250 A, a single conductor with just one water circuit was sufficient. Ceramic insulators have been used here, and the magnets have been trouble-free in operation.The custom-built power supplies have been very reliable, though subject to occasional water leaks. ISIS can operate at 100 fiA with an F- or a D-quad off; if one set goes off, its partner in the superperiod is usually switched off also.Correction dipolesAlthough spaces exist in the synchrotron for complete sets of horizontal and vertical correction magnets, lack of funds has meant that only 7 H and 4 V dipoles have been installed. They are powered by the same type of supply that feeds the trim quads. The vertical dipoles are positioned so as to create an orbit bump in the region of the extraction septum magnet, in order to reduce the demands on the fast kicker magnets.21The horizontal dipoles are used primarily to correct the closed orbit in the region of the injection straight.The dipoles (100 mm long) consist of dry-stacked laminations clamped with stainless steel end-plates and throughbolts. The air-cooled coils surround the return legs, which leads to large fringe fields. As a result, the mild steel supports became unexpectedly hot when the magnets were tested at full power, and were replaced by stainless steel.Skew quadrupolesSuperperiod 5 contains two skew quads to remove any excessive coupling in the betatron motion between the two planes. They are just two trim quads rotated through 45 deg.; their power requirements are so modest that two hi-fi amplifiers provide the drive current, governed by the standard function generators. They are not used in normal operation, but may be required at higher beam intensities.Operating experienceThe first turn round the machine was easily achieved, and subsequent measurements showed that without correction, the H and V closed orbit errors were a few millimeters rms and the tunes within 0.05 - 0.10 of the design values (4.31 and 3.83 in H and V). It must be pointed out that the surveying of the whole complex was excellently done, since the beam inthe complicated injection and extraction lines can be aligned for themost part to 1% of the quadrupole diameters.Problems with the ring magnets have been confined to coil clamps andpacking (already mentioned) and the occasional water trip, caused by unreliable water-flow monitors. The rotary make-up power supply for some time suffered from a fault which was eventually traced to the shaftencoder coupling, and recently the sliprings were damaged as a result ofan incorrect set of brushes having been fitted.A 10 MHz crystal oscillator controls the frequency of both the AC power supply and the neutron choppers. In normal operation beating with the 50 Hz mains does not cause problems, although it is suspected that occasional timing jitter of the fast extraction kickers could be due to the fact that the thyratron heaters are fed from the mains.The ceramic vacuum chambers are supposedly clear of all contact with magnet poles; this clearance is routinely tested, and the chambersexamined for excessive vibration. Increased vibration of one of thedipoles and its vessel is currently being investigated: this may beassociated with a large amplitude of vibration discovered on aneighbouring pillar which supports correction magnets.AcknowledgementsThanks are due to the following, all of whom made majorcontributions to this work: A Armstrong, R Bennett, R T Elliott, W REvans, H J Jones, J Lidbury, A Slater, A Wardle.22References[1] Ed. B Boardman. 'Spallation Neutron Source: Description ofAccelerator and Target.' Rutherford Lab Report RL-82-006, March 1982[2] R T Elliott, J A Lidbury, M R Harold. 'Synchrotron Magnets for the SNS.' IEEE Trans, on Nuc. Sci., Vol. NS-26, No. 3, June 1979.[3] M Awschalom et al. 'Measurement and shaping of the fringing fields of the Princeton-Pennsylvania accelerator magnets.' IEEE Trans, on Nuc. Sci., Vol NS-12, No. 3[4] A Armstrong et el. 'New developments in the magnet design program GFUN.' Rutherford Lab Report RL-75-066, 1975[5] M J Newman, C P Riley. 'BIM2D User Guide.' Rutherford Lab Report RL-79-088, 1979[6] C S Biddlecombe et el. 'PE2D User Guide.' Rutherford Lab Report RL-81-089, 1983[7 ] R Perin. 'End effects compensation in the auxiliary magnets of the CERN intersecting storage rings.' Proc. Int. Conf. on Magnet Technology, Hamburg, 1970jjSR U 238Figure 1 The ISIS facility23DIPOLE DOUBLET QUADS SINGLET QUADTRIM QUADSFigure 2 An ISIS superperiod24E O T  10(4± s _ s  ; ;j3H.fl£L-B-QjjALLI>Figure 6 The doublet quadrupole measured gradient integralFigure 7 The dipole25Figure 8 The dipole coil clampFigure 9 The singlet quadrupole measured gradient integral26DIPOLEAC P/5/BACK LEG WINDINGSDIPOLEFigure 4 The White circuitFigure 5 The doublet quadrupoleDESIGN AND MEASUREMENTS OF A .C . -  MAGNETS AT DESY I I27G. HemmieDeutsches Elektronensynchrotron DESY Notkestrasse 85, D--2000 Hamburg 52, W, GermanyABSTRACTThis paper summarizes the content of a talk given at the MAGNET DESIGN WORKSHOP at TRIUMPF on October 3-5, 1988. After a short introduction to the DESY 2 Synchrotron lattice and magnet parameters the choice of steel will be discussed. The influence of eddy current effects in the core and endfield design will be described. Some fabrication and assembly details are given. Finally the technique of the AC- measurements will be explained and a few results will be given.INTRODUCTIONThe DESY I Synchrotron which was built in 1957-19652 was designed to accelerate electrons from 4o MeU/c to 7.5 GeU/c at 50 Hz repetition rate. Oing to high intensity operation the coils werde damaged by radiation. Furthermore the 50 Hz operation caused strong vibration of the ceramic chamber and this often lead to vacuum breakdown. In order to reduce down-time and to save costs for repairs the new DESY II synchrotron wasconstructed [11,12]. Whereas the old DESY I machine was directly servinghigh energy physics experiments at 50 Hz repetition frequency the maintask for the DESY II synchrotron was changed to act as an injector syn­chrotron for the storage rings PETRA and DORIS. Therefore the repetition rate could be reduced to 12.5 Hz and the usage of all metal vacuum chambers became possible [31.The main characteristics of the magnets are that they run at 12.5 Hz cycling frequency with a momentum swing from 55 MeU/c to 10 Gev/c. But since the existing beamlines to PETRA can currently only handle 7 GeU/c particles the installed rf-power is only sufficient to accelerate to about 8 GeU/c. In the meantime the electron energy of LINAC I has been pushedup to 200 MeU to get to better injection efficiency.LA TT IC E  AND MAGNET PARAMETERSIn order that the transverse and longitudinal particle motion bedamped a combined function lattice was chosen [41. Extra sextupoles are included in the lattice to compensate the chromaticity. This is partly caused by eddy currents in the metal vacuum chamber [53. The aperture re­quirements are determined by the beam emittance at injection <10 mrad*mm,dp/p = 1 % at 55 MeU/c). This .leads to a good field region requirement of b * H = 80 mm * 40 mm, where dB/B does not exeed ± 5*10A~4, The final de­sign of the DESY II magnets is a result of many special constraints as for example:1. The prior existence of stamping tools (for quadrupoles and sextupoles) and the existence of the White circuit for the dipole powersupply which could easily be changed from 50 Hz resonant frequency to 12.5 Hz.2. The availability of steel and3. the annual operation time of the synchrotron which affects the ratio of investment costs to the power consumption costs.28L IS T  OF PARAMETERS FOR MAGNET DESIGNRepetition frequency 12.5 HzMax. energy <for magnet design) 10 GeV1st completion stage 8 GeVLattice: sep. function 8 superperiodsS BM + 3 QF + 3 QD + 2 SD + 1 SFfocussing strength: KF = .365 iV'-Zi KD = -.328 m A-Z Acceptance (dp/p = ± 1 %> 10 mrad*mmMagnets (ratings refer to 10 GeV/c excitation)48 dipoles length 3.55 mgap 160 * 45 mm ''2bending radius 27. 12 mmax. field in gap 1 .23 Tpeak current 1 147 Ainductance 33.2 mHpower losses 30 KWgood field region (dB/B <=5*10'-4) ±40 mm48 quadrupoles core lenght 0.58 mgap radius 50 mmmax field 14.7 T/mpower losses 6 KW24 sextupoles core length 0.18 mmax. field 77.8 T/m"2power losses 1 . 1 KWMomentum vs. time waveformP [MeV/c 1 = 3587.5 - 3412.5 * cos<25>« t)Pmin=175 MeV/c; Pmax=?000 MeV/c; dE=30 KeV/turn (200 MeV/c Inject.)THE CHOICE OF STEELIn order to avoid severe eddy current effects in the magnet core it is clear that laminated iron had to be used. But what about the thickness of laminations? Various factors determine the choice:1. Eddy current losses in the core that causes heating.2. Magnetic field distortions that might cause acceptance re­duction and/or tracking problems between dipole and qua- quadrupole field excitation.3. Saturation effects.4. The availability of steel on the commercial market and the cos15 for stamping and machining.It is well known that the magnetic flux within the individual lamina­tion is pushed out to the surface and thus that the flux density on the surface is increased above the average flux level 161. Consequently a part of the high magnetic field inside the laminations runs into saturation if the thickness of the lamination is not properly chosen.The ratio of the AC- component of the peak flux density Bs on the surface of the lamination to the average flux density Bav is given by:J  /2>■ ewith |3 = ^ / 4  - arctg (sinx/Sinx) , x = d/D and D = skin depth = l/fifFT/TFor the characteristics of the DESY II ironconductivity = 3* 1 0 ^B m/ m A3repetition frequency f ~ 12.5 Harel. permeability /iy'= 3000we calculate the following numbers:d tmml .5 1.0 1.5 2.0 2.5 3.0 3.5 4.0Bs/Bav 1.0002 1.004 1.019 1.059 1.137 1.2B5 1.443 1.BBGfb [*] 1.0G 4.23 9.39 1G.1 23.7 30.9 3 G . 9 41.3From this we estimated the thickness of lamination d = 1mm and the maximal flux density in the dipole gap to be about Bo = 1.23 T.The cutoff frequency fg given by x = 1 (skin depth D = thickness d of la­mination) characterizes the uniformity of the magnetic flux and should be chosen to be well above the operation frequency.For DESY II iron we find fg = 28 Hz.The specific eddy current losses are given by [6],[71:N ’ = i - . a t .  w 2 -d*Bm2 U F ( x )2-h~Z y — Sxv. Xwith F(x) = —  --------------- and Bm = .42 T (effekt. AC magnetization)we find N ’ = 140 W/mA329This number is rather small and it is obvious that no water cooling is needed for the core. Since in rapid cycling machines the ratio of eddy current power loss to eddy current field scales proportiona1ly to the de­rivative of the magnetic field dB/dt the heating may dominate over the ef­fect of field distortion.Since we wanted to operate the magnets over the full excitation range between G2 Gauss and 1.23 Tesla corresponding to a 55 Mev/c to 10 GeV/c momentum range the following specification on flux density B vs. magneti­zation force H was required:H (A/m) 20 30 50 100 500 1000 5000 30000B (T) .06 .13 .34 .82 1.36 1.34 1.63 1.995y  X  +  C o S  XBs/Bav = ~ 1rz COS & y -  COS y(Calculation performed with magnetostatic computer code MAGNET 80 [81)30It is obvious that in order to achieve good field quality at lowexcitation level the remanent field Br caused by the coercive force He should be as small as possible. At He = 40 A/m we expected about 5 to 7 Gauss remanent field within the dipole gap. We had measured Br= 5.5 to B.5 Gauss after full DC excitation (1300 A *> 1.4 T). There is a slighttransverse gradient in the remanent field caused by the various length of the fluxlines in the C-type dipole. Low coercive force implies that the iron should be carbon-free (<100ppm). Pure iron heated in a hydrogen atmosphere has Hc= 4 A/m and the relative permeabitity is quite large over the full range of magnetization. However the electrical conductivity is comparativi1y large ( X.= 10A7 m/jvmA2) so that the thickness of thelamination has to be small in order to keep the eddy current effects low enough. Since this material is rather expensive and is difficult to obtain the more economic way is to use iron containing a few percent of silicon ( 1 ... 4 %). The si 1 icon content helps precipitaion of carbon into graphite and thus the aging process, increases the electrical resistance and the permeability at low flux densities but reduces the saturation level ofinduction by about 500 Gauss per percent of silicon. The steel was pre­pared under careful supervision.Only a few firms were able to offer steel according to our specifica­tion. The steel that was finally delivered had the following guaranteed propert ies:Type of material 300_.100 coercive force Heconductivityrel. permeability (max) /**• saturation induction Bslamination thickness dinsulating phosphatecoating (Stabolit 50)(EBG Bochum > 40 ± 8 A/m 2 . 3  * 1 0 " 6  m/n.m' B550 1 .995 T 1 mmChemical analysis [%]2y-.CSiMnPA1N.003 . 1 Magnetization curve (measured)MAGNET CORE AND END F IE L D  DESIGNFor ease of access to the vacuum chamber, a C- type cross section for the dipole was chosen. The metallic vacuum chamber allows the vertical beam profile to use nearly the complete gap height. Consequently the pole width could be made small and therefore the stored energy in the gap is minimal. The installation costs as well as the costs for powering the mag­nets became small too. The dipoles were of completely new in-house design. Fabrication and assembly were made in close cooperation between various firms and the DESY workshop.31The laminations were shuffled in order to equalise the remanenteffects. They were stacked on a segment of circle in order to avoid thesagitta. The 30 mm end field blocks were glued together and the completestacks were surrounded by strengthening plates and welded together. The end fields of the dipoles and quadrupoles were carefully shaped in order to avoid eddy current in these core regions.Because DESY II is mainly an injector synchrotron the total operation time is only about 25 % of that of the storage rings which are to befilled with par tides. Therefore the current density of the ex citation coils could be made comparatively high (5.5 A/mmA2>. This allowed us to use flat pancake coils for the dipoles with 20 turns of copper 16.5 x 9 mm''2 each with 5.1 mm cooling hole. The eddy current loss contribution has been estimated to be less than 15 % of the RMS power losses. Some di­poles are equiped with backleg windings for high energy orbit bumping and horizontal orbit correction at injection.The design and construction of the DESY II quadrupoles and sextupoles are similar to those of the PETRA storage ring. In order to save on tools and construction the cross sections are identical. But some modifications have been made in order to get a better fit to the DESY II parameters. In particular copper coils instead of aluminum and a shorter core length have been chosen. The core length of the quadrupoles is determined by the aim to match the saturation characteristics of dipoles and quadrupoles. Thisallows us to keep the tune of the machine within the given limits during acceleration without having additional control on the excitation current.MECHANICAL AND MAGNETIC MEASUREMENTSThe permissible tolerances on the magnets have been discussed in [41. During steel fabrication tests on chemical composition, electrical re­sistance, solidity, flatness, inner tension, filling factor and insulation were performed. The geometric dimensions were checked after stamping and stacking. Magnetic measurements with various probes on coercive force and remanent induction were made. For fabrication tests two dipoles were stacked using spare PETRA iron which was in stock. A prototype dipole with new steel and equiped with removable end field pieces then followed in order to check various end field shapes. The inner field was measured with a Hall probe under DC excitation.AC MEASUREMENT METHODIn order to measure the magnetic properties of the DESY II dipoles and quadrupoles under realistic AC excitation a new method of field measure­ment had been developed [91. Movable coils of various length were usedand the AC induced voltage was compensated by a reference coil positi­oned in the same magnet or in an extra magnet powered in series:1. A short coil ( 100 x 10 mmA2 , driven by a crank shaft ) for end field measurements on dipoles with a reference coil positioned at the center of the sane magnet and2. A long coil (3600 x 10 mm''2 , also driven by a crank shaft) for integralmeasurements on dipoles with a reference coil in an extra dipole po­wered in series (determination of relative magnetic length of dipoles).3. Two simultaneously rotating coils for the quadrupoles mounted in the same plane at different radius with different numbers of turns for pro­per compensation.The output of the coils were preamplified with high common mode re­jection, then integrated using chopper stabilized amplifiers ( ICL 7650, Intersil) and digitized via high resolution sample & hold with analog - digital conversion (MP260 and MP8016, Analogic).During the ramp, up to 11 triggers in sequence were derived from back- leg coils on the reference magnet. These triggers start data processing at 11 different excitation levels. Thus within one turn (about 2 sec.) of the crank shaft (dipoles) or rotation (quadrupoles) one gets a complete field mapping of the whole range of magnetization. The small integrator drift which was assumed to be constant during measurement was compensated nume­rically. Polynomial fits (dipoles) or Fourier analysis (quadrupoles) fi­nally gave the coefficients of the harmonics of field errors.32MEASUREMENT RESULTSIt is obvious that AC measurements are much more difficult to perform and the results normally are not as precise as those obtained with DC. But our measurements show that the DESY II magnets satisfy the field quality requirements over the whole energy range. We did find no indications for dangerous eddy current effects. Figs. 2,3 show typical computer results from dipole- and quadrupole measurements.t y p ic a l  v a lu e  lo n g  c o il I c m  . 3 6 0 cmFig.2 Integral dipole field at 4 di f ferent excitations(d e g re e )Fig.3 Typical normalized quadru­pole dataREFERENCES[1] G. Hemmie, DESY II, A new injector for the storage rings PETRA and DORIS , IEEE NS-30 , (1983 )[21 G. Hemmie, Status Report on the DESY II Synchrotron, IEEE NS~32,(1985) [31 J. Koupsidis, R. Banthau, H. Hartwig, A novel fabrication technique for thin metallic vacuum chambers ..., IEEE NS-32 , (1985 )[41 J. Ro|3bach, F. Willeke, DESY II Optical design of a new 10 GeU elec­tron positron synchrotron, DESY M-83-03 [51 G. Hemmie, J. Ro(3bach, Eddy current effects in the DESY II dipole va­cuum chamber , DESY M-84-05 [61 For example: K. Kupfmuller, Theoret-ische Elektrotechnik , p. 262 - 265, Springer, Gottingen 1955 [71 G. Hemmie, Design, construction and performance of the DESY II magnets, IEEE NS-32, (1985)[81 C. Iselin, CERN Prog. Lib. T600, (1971)[9] U. Berghaus , W. Kriens , S. Paetzold, Magnetic field measurements ofthe DESY II magnets, IEEE NS-32, (1985),33A DC BIASED RAPID-CYCLING MAGNET SYSTEM OPERATING IN A DUAL FREQUENCY MODEH. Sasaki, T. Adachi, H. Someya and I. Sakai National Laboratory for High Energy Physics,Oho 1-1, Tsukuba-shi, Ibaraki-ken, Japan, 305ABSTRACTA prototype rapid-cycling magnet and power supply of a high intensity proton synchrotron for condensed matter research have been constructed and tested in DC biased and dual-frequency mode at practical operation level. This report describes the design and features of the DC biasing device and the performance of the system.1.INTRODUCTIONA rapid-cycling magnet and its exciting system are under development for a synchrotron as an intense pulsed spallation-neutron and muon source for condensed matter research at K EK .l) The development of a rapid-cycling magnet system operating in a dual frequency mode is very important and indispensable to the realization of an intense rapid-cycling proton synchrotron for the reduction of load of the RF accelerating system.A prototype rapid-cycling magnet system was constructed and successfully operated at a practical level in a dual frequency mode without any DC biasing field. The performance of this system was reported in detail in the international workshop on Hadron Facility Technology held at Santa Fe, February last year.2 ) After that, further development is still in progress. Namely, in order to bring the prototype magnet system into more realistic operational condition, a choke transformer, which provides the pass of a DC current biasing the magnet, has been fabricated and introduced into the resonant network. This report will describe the design and performance of the choke transformer and the operation of the prototype magnet system with a DC biasing field in a dual frequency mode.2. MAGNET, PULSE POWER SUPPLY AND RESONANT NETWORKDetails of the prototype magnet system which includes the resonant network, GTO thyristor switching system for dual-frequency-mode operation and pulse power supply, were already reported in the reference of 2). For the sake of convenience, however, the main features of those devices are summarized here. Parameters of the system are listed in Table I.magnetConfiguration of the magnet is shown in Fig. 1. In order to reduce the eddy currents circulating within laminations of core around magnet end, 30mm-spacing slits are introduced at the magnet end over the pole width parallel to beam orbit. Those slits are extremely effective to suppress the eddy current heating of the magnet end without reducing the effective magnet length. A special stranded cable was developed as a new conductor material of rapid-cycling magnet coil, whose cross section is 30mm x 30mm including a 14mm outer diameter and 1.5mm thick copper pipe. The cable consists of 84 aluminum wires of 3mm in diameter, carrying 1.65 kA DC and 0.88 kA peak AC current. Any special process was not applied on the surface of the aluminum wires to insulate each other. Such a stranded cable is considerably effective to the reduction of eddy-current loss. However, there exists still plenty room for improvement.Resonant network and pulse power supply34Table I. Parameters of prototype magnet and its AC excitation systemMagnetmaximum field DC bias field amplitude of AC field gap height pole width magnet length core material number o f turns per pole exciting current conductor of coilweight o f iron core weight o f coil conductorPulse power supplyDC power supply for pulse excitation charging reactor discharging reactorenergy storage capacitorResonant networkinductance o f magnetmax. AC current max. applied voltage resonant capacitor C j and C2 switching element max. switching current resonant frequency0.697 T 0.455 T 0.243 T 164 mm 540 mm 1,800 mm0.5 mm thick silicon steel (non-oriented) 181.65 kA DC and 0.88 kA peak AC30 x 30mm Al stranded cable with 14 mmdia. cooling copper pipe12 tons0.3 tons1.1 kV/25A 1.73 H 12.0 mH93.4 pF2.95 mH (top and bottom coil connected in parallel)1.76 kA peak3.3 kV peak0.859 mF and 6.87 mFdiode, GTO thyristor and SCR thyristor1.56 kA peak100/3 Hz and 100HzA system of pulse power supply and resonant network was constructed for exciting the prototype magnet in AC mode. The special interest was in the operation of this system in a dual frequency mode of 100/3 Hz in acceleration period and 100Hz in reset period by using GTO thyristor switch on a practical operation level. In the early development stage lacking in choke transformer as shown in Fig. 2a, the top and bottom coil of the magnet were connected in parallel, and the timing of the energy transfer from the energy-storage capacitor Cf to the resonant network was shifted from the instant of the magnet AC current zero in the reset period of 100Hz to that in the acceleration period of 100/3 Hz. By this procedure, the output voltage of the rectifier of the pulse power supply was set at the economical voltage level of about lkV. The resonant network system was successfully excited in a dual frequency mode. The GTO thrystor was proven to be quite suitable for a high power switching element of resonant network operating in dual frequency mode in combination with diodes and SCR thrystors.3. CHOKE TRANSFORMERDesign of choke transformerResonant network exciting ring magnets of a large-scale rapid-cycling synchrotron will consist of many identical meshes containing some of ring magnets, choke and resonant capacitors. In such a case, it is difficult in its scale and complexity to fabricate such a common large choke that the mutual coupling among chokes allocated in every mesh is expected to be as close as to unity, while it was realized in the choke in the 3-meshes resonant network of the KEK booster synchrotron magnet system.3) Therefore, the chokes allocated in meshes were designed so as to be completely separated in magnetic coupling from each other.35The inductance o f the choke should be determined from the economical viewpoint. Assuming the cost of the choke and capacitor proportional to th e ir sto red  energy , and denoting a capacitor-to-choke cost ratio per unit stored energy by r, the inductance ratio t| o f the choke to the m agnet inductance giving the optimum cost is determined byIac r r  = I d ^ + rThe ratio r is estimated to be 0.6 to 1. This leads to the fact that the inductance of the choke is0.7 tim es as much as the inductance of the prototype bending magnet only for the purpose of the magnet test, and twice for a prototype of the choke to be set in each mesh of the resonant network o f the designed synchrotron. For the sake of simplicity, finally, the inductance of the choke has been determined to be equal to that of the prototype magnet.As for the type of choke, an iron-clad air-core choke transformer with primary and s e c o n d a r y  w i n d i n g s  surrounding a common magnetic a ir gap, w as chosen in preference to alternative air-core or multi-air-gap iron-core choke.a)Prototype magnet b)0:30Stranded CableDC POWER SUPPLY fo r  PULSE EXCITATIONFig. 2 Resonant network and its excitation system36The main reason for this choice is simplicity of the mechanical structure, well-known magnetic field distribution, low cost and experiences on the construction and operation. The dissipated AC power in the resonant network is supplied from a energy-storage capacitor Cf through the primary winding of the choke transformer as shown in Fig. 2b. The configuration of the choke transformer is shown in Fig. 3.As the winding of such a type of choke is exposed to an appreciable magnetic field, it should be made of stranded or transposed cable to cope with AC power dissipation due to eddy current induced in the conductor. However, application of usual stranded or transposed cable requires presumably a large oil tank housing the choke for forced cooling. This leads to a large size of the system. Fortunately, we can apply a stranded cable with water-cooling pipe to the coil of choke, which have been developed for a magnet coil material of rapid-cycling magnet, and as a result we can simplify and keep the system in a small size. Two kinds of stranded cables were prepared for the material of the secondary windings of the choke to make a comparison on their electric and mechanical properties. One of them is the same one with the cable used in the prototype magnet, which consists of aluminum wires without any insulation process on their surface as already described. Another one is a newly developed copper stranded cable with a copper cooling pipe of 15mm in diameter and 1.5mm in thickness. The latter is in the form of a 25mm x 25mm square including 58 copper wires of 2.6mm in diameter, each of which is processed on the surface by coating ester-imide film for insulation.The secondary winding has to carry the same amount of current with that of the magnet, namely, I^c = 1.65kA and Iac = 0.88 kA peak because of Lch = Lm.On the other hand, the average current in the primary winding fed from the pulse power supply is very low compared with the secondary AC current as given by iav = n Iac/Q, where n is the step-up ratio of the choke transformer and Q the quality factor of resonant network. Therefore, any special device is not necessary to be prepared for cooling. In order to realize a sufficiently high coupling coefficient between the prim ary and secondary winding, the primary winding is in the form of strip and wound close together to the secondary winding as shown in Fig.4. The step-up ratio is adjusted by connecting the primary windings at the output terminals of pancake in series or parallel. The total number of turns of the secondary winding is 84 turns (7 turns x 12 layers), which is divided into identical top and bottom coil by an air gap for field-monitoring space at median plane.Grain-oriented steel is used for the core material. Application of such a material is useful to reduce the weitht of the system.Design parameters of the choke transformer and DC bias power supply are given in Table II.correspond to those of the aluminum cableFig. 4 Coil of choke transformer37Measurements of inductance. AC power dissipation and magnetic fieldThe measurements of the inductance and AC power dissipation of the secondary winding were carried out by applying directly a 50 Hz sinusoidal voltage to the secondary winding, which gives a nominal AC current of 0.88 kA peak. Results were as follows: with withaluminum stranded cable copper stranded cableinductance 11.76 mH 12.11 mHAC power loss 67.0 kW 27.5 kWIn addition to those, the coupling coefficient between the primary and secondary winding was measured, which showed a very high coupling coefficient of 99.58%. The inductance of the secondary winding of the transformer with the copper cable is somewhat higher than the design value of 11.8 mH, while the design of the choke transformer was made on a simple case such as one without air gap separating the top and bottom coil. A remarkable difference between the transformer with the secondary winding made of aluminum and copper stranded cable is in the AC power loss. The dissipated AC power in the former amounts to by 2.5 times of the dissipation in the latter. Of course, this comes of the difference in the insulation process of wires, that is, the former without any insulation process on aluminum wires and the latter processed by coating insulation film on the surface of copper wires. In order to clear the sources of the AC power dissipation, ohmic, eddy current and iron loss were estimated on the latter case, which is in a definite condition in viewpoint of insulation. Then,Ohmic loss in conductor 5.00 kWiron loss in yoke 0.35eddy current loss in cooling pipe 16.12eddy current loss in copper wires_________8.19sum 29.66 kWTable II. Design parameters of choke transformer and DC bias power supplyInductance of secondary winding Max. current in secondary windingIron coretypecore materialthickness o f lamination dimensions o f core width of yoke weight o f iron11.8 mH1.65 kA DC and 0.81 kA peak ACCoilnum ber o f turnsconductor material primary secondarydimensions of coilgap between top and bottom coil step-up ratio weight o f coilcooling of secondary winding number o f circuit pressure drop/circuit total water flowiron-clad air-core typegrain-oriented silicon steel, Nippon SteelCorp. Z9H0.3mm648 mmH x 1,460 mmW  x 1,200 mtnL 125 mm 4.35 tons7 turns x 6 layer x 2 for both o f primary and secondary windingcopper strip stranded aluminum wireOR 595 xIR H339 x 192 (mm3)50 mm1 : 6 or 1 : 3 0.41 tons25kg/cm2 21 1 /minstranded copper wire OR IR H 575 x 351 x 162 (trim3)110 mm1 : 3 or 1 : 2 1.08 tons25kg/cm2 28 1/minMagnetic field at center 0.401 Tesla DC and 0.214 Tesla peak ACDC bias power supply bypass capacitor2kA/50V2F/100V38Thus, the total estimated AC loss amounts to about 30kW, which is consistent with the measured one within a error of 10%. Of those sources the eddy current loss in the copper cooling pipe is rather remarkable than that in the copper wires of stranded cable. The same calculation on the case of the tran sfo rm er w ith alum inum  stranded cable was done except for the eddy current loss in the alum inum  w ires, w hich is indefin ite  in the insu lation  condition. This gives the eddy current loss o f 49kW  in the aluminum wires in comparison with the m easured loss. By making the insulation complete with an appropriate insulating film processed on the surface, such an amount of loss is cut down by a factor of several as seen from the case of stranded cable with insulated constituent wires.We are now in around the goal of the developm ent of a conductor material for rapid-cycling magnet or choke. If there still exists a room for improvement, it is to replace the copper cooling pipe with a pipe made of stainless steel. By this, the eddy current loss in the pipe will be reduced by a factor of 30, i.e., the conductivity ratio of copper to stainless steel.DC biased magnetic fields of the choke were measured by using a set of search coil and voltage-to-frequency converter. The results are shown in Fig. 5. The curve a) is the distribution of AC component of DC biased field in a dual frequency mode. For the sake of comparison, the well-known field distribution in an ideal case, which has no gap separating the top and bottom coil, is indicated by a dotted line. The curve b) shows the distribution of the field induced by eddy current circulating in the coil conductor in a DC biased single resonant frequency mode of 23.5 Hz, Idc=1.65 kA and Iac=0.35 kA peak. The peaking-strip method, which was used for the measurement of the field induced by eddy current in the previously reported case with no biasing field, can not be applied to the eddy-current field measurement because of saturation of the peaking strip due to a high DC biasing field Therefore, the measurement was done by shifting the integration gate timing of the search coil signal by rc/2 radian from that of inphase-field measurement.^)Fig. 5 AC component of DC biased field a), and eddy- current induced field b).4. OPERATION OF THE DC BIASED MAGNET SYSTEM IN DUAL FREQUENCY MODEIn order to excite the magnet system in an AC level as high as possible with a limited capacity of the pulse power supply and for saving the expense for conversion, the connection of the top and bottom magnet coil has been restored from parallel to series and the resonant capacitors have remained with no change in their capacitance. An only change of the existing equipment is the increase of capacitance of the energy storage capacitor to 176mF in consideration of a mismatch to dissipated power in the resonant network in the previous AC operation and of addition of a new source of power dissipation, i.e., the choke transformer.39Thus, the resonant frequency are lowered by a factor of I/V2 from those of the previous operation. With such a dual frequency of 23.6 and 70.7 Hz, it is possible to operate the system at around the nominal excitation level within the power limitation of the pulse power supply.The operation of the system showed that there was essentially no change in the behavior of the system for the introduction of DC biasing current into the resonant network. Some parameters were slightly different from those of designed one, e.g., a dual frequency of 23.5 Hz and 66.7 Hz. Fig. 6 shows the voltage and current waveforms at various points of the system. A rapid oscillation of capacitor current is also observed at the instant opening the circuit for C2 current through the diode as like as in the previous operation without DC biasing device. The time corresponds to the injection time in the accelerator. Such an oscillation originates from a resonance of Ci and C2 with a stray inductance in a loop including thoseb) to ta l  c a p a c ito r  c u rre n tCi c u rre n t 8 3 3 A /d ivC2 c u rre n t 8 3 3 A /d iva) magnet te rm ina l v o lta g e  2 .5 k V /d iv  magnet c u r re n t 5 0 0 A /d ivI  C T O  —  0c) to ta l  c a p a c ito r  c u rre n t 8 3 3 A /d iv  diode c u rre n t 8 3 3 A /d ivGTO-SCRg c u rre n t 8 3 3 A /d ivFig. 6 Voltage and current waveform in the resonant networkd) Cf v o ltag e  l k V / d i vmagnet terminal  vo l t age  2 . 5 k V / d i v40capacitors. Even though the oscillation amplitude is high, it does not affect on the magnet current because of very high inductance of magnet. In fact, we can not find any indication of oscillation in the magnet current while a small oscillation takes place in the total capacitor current as seen from Fig. 6.5. CONCLUSIONSIn order to operate the prototype rapid-cycling magnet system in a more realistic operational condition, a choke transformer for DC biasing was designed and constructed. An iron-clad air-core choke transformer, whose primary and secondary windings surround a common magnetic air gap, was adopted as to the type of the choke by reason of simplicity of mechanical structure, well-known magnetic field distribution, and so on. An only drawback of this type of choke, which operates in rapid cycling, is the coils exposed to higher magnetic field compared with the alternatives, e.g., multi-air-gap iron-core choke. From the viewpoint of the AC power loss due to eddy current, therefore, it should be more careful in the design of coil structure and the selection of conductor material in comparison with the coil of rapid-cycling magnet because of higher imposed field upon the coil. The choke transformer, whose exciting coil is made of copper stranded cable processed for insulation of constituent wires, showed satisfactory features, especially, in AC power loss in contrast to the one with aluminum stranded cable without any insulation process. Although such an aluminum cable is effective to reduce eddy current loss, it is considerably difficult to estimate the power loss due to eddy current induced in the cable in an indefinite insulation condition of the surface of constituent wires. Through the tests carried out so far, we have come to the conclusion that stranded cable equipped with a stainless-steel cooling pipe is an indispensable material for the coil of rapid- cycling magnet or choke, whose constituent wires is processed by insulation film on the surface.The system with the DC biasing device has successfully demonstrated that the operation of the system in dual frequency mode is very stable and reliable by making use of GTO thyristor as a switching element operating at a practical current level in accelerator magnet in dual frequency mode. The magnet biased by a DC field seems to behave like an ideal accelerator magnet. As pointed out in the previous paper, further investigations should be made on a stable operation of GOT thyristor in higher voltage in off-state and also it is desirable to study the collective operation of many GTO thyristor switches distributing among the meshes of an actual ring resonant network of synchrotron.At present, further development is in progress for the introduction of a flat top field into the magnet field in dual frequency mode by using GTO thyristor.REFERENCES1. H. Sasaki et al., Proc. the seventh meeting of the international collaboration on Advanced Neutron Sources, Chalk River Nuclear Lab., Sept. 13-16, 1983, p. 502. H. Sasaki et al., Proc. Int. Workshop on Hadron Facility Technology, Santa Fe, Feb. 2- 5, 1987, p. 4083. H. Sasaki, K. Takikawa and M. Kumada; The resonant network for the KEK booster synchrotron magnet, KEK-73-2, 19734. M. Kumada, H. Sasaki, K. Takikawa, H. Someya, T. Kurosawa and Y. Miyahara, The Magnetic field measurements of the booster synchrotron magnet, KEK-77-30, March 197841ESRF BOOSTER SYNCHROTRON MAGNET & POWER SUPPLY DESIGNN. MarksESRF, BP220, 38043 Grenoble Cedex, France.&Daresbury Lab., Daresbury, Warrington WA4 4AD, U.K.ABSTRACTThe parameters of the dipole, quadrupole and sextupole magnets in the 10Hz Booster Synchrotron of the ESRF are detailed. A number of possible power supply systems are considered, and the advantages of the classical 'W hite Circuit' over other variants are presented. The paper then concentrates on the dipoles and quadrupoles, considering the interaction between the power supply and magnet design. The interaction of coil eddy current loss and magnet voltage is examined, and end 'roll off' geometries for both sets of magnets are presented.THE ESRF BOOSTER SYNCHROTRONThe European Synchrotron Radiation Facility (ESRF) will have a 6 GeV electron/positron Storage Ring for generating the radiation. To inject into this ring, a 6 GeV, Booster Synchrotron is also necessary, and as an injection rate of 10Hz is required, the synchrotron magnets and power supplies must be designed for AC operation.The Booster Synchrotron will have a separated function lattice, with dipole bending magnets, two separate fam ilies of quadrupoles (F and D types), and also two sets of sextupoles - a total of five  systems. Each system will require independent control of the AC and DC excitations in the magnets, and also of the relative phases of the AC com ponents. They w ill be locked in frequency, with the single dipole c ircu it as the standard, probably free-running at its natural resonant frequency.The param eters of the Booster Synchrotron magnets are given in Table 1.SUITABLE POWER SUPPLY CIRCUITS.At an early stage in the project, before any major magnet design work had been carried out, the possib le  pow er supply c ircu its , su itab le  fo r m agnet exc ita tio n , w ere considered . The standard  approach in the past had been to use a resonant network producing a magnet current having the form of a biased sinwave with the sources42Table 1. Booster Magnet Parameters.Number of Dipoles 66Maximum field T 0.91Magnetic length m 2.094Horizontal good field region mm +20.0Required vertical aperture mm 32.0Type 'H' magnet;Number of Quads 78Maximum gradient T /m 15.0Magnetic length m 0.6Good gradient region mm ±26.0Required bore radius mm 30.0Number of sextupoles 54M agnetic length m 0.14of AC and DC separated. This so called 'W hite Circuit' is shown in Fig 1.The circuit has many favorable features and is well understood, having been used for many fast cycling synchrotrons, including DESY, CEA and NINA. It is possible to segment the magnets into separate cells, w ith the resonant capacitors located between each cell, so that a series connection is obtained, but the total series voltage does not accum ula te. This segm entation leads to the auxilia ry inductor (L C(1 in Fig 1) being split into separate windings which, in order to s ta b ilise  the c ircu it resonance, m ust be very c lose ly  coupled  m agnetically. In com mon with other c ircu its  used for fast cycling  magnets, the system also suffers from the presence of a 'delay line resonance', in which the stray capacitance to earth oscilla tes with the m agnet series inductance, leading to d is-s im ilar currents in the magnet chain. A further major problem encountered in the use of this circuit is the design of the invertor system required to excite the 10 Hz oscillation; in spite of the utilisation of the basic circuit in many different machines, there has been no common accord on the optimum  invertor circuit, and many alternatives have been built and operated,43ChokeLchMagnetstringLmMagnetstringLmFig 1 Standard White CircuitFig 2 ModifiedWhite Circuit.with varying degrees of success, in the past.A variation of this circuit has been developed at Fermi Lab, and used at other accelerator laboratories. This is shown in Fig 2. The circuit has a number of major advantages over the standard circuit, the most im portant of which is the use o f .a  conventional phase controlled rectifier as the single power source. Providing the choke has a higher inductance than the magnet string, the current in thepower source does not change sign, and the required d irect andalternating currents are generated by a voltage output waveform that includes both a d irect and alternating com ponent. However, if the power ratings of the respective supplies in the two c ircu its  are exam ined, and the variations of power during a m agnet cycle are plotted (see Fig 3), a m ajor d isadvantage is encountered. The m odified c ircu it has a sub stan tia lly  h igher peak pow er rating. Furthermore, the loading on the supply fluctuates by over 100% (ie during a small part of the cycle, power is returned to the supply), andthe fluctuation is at the same frequency as the magnet alternatingc u rre n t. By c o m p a ris o n , the  c la s s ic a l W h ite  c irc u it  has approxim ately a 30% power fluctuation at a frequency that is double the m agnet excita tion rate. This d ifference is very s ign ificant at 10Hz, for the human eye has a maximum 'flicker' perception at that frequency, and hence the public supply system s place stringent tolerances on such load fluctuations. The high power drawn at 10Hz therefore makes the modified circu it an unattractive proposition.A num ber of possib le  pulse c ircu its  were also exam ined. However, during an early stage in the ESRF project, it was decided that the standard W hite Circuit presented the most favourable set of operating parameters, and the examination of a possible magnet44Fig 3 Variation of power drawn by the Classical and Modified White circuits from the public supply over one magnet, to determ ine more clearly the power supply requirem ents, was carried out.DIPOLE MAGNET DESIGNCoil Eddv Loss.In the past, fast cycling magnets with repetition rates of 50Hz had utilised stranded conductor to reduce coil eddy current loss to an acceptable level. This produces technical d ifficu lties concerning the cooling method. However, such com plications are not regarded as necessary at 10Hz, providing the cross section of the individual turns of solid copper is reasonably small. This, however, w ill result in a large number of turns and consequential high voltage on the coil. As it was hoped to use a simple, single cell power supply circuit, this inter dependence between loss and voltage was studied.A current density of 3.8 A /m m ^ was used to fix the total cross section of copper required in the coil, and this was distributed in two pancakes, above and below the gap, each pancake having two layers of conductor. The height per turn was kept fixed at 13.7 mm, and the number of turns and hence the breadth per turn used as the independent variable in the study. The fie ld d istribution in the coil region and vector potential across the com plete gap was predicted  using the two dimensional code POISSON. The resulting eddy loss was45Number of turns per magnetFig 4. Variation of coil eddy current loss and total magnet 10 Hz alternating voltage with number of turns per magnet.calculated using a simple model in which the field was normal to the coil and the eddy currents did not modify the m agnetic field. The results are shown in Fig 4, where the variation in coil eddy loss per magnet and total magnet alternating voltage is shown as a function  of number of series connected turns per magnet. The eddy loss does not become negligible compared to the expected resistive loss in the coil until the number of turns is above 30, corresponding to circuit voltages in excess of 10 kV. W hilst such voltage levels are practical, it would be cheaper and result in less bulky insulation on many components if a lower voltage could be used.It was therefore decided to use a series/parallel connection for the Booster dipoles, so that there are 32 turns in total, of 8mm wide conductor. There will be two pancakes on each magnet, but they will be connected to present 16 turns to the power supply circuit. The parallel connection will be made externally to the magnets, with two separate circuits around the ring. Such parallel connections result in equal alternating flux coupling the two circuits, and if this condition  is not met by external geom etric conditions, c ircu la ting  currents  will flow. Care w ill therefore be taken to balance the alternating  fluxes coupling the circu its before paralle ling occurs. Direct current balance is dependent on' the DC resistances of the circuits, and it will also be necessary to ensure that there is a close resistance  match between the two circuits. These two separate m easures will m inim ise c ircu la ting  current.46M ag n e t E n d s .At a later stage in the magnet design, the problem of how to term inate the end of the dipole was considered. It is necessary to increase the gap at the magnet end, in a smooth way, so as to prevent excessive eddy currents due to flux com ponents normal to the end lam inations and also to limit non linear effects produced by high flux densities at the pole corners. From experience gained during the design of the SRS Booster Synchrotron at Daresbury, the dimensions of a suitable linear end role off, which would minim ise 10 Hz eddy effects, were established. The resulting azim uthal section through  the dipole end is shown in Fig 5. It was also known that this endFig 5 Azimuthal section Fig 6 Cross section throughthrough dipole end dipole mid region(half gap). (half gap).profile would give a magnetic length close to the magnet's physical length, although this has no great design significance.As the gap dim ension increases, the resulting dipole fie ld  w orsens in qua lity , w ith strong negative sextupo le  com ponents  developing. It is possible to correct for this by increasing the size of the shim at the pole edge of the end laminations with the larger gap. This leads to a complex three dimensional shapes in the end region, and resulting high engineering costs. It was therefore decided to specify a planar roll-off region, with no shim present in this area, to accept the resulting high negative sextupole field in the end region, and to correct the total dipole integrated field quality by means of a positive sextupole field in the central region of the magnet. The mid magnet pole shape needed to develop this correction sextupole field is shown in Fig 6. The resulting fields in the different gap regions of the dipole were computed using the two dimensional code MAGNET,47the calculation being made for a set of transverse slices at different azim uthal positions. These were norm alised against the predicted  two dim ensional d istribution of vertical fie ld through the magnet in the azimuthal plane. The resulting three dim ensional fie ld array was in tegrated num erically, to give the variation o f in tegrated vertical field with radial position shown in Fig 7. It can be seen that nearly full com pensation of the end effects is predicted out to 20mm from  the beam center line.x (mm)Fig 7 Variation of integrated field through the Booster Dipole magnet as a function of horizontal position.QUADRUPOLE MAGNETIn the case of the quadrupole, the stored magnetic energy is much less than in the dipole, and hence there was no difficulty in designing a coil with small conductor dimensions (8.8 mm square) to limit eddy curren t loss, w ith less than 2 kV across the com ple te  seriesconnected quadrupoles.Quadrupole End Desinn.In spite of the quadrupoles having lower pole fie lds than the dipole, it was thought still to be necessary to cut back the pole ends. Magnet engineering indicated that this would be best accomplished by utilis ing  sets of end lam ination packages, each w ith a profiledescribing the arc of a circle. As gradient and fie ld varies as theinverse square of the inscribed radius, the size of steps between thepackages can vary as the square of the radius. The profiling4820mm_ V t ‘^  50mm 30mm_ i _ _______ LFig 8 Azimuthal section Fig 9 Schematic diagram of endthrough quadrupole roll-off packets in quad:end on 45° axis R in s c r ib e d  radius of each packet;r =pole curvature radiuscom m enced with a step size of 1.2mm, which would give negligible 10 Hz eddy currents. Ten successive steps, each of 2mm width, were then used to increase the inscribed radius to 50mm, at which value, it was judged tha t the ro ll-o ff cou ld  te rm ina te . The resulting  azimuthal pole profile along a 45° plane is shown in Fig 8.The pole curvature radius of each end packet, shown as r in thex (mm)Fig 10 Integrated grad ient quality  through the Booster quadrupole  magnet, as a function of horizontal position.49schem atic diagram  of Fig 9, was established by two dim ensional com putations using MAGNET, and the resulting integrated gradient through the quadrupole, at differing radial positions, was determ ined  num erically. The predicted integrated grad ient quality  is shown in Fig 10.ACKNOWLEDGEMENTSI would like to acknowledge and thank: Prof. G.Mulhaupt, deputy project head of the ESRF, with special responsibility for the Booster Synchrotron, M. M .Lieuvin, of the M agnet Group, ESRF, and M. J.F .Boute ille  of the Power Supply Group, ESRF, for many useful discussions and help in the design work reported in this paper.50PO W E R  SUPPLY CONSIDERATIONS FO R M AGNET DESIGNK. ReinigerTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V 6T 2A6For the Booster and Driver rings of the TRIU M F KAON Factory, the  resonant supply configuration dictates certain constraints on the m agnet design.As the  m agnets are excited in series, the uniformity of the m agnets is a prim ary concern. The inductance of the m agnets m ust be carefully controlled to  assure th a t the stored energy in the inductive com ponents are the same. As the resonant system is comprised of a  num ber of resonant cells, it is im portan t th a t these cells be m atched as closely as possible to  assure the m agnetic field as a function of tim e in the various families of m agnets is the same. This implies careful control of the mechanical configuration as well as the am ount of steel in each m agnet.For the Booster ring which operates a t a repetition ra te  of 50 Hz w ith frequency com­ponents of 33.3 and 100 1IZ the lam inations should be chosen to  minimize the  ac losses due to  hysteresis and eddy currents. Probably one would want to  go to  silicon steel with a  lam ­ination thickness of 0.35 mm. This is of im portance in term s of minimizing the am ount of energy which needs to  be replenished during each cycle. This has a direct effect on operating cost for the facility, as it represents a continuous dissipation. M inimizing the ac losses also results in fewer high voltage, high current supplies required for the pulse forming network.W ith  a view to lim iting the voltage to  ground of any m agnet to  less than  10 kV the inductance value m ust be optimized to  a minimum value to  result in e = l* d i /d t  <  10 kV at the highest excitation frequency IE. During the m agnet reset interval. If the m agnet inductance is too high to  accom m odate 2 m agnets in one resonant cell, the num ber of resonant cells double. The existing Booster design has 12 cells w ith 2 m agnets per cell while the Driver presently has 36 cells w ith 4 m agnets per cell. If the number of cells could be reduced beyond this point, the system  complexity is reduced.Each resonant cell has the a ttendan t dc bypass choke for the resonant capacitor banks. As this choke is effectively in parallel with the m agnets in each cell and its inductance affects the resonant frequency, the design of the choke and m agnets are interrelated . The chokes m ust be designed to  close tolerances in term s of their inductance. It is im portan t to  realize th a t the peak voltage across the choke is about twice th a t of any individual m agnet so th a t tu rn  to  tu rn  insulation is a consideration, while peak-voltage to  ground is the same as the individual m agnet, assuming two m agnets per cell.The optim um  choke inductance is about equal to  1.5 tim es the to ta l m agnet inductance in the individual cell. Based on the am ount of ac power lost in the ring due to  eddy current losses etc. one may consider a centrally located multiple winding choke or a  distributed choke scenario. In either case there m ust be good coupling to  the dc bypass secondaries to  allow for system synchronization via the pulse forming network. The ac losses in the choke should also be kept too a minimum as they directly add to  the to ta l ring losses which need to  be make up at high voltage. Provision should be made to  center tap  the dc bias secondary with individual access to  bo th  halves of the winding. In the event of a d istribu ted  choke scenario a back leg winding would have to  be provided to  tie all the chokes together. Backleg windings should also be considered for m agnets.51FERM ILA B M A G N ET CO NSTRUCTIONJam es C. Hum bert Ferm i N ational Accelerator Laboratory*P. O. Box 500, M.S.-314 B atavia, Illinois 60510 USACORE STACKINGWe use three basic core stackers: hydraulic, screw machine and boltedtype. All operate equally well bu t one may prefer one type over others. For very small jobs a  bolted type stacker may be preferable and for long m agnets a screw stacker is preferred. However, for large cross section m agnets a hydraulic stacker would be preferred. We a t Ferm ilab prim arily use a  screw stacker since m ost of our m agnets are of the 10 ft to  25 ft lengths. We do also use some hydraulic stackers. W hen stacking cores we try  to  develop approxim ately 120 psi for proper stacking pressure. Stacking rails are used to keep the lam inations in line and to  keep the m agnet s tra ight.We notch the lam ination on an outside edge. The notch is usually a small “ v”  shaped notch and is used to  keep track of the taper in the m aterial due to  rolling of the steel. W hen building long m agnets, the variation  in m aterial thickness s ta rts  to  enter into the overall length of the core. W hile stacking cores we flip lam inations about every 3 to  4 inches. This compensates for the variation in m aterial thickness.W hile stacking lam inations for ac or pulsing operating, make certain  the lam inations are phosphate coated or have other sim ilar insulating m aterial to reduce eddy currents.There are two types of layups of cores, wet layups and dry. Dry stacking simply means, stacking lam inations, compressing them  and welding the core together using tie bars.Before stacking a  wet layup core, wash the lam inations. This will elim iniate any cu tting  oils left on the lam ination after stam ping. We use a machine called a C leanom at1 w ith solvent. W hile stacking wet layup cores be careful to  use a  mold release agent2 on all stacking parts. This allows easy removal of the core from  the stacking fixture after curing. In addition, a bolted type stacker is usually used in com bination w ith a hydraulic or screw stacker. It is implied th a t epoxy is being used to  hold the core together during a  wet layup. W elding may or m ay not be done on wet layups, bu t it has been our experience th a t  you should weld on the core when the epoxy is still uncured; this will allow a  much be tte r weld. If welding is done after the coil is cured, you will get more porosity in the weld.We use heat cure epoxy3 and cure for 5 hours a t 300°F after the core reaches 300°F.Shear tests should also be m ade on the lam inations and epoxy before stacking. W ash several lam inations, cut them  into  1/2 inch strips and glue them  together using the epoxy th a t is going to  be used in the w et layup.This should be cured and sent to  the test lab. Shear strength  should range from 1000 psi to  1200 psi. If you get in th is range of figures, then  the cleaning was successful. A Roller C oater4 is used to  apply epoxy on one side of the lam ination. A note of caution, we have found th a t  some phosphate*Operated by Universities Research Assn. Inc., under contract with the U.S. Department of Energy.52coating does not lend itself to epoxy gluing. You can use room cure epoxy for wet layups but you must consider pot life and bonding strength. We have found that we achieve better bonding by using heat cured epoxies.When core stacking, you will need end packs which can be laminated or solid steel. Solid steel end packs are expensive. We have been very successful using laminated end packs, that are a wet layup of heat cure epoxy. They can be machined for contouring as long as the shop is warned as to what direction the tool is turning. The tool should rotate in a direction so as not to peel off laminations. This type of end pack is very useful in prototyping.It can be made in such a manner so it can be removed, machined and installed back onto the magnet.COIL WINDINGWhen winding copper conductor for magnets, the following are some guidelines and procedures that should be considered. When ordering copper, make sure copper is specified as dead soft, or Rockwell <40 HRF. This very soft copper will make it easier to wind coils.Tension must be applied to the copper when winding. The tension can take two forms, friction blocks or friction on the spool of copper. Friction blocks are installed between the winding table (winding fixture) and stationary platform. The friction blocks supply the proper amount of tension on the copper to produce a good coil package. We use tension in the range of 1000 psig. Friction blocks can be made of phenolic, nylon, G-10 or other materials.We also use an air break5 type tensioner, when winding coils from spools of preinsulated wire. This is usually in the form of heavy ML6 or ML and Daglas.7 A teflon guide is used to guide the wire to the winding fixture.When forming insulated wire, be very careful not to chip or gouge the wire with forming tools (hammer and block). A small block made of nylon or teflon or some medium hard wood, is commonly used.When winding coils, a space between turns must be added during winding. This space will take the place of the insulation to be installed after winding. We usually have 28 mils between turns so a 1/32 shim works out.A shim between layer to layer is also required. It is important to note that the winding table must be able to go in reverse. After the bottom  layer is wound, a joint must be made in order to wind the top layer, unless you back wind the layer to be used later. This can usually only be done using small cross section copper.We use a sleeve joint when joining conductor. We find this superior than just using a butt joint. This, of course, is only used on water cooled conductor. On solid and small cross section copper an overlap joint is made. When making a joint in copper a brazing fixture is usually needed. This brazing fixture holds the conductors together and maintains pressure on the joint through the use of die springs. Silflos8 solder is used in making our copper joints. After the joints are made, the area is filed flush to the original copper dimensions.After the coil is wound it is prepared for sandblasting (uninsulated type of copper) on all outside surfaces. The sandblasting does two things; it gets rid of any oxides on the surface and allows our epoxies to stick better to the surfaces. After sandblasting the coil is inspected for leaks in any joints. It is then dekeystoned on all corners. If the coil is not dekeystoned, it may cause local pressure when clamped in a curing fixture and cause a turn-to-turn short. The coil may be dekeystoned before it is sandblasted; however, this is usually a decision made on a case-by-case situation.53COIL INSULATINGWhen insulating bare copper conductor, insulating tapes must be chosen and there are a variety of tapes available on the market. At Fermilab we use a variety of Mica,9 glass10 and glass/polyester11 type of B-stage tapes. We also use glass tapes in potted magnets.B-stage tape is impregnated with epoxy and partially cured. The percentage of epoxy in the different tapes ranges from 35% to 75%. These tapes range in size, the most common being 1 inch wide and usually 36 yards on a roll. The normal conductor tapes are usually 7 mils thick. On very large conductors, we have used a Scotchply tape.1 This tape comes in 10, 20 and 30 mil thickness and rolls of different widths.When insulating a coil you must spread the coil apart so one can insulate the individual conductors. We usually use one layer of half-lapped insulation. This gives us 28 mils of insulation between turns and between layers. Butt lapping is also possible; this will have to be done twice on each conductor in order to get 28 mils. The coil is nested back together after it is insulated or can be nested back together during insulating. This is dependent on the complexity of the coil being wrapped.We use Mica-B-Stage tape along with glass tape for ground insulation.The coils/m agnets are then potted with epoxy13 These magnets are usually used in high radiation areas. When using plain glass cloth tape make sure it has a saline or volan finish. This allows the glass cloth to more easily absorb the epoxy during potting.Ground insulation is the additional insulation used to hold the package of conductors together and provide additional insulation between the coil and magnet core (ground). We usually put an additional 30 mils of ground wrap insulation on the coil. This ground insulation can take the form of several tapes. We have found Scotchply and Armorflex14 to be a very good combination. Before installing the completed coil into a curing fixture, we wrap an additional layer of Tedlar15 tape on the coil. This tape is called a stripping tape. This type of tape does not stick to epoxy, and acts as a mold release. After the coil is cured it is removed from the curing fixture.The Tedlar tape is removed and any epoxy flash is removed from the cured coil.We have several methods for curing coils at Fermilab. They consist of oven curing, Dowtherm16 and resistive heating. Oven curing is the easiest, but sometimes the coils are larger than the oven. When coils are larger, a system of using hot liquid (Dowtherm) or resistive heating can be used. Dowtherm heating will require the coil have a water passage so the liquid can pass through the coil and be heated internally.Resistive heating can also be used to heat the coils. A power supply can be hooked up to the coil and heated using the power supply. On very big coils the power supply may not have enough power to cure the coil.When curing coils, the coil is cured for 3 hours at 300°F; that is, when the coil reaches 300°F the curing cycle starts.You can also pot coils, but we have found that the various B-stage tapes we use are just as good as potted coils. However, there may be some cases that a potted coil would be preferable. Again this must be decided on the particular coil or magnet design.We also use a heat cure epoxy17 that is made in such a way that it has the consistency of a soft jell. This is useful when winding solid conductor that has been pre-insulated (heavy ML, Daglas). It provides extra epoxy between conductors and other void areas. The coil is then ground wrapped and cured.54We do some potting of magnets here at Fermilab. We insulate the conductors (usually with glass tape) and ground wrap the coil with glass tape. The coils are then installed into the core and the core is sealed. The magnet is installed into our vacuum oven, heated, and out gassed. The magnet is back filled with heat cure epoxy until it comes out of a riser at the other end. After the magnet is filled with epoxy, it is pressurized to 30 psig and the vacuum oven is brought up to air. After that point the magnet is heated internally with Dowtherm for the final cure. The riser is maintained with epoxy until the epoxy is cured.FINAL ASSEMBLYFor the most part we braze on our flags onto the coil conductor using Silfos silver solder. The water fittings are usually 304 stainless steel and brazed onto the copper using B -l flux18 and Easy Flo 319 silver solder. The insulators between conductor water paths are the bolted-on type and take two forms. One type is made of ceramic and has a flare type fitting pressed20 onto the ends. The other type is a compression type ceramic21 and is held in place using compressing type fittings. The fittings we have used are Swagelok.INSPECTIONWe do several types of inspections on our coils/m agnets produced at Fermilab. As the copper arrives from a vendor, it is checked dimensionally and a ball is passed through the water hole to make sure it has no restrictions.When winding a coil, a joint may have to be made in the coil. After the coil is wound and sandblasted the joint is pressurized with He to 200 psig and sniffed with a He leak detector having a sensitivity of 2 x 10'10 atm  cm3/sec He. No detectable leaks are allowed. A water flow test of the coils is made to check the gallons per minute flow rate of the coils. After the coil passes these tests, it is pressurized to 1000 psig hydrostatically and held there for 1 /2  hour. Again, no detectable leaks are allowed. When the coil is ready for wrapping it is checked for nicks, burrs and proper dekeystoning. If it passes all these checks it is ready to be wrapped. After the coil is wrapped it is tested for resistance value (R), inductance (L J, Q and Ring. All through the assembly process, the coils are continually tested for R, L , Q and Ring. After the coils are installed in the core one additional test is also performed, this is a dc hipot to ground. Our standard hipot is usually 3kV and < 5 f la  leakage to ground. However, this hipot may not work before the m agnet/coils are cured/potted. Our hipot is reduced to 1 kV, and £ 5 fia. leakage. If the area is not air conditioned you may not be able to obtain <5 Ha leakage.After the magnet is completed, bussed and electrically hooked up, we do our final checks. We flow check the entire assembly and then pressurize to 600 psig. The magnet is also surveyed for straightness and flatness and then we do additional R, L , Q and Ring.These procedures will help to insure that a quality magnet is produced.55REFERENCES1. Cleanomat: Division of Data Metalcraft, Inc., 7901 Fuller Road, Eden Prairie, MN 553442. Mold Release Agent: Vydax 500, FNAL Drawing Number 2856-MA-116537.3. Heat Cure Epoxy: lOOg Shell EPON 826, 90g Copps NMA, lg  BDMA,Miller Stephenson Chemical Co.4. Roller Coater: Atlas Co., 1733 N. Milwaukee Ave., Chicago, 111.606475. Air Break: Model LKV-108 Industrial Clutch Corp., P. O. Box 118, Waukesha, Wis. 531866. Heavy ML: Heavy Aromatic Polyimide Coated Wire, Class 220, NEMA Section 20-C.7. Daglas: Polyester Glass Fiber used to cover bare or coated wire.8. Silfos: Brazing Rod for brazing copper to copper. Handy and Harman,850 Third Ave., N.Y.C., N.Y. 100229. Mica Tape: B-Stage type tape. FNAL Drawing Number 2856-MA-116569.10. Glass Tape: B-Stage type tape. FNAL Drawing Number 2856-MA-116567.11. Glass/Polyester: B-Stage type tape. FNAL Drawing Number 2856-MA-116568.12. Scotchply: Non-woven Fiberglass reinforced epoxy resin material. 3M Structural Products, 220-7E 3M Center, St. Paul, Minn. 5514413. Epoxy to Pot Magnets. Batch 1, 45g 826, 38g NMA, .6g DMP30.Batch 2, 45g 826, 38g NMA, 40g Alumina, 50g Micro Glass Beads and .6g DMP30.14. Armorflex: B-Stage type tape. FNAL Drawing Number 2856-MA-225545.15. Tedlar Tape: E. I. Dupont Co., FNAL Drawing Number 2856-MA-116529.16. Dowtherm SR-1: Dow Chemical Co., Specialty Chemical Dept.,Midland, Mich. 4864017. Heat Cure Epoxy Paste: 50g 826, 16.7g 732, 60g NMA, .7g DMP30, 2g Glycerin, 8g Cab-O-Sil.18. B -l Flux: Handy and Harman Co., FNAL Drawing Number 2856-MA- 116549.19. Easy Flo 3: Handy and Harman Co., FNAL Drawing Number 2856-MA- 116560.20. Ceramic Fitting: FNAL Drawing Number MC-22304.21. Ceramic Compression Fitting: Diamonite Products, FNAL Drawing Number 2858-MA-225147.22. Compression Fitting: Swagelok Tube Fittings, Crawford Fitting Co.,29500 Solon Road, Solon, Ohio 44139.56TH E STATUS OF TH E BOOSTER D IPO LE DESIGN AND PLANS FO R TH E PR O JE C T  DEFIN ITIO N STUDY YEARA. O tterTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V6T 2A6ABSTRACTThe present s ta tus of the booster dipole design is presented together w ith a list of the m agnet development tasks during the project definition study.INTRODUCTIONThe KAON Factory will require different types of m agnets, some estim ates show th a t as m any as sixty different designs will be needed. These will be bo th  dc and ac m agnets and ac m agnets are a new technology for TRIU M F. During the P.D.S. we have to concentrate on a  few designs in detail. We have chosen to  complete the design and build a prototype of the booster dipole because it presents the most challenge. This will give us experience in designing an ac m agnet and m easurem ents on the prototype will give us a calibration of our design m ethods.BO OSTER DIPOLE STATUSThe Booster Dipole param eters listed in the KAON Factory proposal1 are as follows:Bend Angle, 3 GeV protons 15°Effective Length 3.18 mVertical aperture 10.68 cmGood field width 11.76 cmMaximum field 1.05 TMinimum field 0.277 TExcitation rise tim e 33.33 HzExcitation fall tim e 100 HzThese values are all subject to  revision and Uli W ienands2 is actively considering a combined function dipole w ith gradient of 3.9 T /m , using the same central field, bu t a reduced vertical aperture. Most of our work has been concerned with the uniform field dipole. During the preparation of the proposal Keith Lacey of DSMA-Acton prepared basic designs for all the ring m agnets using a prelim inary costing program . This was similar to but not as extensive as the program m e used a t Argonne by Ken Thom son.3 It allowed us to prepare cost estim ates and m aterial quantities for both  quadrupoles and dipoles. The dipole estim ated by Lacey is shown as Fig. 1, it was not intended to  be the final design.A further study was undertaken by E.M. Gibson a summ er student. The aims of this study were two fold:• To modify the pole shape to  achieve good field uniformity over the required aperture and;• to  investigate eddy current losses in the coil.57POLE GAP POLE WIDTH YOKE FLUX DENSITY YOKE THICKNESS10.68 cm2 7 .8  cm  (2 .6  x GAP) 1.4 T13.9 cmCONDUCTOR OUTER DIMENSION CONDUCTOR COOUNG HOLE ARRAYTURN ARRAY COIL WIDTH COIL HEIGHT6 m m  SQUARE 3 .4  m m  ID4  WIDE x 3 HIGH /  TURN4 x 411.5 cm8.8  cmDC 1 ,973 A1 R M S  2 .1 3 5  AMAXIMUM VOLTAGE 9.2  kVINDUCTANCE 3 .17  mHl2R LOSS (COIL) 5 2 .5  kWEDDY CURRENT LOSS (COIL) 52.7  kW IRON LOSSES 11.5 kWMAGNET WEIGHT 8 00 0  kgFig. 1. B ooster d ipo le in itia l design concep t.Figure 2 shows the m agnet profile which resulted from this study. The m agnet w idth and height were increased and the pole is now tapered. Flux densities in the yoke vary from 1.24 to  1.39 T  at maximum excitation. The field profile is shown in Fig. 3 and it is not quite good enough at a central field of 1.05 T. We need to  make the pole wider and changeFig. 2. M agnet profile w ith  in itia l pole sh im .58the shape of the  shim. The field uniformity a t low excitation (injection) was not considered by Gibson and Fig. 4 shows th a t it is far from satisfactory. The effect of increasing the m agnet size to  reduce the average yoke flux density to  below 1 T to  reduce the  iron am pere turns does not solve this problem as is also shown in Fig. 4. We need to  change the shim to establish a satisfactory pole profile which will give adequate field uniform ity a t bo th  injection and extraction energies.FLUX vs. X (FINAL DIMENSIONS)X (cm )Fig. 3. F ield  profile a t  ex tra c tio n  energy.F ig . 4. F ie ld  u n ifo rm ity  a b o u t cen tre line  a t  in jec tion  an d  ex tra c ­tio n  energies for various yoke flux densities.59Eddy current losses in the coil were studied by estimating the loss due to transverse magnetic field4 over the coil region for various conductor sizes. The program Poisson was used to obtain the field and parameters were adjusted for each conductor size to get the field at the centre of each conductor. The individual losses were then summed over the coil. Figure 5 shows the losses per metre as a function of conductor size.CONDUCTOR SIZE (cm)Fig. 5. Eddy current loss/metre over coil for various conductor size. Square hollow conductors.The shape of the coil was varied and it appeared that the losses were minimized for a rectangular coil section as shown below.Coil Section Loss/mwidth height8 cm 18 cm 4490 W11 cm 13 cm 3912 W18 cm 8 cm 4930 WFinally losses calculated from the harmonic analysis of the dual frequency waveform were compared with those of the average frequency of 50 Hz and found to be 29% higher. It is not accurate enough to use the average frequency to design these magnets.We needed a quicker way to estimate eddy current losses for initial design reviews and optimization studies. Ed De Vita5 looked into this and pointed out that if the coil is inserted into the slot as shown in Fig. 6 the field in the coil region is horizontal and increases uniformly from zero at the top of the coil to ^ — and it is easy to show that the average value° f  over the coil region is 12s^ 2 where B g is the field at the gap centre. The expression allows a very quick initial estimate of coil losses. A rectangular conductor with a rectangular hole was considered to be better than the more normal square hollow conductors. With this arrangement the losses are reduced as the coil slot width is increased. It was shown that60with a standard  rectangular conductor from Hitachi th a t the cooling considerations could be resolved and the coil losses reduced to  10.8 kW /m agnet.COM BINED FUNCTION D IPO LE DESIGNWe are ju st s tarting  to  look at the gradient m agnet design requested by Uli W ienands. The m agnet increases the peak fields in the gap and in the steel bu t to  first order the stored energy in the gap is the same for bo th  types of dipoles, if they have the same aperture. We will make a comparison of the two designs w ithin the next two m onths and then  decide upon which to  build as a prototype. Our schedule shows th a t we expect to  call for tenders on this m agnet early in 1989.PR O JE C T  DEFINITION YEAR PLANSBuilding a prototype Booster dipole will be one of our m ajor projects, it is not our only one, we also expect to  design and build a prototype quadrupole. Our o ther m ajor aims are to  do optim ization studies to  determ ine the best flux density to  use in the re tu rn  yokes and to  determ ine its effect on:• Power supply cool, m agnet cost and size, tunnel environm ent considerations such as cooling and handling.• Review the costing and basic designs, and set up a d a ta  base w ith CAD sketches and prelim inary designs for all the KAON Factory m agnets.• M easure core losses in electrical steels and eddy current losses in conductors a t both sinusoidal and dual frequency waveform excitation.• Make C anadian industry  fully aware of our plans and involve them  in our studies wherever possible.In order to  complete this work we are looking for visitors from other laboratories with m agnet experience to  come and work at TR IU M F with us. We also will try  to  set up an industrial program m e in which engineers from Canadian industry  will come and help us. We feel th a t we can learn a lot about calculating ac losses from engineers in the electrical m anufacturing industry.61REFERENCES1. KAON Factory Proposal, TR IU M F, 1985.2. U. W ienands, KAON Factory M agnet Requirem ents, this workshop.3. K.M. Thom son, An interactive com puter program  for the design and cooling of m agnets. 8th International Conference on M agnet Technology, Journal de Physique, FASC. 1 Cl 1984.4. G.W . C arter The electrom agnetic field in its engineering aspects, 2nd edition, Longmans 1967.5. E. DeVita, Eddy current losses in the excitation coils of ac dipole m agnets. TRI-DN- 88-38.62KAON FACTORY MAGNETS V. VermaTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V6T 2A6A BSTRACTThe KAON Factory a t TR IU M F will require over 1,200 m agnets of up to  60 different designs. The to ta l cost of the KAON Factory is estim ated to  be $571 million (in 1986 dollars) and out of this approxim ately $75 million is for these m agnets. This paper will describe the plans for design, procurem ent and installation (excluding kickers, sep ta and bum p m agnets) under a  very tight scheduling constraint. Some logistics problems for testing and installing all the m agnets will be outlined. We expect th a t discussions a t this workshop will help us find solutions to  some of these problems.INTRODUCTIONBasic param eters for the accelerating and storage ring m agnets are alm ost defined and the conceptual designs for estim ating the costs were completed while w riting the KAON Factory Proposal in 1985. Some conceptual designs for the m agnets in the experim ental halls and beam  switchyards are available. TR IU M F has considerable experience in the design, m anufacturing and installation of dc beam  transport m agnets bu t has no experience in the design of ac accelerator m agnets. In order to  produce the large quantity  of m agnets required for the KAON Factory under a tight schedule, TRIU M F will have to  acquire a good level of competence in the design of ring m agnets. The design and calculation of losses in the ac m agnets will have to  be b e tte r understood in order to  effectively optim ize the m agnet designs. Moreover the services, alignment and installation techniques m ust be designed and im plem ented to  achieve a very reliable operation. A t target locations extrem ely high radiation fields of up to  108 R em /hour will be experienced. The design m ust consider the repairs and m aintenance costs in term s of downtime, m an-dose and plain dollars.O B JECTIV ES DURING PR O JE C T  DEFIN ITIO N STUDY PHASETR IU M F has received a  funding of 11 million dollars for Project Definition Studies (PD S) of the KAON Factory. Out of this approxim ately 1.2 million has been allocated for design and prototyping of ring m agnets and kicker m agnets. The project definition study (PD S) phase is planned to  last from 12-15 m onths as shown on a schedule (Fig. 1). The prim ary objectives during this period are as follows:1. Develop the design procedures and cost estim ates.2. Design and build prototype ac m agnets for the dual frequency excitation booster rings.3. Design and build prototype kicker m agnet.4. S ta rt an industrial involvement program  for the fabrication of m agnets.5. Continue program s to  advise industrial companies in C anada for the the production of the large quantities of m aterials (steel and copper conductor) required for all m agnets.6. Determ ine the m easurem ent techniques and systems th a t will be required.7. Finalize design param eters and layouts of the experim ental halls, target areas and switchyards.63648. S tart to  look at the m agnet requirem ents in experim ental halls and in the  target areas. PR ESEN T PLANS AND INDUSTRIAL DEVELOPM ENTSWe are not aware of many Canadian companies or consulting engineering companies th a t specialize in the design and fabrication of accelerator m agnets from a list of param eters or specifications developed from theoretical optics design. One of the im portan t aims to be m et in the PDS phase is to  s ta rt to  transfer this technology and expertise to  industry. At TR IU M F we have generally done the engineering design and invited o ther companies to fabricate the m agnets for us. Fifteen years ago, all of our m agnets were m ade outside C anada and today more than  90% of our m agnet requirem ents are m et in Canada. A t the present tim e we deal w ith about four coil m anufacturers and about eight companies who m ake steel assemblies for us.In order to  develop and enhance industrial capabilities in C anada, we would design the m agnets and prepare engineering drawings and specifications. During the PDS we plan to  invite visitors from other laboratories and engineers from industry, wherever possible, to  assist us and work w ith physicists and engineers a t TR IU M F on a sabbatical basis. This would apply to  synchrotron m agnet design and the fast kicker m agnets. Two to  three people may be brought in on such a basis. Canadian m anufacturers will bid on the fabrication and supply of m aterials and other components. If Canadian industry  does not respond to  the challenge we will have to  look to  offshore suppliers and m anufacturers.SCHEDULEAssuming a five year construction period for the KAON Factory prior to  the first 30 GeV beam  out, it is anticipated th a t m ajor contracts for fabrication can be awarded only after 1 1 /2  years of design work and qualifying the suppliers and contractors. Table I shows the estim ated m agnet quantities for the KAON Factory. The first block of orders will consist of m agnets for the 450 MeV transfer line and rings ‘A ’ and ‘B ’ which will be fabricated installed first, followed by the m agnets for rings ‘C \  ‘D ’, and ‘E ’. The m agnets required for the beam  switch yard and experim ental halls will be a continuing requirem ent, and will be installed in the 5th and 6th years respectively after installing the m agnets in all five rings.According to  the overall schedule in Fig. 2, the tim e allowed for fabrication of the first block of orders for the 450 MeV transfer line and rings ‘A ’ and ‘B ’ (235 m agnets) is only 15-18 m onths which is so tight th a t it will be necessary to  spread this contract among five or more contractors hopefully including the companies which have worked with TR IU M F staff during the design phase. It should be noted th a t the size of the subsequent order for ‘C \  ‘D ’ and ‘E ’ rings is significantly larger (792 m agnets) but the tim e allowed is still 18-24 months. It can be m et only by increasing the number of contractors, perhaps up to  ten or more, and enhancing the m anufacturing techniques based on the experiences of those involved earlier.The testing and installation schedule is also very challenging. The m agnets for the 450 MeV transfer line and rings ‘A ’ and ‘B ’ are to  be tested in 12 m onths and then installed in another 12 m onths. Similarly the m agnets for rings ‘C ’, ‘D ’ and ‘E ’ are planned to  be tested  in 15 m onths and then installed in another 12 m onths only. This would complete the installation of m agnets in rings ‘A ’ and ‘B’ after 3 and 3 1/2 years respectively and in rings ‘C \  ‘D ’ and ‘E ’, after 4 1/2 years thereby allowing a period of six m onths for overall commissioning to  get 30 GeV beam  out.65Oooo< r•VV -UiUJ>HOZ<smE-*Wc<s> -o zoE -O<& Ho<tr<5it!? §o a:jSUjell i* s> isi tas<-S55£ aas|II es |ujhd*$> 5li?* s -ss itak ®!CZ UJtrUJ5!UJ3\<3OC2< 3UJoc.<c/fUjjj§1ft2<e>V3O p  5oz CJz z9'« 9«-H9^3S 3 ; 35! * ! * 5SP 31°x<N <NQW«N <NQ»i S. <5< P b9 *I ;5 shISa w  —  wC* W'ozs• K9 «35S331i  X  tNf\2Qcc<I I3*yfi s2Z<X< r<QZo3£ss</>K  3  If) <ZO<CLCL3o0.00d£zooFig.66Table I. Estim ated m agnet quantities for KAON FactoryDipoles Quads Sextupoles CO D ’s Total450 MeV Transfer 4 10 __ 6 20‘A ’ Ring 24 48 12 24 108‘B ’ Ring (AC) 24 48 12 24 108lC ’ Ring 72 96 24 48 240‘D ’ Ring (AC) 144 96 24 48 312‘E ’ Ring 72 96 24 48 240Injector & E xtraction  12 kickers, 15 Septa, 10 Bump (all Dipoles) 37Switch Yard & 25 50 — — 75Prim ary BL’s Secondary Beamlines 25 50 10 24 109427 494 106 222 1249The highlights of design, m easurem ent, testing and installation phase are outlined as follows:Design PhaseAfter finishing the design and fabrication of prototype ac m agnets for the booster ring in the PDS phase, the detail design for approxim ately 1,200 m agnets will have to  be done after full funding is received. It may require up to  60 different designs and will take 1 1/2 to  2 years w ith a proper mix of manpower :(physicists, engineers and designers) before the m ajor contracts for m aterials and fabrication can be awarded. Assuming an average of three m an m onths for a typical m agnet design, we would require a team  of 8-10 designers working over a period of two years for these m agnets.To m eet the schedule, TR IU M F plans to  invite companies who design and build ac electrical machinery and components to  contract out to  us engineers who can assist in the design and calculations of the m agnets. These engineers would work a t TR IU M F so th a t we can teach them  about accelerator m agnets and we can learn the m anufacturing techniques, problems and ac design approaches from them . This would be especially useful because we would be expecting them  to assist us in building the m agnets a t their home factories. This approach would help in m eeting the engineering specifications and requirem ents very effectively along the transferring the technology to  the C anadian industries.Procurem ent PhaseThere are good opportunities for m aterial suppliers also, for KAON Factory m agnets we will require up to  320 tons of conductors and up to  4250 tons of m agnet steel. At the present tim e, there is no Canadian company th a t has the equipm ent to  produce hollow cop­per conductors required for our m agnet excitation coils. However, due to  the large quantities involved it is im portan t to  explore the possibilities of helping industries in acquiring such67equipm ent so th a t the Canadian companies can supply the conductor. Pyrotenax has shown a very keen interest in expanding their capabilities to  m anufacture copper conductors. Ini­tial discussions w ith Dofasco indicate no production problems in acquiring steel for all the m agnets.To expedite the schedule for fabrication of all m agnets, the potential suppliers of m agnet steel and copper conductor should be kept in touch while design is in progress. They will be informed as soon as possible regarding the quantities and specifications of m aterials and o ther m ajor design features to  allow sufficient tim e for developing tools, jigs, etc.Realistic quality control specifications should be prepared and distribu ted  before or along w ith the Request for Quotes. TRIU M F engineers and physicists will have to  work hand- in-hand with the suppliers and contractors to  provide any technical expertise and resolve technical problems as necessary. A utom ated m agnet m easuring system s will have to  be designed and built to  expedite the acceptance tests and wherever necessary consultants will be engaged to  carry out these tests to  avoid bottle-necks a t the factories and to  expedite the overall schedule.Logistics in Testing the Installation PhaseDue to  m anpower, logistics and space constraints, testing and installation will have to  be staggered for different rings. F irst of all, the m agnets for the 450 MeV transfer line will be installed, followed by the m agnets in ‘A ’ and ‘B ’ rings. Most of the services (water, electrical and interlocks) for m agnets should be installed up to  closest proxim ity and checked out before m agnets are received at TRIU M F. All the alignment services and equipm ent should also be installed and tested before hand. It may be necessary to  carry out some quick preinstallation tests which may require up to  8-10 test setups to  prepare m agnets for installation. M anpower for testing the installation will have to  be very carefully planned and scheduled. For example, to  test the m agnets for the 450 MeV transfer line and for rings ‘A ’ and ‘B ’ in one year and install in the following year, we will have to  test one m agnet per day and during the installation one m agnet per day should be installed, connected and tested. Similarly, for ‘C ’, ‘D ’, and ‘E ’ rings, more than  three m agnets will have to  be tested per day and more than  three m agnets per day will have to  be installed, connected and checked out.SUMMARYThere is a lack of experience at TR IU M F in the design of ac accelerator m agnets. To meet an extrem ely tight schedule to  procure and install over 1,200 m agnets for the KAON Factory, it will be essential to  achieve the active participation of Canadian industries. We would like engineers from industry to  work with TRIU M F staff from the design through the installation phase. The tight delivery schedule for fabrication and assembly of m agnets emphasizes the need to  spread the orders among ten or more contractors and complete the acceptance tests before m agnets are shipped to  TR IU M F. At tim es, we have to  test up to three m agnets per day and also install and connect up to  three to  four m agnets per day. It will necessitate a combination of careful and detailed planning, shift work, autom ated testing and efficient installation procedures.ACKNOW LEDGEM ENTSThe au thor wishes to  acknowledge the helpful comments from Mr. A .J. O tter and the help received from Mr. M ark Keyzer for preparing tim e schedules.68K ICK ER REQUIREM ENTS FO R TH E KAON FACTORY Ulrich W ienandsTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C ., C anada V6T 2A6A BSTRACTAn overview of the kickers needed for the fast injection and extraction systems for the KAON Factory rings is given. The specifications are based on the  newly developed racetrack lattices. Rise tim e of the field and lim its on field variations (ringing) are given.INTRODUCTIONGiven the num ber of 5 rings in the KAON Factory, there will be 8 fast in jection / ex traction  system s. A typical transfer from one ring to  the next will include an extraction kicker and septum , the transfer line, and an injection septum  and kicker. Given the flexibility of the racetrack design, we can choose a somewhat higher value of the b e ta  function in the extraction section, allowing to  trade  aperture  of the kicker for a reduction in kick angle. For the small rings, a m aximum  of about 37 m has been chosen for the horizontal b e ta  function, while for the large rings, the b e ta  function rises up to  100 m.Table I gives the requirem ents for the kickers. All angles are based on a horizontal extraction scheme and a t least 1 cm clearance between kicked and circulating beam . The em ittances used are as given in the proposal, 140 n m m-mrad horizontal and 62 n m m-mradTable I. Kicker param eters for the five KAON Factory ringsRing fArep(Hz)mode Angle(m rad)M omentum(G eV /c)t<(us)t fall(ns)A perture^(cm)Aperture^(cm)A 50 extr. 4.0 1.01 108 0.87 108 15 8As it 50 extr. 4.0 1.01 10 0.87 25 15 8B 50 inj. 4.0 1.01 5 ms 0.87 108 15 8B 50 extr. 2.5 3.82 82 0.66 5 ms 15 8C 50 inj. 1.5 3.82 22 us 0.66 82 15 6C 10 extr. 1.5 3.82 82 3.60 1 ms 15 6D 10 inj. 1.5 3.82 25 ms 3.60 82 15 6D 10 extr. 1.0 30.90 80 3.50 25 ms 15 6E 10 inj. 1.0 30.90 80 3.50 80 6 4vertical a t injection into the Booster. The length available for the kickers is 3 m in the  small rings (A,B) and 9 m in the large rings (C ,D ,E). In addition there will be orbit bum ps needed to  steer the adiabatically shrunk beam  in the accelerator rings towards the septum . Figures 1 and 2 show the lattice  functions of the small rings (Booster) and the large rings (Driver). Positions of kickers and septa are indicated.691 , 1 , ■ , '*  -  " ° 75 D IS T A N C E  ( m )Fig. 1. Lattice functions of the Booster ring. The dipoles are vertically focusing.D I S T A N C E  ( m )Fig. 2. Lattice functions of the Driver ring.70The rise tim e requirem ents are based on the presence of a  kicker gap of 5 buckets length in the  beam . E xtraction  kickers need to  rise to  full field strength  w ithin passing of the gap, while for injection kickers the fall tim e is defined this way. The other edges of the pulse can be longer, on the order of milliseconds. At present, one of the possibilities to  create th is gap relies on an ultra-fast kicker for extraction from the A-ring, which would have no kicker gap in the beam , and an increase in length of the B-ring by about 11% to  a harm onic num ber of 50, which leaves 5 buckets em pty on the transfer from A to  B. For extraction from the A-ring to  be lossless, the kicker has to  have a rise tim e of about 10 ns; alternatively the cyclotron can be operated  in a 4 out of 5 mode leaving one bucket in the cyclotron em pty and allowing for a rise tim e of 25 ns for the A-ring extraction. Even this is quite a challenging requirem ent th a t can alm ost certainly be m et only by either electro-static kickers or by kickers working in transm ission line mode (stripline kickers), whose param eters are given in row 2 of Table I.However, in th is way, any need for a beam  chopper in the A-ring injection line is obviated, which would otherwise be needed since the kicker gap cannot be created in the cyclotron due to  the m ulti-turn  extraction of H~ ions.FIELD QUALITYThe field quality includes uniformity of the field as well as stability  during flattop (ringing). Given th a t the kick to  the beam  has to  be a t least y/2(k ,  an error in the kick angle of 1% leads to  an em ittance increase of 2% per transfer. For 8 transfers, a worst case scenario would therefore give a factor of 1.02®= 1.17 or 17% in em ittance blowup, which is not negligible but tolerable. More optim istically one would assume the angle error to  be gaussian d istribu ted  and therefore get quadratic  addition of the errors and a value of only 4% em ittance blowup. This is certainly optim istic. We therefore set a  lim it of 1% for ringing and field uniformity.71K I C K E R S  A N D  S E P T A  A T  T H E  PS COMPLEX, C E R ND. Fiander, K-D. M e t z m a c h e r , P. P e a rce  P S  D i v i sio n, CERN, 1211 G e n e v a  23A B S T R A C TThe s t o r y  of k i c k e r s  and septa of the PS c o m p l e x  s t a r t e d  in M a y  1963 when the f ir st fast e x t r a c t e d  beam was o b t a i n e d  fro m the 28 G e V  s y n chro tron. Since those e a r l y  days of the p l u n g i n g  small a p e r t u r e  k i c k e r  and its a s s o c i a t e d  b e n d i n g  m a g n e t  there has been a c o n s t a n t  e v o l u t i o n  and i n c r e a s e  in c o m p l e x i t y  as r i n g after r i n g  has been add ed and the p a r t i c l e  spe cies widened. T o d a y  the c o m p l e x  has 20 k i c k e r  s y s t e m s  c o n t a i n i n g  n e a r l y  100 m a g n e t  m o d u l e s  and r e q u i r i n g  no fewer than 200 t h y r a t r o n s  for P F N  switching. The s e p t u m  m a g n e t  p o p u l a t i o n , w h e t h e r  d.c. or pulsed, a p p r o c h e s  40.The d es ign of the k i c k e r  s y s t e m s  has been i n f l u e n c e d  by the s h o r t a g e  of s t r a i g h t  s e c t i o n  l e n g t h  and the r e l a t i v e l y  sho rt i n t e r ­bunch i n t erva l of all the PS m a c h i n e s . T y p i c a l l y  n e e d e d  field rise and fall times are 30 - 100 ns and the i m p e d a n c e  l e v e l s are 8 - 30 ohms.The p a p e r  r e v i e w s  the c u r r e n t  f a m i l y  of kickers, i n c l u d i n g  their pulse g e n e r a t o r s , tries to j u s t i f y  the d e s i g n  o p t i o n s  w h i c h  we r e  taken and re l a t e s  the p o s i t i v e  and n e g a t i v e  a s p e c t s  of o p e r a t i o n a l  e x p e r i e n c e  e x t e n d i n g  over 15 years in c e r tain cases. A  s i m i l a r  but b r i e f  a c c o u n t  of the septa is also given, but l i m i t e d  to m a g n e t  c o n s i d e r a t i o n s .I N T R O D U C T I O NK i c k e r  and s e p t u m  m a g n e t  s y s t e m s  have b e c ome a w a y  of life in the PS complex, a c c e p t e d  as mundane p i e c e s  of e l e c t r o t e c h n i c a l  e q u i p m e n t  and e x p e c t e d  to have the same r e l i a b i l i t y  as a n y  o th er a c c e l e r a t o r  c o m p o n e n t . To a lar ge e x t e n t  this l a t t e r  wis h has been met, w hi ch is not to say that k i c k e r s  and septa do not fail but r a t h e r  that they do not fail too often. When m i s f o r t u n e  s t r i k e s  the d a m a g e  can often be s p e c t a c u l a r , w h i c h  s er ves to r em ind that m u c h  of the d e s i g n  is often at the l i m i t s  of the t e c h n i c a l l y  p o s s i b l e .Ove r the p a s t  25 years l e s sons have been l e a rnt from the p r o ­lo n g e d  and a r d u o u s  o p e r a t i o n  to w h i c h  some of the s y s t e m s  h a v e  been subjected. These, t o g e t h e r  wit h the n o w  a v a i l a b l e  p o w e r f u l  and user f r i e n d l y  c a l c u l a t i o n  p r o g ram s, hav e a l l o w e d  a continuous e v o l u t i o n  in d e s ign such that it has become m u c h  e a s i e r  today to build r e l iabl e  s y s t e m s  of p r e d i c t a b l e  p e r f o r m a n c e . This is n o t  to say that at the o u t s e t  the d e s i g n e r  is still often faced with a ran ge of daunting questions typical of w h i c h  are w h a t  v o l t a g e / i m p e d a n c e  level to choose, wha t m a g n e t i c  c i r c u i t  [if any) to use, w h e t h e r  to p l a c e  the k i c k e r  in m a c h i n e  vacuum or not and in the case of septa, w h e t h e r  to use d.c. or pulsed. To all these q u e s t i o n s  there is no s im ple or s in gle a n s w e r  b e c ause they take on a s i g n i f i c a n c e  w h i c h  d e p e n d s  v e r y  m u c h  on the m a c h i n e  e n v i r o n m e n t . S u f f i c e  to say that at the PS we hav e s t r iven to k e e p  the ZQ of our f a s t e s t  k i c k e r s  as hi g h  as p o s s i b l e  [ t y p i c a l l y  25 - 30 Q) and n e v e r  c o n s t r u c t e d  a t r a v e l l i n g  w a v e  k i c k e r  b e l o w  12,5 Q. We hav e also c o n s i s t e n t l y  o pt ed for p l a c i n g  our k i c k e r s  in m a c h i n e  vacuum, even whe n this has m e a n t  p u t t i n g  h u n d r e d s  of k i l o g r a m s  of72f e r r i t e in the ultra h i g h vacuum s y s t e m s  of the A A  and LEAR. We h a v e t e n ded to c h o o s e  two o p e r a t i n g  v o l t a g e  levels, 80 k V  and 40 kV, for c h a r g i n g  our p u l s e  g e n e r a t o r s , the h i g h e r  v o l t a g e  b e i n g  r e s e r v e d  for s i t u a t i o n s  w h e r e  we are d e s p e r a t e l y  sho rt of s t r a i g h t  s e c t i o n  length and w h e r e  k i c k  is m o r e  i m p o r t a n t  than cost. All our g e n e r a t o r s  hav e  use d p u l s e d  r e s o n a n t  c h a r g i n g  b e c a u s e  we q u i c k l y  d i s c o v e r e d  the e n o r m o u s  b e n e f i t  that this type of r a p i d  P F N  c h a r g i n g  h a d  on the rise time of our t h y r a t r o n  s w i t c h e s . The m a g n e t i c  c i r c u i t s  of our k i c k e r s  h a v e  a l w a y s  been bui lt wi t h  n i c k e l  zinc ferrites but we h a v e  use d both C - c o r e  and w i n d o w  frame c o n s t r u c t i o n ,  the f o r m e r  for t r a v e l l i n g  wave m a g n e t s  and the l a t t e r  for l u m p e d  i n d u c t a n c e  magnets. On the s e p t u m  front the u n i f o r m i t y  of a p p r o a c h  is even less obvious, the on l y  two r u l e s  being that septa s h o u l d  n o t  be p l a c e d  in the U H V  machines (where e x t e r n a l  to v a c u u m  m u l t i t u r n  d.c. septa are e m p l o y e d ) and that the h i g h  r e p e t i t i o n  rat e septa for the l e p t o n  p r o g r a m m e  s h o u l d  be d.c.. O t h e r w i s e  we ha v e  a m i x t u r e  of p u l s e d  and d.c. septa i n s t a l l e d in m a c h i n e  vacuum.H i n d s i g h t  a l w a y s  p e r m i t s  the r e d e s i g n  and i m p r o v e m e n t  of e v e r y ­thing that has eve r been m a d e  and the k i c k e r  and s e p t u m  s y s t e m s  of the PS  are no e x c e p t i o n . H o w ever , it is a fac t that o n l y  one of the m o d e r n  [post 1970] s y s t e m s  has had to be r e p l a c e d  b e c ause of p e r f o r m a n c e  w e a k n e s s ; all o th ers are s u b s t a n t i a l l y  still in the f o r m  in w h i c h  they wer e created. The fault s t a t i s t i c s  for 1 9 8 7 ,  the last a v a i l a b l e , s h o w  that i n j e c t i o n  and e j e c t i o n  systems, p r i n c i p a l l y  but n o t  e x c l u s i v e l y  k i c k e r s  and septa, were r e s p o n s i b l e  for o n l y  6X of the total fault time ( i t s e l f  o n l y  81  of the s c h e d u l e d  r u n n i n g  time]. This is not a bad r e c o r d  whe n c o n s i d e r i n g  h o w  close to the limit some of the e q u i p m e n t  is operated.K I C K E R  S Y S T E M  D E S I G N  C O N S I D E R A T I O N SM a g n e t sK i c k e r  m a g n e t s ,  by the v e r y nature of their task, are fast m a g n e t s  h a v i n g  o n l y  s i n gle turn e x c i t a t i o n . M u l t i - t u r n  m a g n e t s  are n o t  a d e s i g n  o p t i o n  w h e r e  r i s e - t i m e s  b e l o w  a f e w  hundreds of n a n o ­s e c o n d s  are c o n c e r n e d .D e s i g n  o p t i o n s  do p r e s e n t  t h e m s e l v e s  w h e n  it comes to c h o o s i n ga] e n v i r o n m e n t  - i n s t a l l e d  in m a c h i n e  v a c u u m- e x t e r n a l  to machine v a c u u mb) type - transmission line of s p e c i f i c ZQ- l u m p e d  i n d u c t a n c ec] a p e r t u r e  - w i n d o w  frame- c l o s e d  C -c ore- open C -c ored) t e r m i n a t i o n  - m a t c h e d- s h o r t - c i r c u i t e d .The a d v a n t a g e  of i n s t a l l i n g  k i c k e r s  in m a c h i n e  v a c u u m  is that a p e r t u r e  d i m e n s i o n s  are m i n i m i s e d , r e d u c i n g  both v o l t a g e  and current for a g iv en k i c k  and rise time. M e t a l l i s e d  c e r a m i c vacuum c h a m b e r s  for c o n t a i n m e n t  of m a c h i n e  vacuum and p r o v i s i o n  of ima ge c u r r e n t  pat h are73avoided. H o w e v e r  the in v a c u u m  m a g n e t  is costly, both in terms of its c o n s t r u c t i o n  and its v a c u u m  tank and pumping. Mo r e  wor ryin g, p a r t i c u ­l a r l y  in machines h a v i n g  s h o r t  h i g h  d e n s i t y  bunches, is the c o u p l i n gim p e d a n c e  whi ch the k i c k e r  p r e s e n t s  to the beam. M a c h i n e  v a c u u m  is a r e l i a b l e  d i e l e c t r i c  and P S  e x p e r i e n c e  has s h o w n  that even w i t h  thever y large c a p a c i t o r  p l a t e  s u r f a c e s  i n v o l v e d  (several tens of m Js t r e s s e s  of 70 k V / c m  can be s a f e l y  a d o p t e d  w i t h o u t  r u n n i n g  into v o l tage c o n d i t i o n i n g  problems. E x t e r n a l l y  m o u n t e d  m a g n e t s  leave open the c ho ice of d i e l e c t r i c ; c o m p a c t  d i m e n s i o n s  can be o b t a i n e d  for t r a n s m i s s i o n  line m a g n e t s  bui lt wit h h i g h  t sol id or l i q u i d  d i e l e c ­tric. B r e a k d o w n  of these d i e l e c t r i c s ,  if it occurs, p r e s e n t s  g r e a t e r  r i s k  than a vacuum f l a s h o v e r  - the sol id d i e l e c t r i c  is l i k e l y  to suffer fatal f a i l u r e  on the first b r e a k d o w n  and the l i q u i d  d i e l e c t r i c , even if s e l f  hea ling , can p r o p a g a t e  s h o c k  w a v e s  w h i c h  can s h a t t e r  the c e r amic v a c u u m  chamber. P e r h a p s  through lack of c o u r a g e  the PS has o pt ed for the "in vacuum" solution.The d e c i s i o n  b e t ween a t r a n s m i s s i o n  line m a g n e t  of sp e c i f i c  Z and a l u m p e d  i n d u c t a n c e  m a g n e t  is m a i n l y  a q u e s t i o n  of rise time. The l a t t e r  has r e a c h e d  o n l y  87X of its ki c k  s t r e n g t h  wh e n  the f o r m e r  is f u l l y  e x c i t e d ; this d i s a d v a n t a g e  can be p a r t i a l l y  o v e r c o m e  by a d d i n g  a s hu nt c a p a c i t o r  to the l u m p e d  m a g n e t  w h i c h  s t e e p e n s  the final sta ge of the rise, p r o v o k i n g  some o v e r s h o o t  it the c a p a c i t o r  is a bo ve the c r i tica l value. H o w  v a l i d  is this a p p r o a c h  d e p e n d s  on the d e g r e e  of o v e r s h o o t  w h i c h  is a c c e p t a b l e  and on the k i c k  l i m its d e f i n i n g  the rise time. L u m p e d  i n d u c t a n c e  m a g nets , u n l ike t r a n s m i s s i o n  lin e ones, are t r a d i t i o n a l l y  i n s t a l l e d  a f t e r  their t e r m i n a t o r , in w h i c h  p o s i t i o n  they are o n l y  s u b j e c t  to v o l t a g e  d u r i n g  the ris e and fall of the c u r r e n t  pulse. This h e l p s  their v o l t a g e  hol d-off, p a r t i c u l a r l y  for long p u l s e s  but e x p o s e s  the m to b i - p o l a r  voltage. A n o t h e r  c o n s e q u e n c e  of l u m p e d  i n d u c t a n c e  m a g n e t s  is that they p r e s e n t  a t r a n sie nt m i s m a t c h  to the pul se g e n e r a t o r , w h i c h  r e s u l t s  in p o s t - p u l s e  r e f l e c t i o n s  at the m a g n e t  if s p e c i f i c measures are n o t  taken to a b s o r b  them at the r e m ote end of the PFN. The P S  p o l i c y  has been to use l u m p e d  i n d u c t a n c e  m a g n e t s  onl y where the p e r m i s s i b l e  rise time has e x c e e d e d  150 ns.The c h o i c e  of Z  f o r  a t r a n s m i s s i o n  line m a g n e t  is a q u e s t i o n  of the a v a i l a b l e  s t r a i g h t  s e c t i o n  length. In g e n e r a l  the h i g h e s t  i m p e ­d a n c e  (say up to 50 Q] s h o u l d  be used c o n s i s t e n t  w i t h  the l e n g t h  a v a i l a b l e  and the c ho sen level of P F N  charging voltage, w h i c h  in turn d e t e r m i n e s  the n u m b e r  of p a r a l l e l  m o d u l e s  needed. The h i g h e r  the Z , the lower the s w i t c h e d  c u r r e n t  and the h i g h e r  the m a g n e t  c u t - o ^ f  frequency. B o t h  are p a r t i c u l a r l y  i m p o r t a n t  e l e m e n t s  in o b t a i n i n g  clean, fast, l o w  f l a t - t o p and p o s t - p u l s e  r i p p l e  kic k pulses. The cut o f f  f r e q u e n c y  f (Fig. 1] d e p e n d s  on the cell i n d u c t a n c e  L and c a p a ­c i t a n c e  C and als o on the e q u i v a l e n t  s e r i e s inductance I of the c a p a ­citor, due in p a r t  to the p h y s i c a l  i n d u c t a n c e  in the c a p a c i t o r  branch and in p a r t  r e p r e s e n t i n g  the m u t u a l  c o u p l i n g  b e t ween cells. W h i l s t  in theory the fall of t with r e d u c i n g  Z can be a r r e s t e d  by r e d u c i n g  the cell d i m e n s i o n s , in p r a c t i c e  this 2ea§s to wafer thin c a p a c i t o r  p l a t e s  and an i m p o s s i b l e  m e c h a n i c a l  c o n s t r u c t i o n  as two v a c u u m  gap s m u s t  still be m a i n t a i n e d  per cell. T y p ical f values as a f u n c t i o n  of Z for the type of c o n s t r u c t i o n  used in the Ps are l i s t e d  in Fig. 1. 0A p e r t u r e  c o n s i d e r a t i o n s  are v e r y  m u c h  i n f l u e n c e d  by w h e t h e r  or not a hig h p m a g n e t i c  c i r c u i t  is used. P S  p r a c t i c e  is a l w a y s  to use a74l /2 . L/2. J im s.— *— mat—*-iI: c------0 KitvCE*. JTe.■f -  ' 1T  J(L+^L,)C 2 4 5*oIS 1 4 - 2 2----- ° *Fig. 1 E q u i v a l e n t  c i r c u i t  of s i n gle m a g n e t  cell.m a g n e t i c  circuit bec ause failure to do so c o n s i d e r a b l y  i n c r e a s e s  thee f f e c t i v e  v e r t i c a l  ape rture, l o w e r i n g  in c o n s e q u e n c e  the Z q and m a k i n gc o n trol of the field q u a l i t y  in the useful a p e r t u r e  m o r e  difficult.H a v i n g  d e c i d e d  on a m a g n e t i c  c i r c u i t  it r e m a i n s  to c h o o s e  its f o r m  andthe s p e c i f i c  mat eria l. Both w i n d o w  frame and C -c ore s e c t i o n s are usedin the PS, the f o r m e r  r e s t r i c t e d  to l u m p e d  i n d u c t a n c e magnets and thel a t t e r  a l w a y s  a d o p t e d  for t r a v e l l i n g  wav e d e s i g n s  and o c c a s i o n a l l y  forthe others. C - c o r e  c o n f i g u r a t i o n s  u s u a l l y  h a v e  the a p e r t u r e  c l o s e d  bythe return c o n d u c t o r  but this is not p o s s i b l e  where the bea m has to bes w e p t  into or out of the a p e r t u r e  by R F  g y m n a s t i c s . In this case ther et urn c o n d u c t o r  is l o c a t e d  abo ve and b e l o w  the aperture or evenp l a c e d  a r o u n d  the b a c k l e g  but wit h a 10X or so p e n a l t y  in i n d u c t a n c e .The m a g n e t i c  m a t e r i a l  is i n v a r i a b l y  n ic kel zinc ferrite, but thiscomes in a v e r y  wid e v a r i e t y  of g r a d e s  w i t h  d i f f e r e n t  m a g n e t i c  ando u t g a s s i n g  p r o p e r t i e s . S l o w l y  a c q u i r e d  e x p e r i e n c e  in the PS has shownthat for the c u r r e n t  p u l s e  rise times w h i c h  we e m p l o y  (in turn l i m i t e dby w h a t  we can get out of the t h y r a t r o n  s w i t c h e s J  and t y p i c a l l y  nof a s t e r  than 17 ns (10-90X), f e r r i t e  wi t h  a p of a r o u n d  1000 can trackthe e x c i t a t i o n  pul se wit h a n e g l i g i b l e  d e l a y  of a n a n o s e c o n d  or so.G r a d e s  w h i c h have found f a v o u r  are I n d i a n a  H 2 , P h i l i p s  8C11 andC e r a m i c  M a g n e t i c s  CMD 5005. The one m o s t  e x t e n s i v e l y  used is that ofPhilips, and e x c l u s i v e l y  so for the U H V  mac hines. As a m a t t e r  ofr o u t i n e  all f e r r i t e  ( f u l l y  m a c h i n e d )  is n o w  v a c u u m  fired at 1000°Cp r i o r  to a s s e m b l y  into magnets. The as i n s t a l l e d  o u t g a s s i n g  rat e is10 Torr l i t r e / s e c / c m  . L o w  H  is r e q u i r e d  to minimize the r e m a n e n tfield, t y p i c a l l y  0,2 O e r s t e d s  for B  of 300 0 Gauss. This h o l d s  thefB dl of m o s t  of our k i c k e r  s y s t e m s  b e l o w  0,5 Gauss-meter. remThe final option, not so m u c h  of magnet de s i g n  but r a t h e r  of system c o n f i g u r a t i o n , is w h e t h e r  or not the m a g n e t should be t e r m i ­n a t e d  or short circuited. S h o r t - c i r c u i t i n g  is the u l t i m a t e  m e a s u r e  for d e a l i n g  wit h a space shortage. It a l l ows the same k i c k  with the same ris e time to be obtained f r o m  h a l f  the space or a l t e r n a t i v e l y  the Z to be d o u b l e d  in the same space. The pri ce to p a y  is in pul se g e n e ­r a t o r  c o m p l e x i t y  b e c a u s e  d o u b l e  end ed P F N  s w i t c h i n g  is required, one s w i tch h a v i n g  to be b i - d i r e c t i o n a l . The i n a b i l i t y  of the s w i t c h e s  to r a p i d l y  t r a n s m i t  the m a g n e t  g e n e r a t e d  c u r r e n t  wav e u s u a l l y  r e s u l t s  in some small post-pulse reflections, r e n d e r i n g  this a p p r o a c h  more, a t t r a c t i v e  for e j e c t i o n than i n j e c t i o n  schemes. N e v e r t h e l e s s  it has been s u c c e s s f u l l y  a p p l i e d  to both in the a n t i p r o t o n  c o l l e c t o r  of the PS, spa ce s h o r t a g e  obliging.75P u l s e G e n e r a t o r sD e s i g n  c o n s i d e r a t i o n s  for p u l s e  g e n e r a t o r s  fall b r o a d l y  int o two classes: c o n f i g u r a t i o n  and c o m p o n e n t s . The c o n f i g u r a t i o n  p o s s i b i l i t i e s  often d e p e n d  on the m a g n e t  o p t i o n s  and the k i c k  g y m n a s t i c s  required.The s i m p l e s t  c o n f i g u r a t i o n  is that of a c ab le PFN, c h a r g e d  to d ou ble the needed p u l s e  voltage, s w i t c h e d  int o a t e r m i n a t e d  t r a n s m i s ­sion line magnet. The s w i t c h  is f u l l y  f l o a tin g; if an o r d i n a r y  thyra-tron it can be d a m a g e d  by i n v e r s e  c u r r e n t  f r o m  a loa d short. A variant, f r e q u e n t l y  used at the PS, is to add a r e m o t e  end t e r m i n a t o r  and s wi tch w h i c h  has the a d v a n t a g e s  of p e r m i t t i n g  p u l s e  l e n g t h  control (partial e x t r a c t i o n ) , i m p r o v i n g  the f a l l - t i m e  (by p r e - d i s t o r t i o n  if n e c e s s a r y ]  and l i m i t i n g  s w i t c h  d am age f r o m  a loa d short. A  l e s s e r  a l t e r n a t i v e  is to d ou ble the P F N  l en gth and c o n n e c t  a m a g n e t  load to each end. Cab le P F N ' s  f u r n i s h  r i p p l e  free p u l s e s  but l o w  a t t e n u a t i o n  is e s s e n t i a l  if d r o o p  and "cable tail" are to be w i t h i n  a c c e p t a b l e  l im its for l o n g  pulses. Attentuation is a d v e r s e l y  a f f e c t e d  if semicon­ductors are a d d e d  at the d i e l e c t r i c  b o u n d a r i e s  to i m p r o v e  vol tage w i t h s t a n d . The PS s o l u t i o n  has been to use S F  p r e s s u r i s e d  P E  tapecables w i t h o u t  s e m i - c o n d u c t o r s  for the 80 k V  systems. T y p i c a l l y  droop is l i m i t e d  to IX on a 2.7 ps 26 Q pulse, w i t h  no H V  f a i l u r e  in 10 years service.The cable P F N  c ea ses to be a t t r a c t i v e  for p u l s e s  e x c e e d i n g  abo ut 3 ps on a c c o u n t  of cost, bulk and the d r o o p / t a i l  problems. The a l t e r ­n a t i v e  is the lum ped e l e m e n t  P F N  with R - L - C  head cell to i m p r o v e  the ini tial rise. So e q u i p p e d  the lumped element line can equal the cable for rise but the fall can n e v e r  be m a d e  fast e no ugh for injection a p p l i c a t i o n s  w i t h o u t  r e c o u r s e  to a s h o r t i n g  c l i p p e r  s w i t c h  on the ki c k e r  t r a n s m i s s i o n . Such a c l i p p e r  c r e ates m u l t i p l e  r e f l e c t i o n s  w i t h i n  the l u m p e d  e l e m e n t  lin e w h i c h  in turn m a y  r e q u i r e  a thirdswi tch and t e r m i n a t o r  at the r em ote end for their a b s orpt ion. Thus w h i l s t  for e j e c t i o n  a p p l i c a t i o n s  a l u m p e d  e l e m e n t  line w i t h  s in gle swi tch is p e r f e c t l y  s a t i s f a c t o r y , at lea st two if not three sw i t c h e s  are n e e d e d  w h e n  the task is i n j e c t i o n . C l i p p e r  s w i tch rise time n e e d s  special care b e c ause it has to h a n d l e  twice the m a g n e t current. A f e w  ex a m p l e s  of both one and three s wi tch l u m p e d  e l e m e n t  P F N ' s  e x i s t  in the PS, r e s e r v e d  for l o w  Z ( d o w n  to 8 Q) and lo n g  p u l s e  (up to 24 ps ) a p p l i c a t i o n s .An a l t e r n a t i v e  f o r m  of e n e r g y  sto re is the Bliimlein a r r a n g e m e n t  w hi ch has the v i r tue of g e n e r a t i n g  p u l s e s  of v o l tage equal to the P F N  c h a rge voltage. H o w e v e r  it r e q u i r e s  c a b l e s  of h a l f  the Z of the loa d and the c l o s i n g  swi tch has to h a n d l e  twice the m a g n e t  current. It is a current a g a i n s t  v o l t a g e  t r a d e - o f f  w i t h  r e s p e c t  to the s i m p l e  cable PFN. It is not used at the PS b e c ause of the i n c r e a s e d  thy ratr on s w i t c h i n g  time w h i c h  w o u l d  result.Some of the a d v a n t a g e s  of the Bliimlein s y s t e m  but w i t h o u t  its d i s a d v a n t a g e s  can be o b t a i n e d  by incorpora ting the t r a n s m i s s i o n  line m a g n e t  as p a r t  of a cable P F N  w i t h  a s h o r t i n g  s w i t c h  at one end and a te r m i n a t o r  and s e c o n d  s wi tch at the other. The rise time is that due to a single m a g n e t  p r o p a g a t i o n  and the p u l s e  v o l t a g e  is the P F N  c ha rge voltage. The fall time c a n n o t  be s h o r t e r  than two m a g n e t  p r o p a g a t i o n s . A n  a d d i t i o n a l  p e n a l t y  is that the magnet must withstand the P F N  charge76v o l t a g e  p r i o r  to the p u l s e  and s u f f e r  p a r t i a l  v o l t a g e  i n v e r s i o n  d u r i n g  it. This a r r a n g e m e n t  is p a r t i c u l a r l y  a t t r a c t i v e  whe re spa ce is l i m i t e d  and w h e r e  o n l y  ri s e  time is important. It is used in the P S  for theB o o s t e r  e x t r a c t i o n  and r e c o m b i n a t i o n  kickers.C e r t a i n  e j e c t i o n  s c h e m e s  at the P S  h a v e required the g e n e r a t i o n  of s t a i r c a s e  w a v e f o r m s  and sho rt i n t erva l p u l s e trains. T h e s e  are not of i n t e r e s t  for the K A O N  f a c t o r y  and will n o t  be f u r t h e r discussed e x c e p t  to say that e n t i r e l y  s a t i s f a c t o r y  r e s u l t s  in both cases can be o b t a i n e d  by the d i s c h a r g e  of s e r i a l l y  c o m b i n e d  P F N ' s  and t h y r a t r on switches.On the c o m p o n e n t  side, p r i n c i p a l  a t t e n t i o n  has to be p a i d  to the c ho ice of hig h v o l t a g e  switches, r e c h a r g i n g  p o w e r  s u p p l i e s  and t e r m i n ­a t i n g  r e s i s t o r s . P r e s e n t - d a y  p r a c t i c e , f u l l y  j u s t i f i e d  by results, is to use t h y r a t r o n  s w i t c h i n g throughout. Care m u s t  be taken in tubes e l e c t i o n , p a r t i c u l a r l y  in circuits p r o n e  to inverse c u r r e n t  in oftr e p e a t e d  f au lt c o n d i t i o n s . T o d a y  there e x i s t s  an e x t e n s i v e  range ofb i - d i r e c t i o n a l  t h y r a t r o n s , e i t h e r  of the double c a t h o d e  or h o l l o w  a no de type, c a p a b l e  of s a f e l y  h a n d l i n g  i n v e r s e  current. The small cost i n c r e a s e  w h i c h  the y r e p r e s e n t  is oft en an e x c e l l e n t  i nv estm ent. Tube r a t i n g s  nee d to be r e g a r d e d  wit h a c e r t a i n  con ser v a t i s m ,  p a r t i c u l a r l y  the v o l t a g e  r a t i n g  of m u l t i s t a g e  tubes w h i c h  h a v e  to be p u s h e d  to m a x i m u m  dl/dt. PS e x p e r i e n c e  is that in goo d h o u s i n g s  w i t h  c o r r e c t l y  d e s i g n e d  c i r c u i t r y  and t r i g g e r i n g an a v e r a g e  d l / d t  of 100 A / n s can be r e a d i l y  o b t a i n e d  and hel d b e t w e e n  the 10 and SOX points, even for the 80 k V  a p p l i c a t i o n s . As m u c h  as 150 A / n s  is p o s s i b l e  in 6.25 Q c i r c u i t s  o p e r a t i n g  at 40 k V . R e p e t i t i o n  rates of 100 Hz hav e shown no a d d i ­tional pro blems. Tube l i f e t i m e  in our m o d e s t l y  {low Hz) rep. rated s y s t e m s  a v e r a g e s  mo r e  than 2 0 0 0 0  f i l a m e n t  hours. J i t t e r  is g r e a t e s t  on m u l t i - s t a g e  tubes but still under 5 ns, i n c l u d i n g  the t r i g g e r i n g  system. S l o w  d r i f t  is e a s i l y  s t a b i l i s e d  by s u i t a b l e electronics. The most used tubes in the PS s y s t e m s  are the C X  1171 and its v a r i a n t s  forthe 80 k V  s y s t e m s  and the C X  1154 and its v a r i a n t s  for the 40 k Vsystems. G l a s s  C X  1158 tubes are use d up to 30 kV. A l m o s t  w i t h o u te x c e p t i o n  the tubes are oil i m m e r s e d , o ft en f o r ced cooled.P o w e r  s u p p l i e s  are, w i t h o u t  e xc epti on, of the p u l s e d  r e s o n a n t  type p e r m i t t i n g  P F N  r e c h a r g e  in a f e w  ms. Fas t r e c h a r g i n g  has a ver y f a v o u r a b l e  i n f l u e n c e  on the se l f  f i r i n g  f r e q u e n c y  of t h y r a t r o n s  at any g i v e n  r e s e r v o i r  s e t t i n g  and is a n e c e s s i t y  if the p r e v i o u s l y  q u o t e d  d l / d t  v a l u e s are to be o b t a i n e d . The P S  s u p p l i e s  use a ste p up t r a n s ­f o r m e r  as the r e s o n a t i n g  inductor. Core bias, which i n f l u e n c e s  r e c o v e r y , is a p p l i e d  t h r ough a t e r t i a r y  winding. H V  d i o d e s  w h e n  f i t t e d  b e t w e e n  t r a n s f o r m e r  o u t p u t  and PFN, i m p r o v e operational f l e x i b i l i t y  in d e c o u p l i n g  the p o w e r  s u p p l y  and t h y r a t r o n  t r i g g e r timing. This is a l w a y s  done at the PS. The p r i m a r y  e n e r g y  store is oft en a large e l e c ­tr o l y t i c  c a p a c i t o r  r u n n i n g  at a b o u t  200 V.T e r m i n a t i n g  r e s i s t o r s  can be of the e l e c t r o l y t i c  or c a r bon m a s s  type, the l a t t e r  of disc or t u b u l a r  form. A t  the P S  we h a v e  s t a n d a r ­di s e d  on the disc c a r b o n  m a s s  type, oil i m m e r s e d  wit h f or ced oil cooling. S t a b i l i t y ,  p a r t i c u l a r l y  in a r a d i a t i o n  e n v i r o n m e n t , is not e x c e l l e n t  and r e b u i l d i n g  of r e s i s t o r  s ta cks r e p r e s e n t s  a m a j o r  p r o p o r ­tion of our m a i n t e n a n c e  effort. P r o b a b l y  the t u b u l a r  type is m o r e  sta ble but c o n t a c t  p r o b l e m s  are m o r e  severe. The ideal t e r m i n a t o r  is yet to be found.77P S  K I C K E R SH i s t o r i c a l  b a c k g r o u n dThe f i r s t  small aperture, hydraulically a c t u a t e d  k i c k e r  became o p e r a t i o n a l  in the PS ring in M a y  1963. It had an aperture of 5 x 3 cms and ris e time fast e n o u g h  to eje ct c l e a n l y  one bunch. E x c i t a t i o n  was fro m e i t h e r  sho rt or lon g spark gap s w i t c h e d  l u m p e d  e l e m e n t  lines, l o c a t e d  i n s i d e  the m a c h i n e  tunnel. It remained operational until 1968 whe n it was r e p l a c e d  by a n o t h e r  p l u n g i n g  m a g n e t  of use ful a p e r t u r e  2,0 x 2,2 cms e x c i t e d  f r o m  r e m o t e l y  positioned lumped element l i n e s  with main, dum p and c l i p p e r  spark gap s for full control of pul se length. In 1974 this e q u i p m e n t  was r e p l a c e d  by the present full aperture k i c k e r  system. D u r i n g  1964-9 a sin gle m o d u l e  p u s h - p u l l  e x c i t e d  (± 120 kV)full a p e r t u r e  d e v ice was d e v e l o p e d . Its f e r rite circuit, s u i t a b l y  i m p r e g n a t e d  wit h e p o x y  resin, f o r m e d  p a r t  of the c o n t a i n m e n t  for m a c h i n e  v a c u u m  and also furnished c a p a c i t a n c e  for the d e l a y  line magnet. W h i l s t  it was s u c c e s s f u l l y  t es ted w i t h  bea m it was a b a n d o n e d  in 1969 b e c ause it cou ld not s a t i s f y  the t i g h t e n i n g  PS v a c u u m  s p e c i f i ­cat ion and was c o n s i d e r e d  an oil h a z a r d  in the eve nt of h i g h  v o l t a g e  breakdown. At this p o i n t  d e v e l o p m e n t  s t a r t e d  on the present PS full a p e r t u r e  kicker, f r o m  w h i c h  m o s t  of the oth er k i c k e r s  h a v e  since evolved.P r e s e n t  s i t u a t i o nTable 1 lis ts the r a t i n g s  of the present k i c k e r  p o p u l a t i o n  of the PS complex. C o m m e n t  will be r e s t r i c t e d  to the o l d e s t  hig h v o l tage s y s t e m  w h i c h  is the 12 m o d u l e  full a p e r t u r e  k i c k e r  of the P S  ring. C o m m i s s i o n e d  in 1973, this k i c k e r  was used i n i t i a l l y  for m u l t i p l e  pa r tial e x t r a c t i o n s  for b u b ble c h a m b e r  p h y s i c s ; in l a t e r  times it has  s e r ved as the e j e c t i o n  k i c k e r  for p r o t o n s  for p p r o d u c t i o n  and SPS  f ix ed tar get physics, and m o s t  r e c e n t l y  for leptons. It also r e i n j e c t s  p into the PS r i n g  f r o m  the A c c u m u l a t o r . The k i c k e r  has a six shot per cycle c a p a b i l i t y  with m i n i m u m  i n t erva l of 30 ms. Bot h kic k a m p l i ­tude and d u r a t i o n  can be f r e e l y  v a r i e d  f r o m  shot to shot. The p u l s e  g e n e r a t o r  is a gas p r e s s u r i s e d  P F N  cable with CX1 171 s w i t c h i n g  at e i t her end. T r a n s m i s s i o n  d i s t a n c e  b e t w e e n  g e n e r a t o r s  and m a g n e t s  is 170 m, m a i n l y  in gas p r e s s u r i s e d  cable but wit h final c o n n e c t i o n s  in solid P E  cab le to f a c i l i t a t e  m a i n t e n a n c e . To date each m o d u l e  hasQp u l s e d  well in e x c e s s  of 10 times, the s t a n d a r d  c ha rge v o l t a g e  being80 kV. There h a v e  been no s e r i o u s  hig h v o l t a g e  f a i l u r e s  in the p ul seg e n e r a t o r s  and the m a g n e t  v a c u u m  tanks h a v e  n e v e r  been opened. P r i n ­cipal w e a k n e s s e s  have been the f l e x i b l e  coa xial cab les for final c o n ­n e c t i o n  of the t r a n s m i s s i o n  l i n e s  and t e r m i n a t o r  stability. R e c h a r g i n g  po w e r  s u p p l i e s  u s i n g  315/1 step up t r a n s f o r m e r s  and 210 V e l e c t r o l y t i c  p r i m a r y  s t o r a g e  h a v e  been t o t a l l y  t r o u b l e - f r e e .PS S E P T AT ab le 2 l i s t s  the r a t i n g s  of the septa c u r r e n t l y  in use in the PScomplex. The m i x t u r e  is m a i n l y  of m u l t i t u r n  d.c. and s i n gle turn78p u l s e d  m a g nets , wit h a p r e d o m i n a n c e  for "machine v a c u u m " i n s t a l l a t i o n  d e s p i t e  the c o n s i d e r a b l e  gas load w h i c h  results. M a g n e t i c  c i r c u i t s  are f e r r i t e  or l a m i n a t e d  steel for the p u l s e d  m a g n e t s ; certain d.c.Tab le 1. R a t i n g s  of p r e s e n t  PS c o m p l e x  kickers.Machine Application Type andZn (Q)0Aperture w x h(mm)N o . of ModulesTotal jBdl (Gauss-m)PFN Voltage (kVJKickRise(ns)(5-95X)Fall(ns)(ns)Flat top (ps)Booster Ejection Delay line 25 Q115 x 7V 4 498 40 S3 55 38 0,625Transfer Delay line12.5 Q70 x 115 2 5 67 40 52 54 36 0.6251.25PS Ring InjectionipiDelay line 26.3 Q150 x 53 4 304 80 _ 39 30 2. 65Injection (e* e )Delay line 15. 7 Q112 x 74 1 283 80 87 90 78 1.80Fast ejection Delay line 15 0147 x 53 12 1680 80 68 70 55 0.10-2.10Fast ejection p LEAR onlyLumped L25 Q153 x 70 1 210 33 240 - “ 0.60-6.81ContinuousTransferLumped L 25 Q and 8.3 Q158 x 52 1 * 1 543 Up to 40400 12.0Ditto Delay line 15 Q150 x 53 1 450 80 160 ~ 165 4.20LEAR Inj action Delay line 15. 7 Q150 x 66 2 430 80 97 100 93 0.85Injection Bumper (H )Lumped L 15 Q135 x 85 1 120 33 860 2800 ~ 0.20-28. 0AntiprotonRingsCollectorInjectionDelay line 15 Q Short circuited140 x 72 140 x OS- 140 x 352222642 80 185 1868272830. 60CollectorEjectionDelay line 15 Q Short circuited250 x 100 232 x 30221654 80 2202502202501031200.56AccumulatorInjectionDelay line 15 Q132 x 45 4 1250 60 120 120 106 0. 56AccumulatorEjectionLumped L 15 Q60 x 23 1 780 80 165 ~ 0. 56Electron Positron AccumulatorInjection Delay line 30 Q110 x 35 2 90 40 36 38 24 0.05 (100 Hz)Ejection Delay line 30 C100 x 35 1 45 40 36 23 0.05(x8)m a g n e t s  use sol id cores, o t h e r s laminated, p a r t i c u l a r l y  for s t o r a g erings. Until a b o u t  5 yea rs ago l a m i n a t e d  c i r c u i t s  c o m p r i s e d  e p o x yg l u e d  stacks. R e c e n t  p r a c t i c e  has been to use tra nsil steel, an i n o r g a n i c insulated m a t e r i a l ,  in p r e s s u r e  h e l d  s t a c k s  w h i c h  can be v a c u u m  bak ed in sit u up to 150°C. The u l t i m a t e  o u t g a s s i n g  of this n e w e r  c o n c e p t i o n  is lower but b a k e o u t  is essential for r e a s o n a b l e  pump down time.M o s t  P S  sep ta h a v e  p e r f o r m e d  e x t r e m e l y  well and the a v e r a g e  se r v i c e  life of a m a g n e t  is p r o b a b l y  a r o u n d  10 years. The m o s t  s i g n i ­f i c a n t  d i f f i c u l t y  has been b l o c k a g e  of d.c. septa by c o p p e r  oxided e p o s i t s from the i n t e r a c t i o n  of d i s s o l v e d  o xy gen in the d e m i n e r a l i s e d  c o o l i n g  w a t e r  w i t h  their c o p p e r  c o n d u c t o r s . A  s o l u t i o n  h a s  been found79by c o o l i n g the m o s t  d i s s i p a t i v e  d.c. septa wi t h  l o w  o x y g e n  (< 100 ppb) d e m i n e r a l i s e d  water. T y p ical c u r r e n t  d e n s i t y  and w a t e r  v e l o c i t y  in these m a g n e t s  are 70 A / m m  and 12 m/s r e s p e c t i v e l y . There is o n l y  one r e c o r d e d  case of coil f a i l u r e  fro m cavitat ion, p e r h a p s  3 or 4 from oxide b l o c k a g e  a l t h o u g h  p r i o r  to the i n t r o d u c t i o n  of l o w  o xy gen w a t e rTable 2. R a t i n g s  of p r e s e n t  PS c o m p l e x  septa.Machine A ppl i c a t i o n Type No. in s e r vi c eSBdl IT.mlMax.I(kA)No.ofturnsLength(mm)Gap w (mm xx h mm]Booster D i s t r i b u t o r P.V.F 4 0.004 0. 5 1 354 98 50I n j e c t i o n P.V.F 1 0.172 20 1 860 32 112P. V.F 2 0. 136 15 1 860 32 112P. V.F 4 0.071 4 1 710 60 40E j ec t i on DC.V.SS 4 0. 220 4 1 11 70 80 24.5Transfer DC.V.SS 3 0.3S0 2.0 12 950 100 60P.V.F 2 0.013 2. 1 1 400 70 60PS Ring I n j e c t i o n DC.V.SS 1 0.2 5 1.6 12 700 100 60I n j  ect i on e P.V.TS 1 0.388 13. 3 1 400 70 18I n j e c t i o n  e P.V.TS 1 0.388 13. 3 1 400 70 18Slow Ext r ac t i o n P.V.LS 1 0.4 IS 12 1 831 59 30P.V.LS 1 1. 01 12 2 1010 39 30P.V.LS 1 0. 1 2. 1 1 931 50 25Ej  ect ion P.V.LS 2 1. 266 27 1 1120 53 30E j e c t i o n  p P.V.F 1 0. 118 3 1 700 70 60f o r  LEARE j e c t i o n  p P.V.LS 1 1. 482 38 1 988 62 25f o r  SPSLEAR I n j e c t i o n / DC.A.LS 1 0.363 1.8 10 900 155 50Ej ect ion DC.A.LS 1 0.244 1.8 20 400 135 74Antiproton C o l l e c t o r P.A.LS 1 1. 665 38 2 1714 119 90Rings I n j  ect ionC o l l e c t o r P.V.TS 2 0. 77 27 1 85 3 75 30Ej ect ionAccumulator DC.A.LS 2 0.6174 4 10 1001 296 76I n j  e c t i o n /E j e c t i o nElectron I n j  ect i on DC.A.LS 2 0.2862 2. 72 4 519 91 25Posi tron DC.A.LS 2 0. 7842 2. 72 12 519 91 25Accumulator E j ec t i o n P.V.TS 1 0.53 18.6 1 400 70 18P = Pulsed A = A i r  mounted F = F e r r i t e  LS = Laminated s t eelDC = Continuously powered V = In machine vacuum SS = S o l i d  s t e e l  TS = T r a ns i l  s t e e lf r e q u e n t  r i n s i n g  wit h s u l f a m i c  aci d was needed. All p u l s e d  septa areo p e r a t e d  at l o w  r e p e t i t i o n  rat e (< 1 Hz) and p r e s e n t  no c h a r a c t e r i s t i cw e a k n e s s e s  a l t h o u g h  those wit h m o s t  w a t e r  c i r c u i t  j o i n t s  w i t h i n  the va c u u m  e n v e l o p e , a p r i n c i p l e  to be a v o i d e d  if p o s s i b l e , have been l e a s t  reliable.C O N C L U S I O NThe range, p e r f o r m a n c e  and r e l i a b i l i t y  of the C E R N  P S  septa and k i c k e r s  m a y  s er ve as e n c o u r a g e m e n t  for the k a o n  f a c t o r y  project. H o w e v e r  the PS c o m p l e x  is a s l o w  c y c l i n g  facility.80New stripekicker in the injection chain of HERAJ.Riimmler DESY The dats of this paper were presented at the Magnet Design Workshop at Triumf in Vancouver.1. Summary2. New stripekicker at DESY3. cross-section of an stripekicker4. Stripekicker for rectangular pulses and shortpilot pulses for tests.5. Stripekicker for sinewave pulses.The injectionchain into HERAe-Kicker 3e-K icker 2(figure 1) (figure 2) (figure 3)(figure 4) (figure 5)811. Summary (figure 1)In the HERA injectionchain for the inj- and ejections many stripekickers are used to switch the beam. The kicker in PETRA must switch in an very short time. Ferrite kickers with stripes meet all the requirements.During my talk I showed a lot of kickers and septa in detail. Many aspects of the constructions can give problems in the complete injektion or ejektion system.I concentrate here on good stripekicker technic.2. New stripekicker at DESY (figure 2)Stripekickers work in an wide frequence range. Thyratrons can change rectangular pulses (in PETRA ) from 7.5uS, the PETRA beam, to single short pulses for the pilot bunch ejection to test the transport lines.All the kickers in the HERA elektron injektionchain have been tested and work well.The kicker idea is to have a kicker chamber which leads the rf like a vakuum chamber.The wall of the chambers is horizontal with the outer sideclosed; metal leads the rf through the kicker without re­flections and can also absorb synchrotronradiation.Longitudinal stripes on the top and bottom of the chamber lead- the rf and protect the ferrites from the rf.Stripes are connected alternately on the left and right and the parallel capacities between the stripes close the chamber for the rf.But for the quick rising kickerfeld the stripe capacities are in series, so that the kickerfeld is not shorted.The current lead of the kicker is tapered at both ends to minimize rf reflection.There is much more to see.HV betwen the stripes.HV betwen stripes and ferrite.Capacities betwen the stripes.Rf reflection mesurements over a wide frequency range.And there are stripekickers for sinewave pulses, and rectangular pulses.After HERA is already build I give more information.8S&8C82f i t r i  n p k icross- section of a stripekicker figure 3Pefra  e~ EjeKtion 3 Kicker mif Pulser84J.RummlerDATE :jun 29/8785J-3 S87OVERVIEW OF KICKER MAGNET SYSTEMS AT THE SPS AND LEPV.Rodel, G.H.Schroder CERN, Geneva, SwitzerlandABSTRACTThe 450 GeV Super Proton Synchrotron (SPS) and the Large Electron Positron Collider (LEP) at CERN are equipped with more than 40 kicker magnet systems containing about 100 high power thyratrons for injection, ejection or dumping of the various types of accelerated particles (pro­tons, antiprotons, heavy ions, electrons and positrons). Depending on their particular application the magnets and pulse generators are built according to different design criteria.After a short overview of the different systems the largest device, the SPS proton and positron inflector (12 independent kicker magnets, kick strength 0.43 Tm, kick rise time 145 ns, pulse duration 1 to 12ps) is described in more detail.INTRODUCTIONThe Super Proton Synchrotron (SPS) at CERN accelerates proton beams of about 3.1013 particles to 450 GeV for fixed target physics. It also operates as a proton-antiproton collider at 315 GeV with a lumi­nosity of >1030 cm-2s- 1 . The SPS is furthermore used as pre­accelerator of the Large Electron Positron Collider (LEP), accelerating short electron and positron bunches from 3.5 GeV to 20 GeV. LEP will start operation in July 1989 and will initially run at an energy of about 55 GeV per bunch.The SPS with its different operation modes needs various kicker mag­net systems for injection, ejection, beam dumping and tune measurements. LEP requires 2 injection systems for electrons and positrons consisting each of 3 kicker magnets and a fast septum kicker magnet. All kicker sy­stems differ considerably in their specifications with respect to de­flecting power, repetition rate, kick flat-top duration, kick rise and fall times and beam apertures resulting in systems of rather different design.This paper gives an overview of the kicker magnet systems at the SPS and LEP. The SPS proton and positron inflector is then treated in more detail as it comes closest to the future requirements of TRIUMF. Detailed descriptions of the various systems can be found in references1-8.OVERVIEWThe kicker magnet systems at the SPS and LEP can be divided into 2 groups employing rather different technologies: hadron kickers and lepton kickers.Table I . Hadron KickersProtonInflectorAntipro­ton In­flectorFastEjectionBeam dumping vert. hor.Q-measurement vert. hor.Number of magnets12 5 7 2 3 1 1Kick strength per magnet (mT.m)36 39 197 381 667 40 40Magnetimpedance(Q)12.5 12.5 10 2 (0.24) 6.25 6.25Kick flat- top duration (Vis)1 to 12 0.2 1 to 23 23 (quasi-sinu-soidalpulse)23 23Kick rise time (2to98%) (Vis)Kick fall time (98to2%) ( p s )type S: 0.145 type L: 0.2200.6900.2200.2201.111 23 1 1 1 1Repetitionrate5 pulses 0.65 s apart within 15 s2.4 s 3 pulses 0.1 s apart within4 s15 s 15 s 4 pulses 0.1 s apart within 4sMax. genera­tor voltage (kV)60 60 60 60 12 30 30Max. magnet current (kA)2.4 2.4 3 15 46 2.4 2.4Max. stored energy per pulse (kJ)Switch type:1.9 0.03 2.3 11.3 3.7 2.0 2.0Thyratrons CX1171B CX1171A CX1171B CX1171B - CX1154Ignitrons - - - BK7703 BK488 —Number of magnet cells22 22 7 5steel 0.2 mm1 2Yoke material ferrite ferrite ferrite ferrite ferriteMagnetcapacitorlocationvacuum vacuum oil oil oil oil89The (older) hadron kickers have large oil insulated pulse generators charged up to 60 kV producing quasi-rectangular pulses of fast rise and/or fall time and low repetition rate. The ferrite magnets are of the delay line type and housed in vacuum tanks. The stored energy per pulse is considerable, amounting to more than 11 kJ in the case of the vertical beam dumping system. The main parameters are listed in Table I.The (newer) lepton kickers operate with bursts of half sine wave pul­ses of very short repetition time produced in complex air insulated ge­nerators and resonant charging systems. The magnets are unmatched in­ductances either housed in the accelerator vacuum (in the SPS) or in air around metallized ceramic vacuum chambers (in LEP). The main parameters are given in Table II.Hadron KickersFor fixed target proton operation the whole SPS machine circumference is filled with particles except for 2 short gaps required for the magnetic field rise of the kicker magnets. As the hadron kickers shall give a constant deflection during up to 1 turn of the circulating beam, they must produce quasi-rectangular pulses with short rise and fall times. As a consequence these systems are built as matched travelling wave structures including delay line type kicker magnets. Depending on the rise times, different methods for matching of magnets and generators have been used.a) For short rise times (< 150 ns) the pulse forming networks (PFN's) and the magnets are composed of many cells. In the case of the proton in- flector which produces a kick rise time of 145 ns the PFN is built of 35 cells and the magnet of 22 cells. As a short rise time requires a high characteristic impedance of the system, the matching capacitors of the magnet are sufficiently small enough to be formed by parallel plates interleaved with the ferrite blocks using the vacuum as a dielectric. The plate size is still such that the vacuum tank has an acceptable diameter1 . This type of construction limits stray inductances in series with thecapacitors and results in a high cut-off frequency of the magnet.b) For longer rise times (< 1.5 ps) the magnets are subdivided in only 5 to 7 cells and their characteristic impedance is chosen lower re­sulting in a high kick strength for a given generator voltage3. Examples of this type are the vertical beam dumping and the fast ejection systems. The beam dumping system has a rise time of 1 ps and an im­pedance of 2 Q with 5 cells per magnet. The matching capacitors are nowrather large and therefore housed in oil-filled boxes under the vacuumtank. The unavoidable stray inductances in series with the capacitors produced by the vacuum feedthroughs determine to a large extent the fre­quency response of the system.c) For rise times of about 50 ps, as in the case of the horizontal deflectors (the "sweepers") of the beam dumping system, the magnets are built of 0.2 mm thick silicon steel laminations insulated by a thin oxide90Table II. Lepton KickersSPSPositron-ejection,Electron-injectionSPSElectron- ej ectionLEP Injection Kicker Kicker septumNumber of magnets 3 1 6 2Kick strength per magnet (mT.m)56.7 80 56 374.5Kick rise time (0 to 100%) (|is)0.7 0.8 3.2 3.2Kick fall time (100 to 0%) (yis)1.5 1.5 6 6Max. generator voltage (kV)30 30 24 24Max magnet current (kA)3.3 5.6 1.6 5.1Repetition rate >65 ps between pulses, 8 burst repetition rate 1.pulses per burst, 2 sThyratrons CX1154, CX1159, CX1181DConstruction ferrite magnet in vacuum tank2-turn airinsulatedmagnet,ceramicvacuumchamberferritemagnetinvacuumtanklayer and housed under vacuum. Steel is preferred for this application because of its higher saturation induction. Despite the large number of laminations (>6000) an acceptable pressure of < 10 ~ 8 Torr is ob­tained. No organic insulation material is used. The magnets are unmatched inductances and excited by half sine wave pulses generated with air in­sulated capacitor banks4 .Leuton KickersFor lepton operation the requirements on the kicker magnets are con­siderably different: Instead of a machine circumference homogeneouslyfilled with particles (as in the case of proton operation on fixed tar­gets), the leptons are grouped in a few short bunches (4 or 8 in the SPS, 2x4 in LEP) of < Ins duration with a bunch separation in the SPS of911.6 ps and in LEP (at the kicker magnet position) of 3.2 p s . The bunches must be injected and ejected individually. The time between bunches can be used as kick rise time and the flat-top time can be very short, now determined mainly by the jitter of the discharge switches. Be­cause of the long rise and the short flat-top times a quasi-sinusoidal excitation is used. The magnet is then a pure inductance oscillating for half a period with a pulse capacitor.An LC oscillation circuit has many advantages compared to a matched travelling wave system: In addition to the low cost unmatched magnet, the operation voltage is comparably low (< 35 k V ) . This allows air insulation and the use of low cost switches. However, the pulse generator must be located close to the magnet and is therefore not accessible for main­tenance during operation. Furthermore all components must be radiaton resistant.Contrary to the low repetition rate of the hadron kickers a fast burst operation is required for the lepton kickers. Up to 8 pulses (one burst) must be produced with a repetition time < 98 ps in the SPS and < 65 ps in LEP. The repetition time of the burst itself is about 1 s. Only one-gap thyratrons with their inherently short recovery time can handle the high repetition rate. Their operation voltage is limited to about 35 kV, just sufficient for this application.The high repetition rate generators with their fast recharging sy­stems located at up to 1 km distance from the magnets are described in references5 » 6 .For the magnets two different technologies have been applied:- Magnets which deflect only the low intensity bunches are housed in vacuum tanks. This is the case of the SPS injection and ejection kickers and the pulsed injection septum magnet of LEP7 .- The 6 LEP inflector magnets through which the short high intensity bunches of the full LEP current circulate use metallized ceramic vacuum chambers. The ferrite magnets are in air and have 2-turn mica epoxy insulated excitation coils. The ferrite is screened from the bunches by the metallized chamber. This construction is ne­cessary to avoid heat-up of the ferrite by bunch induced gyro- magnetic resonances.Contrary to the usual practice, the LEP injection septum magnet has the same fast pulsed excitation system as the LEP kicker magnets. A d.c. or a slowly pulsed excitation system would have been uneconomic, either because of high power losses on the 1 km long transmission line, or the necessity to provide low radiation space for a generator in the ac­celerator tunnel close to the magnet. The septum is designed as a simple metallic plate acting as eddy current screen and requiring neither cooling nor connection to the excitation circuit.92PROTON INFLECTORWe give now a more detailed description of the SPS proton and posi­tron inflector as its design comes closest to the future requirements of TRIUMF. We follow the presentation given in reference 1.The proton inflector is used to inject into the SPS:- protons at 14 GeV/c for fixed target operation, andat 26 GeV/c for proton-antiproton collider operation- positrons at 3.5 GeV/c for LEP pre-acceleration.GeneralAs for all other hadron kicker magnet systems of the SPS, the maximum generator voltage is limited to 60 kV mainly for two reasons:- It allows reliable operation of the thyratron switches at pulse durations of up to 12 ps with current levels of up to 5 kA.- Standard low cost coaxial cables RG 220/U can be used for the 250 m long transmission lines between pulse generators and magnets. These cables are readily available from industry. Higher voltages would require costly nonstandard cables.To achieve a kick rise time of 145 ns, the magnet module length mustbe limited to 0.7 m. Twelve separately powered modules are then requiredto provide the kick strength for collider injection at 26 GeV/c.For beam optical reasons the first 8 modules (type S) have a smalleraperture ratio and hence a shorter kick rise time (145 ns) than the 4 re­maining ones (type L, 220 ns rise time).Only the 8 S-modules are used for fixed target operation at 14 GeV/c injection momentum where the short rise time is important.For injection of a burst of 4 positron bunches at 3.5 GeV/c the 8 S-modules are grouped in pairs and successively powered at 98 ps in­tervals. The pulse generators are then resonantly recharged within 30 ms and the operation is repeated to inject a further burst of 4 bunches. These 8 bunches are then accelerated together to 20 GeV and ejected towards LEP.The magnet modules are built as delay lines terminated by matched re­sistances. The characteristic impedance of the magnets is about 5% lower than that of the remaining system to compensate the attenuation of the travelling wave front in the 250 m long transmission line.The pulse generators are lumped element pulse-forming networks, equipped with 3 thyratron switches to produce pulses of adjustable flat-top duration and short fall time. The modest contribution of the ge­nerator rise time to the total kick rise time permits connection of 293magnets in parallel to 1 pulse generator which has therefore half the characteristic impedance of the magnets. This measure reducesconsiderably the total cost of the generators.Magnet DesignA C-shape ferrite yoke has been chosen. The return conductor is lo­cated at the open C-side about 5 mm from the ferrite. It is therefore non-inductive and can be put at earth allowing the use of coaxial con­nections with earthed outer conductors at the input and output of the magnet. The lower field quality of a C-shape magnet as compared to a win­dow frame magnet is improved by adding shims along the gap . A C-shape yoke eliminates furthermore the risk of flash-over along the ferrite sur­faces since the ferrite is at uniform high potential.The magnet yoke is assembled from 22 C-shape blocks of ferrite, each 26 mm thick with overall dimensions of 195 mm x 136 mm. They are in­terleaved with 5 mm thick Al-Mg plates of about 40 cm x 50 cm formingthe high-voltage side of the capacitors. Earth plates are mounted oneither side of the high-voltage plates at a distance of about 5 mm. The total vacuum capacitor plate surface of the 12 modules exposed to anelectric field of up to 60 kV/cm amounts to about 180 m 2 .Philips Ni-Zn ferrite (type 8C11) is used as magnetic material. It is particularly well suited for kicker magnet applications because of its good vacuum properties (density g > 5.1 g/cm3), its high saturationflux density (B > 0.3 T at 10 Oe and 20°C) and its low coercitive force (Hc < 0.25 Oe). The blocks are vacuum baked at 400 °C prior to assembly.The 1-turn excitation conductor, made of titanium, is pressed against the ferrite and screwed directly onto the high-voltage plates of the ca­pacitors. Bad injection line steering could cause the high intensity proton beam to hit the return conductor. To avoid damage of the magnet under these circumstances, the return conductor is made of beryllium which is nearly transparent to the beam because of its low density.Twelve magnet modules are housed in 3 vacuum tanks of stainless steel 304L, each 3.5 m long. A tank is composed of a base plate onto which the magnet modules can conveniently be mounted and aligned, and a Q-shape cover. The latter is connected vacuum tight to the base plate by means of a diamond-shaped Al-seal of nearly 9 m circumference. A vacuum pressure of 2 x 10-9 Torr is achieved with 4 sputter ion pumps and 4 titanium sublimation pumps mounted under the base plate of each tank.The coaxial low inductance termination resistor is mounted onto the base plate of the tank and connected via a matched coaxial feedthrough to the magnet. The active resistor stack consists of ten ceramic-bound car­bon discs, each 1 inch thick with 3 inches outer diameter, mounted in series, and interleaved with flat metal spirals for forced cooling by silicone fluid.94Pulse Generator DesignThe pulse generator consists of a pulse forming network and 3 high power switches, each mounted in a separate metallic enclosure on top of the PFN. The PFN has 36 cells of 30 nF capacitance and 1.17 pH in­ductance each. The cells are arranged in 2.5 rows separated from each other as far as possible in order to avoid electro-magnetic interferencesbetween different rows. In a PFN of constant pulse length interferencescan be compensated by cell adjustment. This is however not possible when the pulse length is variable. They must therefore be avoided either by sufficient row separation or by screening.Previous experience with a 3-switch generator had revealed inter­ferences between the switches. In particular the clipper switch is sen­sitive to erratic firing if it is connected to the cathode of the mainswitch. In order to avoid these interferences the three switches arehoused in separate tanks and the clipper switch is connected to the anode of the main switch. The connection is done with a matched stripline lo­cated in the PFN container.The switches are three-stage ceramic thyratrons with two cathodes as­semblies (double ended), type EEV CX1171B. Double ended thyratrons are used because of current reversals and higher capabilities, compared to single-ended tubes, to conduct high-current pulses of long duration. The pulse generators are resonantly charged about 10 ms prior to the firing of the main thyratrons. This mode of operation improves the voltage hol­ding capability of the thyratrons. It allows furthermore operating them at higher gas pressure which favours a short current rise time.PerformancesThe proton inflector system has been in operation since 1981 andperforms as anticipated. The reliability of the complex installation with 18 high power thyratrons, more than 200 high voltage coaxial connectors and about 180 m 2 of capacitor plates under high voltage stress isexcellent. The thyratron lifetime is between 15000 and 20000 hours. Thereplacement of a faulty thyratron including the 15 minutes preheatingtime of the new valve takes less than 1 hour.CONCLUSIONThe kicker magnet systems of CERN SPS and LEP machines presented in this brief overview are of proven design. They have operated reliably since their installation; for some of them this is more than a decade. The parameters of the proton inflector come close to the needs of TRIUMF's kaon factory. Its technology is well understood and could readily be applied.95REFERENCES1. E. Frick, H. Kuhn, M. Mayer, V. Rodel, G.H. Schroder and E. Vossen- berg, Fast Pulsed Magnet Systems for Proton and Antiproton Injection into the CERN 400 GeV Proton Synchrotron, 15th Modulator Symposium, IEEE Conf. Record, p. 290, 1982 and CERN SPS/82-14 (ABT), 1982.2. H. Kuhn, G.H. Schroder, High Power Pulse Generators for Fast Pulsed Magnets, Developments and Operational Experience, 14th Pulse PowerModulator Symposium, IEEE Conf. Record, p. 264, 1980 and CERNSPS/80-06 (ABT), 1980.3. P.E. Faugeras, E. Frick, C.G. Harrison, H. Kuhn, V. Rodel, G.H.Schroder and J.-P. Zanasco, The SPS Fast Pulsed Magnet Systems, 12th Modulator Symposium, IEEE Conf. Record, p. 147, 1976 and CERN SPSBT/76-1, 19764. P.E. Faugeras, C.G. Harrison, M. Mayer, and G.H. Schroder, A Laminated Iron Fast Pulsed Magnet, CERN/SPS/ABT/77-16, 1977.5. G.H. Schroder, E.B. Vossenberg, High Tension Burst Pulser for the Electron Extraction Kickers of the CERN Super Proton Synchrotrton, 16th Pulse Power Modulator Symposium, IEEE Conf. Record, p. 103, 1984 and CERN SPS/84-12 (ABT), 1984.6. U. Jansson, G.H. Schroder, E.B. Vossenberg, A Remotely Powered HighTension Burst Pulser for the Injection Systems of CERN's Large Elec­tron Positron Collider (LEP), 17th Pulse Power Modulator Symposium, IEEE Conf. Record, p. 39, 1986 and CERN SPS/86-19 (ABT), 1986.7. G.H. Schroder, J. Bonthond, U. Jansson, H. Kuhn, M. Mayer, G. Vossen­berg, The Injection Kicker Systems of LEP, European Particle Accele­rator Conference, Rome, 1988 and CERN SPS/88-26 (ABT), 1988.8. M. Mayer, H. Kuhn, G.H. Schroder, Metallized Ceramic Vacuum Chambers for the LEP Injection Kicker Magnets, European Particle Accelerator Conference, Rome, 1988 and CERN SPS/88-19 (ABT), 1988.96A KICK ER U PG RA D E FOR LOS ALAMOS PROTO N STO RA GE RING (presentation at KAON PDS M agnet Design W orkshop)H.A. ThiessenLos Alamos National Laboratory, Los Alamos, NM 87545Collaboration! Outline!• Los Alamos- Vern Sandberg- Gary Rodenz- Arch Thiessen• SAIC- Russ Winje• With Lots of Free Advice From- David Fiander, CERN1. Specifications2. Current Required3. Inductance4. Rise Time5. Tricks to Divide the Inductance6. Safety Considerations7. Accelerated Lifetime Tests8. Effect of Ferrite Permeability9. SPICE Calculations10. SummarySpecifications! Current Required!• Magnet Gap 9 cm • Assume Packing Factor = 0.5• Magnet Width 18 cm - 2 meters of magnet• Kick Angle 18 mrad • Assume Magnet Efficiency = 0.85• Rise Time 70 -ns o^'• Pulse Length 400 ns B = ------ —g x eft• Repetition Rate 60 Hz 9 = ?• Uniformity of Kick 5 % P = 0.2998 x Bp• Available Space 4 meters• Coupling Impedance to Beam ? OhmsP0 9 1460x 0.018x9 377 x 10'6 lxeff 377 x 10‘6x 200 x 0.85Inductance! Rise Time |L =L =p0x I xwg x eff4 x 7t x 10~7 x 2 x 0.2 0.09 x 0.85 6.6 pHz = v  45000 2i2L2 x  3690 6 .6 x 1 0= 6.1Q-6.-.must divide into multiple units3690 Ampst(5%-»5%) (n*)97Tricks In Division |1. Power Each Half of Magnet Separately- as at ANL rapid cycling synchrotron2. Use Push-Pull Power Supply- as at Rutherford ISISOur Decision- Use First of Above- Try Second Laterwhen 2 pulsers availableNumber of Units!Approximately 16 Units Needed- To get 70 nanosecond Rise TimePulser Can Handle 2 or More Unitsper ThyratronSPICE Calculation of Rise Time)L/(2Z) (n*)98Kicker & Power Supply Module)Charging Lines 2x3x20 Ohms Thyratron 3.33 Ohms CXI725  in OilTransmission Lines 2x3x20 Ohms 3.33 OhmsKicker In Freon or VacuumHV Power Supply 60 kV 120 Hz in OilTerminations 2x3x20 Ohm Oil Filled IKicker In VacuumTransmission Lines 2x3x20 Ohms 3.33 OhmsKicker in Sulfur HexWater Cooled Ground ShieldHex Left ConductorGround RightPlanes ConductorMagnetic Field Calculatior' iK *y l -  3 s-Safety Considerations]• Fire Prevention Dictates- Silicone Oil (if indoors)- EPR (Ethylene Propylene Rubber)• EPR Cable- more radiation resistant than polyethylene- more ozone resistant than polyethylene- compatible only with silicone oil- more attenuation than polyethylene• Test of EPR Cable Attenuation- to be reported by Russ Winje (SAIC)Left Ground RightConductor Planes Conductor99Lifetime of lnsulatiorT]j20ISIO♦  ♦ 0 V -- ♦ ^  ■ -- x , ® --\ o--X-I d---- 1____ iw  im th __ j. i 'r . ' ° 1 IOOj5.6 kV /m m  (rms)=  8 kV /m m  peak=  80 k V /cm  peak =  200 V /m il  peakIO Jh IO" Ih IO' IO j h IO1 IO"h IO5 t I0 6hFig. 6. Voltage life versus field strength. © Cavities in polyethylene-0.3 mm thick. ■ Cavities in polytetrafluoroethylene-O.lmm thick. +  Cavities in polyvinyl chloride-0.1 mm thick. — Voltage life of polyethylene cable according to Oudin [4],from REUGER: RESISTANCE OF D IELECTRIC MATERIALSAccelerated Lifetime Testing |Polyethelene Lif etime°c V” 877 1.3V « x 10 Lif etime Reduction•  Test at 1.3x45kV=60 kV for 1 year-  to get 10 year lifetime•  LAMPF Runs 6 Months per Year• 3 Months at 120 Hz Test TimeEffect of Ferrite Permeability)• Propagation Time ~ sqrt(p.e)- e~10- therefore use p~35 ferrite• Field Uniformity Unaffected- Conductors Help Maintain Uniformity- in pulsed application• But must allow for lower efficiency of magnet• Tested by Q. Kerns, et al, FermilabFerrite Choice ? |C20K), C2025, C2080, 62078, N40 High Frequency NlekBl-Zlhb ParrttaaTDK group of matrrtatl wm apoeflfcUy OHalHMtad 16 ghw Mgh M iM Ry m accommodating requirement* to 1000 MHz In appOoMMM M p&M IllftpiiS*, linear empOOer*, UHF. and VHP. Thay art available m eize* up lo SO". Our Engtnoartnf OoptttittiHt WOUld bo ptaaeod la aeolat you kt determining which of thaaa larrttat lo bool lor your app*ea»cn.Typical Magnatic and Mtytlbai CharaetarlallaaMTTML PCMMfeUMUtY 1 MHz MAXIMUM PtMMfej&HJtyn D M M n l  rtVTA ITCITOII T*WWWfl rUfwt wHwu—--------- — . . . . i . .  j ««vaatx a uC U M  TPMPVRKftJREi n ;do VOUflflt F*&8ftWY, Otlrtvctt!• «  40 aamoa appMed fluid otrangth0*085/ COOTS N4017* 100 36 151100 900 160 50MOO 9400 2700 1000MO* MOO 1*00 70014 10 7.0 7.5<70 *40 420 510XF 10* 10"lUnlfiil DmAoMI CulfMtti a y ppoooaw a r vwyrv w i  t wopQ ua lity  F acto r *» . FiaqW HW y In itia l Parm aaM W y va. FrequencyQuestions Remaining |Kicker• Vavuum or Sulfur Hex?• 60 kV Vacuum Feedthru• Minimizing Inductance of Magnet Feeds• Spacing for 60kV Holdoff in Vacuum• Construction DetailsPulser• Air Bubbles• 60 kV Termination• Spacing for 60 kV Insulation• Life of Terminating Resistors in Oil• Mods for Push-Pull Operation101A PRELIMINARY DESIGN OF THE LOS ALAMOS FAST KICKER MAGNET PULSER AND POWER SUPPLY*R.A. WinjeScience Applications International Corporation 227 & 230 Wall Street, Princeton, NJ 08540ABSTRACTThe technical design of the Kicker Magnet Pulser and Power Supply is based on the switching of a precharged pulse forming network (pfn) into a matched load. Provisions are made through the selection of the main switch tube to accommodate loads that are not matched to the pfn impedance. The paper includes a description of the major components of the power supply and a summary of the performance parameters.SYSTEM REQUIREMENTSThe design concept for the Kicker Magnet Pulser and Power Supply is based on discharging a pulse forming network (pfn) through a high speed thyratron switch into a matched resistive load. Figure 1 shows an elementary diagram of the system and Table I gives the major performance specifications.6 0 0 V  .50k w itR F S O N A N T CHARGERPULSERI__TERMINATION LOAD  1TERMINATION LOAD 2Fig. 1. Elementary schematic of fast kicker magnet pulser and power supply.FIBER OPTICSTable I. Performance specifications.PFN Operating Voltage, max PFN ImpedanceLoad Impedance/each of two Pulse Flat Top Length, min Pulse Rise Time, max Pulse Repetition Rate, max Pulse Jitter, max Peak Output Power65 kV 3.125 ohms 6.25 ohms 411 nsec 30 nsec 120 Hz 5 nsec 338 MWThe power supply will be controlled by a local control system which serves both to coordinate all of the control and fault protection aspects of the power supply and as the interface point for the Los Alamos remote control system.*This work is funded by Los Alamos National Laboratory under Subcontract No. 9 -XF8-6797L-1.102THYRATRON SWITCHThe criteria used for selection of the thyratron tube were the following:Voltage Hold Off Forward Pulse Current Reverse Voltage Reverse Current di/dtPulse LengthPulse Repetition Frequency Expected Life Time70 kV 7 kA 7 kV 7 kA350 kA/^sec 422 nsec 120 Hz E+09 pulsesThe tube selected for the application is the English Electric Valve CX 1725. The principal advantage with this tube type is its ability to conduct current in the reverse direction. This tube has a special hollow anode design enhancing the ability to carry the reverse current not as a metal arc, but through electrons which have been trapped near the anode during the forward conduction part of the cycle. The tube is otherwise similar to the CX 1525 which would be the tube of choice if the load was matched to the pfn. EEV has reported greater than 2E+09 pulses achieved from a single gap version of the CX 1525 (CX 1625) operating at higher pulse currents and repetition rates.Thyratron Support StructureThe thyratron is positioned in a coaxial mount designed to minimize the stray inductance associated with the tube. Figure 2 shows the coaxial mount. Based on this coaxial geometry, the value of the series inductance is estimated to be 88 nH. In a system where the ZQ of the pfn and the load cables are 3.125 ohms (each), the time constant of the circuit will be r “ Lstray/2Zo = 14 nsec • Assuming a square wave input from the thyratron switch, the pulse rise time of 30 nsec should be achieved.Fig. 2. Support structure for the CX 1725.103The structure has been designed for a voltage withstand capability of 80 kV. As the structure will be in insulating oil and the anode section of the tube is separated from the input section by a cylindrical epoxy-fiberglass (G- 10) tube, the voltage holdoff during the charge cycle should be readily achieved.The tube and support electronics will be mounted from the cover of the enclosure to facilitate maintenance. The unit can be lifted by a crane attaching to eyebolts on the cover.Support ElectronicsThe thyratron filament and reservoir heaters are driven from filtered direct current power supplies to reduce output pulse jitter due to the tube. Taps on the primary of the filament heater isolation transformer will be used to set the filament voltage at the required value (6.6±5% V). The reservoir heater power will be controlled from the ground based control circuit through a voltage variable transformer to permit adjustment of the gas pressure in the tube to obtain optimum turn-on and voltage hold-off characteristics. The heater current and voltage from both circuits will be directly monitored at the local control station utilizing voltage-to-frequency and frequency-to-voltage convertors coupled together with fiber optic links for voltage isolation.Grids G1 and G2 will be pulsed with an unloaded voltage pulse of at least 2 kV (peak) with the leading edge dv/dt of 20 kV/^sec. Figure 3 shows the circuit used for each grid pulser. The trigger amplifiers are thyratron switches (CX 2535) discharging 1 n s e c ,  25 ohm pulse forming networks. The trigger pulses will be transmitted from ground potential via a fiber optic cables. A pulse current of 25 to 50 A will be delivered and maintained to G1 one-half microsecond or so prior to the start of the G2 pulse. The high G1 prepulse current provides a supply of electrons to initiate the main anode current pulse with the required di/dt of nearly 350 kA//isec.PFN1 ps,  2 5 0Fig. 3. Grid pulser schematic diagram.104COAXIAL CABLETwo types of coaxial cables are being investigated for use in the power supply. Both polyethylene and ethylene propylene rubber (epr) based dielectrics are being evaluated for this application. In either case, the main dielectric will be shielded at both the inner conductor and the outer conductor surfaces with a high dielectric constant material (k>10) to reduce the electric field stress on the main dielectric. For a cable having a design lifetime of 40 E+10 pulses at 65 kV, the following construction is being considered.The attenuation of this cable is 4.2 dB/100 ft measured at 50 MHz. The surge impedance of this cable is 20 ohms and with six cables in the pfn, the impedance would be 3.33 ohms. The length of the pfn cables is 33.3 m.A special connector is being designed for the cable. Figure 4 shows a cross section of the connector. Electric field stress control is obtain by the shaping the outer shield terminal and the center conductor terminal. The connector is sealed by a wedge shaped rubber gasket.Central core (rubber)Center ConductorHigh dielectric constant stress relief Dielectric (epr)High dielectric constant stress relief Outer Conductor0.8 in. OD 1.0 in. OD 1.07 in. OD 1.78 in. OD 1.85 in. OD 1.9 in. ODO O CABLE CLAMPFig. 4. Coaxial cable high voltage, oil insulated connector.105CHARGING POWER SUPPLYThe resonant charger is shown in Figure 1 and consists of the charging dc power supply (600 V, 30 k W ) , thyristor, HV step-up transformer, isolating HVdiode and the other associated circuitry. The output voltage from the chargingpower supply is 601 V which is required to produce a peak voltage of 80 kV onthe pfn. To operate at 65 kV, the power supply voltage is reduced to 488 Vdc.The output capacitor was chosen to be 6200 ufd which is about 10 times the primary referred value of the pfn capacitance. It will be made up of a series - parallel array of 3100 ufd, 450 Vdc electrolytic capacitors. The capacitor is recharged during the interpulse period by the charging power supply. At a repetition frequency of 120 Hz, the average charge current is 63.6 A. However, at the prospective operating voltage of 65 kV, the average current reduces to 51 A and the required power from the charging power supply is 25 kw. Likewise, the primary rms current for Vpfn = 80 kV has been calculated to be 204 A and for Vpfn = 65 kV, the rms current is 164 A. The transformer current ratings will be sized to the 65 kV case.The thyristor in the charging circuit is triggered from the control system about 2 msec prior to triggering the thyratron. The thyristor will conduct the half-sine wave of primary current and will naturally commutate on the first current zero. The average current for the thyristor is 64 A which is easily in the range of a wide selection of devices. The pfn voltage will be regulated from changes in line and load conditions through a voltage feedback loop.TERMINATION RESISTORSTwo terminating load resistor banks are required for the power supply. Each resistor bank will contain terminating resistors for each coaxial cable.Table III. Parameters for the termination resistor assembly for 65 kV.Assy Impedance, ohms 6.25Peak Current/Assy, A 5200Load Impedance, ohms 20Number/Assy 15Peak Power, MW 169Average Power, W 8552The basic resistor used in the assembly is the Carborundum type 1038AS which has a power capability of 225 W (in air) and can withstand 65 kV. This resistor is 18 in. long and 1-1/2 in. diameter. These resistors will be mounted in a coaxial cylinder so that each individual coaxial cable is properly terminated. The termination resistor assembly will be an oil enclosure that can accommodate up to three 20 ohm resistor assemblies. The resistors will be immersed in insulating oil and the oil pumped to remove the heat and to promote voltage withstand capabilities.CONCLUSIONA preliminary design of a thyratron based 65 kV pulser and power supply which will be used for driving a fast kicker magnet has been presented. Fabrication work is now beginning with completion projected for March 1989.106M AG NET REQUIREM ENTS FO R  EX PERIM ENTAL AREASE.W . Blackmore and A .J. O tter TR IU M F, 4004 Wesbrook Mall, Vancouver, B .C ., C anada V6T 2A3A BSTRACTThe m agnet requirem ents for the experim ental areas of the TR IU M F KAON Factory are varied and demanding. In the target areas the m agnets will be located in radiation fields up to  10E7 ra d /h  and will also absorb therm al loads from beam  heating up to  several W /cm 3. In this operating environm ent the m agnets m ust be reliable and capable of rem ote installation and servicing. O ther m agnet designs include Lam bertson septum  m agnets for beam  splitting in the proton switchyard and superconducting or superferric m agnets in the secondary channels and detectors. This paper describes some of these m agnet requirem ents and presents some prelim inary ideas on their designs.INTRODUCTIONFigure 1 shows one possible layout of a target area for a kaon factory. Up to  100 f i k  of 30 GeV protons is incident on an interaction length production target and the secondary par­ticles such as kaons and pions are collected in two beam  lines which tran spo rt these particles to  the experim ental areas. The protons which pass through the target w ithout undergoing interactions or large-angle scattering are transported  to a beam  dum p or may be refocused on a second downstream  production target. The kaons are produced in a forward direction w ith a m axim um  production angle of about 10°. The first element in the secondary beam  line is normally a dipole which is located as close to  the production target as possible for m aximum  acceptance. In the arrangem ent shown two secondary beams, one a t low m omen­tum  (<1 G eV /c) and one at high m om entum  (>4 G eV /c) are provided. A nother possibleK4Q3 K4Q 4— 7 K4Q5K4B3TARGETr ™  \  \  \  \  \COPPER PLATES IN WATER BATH \K40I K4Q2TARGETSHIELDPROTONBEAMLINEK101 KIQ2 KIQ3 KIQ4K105Fig. 1. Possib le a rran g e m en t oflow - an d  h igh -m o m en tu m  secondary  channels from  th e  sam e p ro d u c tio n  ta rg e t.107Ep*30GeV, TARGET: W (0 .8cm *< k 6cm)1\ } '\Xt\\\\\\\XXVk\\\\\\XXV^XtX^r ^ i1 " k  1kxxxxxxxxxxxxxxkkxS^I/v ac.1 ------- _ i o  c u |F too ~ ---- ^1 . . v a c .TARGET 4 0  8 0  120 160 2 0 0DEPTH (cm)Fig. 3. C o n to u rs  o f equal energy density  in W /c m 3 for 100 pA  o f 30 G eV  p ro to n s  on an in ­te ra c tio n  leng th  ta rg e t.arrangem ent is the MAXIM concept1 for three independent secondary channels which has been developed for the A IIF and E IIF  proposals. The beam com ponents in the region w ithin several m eters of the production target m ust survive high radiation and therm al fields. Fig­ure 2 shows the radiation levels around a shielded target area for 100 /rA of 30 GeV protons incident on an interaction length platinum  target and Fig. 3 shows the energy density due to beam heating around a similar production target geometry.2 For reference the energy density due to  resistive losses in a typical m agnet coil is about 0.5 W /cm 3.For radiation doses above l x  10° rad conventional m agnet design using fibre-glass epoxy insulation and rubber hoses on water-cooling manifolds is not acceptable and techniques for producing more radiation-resistant m agnets m ust be found. Considerable experience in the design of radiation-hard m agnets exists a t the meson factories3 and this same technology has to be used to produce the m agnets for a kaon factory. The m ajor difference for the la tte r  requirem ent is th a t the m agnets tend to  be larger both  in apertu re  and length as the m om entum  of the particles being bent is much higher. Details of the design of rad- hard m agnets is given elsewhere in this workshop proceedings. Some of the considerations in transform ing a large dipole to a radiation resistant design is presented in the next section.RADIATION RESISTANT MAGNETSAt TR IU M F m ost of the rad-hard  m agnets are fabricated using directly cooled mineral- insulated Pyrotenax conductor.4 All quadrupoles directly downstream  of the production ta r ­gets either in the prim ary beam line or on the secondary channels are m ade using this tech­nique. Pyrotenax coils have also been used on dipoles although the large com bination m agnets at the cyclotron extraction ports use indirectly cooled Pyrotenax coils po tted  in a solder m a­trix  along with copper cooling channels. Another technique has been used at TR IU M F in the fabrication of a high current, 5000 A, septum  m agnet.5 The coil was formed from rectangular copper conductor with a central cooling channel which was held in place by a num ber of ceramic spacers furnace brazed to  the conductor. PSI (formerly SIN) has considerable ex-Fig. ‘2. C o n to u rs  o f equal dose equ ivalen t for 100 / iA o f 30 G eV  p ro to n s  on an  in te rac tio n  leng th  ta rg e t su rro u n d e d  by sh ie ld ing .3 02 5E  2 0p  150  <01 1050108Table I. Comparison of current densities for different methods of fabaricating radiation-resistant mag­net coils. Parameters for Brookhaven 30D72 or 18D72 magnet: B — 1.5 T; Gap =  10 cm; NI =  130,000 At; Length/turn =  20 ft; Ap =  60 psi; AT = 40°C.C o n d u c to r typeC o n d u c to rd im ensions(in .)M axim umcu rren t(A)C u rren t density  C o n d u c to r Coil (A / in .2) ( A / in .2)C om m en tD irec tly  cooled P y ro te n a x0 .53x0 .53 500750@ 120psi4808 17630 .75x0 .75 1500 6321 2453 M u ltip le  cooling c ircu itsIn d irec tly  cooled P y ro te n ax0 .53x0 .53 500 4010 1161 M ost expensiveA ir-cooled b are  conduc to r0 .375x8 .0 3000 1000 521W ater-cooled  b are  conduc to r0 .73x1 .46 2500 3316 1381 L east expensiveF ibreg lass epoxy conduc to r0 .73x1 .46 2500 3316 1905 R eferenceperience using indirectly cooled m ineral-insulated conductor w ith lead or solder as the heat transfer m edium .6A typical large dipole used at Brookhaven National Laboratory was considered as an example of the type of m agnet which would have to  m ade rad-hard  at a kaon factory. Table I lists the relevant param eters for this m agnet and a comparison of the various coil construction techniques in term s of current densities. The much longer lengths for a single tu rn  puts a lim it on the current-carrying capability due to  the pressure drop of the cooling water in one tu rn . It is likely th a t the m agnet steel will also have to  be cooled. For a typical energy density level of 1 W /cm 3 cooling would have to  be placed at 10-15 cm intervals in the m agnet yoke. M agnets m ade from thin lam inations would be precluded due to  the poorer therm al conductivity across the lam inations.TA RGET CELL M AGNET CONSIDERATIONSThe m agnets for the target cell regions must be capable of rem ote installation and servicing. One of the im portan t considerations is the interaction between the vacuum enclo­sure around the beam  and the m agnet design. Figure 4 shows a vertical section through a proposed target shield for the existing 500 MeV facility to  indicate some of these consid­erations. The target is in vacuum to elim inate air activation and the necessity of windows in the high-intensity proton beam. There are three options for the m agnets w ith respect to the vacuum enclosure. The m agnets could be immersed in vacuum, the beam  pipe could be continuous through the m agnet as is shown, or there could be vacuum flanges a t each end of the m agnet and the m agnet is installed or removed along with its section of beam  pipe. This la tte r  option requires the design of a rad-hard , rem otely handleable vacuum flange capable109COLLIMATORSTEEL SHIELDINGCONCRETEPRODUCTION TARGETNON VACUUM-JOINT COVERSFig. 4. V ertical section th ro u g h  a p roposed  ta rg e t sh ie ld  a rran g e m en t for th e  ex isting  500 M eV facility.of operating under varying therm al conditions. A preferred solution is the design of m agnets which can be positioned around a continuous beam pipe such as a  half-quadrupole7 or a split quadrupole as shown in Fig. 5. O ther more exotic designs could also be considered which have the additional advantage of positioning the coil further from the target region and in a more convenient position for servicing as illustrated  in Fig. 6.8PROTO N BEAM SW ITCHYARD MAGNETSThe beam  line m agnets for transporting  the 30 GeV extracted  beams to the experi­m ental areas can be of conventional design and would m ost likely be similar to  the design of the extender ring dipoles and quadrupoles. The present KAON proposal has the slow ex­tracted  beam split into two or more beam lines using an initial electrostatic septum  followed by one or more Lam bertson septum  m agnets as done at Fermilab9 and Brookhaven.10 At the much higher intensities of a kaon factory the design of the Lam bertson septum  m agnet has to be made rad-hard. As the Lam bertson has only m agnet steel exposed to  the high intensity beam  the design of a rad-hard  version should not be too difficult and an indirectly cooled m ineral-insulated conductor could be used.CONCLUSIONThe design of radiation-resistant, large acceptance m agnets for the experim ental areas for a KAON factory will provide new challenges for the m agnet engineer. Much of the expe­rience a t the meson factories will be relevant bu t the requirem ents for much larger m agnets110Fig. 5. Techniques for installing a quadrupole around a fixed beam pipe.BREAKLINENON MAGNETIC SHIELDINGBREAKLINENON MAGNETIC SHIELDINGBEAMTUBEFig. 6. Proposed long pole magnets for operation in high radiation areas.I l land the additional power dissipation problems due to  beam  heating will require some new ideas.REFERENCES1. C. Tschalar, “M ultiple Achrom atic Extraction System ” , Nucl. Instrum . M ethods 249, 171 (1986).2. C. Yamaguchi, “Some Energy Deposition and Absorbed Dose Calculations for Kaon Fac­to ry” , TRI-DN-85-11.3. A. Harvey, “Experience with the LAM PF M ineral Insulated M agnets” , In t. Conf. on M agnet Technology MT-6, (1977) 551.4. Supplied by Pyrotenax of C anada Ltd., Trenton, Ont.5. T. G athright and P. Reeve, “A Radiation Hard Septum  M agnet” , IEEE Trans, on M ag­netics M T-7, (1981) 1714.6. D. George, “M agnets w ith M ineral Insulated Coils a t SIN” , Int. Conf on M agnet Tech­nology M T-5, (1975), 719.7. D. George, R. Abcla and D. Renker, “Half Quadrupoles for Use in High Radiation Envi­ronm ent” , Int. Conf on M agnet Technology MT-9, (1985) 184.8. A. O tter and C. Ilo jvat, “A Prelim inary Concept for the Design of M agnets a t Target Locations” , Proc. Int. Workshop on Hadron Facility Technology, Sante Fe (1987) 541.9. LAV. Oleksiuk cm et al., “The NAL Beam Splitting System ” , IEEE Trans. NS-20, 428 (1973).10. II. Brown et al., “The New AGS Slow External Beam Sw itchyard” , IEEE Trans. NS-28, 2985 (1981).112RADIATION-HARDENING OF M AGNET COILSAlex Harvey Stanford Linear Accelerator Center, etc.1. INTRODUCTIONThe first essential before em barking on the radiation-hardening of electrical insulation-  m ostly m agnet coils -  in any beam  line application is to  obtain  a reliable estim ate of the dose to  the components. There are examples (switchyards a t SLAC and LA M PF) where the degree of hardness specified was much higher than  was required. A lthough experience shows th a t the cost prem ium  for substantial radiation-hardening is of the order of 10%,1 it has also become clear th a t well-designed beam  lines have negligible losses: hardening is required only in the vicinity of targets, collimators or other beam -intercepting devices. W here the beam  is deliberately scraped, local shielding will minimize the associated radiation in the surroundings. E lectron machines have their own special problems due to  synchrotron radiation, so certainly coils and other electrical equipm ent should be kept away from the beam  bend-plane.Because proton beams in teract with thick targets in the meson factories, TR IU M F, LA M PF and PSI (formerly SIN) have examples of very hard m agnet coils near their target cells. The activation th a t is associated w ith these substantial doses requires rem ote handling of the m agnets, and poses the question of w hether it is worth considering repairing a  damaged m agnet when it fails. As disposal of radioactive waste becomes more and more difficult, repair may become more a ttractive, but provision for it needs to  be designed-in from the s ta rt. It is these problems of radioactive handling th a t add substantially  to the cost of radiation- hardened m agnets.An interesting disposal idea originated at PSI in Switzerland -  damaged m agnets were incoporated in to  cast concrete shielding blocks for their target cells.2. LEVELS OF RADIATION RESISTANCEA. “Conventional”A standard  technique for insulating m agnet coils is to  use epoxy resin, reinforced with fiberglass. S tandard  resin system s, such as Novolac, or Bisphenyl-A, w ith NMA hardener, can be expected to  to lerate 109 rad  (107 Gy). Additives to  avoid are the old Carbowax flexibilizer-  in large parts, flexibility can be m aintained with flexible epoxies such as Dow’s DER 732. One advantage of epoxy systems is the color change (darkening) th a t indicates exposure, and allows replacem ent before failure.All these epoxies can have their radiation tolerance enhanced by adding an inorganic filler, glass or alum ina. Some indication of the improvement to  be expected can be found in Refs. 2 and 3. Note th a t Ref. 2 gives some of the few available d a ta  on electrical properties following irradiation: m ost radiation testing is done on the basis of mechanical properties. Table I gives the recipe currently in use a t SLAC.O ther organic possibilities include polyimide (K apton) for which D uPont makes very m odest radiation-resistance claims. However, its resistance exceeds IO10 rad, and it is available as a film coating on m agnet wire (Il-film) as well as in tapes and sheets. As a copper- polyimide com posite, it can be formed by printed-circuit techniques for pole-face winding and Lam bertson m agnets.113T ab le  I. A lum ina-loaded  epoxy m ix for coil p o ttin g , SLA C  1969 (to  m ake ap p ro x im ate ly  one U.S. gallon  o f m ix tu re ).P er cen t RM  by w eightE nglishp o u n d s ouncesM etricgram sD E R -332 B isphenyl-A  epoxy 45.0 1 14 850.5D E R -732 F lex ib le  epoxy 55.0 2 5 1048.9T o ta l 100.0 4 3 1899.4A dditives:NM A H ardener 96.5 4 1 1832.1BD M A C a ta ly s t 1.7 1- 1 /8 31.9Z6040 Silane w ettin g  agen t 1.0 3 /4 21.3C ab-o-S il M ain ta in s  suspension 3.75 2- 1 /2 72.0A D O 2 Inorgan ic , T-61 224.0 9 6 4252.4T o ta l 327.0 13 11 6209.7T o ta l w eight per U.S. gallon 17 14 8109.2N O TE S:D E R -332 and  D E R -732 from  Dow C hem ical C orp.NM A (nad ic  m e th y l anhydride)BD M A  (benzl d i-m ethy l am ine)AI2O 3 iron-free and  soda-free a lum ina , 325 m esh, from  E .V . R o b e rts  o r S ch o o f’s, N oraga, CAOne organic to  be avoided is Teflon, which although an excellent high-tem perature insulation, is very poor in radiation fields. Care needs to  be taken to elim inate its use in hook-up wire, and in factory-installed wiring in protective devices like flow switches.B. InorganicsFor maximum radiation resistance, organics m ust be avoided completely. There are techniques in the electrical engineering field which are largely inorganic, and adaptions of these lead to  the highest levels of resistance.Mica is a traditional insulation th a t w ithstands high tem peratures and corona very well, in addition to  radiation. Unfortunately, its physical form at makes its application difficult in m any circumstances. It has been modified by the electrical industry  -  reconstitu ted  m ica -  to overcome the physical lim itations, but often a t the expense of its radiation resistance. One of the better composites is Mycalex, incorporating mica in a glass m atrix . It can be hot-pressed onto substrates such as 400 series stainless steel (to  m atch its expansion coefficient) and is useful in some applications, feedthroughs and stand-offs, for example.Several insulation systems have been developed for accelerator applications: a hard- anodized surface on aluminum conductors4 and concrete, usually in conjunction with fiber­glass.5,6 However, the m ost widely used system found, for example, in the targe t cells of all the meson factories, LA M PF, PSI (SIN) and TR IU M F), is based on the use of magnesium oxide, “m ineral-insulation” in the trad e .7 Two m ethods of using m.i. cables are possible in coils: direct water cooling and indirect cooling.114In direct w ater cooling, the conductor is fabricated w ith a  central hole, and is cooled by direct contact of the water w ith the current-carrying conductor. This implies th a t there are insulating tubes between the conductors and the water headers, which for high-radiation service, are ceram ic-to-m etal parts (Ref. 7). Figure 1 is a sketch of an insulating assembly produced by industry8 th a t has the advantages of:1. high aspect ratio  of length to cross section of water, giving low leakage current;2. sm ooth in ternal bore -  no diam eter changes or pockets to  accum ulate deposits;3. substan tia l tube  ends to  act as sacrificial electrodes, if they are not adequately passi­vated to  inhibit corrosion; and4. tube ends arranged so th a t installation does not apply tension to  the assembly.These features are a result of operating experience at Los Alamos.9Fig. 1. W ater insulator ceramaseal.The characteristics of two commercially available sizes of hollow m ineral-insulated square cables are given in Table II.Indirectly cooled m ineral-insulated coils avoid the w ater-insulator problem  by keep­ing the coolant in separate tubes, and conducting the Joule heat through the m agnesia and a m etal m atrix  th a t includes the cable sheaths. To facilitate this heat transfer, the copper sheaths are soft-soldered together, or cast in lead. C urrent densities over 20 A /m m 2 have been dem onstrated  by this technique,10 though the power consum ption, of course, is correspond­ingly high. M agnets w ith coils of this construction are in service a t Los Alamos, TR IU M F, and PSI (SIN ).11For all applications where the m agnet is in a high-radiation environm ent, so becoming activated, it  is vital th a t the interlock systems be as radiation-resistant as the coils; th a t be a t least as reliable, and preferably incorporate redundancy and if possible an in-situ test capability.There is considerable experience in industry, world-wide, producing coils w ith mineral- insulated conductors. However, there have also been some disasters, and consultation with an experienced user before com m itting to  a contract would be prudent. Moulds to  contain substan tial am ounts of lead-tin alloy m ust be of adequate strength  to  resist deform ation, and m olten lead-tin dissolves copper, so the potting  tim e has to  be minimized.The characteristics of two commercially available cables, solid-conductor, square cross section, are given in Table III. The 0.53 in .2 cable in particu lar has been developed for the maxim um  current density per unit of overall cross-sectional area.Table II. Mineral-insulated cable, square, hollow conductor.1151 cr1 m E1 E ou E 4-> r^» Eran ° LOT— CM CJ COO cr E cr CJ 1—1 -L> • to .x E to to <a •| CJ o _x a. o Ei £= E G cr E J»C+ E E E G e +1 COE LO CD LO LO -XLO CO r— c o CM cnO to »— CM LO •r— LO CM CO r— E• • • • E • CM •CTt r— o r— r— LO O o CM o' ' '— f\CJrxs>_xcrto LO• 4->c r— •+— co• m toto o 4-> CM o• cr 4- oCL o to cr -*-> ••* 1 E 4-1 to ra o 4J4- o C c 4- s- H--i— T— G £Z cr +(+ *r— E a_ •t— -Q -L>CO o LO o to LO L+-LO LO r-- CO LO G LO cr) CL COO LO o O CO CM 3) o oCOo O O o o LO O CM o CMcrmEEE CMCO {=LO 4J jSCJ CM ra «*T— • E COO cr E cr CJ-*-> 1 to to CJ ra ECJ o -X 1/1 o_ oez E a E E X CD+ E E E G c: +1 -Xc CM CD LOLO CO LO CM CT I— LO • LOr^ LO O CM o LO •r— LO o CM o E• E CMCO T— o O r— ■ r— C o o r— CM' 1 ' COora>-Xcrto LO 4Jc: r— M—O T—r— +->in o ra CM LOcr -M oc. o to 4— cr s: •> Or 4-> to CU 4-> -Lin~ o C2 c E 4- CO L— O ll-I C c oo+ G •*— E ZD r- +1 JO -Li1 CO LO to M-CO LO CTl CO «a- G CO LO CL COLO o CO O o CO o r^- Oo o o O o LO o o o o r>.CME3 raE cz •CX.2X EC3 o *—•r— 22 —racr •> CJTT aj to uE CJ m cr r- Eo cr CJ <a 4-i t o32 ra cr cr. ra E1 -Li -X to c c to ,—^• to o *r— aj c to p—I CO aj *»— **— to 2_ E QJ rO r~I to N1 to JZ CJ -l> o 2; c c 4JO CJ 3 *r— QJ 1— cr to 22 CJ>! *"■* rsi T— go CC LJ E c-J— ■y— *o cr c CJ o QJ< CO C3 L- J- o o •r— aj o 'jz icj OQ o o 2- M H—t—1 r— 4-> 4-i 4-i ■p 4-» l7Li— •— J- O O ra ra CJ CO c~ -Li4 rO <u 3 3 r— t— a s_ •4J £o 2- e *a T2J 3 3 i— oj CD ra crLU <D s_ cr C to to a 4-i QJ x> o o o cr cr *T— o ra sz CJ1 "O CJ CJ CJ Q 322 3 : to 32 S .p—^to O-O -L> Ocr toCJ aj2--Li 3 0-CL -L> • oQJ ra cr.LJ S- • QJX QJ ►—« i~QJ C_L 3~^ f- c toaj -T— COQJ -Li GJr— aj 2-JZ cn C.ra o cCJ CJ ra S-S- JZ cuM— CJ -L>o -L> rara o ■?-c co 2- 1—•I— QJ o rato -L i CJ cS- ra 3 2-GJ s -O ajf- o 4-iH c cC L *“*—^, p—i— CM> .f— a-Li -aCJ -I—3 X"O r— cC2 rao QJ Eo c 3crLO ra toCJ a< i— crf— ra cr.raa Eo s_o QJ “Or— E QJE 4 -i•" o CJs- CJ raaj CLCL »* ECL S- oo QJ CJCJ CLCL *«•* O cS- CJ oo •l—4-i • 4-iCJ JZ ra3 4 -1 r—TJ ra 3C a too JC cCJ to 1—'Shipping: Colled to minimum diameter of 4 ft (1.2 m), I ps1 (3.1 mPa) to produce no Ends sealed against moisture. 1 change in I.R.Table III. Mineral-insulated cable, square, solid conductor.1161173. SPECIAL CASESM ost accelerator beam  lines include a few m agnets w ith special requirem ents th a t challenge the m agnet designer. Probably the commonest “special case” is the current-sheet septum , since it is inevitable th a t beam  spill takes place on the septum  conductor. Some approaches support the conductor only a t its edges, so th a t the insulation is removed from the beam  center-line (BNL); Los Alamos has used polyimide successfully in its P ro ton  Storage Ring. At SLAC, the Dam ping Ring septa have plasm a-sprayed alum ina on copper conductors in m agnets w ith small vertical aperture. Factors in their success are 1) use of stainless-steel tubing for cooling w ater, brazed to  the copper (it is not oxidized by the irrad iated  water), and 2) double-brazing of all the conductor parts for assurance th a t there is adequate filler m etal. These coils, which are inside the vacuum chamber, are run at their full current rating before assembly into the iron.ACKNOW LEDGEM ENTAny account describing the usefulness of m ineral-insulated cables in m agnet coils for accelerator applications has to recognize the contribution of the la te  Sid Walker of Pyrotenax of C anada Ltd. to  its successful development. Sid undertook the challenges of new configu­rations for the m .i. cable th a t the Trenton, Ontario plant produced, and cheerfully pushed their technology to  the lim its th a t the m agnet designers were after. If there were more like him, everyone s industrialization program  would be a pleasure as well as a  success.REFERENCES1. A. Harvey, Radiation-hardened m agnets using m ineral-insulated conductors, Los Alamos report LA-5306-MS (1973).2. G. Liptak et al., Radiation tests on selected electrical insulating m aterials .. CERN re­port 85-02 (1985).3. Many CERN reports by Van de Voorde: CERN 72-7 is probably the best summ ary:CERN 70-10 EpoxyCERN-ISR-M AG/68-59 HosesCERN 68-13 Epoxy Electrical 1968CERN -ISR-M A G /68/44 Epoxy MechanicalCERN 69-12 M aterialsCERN 70-5 Polymers-High Energy AcceleratorsCERN-ISR-M AG/73-36 DosimetryCERN LAB II-RA /72-10 Electronic Com ponentsCERN-ISR-M  AG/67-3 EpoxyCERN-ISR-M AG/PS-6464 PaintsCERN-ISR-M AG/67-19 Glass ReinforcedCERN-ISR-M AG/68-14 LubricantsCER N -ISR -M A G /PS/6455 TextilesC E R N -M PS/66-22 W ater-RadiolysisCERN -ISR/M A /75-38 Polymers a t Cryogenic Tem peratureCERN 72-7 Selection Guide for Organic M aterials inNuclear Engineering4. M.M. Holland and J. Shill, R adiation-resistant m agnet coils from hard-anodized alu­minum conductors, IEEE Trans N S -2 0 , 3, 708 (1973).1185. R .E . Sheldon and G. Stapleton, Construction of m agnets for particle accelerators using cem entatious m aterial, Rutherford report RHEL-R185 (1969).6 . R.L. Kaizer and M. M attier, M ineral-insulated m agnets, C E R N /SPS /E M A /77-3  (1977).7. A. Harvey, Radiation-hardened m agnets using m ineral-insulated conductors, MT-4 (CON F. 720908), Brookhaven, p. 456 (1972).8 . Ceram aseal Inc., New Lebanon, N.Y.9. A. Harvey, Experience with the LAM PF m ineral-insulated m agnets, M T-6 , Bratislava, p. 551 (1977).10. R .J. Grieggs, D .J. Liska, and A. Harvey, Radiation-hardened field coils for FM IT quadru­poles, IEEE Trans. N S -3 0  # 4 , 3617 (1983).11. D. George, M agnets w ith m ineral-insulated coils a t SIN, M T-5, Rome, p. 719 (1975).119SEARCH COIL MEASUREMENTS OF PARTICLE ACCELERATOR MAGNETSK. N. Henrichsen CERN, CH-1211 Geneva 23, SwitzerlandABSTRACTSpecific problems related to search coil measurements in magnets for particle accelerators are discussed. This includes the coil manufacture and calibration as well as the measurement of the magnetic flux.INTRODUCTIONElectromagnets used as beam guiding elements in particle accelerators and storage rings require very tight tolerances on their magnetic fields. Construction techniques and measurement equipment must therefore match these requirements. The length of the magnet is usually smallcompared to the wavelength of the betatron oscillations. This means that only the integrals of the magnetic field and its derivatives along the beam axis are of interest. The magnetic measurements are important at various stages of an accelerator project: design, construction and operation.MEASUREMENT METHODSTwo factors influence the choice of the measurement method and equipment: the required quality of themeasurements and the number of magnets. It is interesting tonote that while the measurement methods have remained virtually unchanged for a very long period, the equipment has been subject to continued development.Figure 1 shows the accuracy which can be obtained in an absolute measurement as a function of the measured field for a few commonly used methods.The most commonly used method for the measurement of accelerator magnets is the fluxmeter method, which is the only one we will discuss in detail below.Hall probes are widely used for field mapping in spectrometer magnets [1] and are also preferred for strayfield measurements around accelerator magnets and in special applications where the use of a search coil mechanism should be avoided.The nuclear magnetic resonance (NMR) measurement is mainly used for calibration purposes, but has also been used for precise measurements of field integrals in long dipoles.120FI EL D ( T e s t a )Fig. 1. Magnetic measurement methodsFor a more complete description of the various measuring methods, reference is made to two classical bibliographical reviews [2,3] .THE INDUCTION METHODBased on the induction law, this method is the oldest[4] of our currently used methods for magnetic measurements, but it can be very precise [5]. It is also the most precise method for the measurement of the direction of the magnetic flux lines which is of particular importance in accelerator magnets. Measurements are performed either using fixed coils in a dynamic magnetic field or by moving the coils in a static field.Very high accuracy may be reached in differential fluxmeter measurements using a pair of search coils connected in opposition, with one coil moving and the other fixed. A large variety of coil configurations are used in magnetic measurements, ranging from the simple flip-coil to the complex harmonic coil systems used in fields of cylindrical symmetry [6,7]. The choice of geometry and method depends on the useful aperture of the magnet. The sensitivity of the fluxmeter method depends on the coil surface and the quality of the integrator.The coil method is particularly suited for measurements with long coils in beam guiding magnets [8], where the precise measurement of the field integral along the particle trajectory is the main problem. In this case the geometries are chosen so as to link with selected field components [9]. The search coil is usually wound on a core made from a mechanically stable material in order to ensure a constant121coil area and the wire is carefully glued to the core. Glass with low thermal dilatation or ceramics are often used as core materials. During coil winding the wire must bestretched so that its residual elasticity can assure a well defined geometry and mechanical stability of the coil.The coil-integrator assembly can be calibrated to an accuracy of a few tens of ppm in a homogeneous magnetic field with reference to a nuclear magnetic resonance probe. Not only the equivalent surface of the coil must be measured, but also its median plane which often differs from its geometric plane due to winding imperfections. In the case of long measurement coils, it is important to meet very tight tolerances on the width of the coil. If the field varies strongly over the length of the search coil, it may be necessary to examine the variation of the effective width of the coil. A hybrid permanent dipole has been developed for this purpose at LURE, Orsay [10]. It has a magnetic length of about 60 mm and a useful width of about 30 mm.THE FLUX MEASUREMENTInduction coils were originally used with ballistic galvanometers and later on with more elaborate fluxmeters. The coil method was improved considerably with the introduction of the classical electronic integrator, the Miller integrator, but it remained necessary to employ difference methods in precision measurements [11]. The advent of digital voltmeters made fast absolute measurements possible and the Miller integrator has remained the most popular fluxmeter. With the development of solid state d.c. amplifiers, this integrator has become inexpensive and is often used in multi-coil systems.Fig. 2. Analog integrator122Figure 2 shows an example of such an integrator. It is based on a d.c. amplifier with a very low input voltage offset an a very high open loop gain. The thermal variation of the integrating capacitor is the most critical problem. The integrating components are therefore mounted in a temperature controlled oven. Another problem is the decay of the output signal through the capacitor and the resetting relay. Careful guarding of these components is therefore essential in order to reduce the voltages across the critical surface resistances.The dielectric absorption of the integrating capacitor sets a limit to the integrator precision. The best choice is a teflon capacitor. It has a temperature coefficient of -40 ppm/degree C, a dielectric absorption of 30 ppm and a very high insulation resistance. A suitable integrating resistor is much easier to find. Most metal film resistors have stabilities and temperature characteristics matching those of the teflon capacitor. Commonly used time constants are between 10 and 200 msec. The sensitivity of the integrator is limited by the d.c. offset and low frequency input noise of the amplifier. A typical value is 0.5 jiV which must be multiplied by the measurement time in order to express the sensitivity in terms of flux. The overall stability of the integrator time constant proved to be better than 50 ppm over a period of three months.A few electronic integrators have been developed by industry and are commercially available. The choice is, however, rather limited.In recent years, a new type of digital integrator has been developed, which is based on a high quality d.c. amplifier connected to a voltage-to-frequency converter (VFC) and a counter.Fig. 3. Digital integrator123The digital integrator shown in figure 3 was developed at CERN [12] and is now commercially available. The input of the VFC is provided with an offset of 5 V in order to provide a true bipolar measurement. This offset is balanced by a 500 kHz signal which is subtracted from the output of the VFC. Two counters are used in order to be able to measure with continously moving coils and provide instant readings of theintegrator. One of the counters can then be read and resetwhile the other is active. In this way no cumulative errors will build up. This integrator has a linearity of 50 ppm. Its sensitivity is limited by the input amplifier as in the case of the analog amplifier.This system is well adapted to digital control but imposes limits on the rate of change of the flux since the input signal must never exceed the voltage level of the VFC. The integration period must be of the order of a second if one wants a reasonable resolution.REFERENCES[1] K.N. Henrichsen, Journal de Physique, Cl-1984. 937[2] J.L. Symonds, Rep. Prog. Phys. 18, 83 (1955)[3] C. Germain, Nucl. Instr. and Meth. 21, 17 (1963)[4] W. Weber, Ann. der Physik 2 , 209 (1853)[5] J.H. Coupland, T.C. Randle, M.J. Watson,IEEE Trans, on Magn. MAG-17. 1851 (1981)[6] C. Wyss, Proc MT5, 231 (1975)[7] O. Pagano, P. Rohmig, L. Walckiers, C. Wyss,Journal de Physique, Cl-1984. 949[8] E.A. Finlay, J.F. Fowler, J.F. Smee,J. Sci. Instrum. 27., 264 (1950)[9] B. de Raad, Thesis, Univ. Delft, 55 (1958)[10] A. Dael, private communication[11] G.K. Green, R.R. Kasner, W.H. Moore, L.W. Smith,Rev. Sci. Instr. 24/ 743 (1953)[12] P. Galbraith, private communication124AC MAGNETIC MEASUREMENTS OF THE ALS BOOSTER SYNCHROTRON DIPOLE MAGNET ENGINEERING MODEL*M. I. Green, E. Hoyer, R. Keller, and D. H. Nelson Lawrence Berkeley Laboratory, University of California, Berkeley, CA 94720ABSTRACTWe made a minimal set of AC magnetic measurements of the engineering model of the ALS Booster Dipole Magnet as part of the process of qualifying its design for production. Magnetic induction integrals over paths approximating electron-beam trajectories were measured with long curved coils connected to an electronic integrator. Magnetic induction was measured with point coils and an integrator and independently with a Hall-effect Gaussmeter. These quantities, and magnet current, were displayed on a commercial digital storage oscilloscope as parametric functions of time.The displayed waveforms were stored, processed and redisplayed as representations of selected magnet parameters. A waveform representing the magnet's effective-length was created by dividing the integral waveform by the magnetic induction waveform. Waveforms of the transfer functions were produced by dividing both the integral waveform and the magnetic induction waveform by the current waveform. Pairs of matched coils, connected in series opposition, provided differential measurements of field uniformity. Quadrupole and sextupole coefficients were derived from the uniformity data.These magnet parameters were measured at 2 and 10 Hz frequencies. Together with measurements of the magnetic field at selected DC levels, the AC measurements demonstrated that the magnet design met specifications and qualified it for production.INTRODUCTIONThe ALS is a third generation, 1-2 GeV synchrotron radiation facility specifically designed to maximize the brightness of the radiation from wigglers and undulators1. This project includes a low-emittance electron storage ring optimized at 1.5 GeV, an injection system which includes a 50 Mev linac and the 1.5 GeV booster synchrotron, and a complement of insertion devices and photon beam lines. Twenty-four dipole magnets will provide the main guide field for the booster synchrotron. The synchrotron is intended to operate at 1 Hz, but the magnets are designed for 10 Hz operation.The booster synchrotron dipole magnet is of the split H type with flat pancake coils as shown in Figure 1. To minimize the stored energy and power requirements, the core is curved to follow the electron-beam trajectory. It is constructed from 0.025 inch, C5 coated, M36 silicon steel laminations. Table 1 gives dipole magnet design parameters2. An engineering model of this dipole magnet has been designed, fabricated, qualified through magnetic measurements, and is now in production.*This work was supported by the Director, Office of Energy Research, Office of High Energy and Nuclear Physics, Division of High Energy Physics, of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098.125Fig. 1 The ALS Booster Synchrotron Dipole Engineering M odelT able I. ALS 1.5 GeV B ooster Synchrotron Dipole M agnet Design ParametersTable II. D ipole Engineering M odel Magnetic M easurementsInjection M agnetic Field .0416TDesign M agnetic Field 1.248 TBend Angle 15 DegreesEntrance/Exit Edge Angles 7.5 DegreesM agnet Bend Radius 4 .0 1 0 7 mM agnetic Length A long O rb it 1.050 mM agnet Vertical Aperture 4.4 cmGood Field Aperture W idth +/- 3 .0 cmGood Field Aperture H eight +/- 1.8 cmField Q uality +/- 1.0 x l0 '3(excluding fringe field)Excitation W aveform  @ 1 H z M odified sawtooth @  10 H z DC biased sinusoid1. DC M easurem ents at selected field levels1.1 Central Vertical-Field1.2 Central Vertical-Field Integral1.3 M idplane Vertical-Field Integral Uniformity2. AC M easurem ents o f param etric functions o f time2.1 Central V ertical-Field U niform ity with/without the vacuum cham ber2.2 Central Vertical-Field Integral2.3 M idplane Vertical-Field Integral Uniformity126MAGNETIC FIELD PARAMETERS FOR QUALIFICATIONThe objective of the magnetic measurement effort for the booster dipole engineering model was to qualify the magnet design for production. Verification of the 2-D magnetostatic design and evaluation of the higher order field terms quadrupole, sextupole, etc., largely generated at the ends were mostly done with DC measurements7. AC measurements were carried out largely to investigate magnet and vacuum chamber eddy current effects. The measurements performed are tabulated in Table II on the preceding page.AC MEASUREMENT TECHNIQUES AND INSTRUMENTATIONFigure 2 shows test equipment configurations used for most of the AC measurements Instrumentation features are tabulated in Table HI. Calibration procedures provided 0.1% absolute accuracy of both the Hall probe3 and the coil sensitivities and enabled us to match the sensitivity of coil pairs to better than 0.02%.4 In operation we minimize the differential signal when one coil is on the magnet centerline and the bucking coil is at an arbitrary, stationary reference position (x=x0). Detailed descriptions of the measurement procedures are contained in a separate report.5Fig. 2 describes the magnet coordinate system used for the AC measurements. The origin of the cartesian coordinate system is defined as the centroid of the volume between the pole tips; +x extends the radius of curvature vector from the origin; +y is the upward normal to the lower pole; s is the distance from the origin along the curve consisting of a 15° circular arc (defined by the magnet bend radius) extended at each end by straight lines.The measurement technique employed a Hall-effect Gaussmeter and point coils connected to an electronic integrator to measure magnetic induction. Magnetic induction integrals were measured with line coils shaped to conform to the nominal beam trajectory and connected to an electronic integrator. Magnet current was measured by use of a current monitoring resistor (shunt) in series with the magnet and its power supply. Data acquisition was with a digital storage oscilloscope (DSO) where signals from the Gaussmeter, integrator, and shunt were recorded and stored as parametric functions of time (t). The DSO allowed two signals to be acquired simultaneously.Fig. 2 AC Measurement System Block Diagram127Table III. Instrumentation FeaturesEquipm ent/M odel No. FeaturesStorage O scilloscope - 500 M Hz bandwidthsTektronix M odel No. 11401 M ultiple 10 Bit (10,240-point) waveform records 10-ps horizontal resolutionFull program ability  through IEEE-488 or RS-232C (not im plem ented) Sim plified acquisition & processing features available by "touch-screen" controlElectronic Integrator - LBL M odel No. 711-M icroVolt Second resolution (over several minute periods)Point Coils & L ine Coils - Absolute accuracy +/- 0.1 %LBL Designs Relative accuracy +/- 0.02 %Linearity lim ited to the linearity o f the integrator resistanceFlux Standard - Absolute accuracy +/- 0.02%LBL M odel No. 43 Relative accuracy +/- 0.01%Range: .01 - 1.0 Vs selectable in 1,2,5 stepsHall Effect Gaussm eter - Absolute accuracy 0.25% o f full scaleF.W. Bell M odel 811 +/- 0.1%  with linearity correction and in field calibration and zero/offset procedures3For determining transfer functions and effective length, a single pair of signals was acquiredi.e., the integrator signal and the shunt signal for determining the transfer function waveform, and two integrator output signals for determining the effective length function.One of the authors (Don Nelson) developed an AC measurement technique^ whereby variationsin effective length (ALeff) are determined from differential measurements as follows:{ tey (°> ° )d s -By {x0, Oto)L (B*)}(t) , .eff "   rETl-------u T \-----------------= ^eff (0 * Le„(B )eri (B y (0,0,0)} (t) ettw  “U ’The numerator is the differential signal obtained by connecting the point coil in series opposition to the line coil. Len (uivalent) (B*) is numerically equal to Lgff (t*), the effective length evaluated at t=t* i.e., when {By (0,0,0)) (t=t*) = B*. Leq (B*) is realized by adjusting the point-coil divider such that its divided signal precisely cancels the line coil's signal at t=t* Dividing this differential signal waveform by the magnetic induction waveform yields the difference in effective length as shown by the equation. To compute Leff (t) the effective length Leff(t*), determined independently,is added to ALeff. This technique overcomes the accuracy limitation in determining the effective length by the ratio of two parametric functions whose resolution is limited by the resolution of the oscilloscope (see RESULTS - Effective Length).For determining the field uniformity on the aperture midplane, a series of differential waveforms was acquired with a pair o f matched coils electrically connected in series opposition. The acquired waveforms corresponded to the field at the x-position of one of the coils (x = x;) with respect to the field at an arbitrary, stationary, x-position of the second (bucking) coil (x = x0) i.e.,{AB (x j )  = B ( x j )  - B ( x 0 )} for the local magnetic induction uniformity and {AjBds ( x j )  =jBds (x j )  - jBds ( x 0 )} for the integral uniformity. Each series started and ended with the first coil located on the aperture centerline (x j  = o). The averaged waveforms at Xj = o were subtracted from each of the other waveforms. This process partially compensated for integrator drift between the first and last measurements, although this drift was negligible.128RESULTSA detailed description of all measurements and results can be found elsewhere7.Effecdve LengthFigure 3 displays the magnet's effective length as a function of magnetic field. The measurements indicate that the effective length is 1.047 ± .001m up to 1.1 T and decreases to 1.044 m at 1.25 T. This decrease in effective length at higher field values is attributed to saturation effects. Although the effective length is somewhat shorter than the design length (1.05m), the resultant displacement of the accelerated electron beam will be negligible.We have included data identified as unreliable in Fig. 3 in order to illustrate a limitation in the 1st method used for determining the effective length. The effective length is defined as:jByds/By(s=o). In the first method By(s=o) and jByds were measured as parametric functions of time over the magnet's operating field range (.04 T to 1.25 T). The accuracy of the acquired data is limited by the 10 Bit resolution of the oscilloscope to 0.1% of the maximum signal size or 1.25 x 10-3 T which is about 3% of the minimum field. By expanding the time scale the acquired waveform obtains increased accuracy at lower values and the computed effective length is more reliable as is illustrated by the curve labeled .04 < B < 0.2 T in Fig. 3.The technique for determining variations in effective length from differential measurements eliminates the uncertainty in the effective length at low field as is indicated by the third curve in Fig. 3.1.07j rszO) 1.06c.CDCD>2  1-05 LU T 3  \—Eo 1.04aCD1.030.0 0.5 1.0Magnetic Field (T)Fig. 3. Effective length of the Booster Dipole from three different measurements at 2 Hz.Field UniformityField uniformity measurements were made with point coils, both with and without vacuum chamber sections and with integral coils in order to distinguish between eddy current induced effects, pole shape effects, and magnet end field effects on the field quality. Within the vacuum chamber (inserted between the poles) measurements were made at only three x-positions due to space restrictions. Data taken at three positions allowed determination of quadrupole and sextupole coefficients as is illustrated in Fig. 4.C=OoCDc.O-+—•oG3i—XLUDesign Lengthj B d s / B0.2T<B<1.25TJ B d s / B0.04T<B<0.2TDifferentialMethod0.04 T<B<1.1T129For the multipole analysis, we assume that a function (x) - either a vertical magnetic field, By, or its integral, jBy ds, - is composed of a linear and a quadratic term, identified as quadrupole and sextupole components, respectively i.e., O(x) = bi * x + b2 * x2 = O^uad (x) + cjjsext (x)Symmetry relations dictate that: O(-x) = bi * (-x) + b2 * x2 = -Oquad (x) + (])sext (x)From combinations of these two equations we find:Oquad (X) = (O(x) - O(-x)) /  2; <£sext (x) = (<£(x) + O(-x)) /  2; bi = (O(x) - <D(-x)) /  2x; and t>2 = (O(x) + O(-x)) /  2x2O VxO  measured sextupole/  quadrupoleFig. 4 Derivation of quadrupole and sextupole coefficients from magnet field data measured at three locations across the magnet.The results obtained from 2 Hz and 10 Hz measurements are displayed in Figs. 5 - 8 .  Data that represent point coil measurements have to be multiplied by the effective length, to be compared to the integral data. Only those results are displayed that appear not to be affected by the discussed data acquisition accuracy problems.Expressing the quadrupole term of the field integral as the linear change in effective length,Leff (x), we define wedge angles, a , at each end of the magnet as:a  = arc tan [bi/2By(o,o,o)] = arc tan [(Leff (x) - Leff (-x)) /  4x]130In Fig. 5 the effective wedge angles derived from the integral uniformity measurements at 2 Hz are plotted for each end of the magnet. The average wedge angle equals -0.3 degrees which means that the magnet is effectively shorter at its outer side (x>o).The sextupole term of the field integral, normalized to the beam rigidity Bp is given by:1 d 20  2b2m = ----------- = -----Bp dx2 BpThe integrated sextupole values (m), measured at 2 and 10 Hz without a vacuum chamber are plotted in Fig. 6. The close agreement of the data taken at 2 and 3 cm means that up to 3 cm there is a uniform sextupole field. The stronger sextupole values at 4 cm may be interpreted as higher harmonics that could not be analyzed with the measurements made. In absolute terms, the 2 Hz data appear to be more accurate than the 10 Hz data.Magnetic Field (T)Fig. 5 Effective wedge angles for each end of the magnet measured at 2 HzMagnetic Field (T)Fig. 6 Integrated sextupole values without a vacuum chamberIn Fig. 7 the local sextupole strengths, measured near the magnet center without the vacuum chamber, are plotted for both 2 and 10 Hz excitations. The curves for 2.3 and 3.2 cm are essentially flat with negligible amplitude, showing slight saturation effects only at the highest field levels investigated. This means that the transverse pole contours, optimized with a 2-dimensional simulation code, fulfill the quality requirements outlined above.Fig. 8 shows local sextupole strengths measured near the magnet center- at 2 Hz with a 0.8-mm thick stainless steel vacuum chamber and at 10 Hz with a 0.3 mm thick chamber. Strong eddy current effects are seen at the lower field levels as expected.When the Booster Ring is operational, with vacuum chambers in place, the sextupole strength will be determined by three effects: eddy currents, end geometry, and saturation. The first and last of these are time-dependant, each one being significantly strong when the other one is negligible. Due to the opposite signs of eddy-current induced and geometrical sextupoles, the absolute value of the total sextupole strength for the entire excitation cycle is moderate.131Magnetic Field (T) Magnetic Field (T)Fig. 7 Local sextupole values without Fig. 8 Local sextupole values inside vacuumvacuum chamber chamberNote: The 2 Hz curve is composed from three different measurement seriesCONCLUSIONThe AC magnetic qualities of the ALS Booster Dipole were measured at 2 and 10 Hz frequencies. Many measurements, taken over the entire excitation range, showed intolerably high errors at low field values due to the resolution of the data acquisition hardware used. However, by combining measurements that were taken over narrower ranges and in one case by employing a differential measurement technique, these errors were compensated for. The magnetic length and field uniformity values obtained demonstrate that the magnet design meets the specifications and qualify it for production.ACKNOWLEDGEMENTSThe authors wish to thank Klaus Halbach, LBL, for the idea of using a storage oscilloscope; Mike Lapolla, Tektronix, for making the Tektronix 11401 digital oscilloscope available to LBL; John Cerino and his staff at SSRL for making the SPEAR injector prototype 10 Hz power supply available; and to various members of the ALS project team for their help. We especially thank Sharon Fujimura and David VanDyke for their assistance in preparing this paper.REFERENCES1. ”1-2 GeV Synchrotron Radiation Source", Conceptual Design Report, LBL PUB-5172, (July, 1986).2. E. Hoyer, "Light Source Booster Dipole Magnet Design Calculation", LBL Engineering Note M6673, (Dec. 1987).3. D.H. Nelson and D.A. VanDyke, "Recalibration of a Hall Effect Probe", LBL Engineering Note MT 387 (1988). 64. D.H. Nelson and D.A. VanDyke, "Calibration of a Curved Integral Coil for Measurments of ALS Booster Dipoles", LBL Engineering Note MT 385 (1988).5. D.H. Nelson, M.I. Green, E. Hoyer, R. Keller, and D.A. VanDyke, "Detailed Procedures for Making A/C Measurements of the ALS Booster Dipole Prototype Magnet", LBL Engineering Note MT 390 (1988). 6 66. D.H. Nelson, Differential Measurements for Determining the Effective Length of a Magnet Excited by AC Current", LBL Engineering Note MT 391 (1988).7. R. Keller, M.I. Green, E. Hoyer, Y.M. Koo, and D.H. Nelson, "Measured Properties of the ALS Booster Synchrotron Bending Magnet Engineering Model", Internal Report LSAP-50, Lawrence Berkeley Laboratory, (1988).132LOSS M EASUREM ENT PROGRAM S AT TR IU M F Alan O tterTR IU M F, 4004 Wesbrook Mall, Vancouver, B.C., C anada V6T 2A3W ashington NevesD ept, of Electrical Engineering, University of British Columbia, Vancouver, B.C., C anadaV 6T 1W5ABSTRACTDuring the KAON Factory Project Definition Study year we plan to  directly m easure m agnet losses a t sinusoidal excitation and at the proposed dual frequency excitation of the booster and driver rings. The losses due to  transverse fields in the conductors will be m easured using the NINA m agnets and core losses will be m easured on steel samples using an Epstein Bridge m ethod which allows dc bias levels to  be applied. The proposed tests are described and some prelim inary findings are presented. The aim of these tests is to allow us to  understand the loss processes and to  alow us to  calculate these losses with greater accuracy and confidence.EDDY CU RREN T LOSSES IN CONDUCTORSThe power supply tests set up by Klaus Reiniger using the NINA dipoles give us an opportunity  to  install varying size conductors in the m agnet apertu re  and to  m easure the eddy current losses directly by the tem perature  rise of the cooling water.For a  square hollow conductor, as shown in Fig. 1, subject to  a transverse a lternating  field Bcosujt ,  the estim ated loss1 isdia = dFig. 1. Square hollow conductor in a uniform alternating field.For standard  hollow conductors in a field of 0.4 T  we estim ate the losses a t 50 Hz tobe133Conductor size Power loss0.162 X 0.090 in. (411 x  2.3 mm)0.3648 X 0.204 in. (9.27 X 5.18 mm)0.650 X 0.363 in. (16.51 X 9.22 m m )22 W /m  580 W /m  5824 W /mIt should be possible to  m easure the power loss to  about 5%. It is planned to  m easure a t both sinusoidal and dual frequency excitation. It is expected th a t calculating losses a t an average frequency of 50 Hz for the booster m agnets will be very inaccurate. We anticipate th a t the dual frequency loss will be given by the average of the 30 Hz and 100 Hz losses.Straight square hollow copper conductors of varying size will be installed in the gap. I hermocouples will be used for tem perature  m easurem ent and flowmeters for the flow. We also intend to  make m easurem ents on a stranded cable conductor which has been loaned to us by Los Alamos. This particu lar conductor was made by Brown Bovari and is in transit.CO RE LOSSES IN STEEL SU BJECT TO NON-SINUSOIDAL EXCITATIONM anufacturers d a ta  on core losses are invariably given at sinusoidal excitation and specific excitation levels. 1 here does not appear to be any simple method described in the literature  of ex trapolating this data  to our proposed dual frequency excitation waveform superimposed onto a dc bias field. We have therefore decided to m easure core losses using an Epstein bridge modified to  accept a dc bias winding. We propose to use an ASTM procedure so th a t we can relate our m easurem ents to m anufacturers published data.. The Epstein bridge is being m ade and in the m eantim e we have tried out the m ethod in UBC’s D epartm ent of Electrical Engineering using a 1 kVA transform er core. The results which we have obtained are prelim inary because some of the details of the transform er are not accurately known. We will describe only our general findings.Conventional electrical engineering splits the to ta l core loss into two components:W  = u>h and cue ,where“ h = k h - B xm - f  cue =  ke B 2m t 2 f 2 ,wherekh =  hysteresis loss coefficient B m — maximum  flux density, T  f  = frequencyke = eddy current loss coefficient t — lam ination thicknessWe can writeW  ~ J  =  kh Bm +  keB 2mt2f  .134So if the loss/cycle is plo tted  against frequency a straightline with a slope of k e B ^ t 2 is expected. The in tercept a t / = 0  gives the value of the hysteresis term  and hence the value of x  can be determ ined. In practice the curve becomes nonlinear below 20 Hz and the interceptis difficult to  determ ine .2Figure 2 (a-c) shows the transform er circuit and the results obtained at various am pli­tudes of the field. The value of the hysteresis coefficient appears to  vary with field am plitude, as shown in Fig. 2(c).-> TO ANALYZERF ig. 2 (a). C irc u it used for ac loss m easu rem en ts .F ig . 2 (b ). M easu rem en ts o f to ta l  loss per cycle for sinuso idal fields.F ig . 2(c). D ete rm in a tio n  o f x exponen t.135We have also used the circuit of Fig. 3(a) to  look a t the effect of superim posing a  dc bias field. It was necessary to  use two back-to-back transform ers to  reduce the value of the induced ac current in the dc bias circuit. The results, Fig. 3(b), show th a t the loss/cycle above 30 Hz is linear and th a t the dc bias increases the losses, so th a t they m ust be estim ated from the peak field ra ther than  only the ac component.Tl A•  / ' \ T0 * b  J* -ANALYZER j-»T 0  ANALYZER^shL VdcFig. 3(a). Circuit used for measurement of total loss due to an ac bias field.FREQUENCY (Hz)Fig. 3(b). Measurement of total loss per cycle for an ac bias field.SUMMARYThese results are prelim inary and are incomplete. In tim e it is our aim to  complete them  with all three conditions:a) Sinusoidal ac excitation onlyb) Sinusoidal ac excitation w ith a dc biasc) Dual frequency sinusoidal excitation with a dc biasWe hope when these m easurem ents are completed to  be able to  ex trapolate  d a ta  from steel m anufacturers to  our dual frequency excitation with a good degree of accuracy.REFERENCES1. G .W . C arter, The Electromagnetic Field in its Engineering Aspects (Longmans, New York 1967), pp. 254-256.2. F. Brailsford, Investigation of the eddy-current anomaly in electrical sheet steels, JIE E  95 P t II, 38 (1948).136M ETHODS OF ESTIM ATING IRON LOSSES AND FIELD ERRORS IN AC MAGNETSP.A. ReeveTR IU M F, Physics D epartm ent, University of V ictoria, V ictoria, B.C., C anada V8W 2Y2ABSTRACTSome semi-empirical formulas for estim ating the iron losses due to  hysteresis and eddy currents are presented. Also a m ethod of estim ating the field error or phase lag caused by eddy currents in components of m agnets is given.INTRODUCTIONAc m agnets have hysteresis and eddy current losses in the steel. In order to  reduce these losses, the iron in the m agnet is generally lam inated silicon steel. M anufacturers of lam inated steels quote the steel losses referenced to  a fixed field, typically 1 T , and a fixed frequency, typically 60 Hz. The m agnets to be designed for the KAON Factory will be operated  at varying fields and frequencies. The formulas presented here enable the m anufacturers d a ta  to be ex trapolated  to  the fields and frequencies to  be used for the KAON Factory m agnets. The formulas are based on very simple models, assume no sa tu ra tion  occurs, the lam inations are th in  and the frequencies are m odest. These formulas can be used to  see if a potential problem exists in the m agnet design. If a problem does exist and is unavoidable, then some more sophisticated techniques must be used to  study it, such as the com puter program  P E 2D.PO W ER LOSSESHysteresis LossesFrom Boyajian and Camilli1 the hysteresis losses can be estim ated usingPh = k hf ( B m)n , (1)where Ph is the power loss, kh is a  m aterial constant, /  is the frequency, B m is the maximum m agnetic field, and n is the Steinm etz exponent, which can vary from 1.6 to  2.6. For example, silicon steel lam inations a t 1 T , 60 Hz, w ith a Steinm etz exponent of 2, would typically have a hysteresis loss of about 0.5 W /kg.Eddy C urrent LossesAgain from Boyajian and Cam illi1 the eddy current losses can be estim ated fromPe = kef 2 B 2t2 , (2)where ke is a constant for the m aterial, /  is the frequency, B  is the rms value of the m agnetic field, and t is the lam ination thickness. For example, for SiFe a t 1 T  rm s, 60 Hz and a lam ination thickness of 0.36 mm, the losses are typically about 1 W /kg . This loss becomes the dom inant one at higher frequencies but the losses can be readily reduced by using thinner lam inations.137Fig. 1. Eddy current phase lag param eters.PHASE LAG AND FIELD ERRORSEddy currents produced in components of a m agnet will produce local m agnetic fields, which have an am plitude of opposite sign to the main field. The net effect is to  produce alocal phase lag in the rise of the field. This phase lag can also be thought of as a local fielderror. The phase lag can be estim ated using the theory of Lam m eraner and S tafal,2 modified to  the formf io h  ab0 7r26( a/ b+b/ a)  ' ^where no is the space perm eability, I is the m agnetic path  length around the component (Fig. 1), 8 is the m agnet air gap size, a is the component dimension norm al to  the field and parallel to  the beam, i.e. into the paper in Fig. 1, b is the thickness norm al to  the field and the beam , 7  is the conductivity, and r 0 is the phase lag. In cgs units /j,0 is OAn X 10“ 8. Therefore, Eq. (3) becomes_ 1.27 X I0~9ljab  0 8(a/b-\-b/a) ^For a linear ram p of length r  and field B  the phase lag is given bySB B—  =  -  (5)T o  T  V 'and asr = h  ■ <6>Therefore,SB  2.55 x 10_9/7 /a 6B  ~  6(a/b + b / a ) ‘ ^138BO OSTER D IPO LE EXAM PLEAs an example for the above equations, a possible design for a booster dipole was done using M AGDES .3 The m agnet had a gap of 10.68 cm, a pole w idth of 25.16 cm, a length of 318 cm, a m aximum  field of 10.6 kG, and a current of 1000 A. W ith  SiFe lam inations, thick and a frequency of 100 Hz, the hysteresis losses are 12 kW  and the eddy current losses are 25 kW . The field errors caused by the lam inations are very small, 2 X IO-5 but if a stainless steel vacuum box were used, which is 1 cm thick, it would introduce an error of about one per cent.REFERENCES1. A. Boyajian and G. Camilli, Transformer Engineering (Wiley, New York, 1967).2. J.M . Lam m eraner and M. Stafal, Eddy C urrents (Iliffe, London, 1966).3. P.A. Reeve et al., Proc. 9th Int. Conf. on M agnet Technology, Zurich (SIN, Villigen, 1985), p. 763.139R. B aartm an , G. Bowden, R. Cassel, G. Clark, D. Fiander, A. Harvey, G. Mackenzie,V. Rodel, J. Riimmler, 0 . Szavits, A. Thiessen, G. W ait, U. W ienands, and R. W injeQuestions considered1 . Can a 10 ns electric field deflector be m ade for the Accumulator?2. B  versus /  rise tim e. Lag?3. Rise, /  B d l , m agnet length trade-offs4. Coupling im pedance questions5. KAON Factory needsa) Are they w ithin existing technology?b) C ost/risk  of shorter than  80 ns rise timesc) How to  build something in a yeard) How much tunnel space needed?e) How can the kickers be m onitored/controlled?f) How does a test facility get set up?6. W hat do the new lattices mean for the kickers?Electric field deflector1. 10 ns looks impossible for 4 m rad. (Would need many >100 kV switches and even then, doubtful!!)2. Em pty bucket scheme allowing ~30 ns rise looks interesting.3. Job easier if deflector can be shorted to  switch beam.4. Precise calculations needed to  determine num ber of sw itches/voltage for 30 ns beam hole.B - I  tracking1. Experience differs CERN /SLA C.2. Need to  verify /  B d l  rise by both  probe and voltage difference m ethods.3. CERN /SLA C differences may result from different ferrite g rades/path  length.Kicker design1. Take as high a voltage as can be comfortably handled.2. Use as much length as possible.3. Keep Z q high -  be tte r frequency response-  lower switch currentREPORT ON KICKERS WORKING GROUP DISCUSSION1404. Avoid complications wherever possible -  use m atched m agnets.5. W atch rise tim e definition for f  B d t  — take 1-99% as lim its (can add 20 ns to  usual rise w ith 5-95% lim its).Coupling im pedance1. Topic where we have least knowledge.2. DESY designs for electron machines cost a lot of vertical aperture (lowers Z 0 \\).3. CERN PS still OK at 2 .4 x l0 13 in 20 bunches, but for how much longer.4. A t least try  to  make kicker retu rn  conductor(s) continuous and make direct connections to  vacuum  tank  flanges.5. Do not believe delay line kickers to  be worse than  lum ped ones.KAON Factory kicker needs1. W ithin existing technology2. Should not go for shorter rise(fall) times! Don’t forget:-  50 Hz-  low losses-  cost-  thyra tron  life3. Building prototypes in a year -  not possible even in labs w ith existing infrastructure. Instead,a) borrow from other labsb) s ta r t calculations, design and m easurem ents on borrowed equipm ent4. Tunnel space -  no problem5. M onitoring/control -  existing systems present no problems; much to  learn6 . HV and m onitoring facilities needed (but also a dedicated team  of staff)The new lattices1 . A kicker designer’s dream2. All is now much easier:-  fewer modules 23 versus 45-  to ta l kV down 1377 versus 2477-  higher Z0 25 versus 15-  stored energy down 31% versus 100%Preliminary proposals for kickers for KAON factory - Version 2141<DgbO ►aX CDCDPFN 2P sX4)CDc3-4-33>r*'Z >G<U <U CD  H CO OSXODXCO<DO  \ Z■S ^ - o8 ^  <3-o C!<u-Xo<v£3 JS.t-4 X3  8.*roso"3 ■£ « ^I  f f ’S s o ^ r,2 T3 bO cS< xG.2ocPm<D.53ocOco CO co X X CO h-Ci o d rH CD CD CD Ho CD rH rH rH rH rH rHrH rH CD rH rH rH OD COod 00 CM CM CM CM 00Tf TT LO LO LO LO LO00 oo CM CM CM CM O oorHoH00 00 00 00 OO 00o o o o o o o oCO CO CO CO CO CO CO CO= 1377)Tf LO LO LO LO CO CO >oo 00 LO LO LO LO LO LO -X3-4-3EhCO CO rH CO CO CO CO aCi CD CO LO LO LO 3o o CD CD CD CD CD CD LOCM©COCD3LO LO LO LO LO LO LO LO XCM1 CM1 CM1 CM1 CMi CM CM CM oGrH1rHICO1CM1CM1CM1001TJ«HCOCM00 00 00 t— h- h- CO TfX X X X X X X XLOrHLOrHLOrHLOrHLO LOrHLOrHCOLO LO LO o o o o oCD c d CD CM CM CM CD CMrH rH CM rH rH rH X CO2.5LOrHLOrHLOrHCDrHCDrHGO GOWo-h33SGoo-<G0GESJ O NJK  tS W „  _£ <2 .22. °  ■' „  to 5 > >» W  -g.r- Ch riN  o  N  O  Nl S  Na  a  vs a  •= a, «  O  g  O  B  O  I  O~ 10 '& *  \ooPQooCQo-23O0)><DXJGV-4^>Xwo-4-3cO+CMCMCOCM>3-4-3o-4-3c fLOGXosCOXCDt-’3crcdG.2 *00 S—I5>rtVersion 2 has only 31% of stored energy of Version 1142B. Berkes, E. Blackmore, G. Clark, C. Haddock, K. Halbach, M.R. Harold,E. Hoyer, D. Lobb, N. M arks, W . Praeg, K. Reiniger, M. Sabado, H. Sasaki,V. Soukup, U. W ienands and M. ZanolliQuestions considered1 . Considerations determ ining peak field in fast cycling m agnets• How to achieve optim um  field quality over a full range of excitation• Relative im portance of factors such as yoke flux density, cost optim ization, acceptable loss lim its and m agnet size2. Considerations on eddy current losses and conductor design• How accurately can iron and copper losses be calculated and how accurate are esti­m ates of inductance, etc. w ith a nonsinusoidal waveform?• W hat type of conductor should be used, directly or indirectly cooled?• Is it b e tte r to  use grain-oriented or nongrain-oriented steel?3. Considerations on fabrication techniques• A dvantage/disadvantages of one-piece lam ination vs. split lam ination with bolted assembly• Procedures for fabrication curved m agnets4. Considerations on combined function m agnets• Variations from flat pole design• G radient tolerances achievable• Practicality  of gradient control with pole face windings5. Considerations on tolerances and tracking• Should dipoles and quadrupoles be excited in series, so th a t they track together, or separately?• W ill quadrupoles with the same current but different field strengths track together?• W hat tolerances can reasonably be expected?• W hat correction or tuning windings should be provided?6 . Considerations on m agnet installation in the tunnel• Location of chokes and other power supply components• Effect of stray  m agnetic field from one ring, e.g. cycling m agnets on dc m agnets.1. Optim um  field quality will be achieved over full excitation range by:1) Using good quality (high fi) steel2) Low, wide shimsAdvantage should be taken of beam  damping. Average yoke flux density a t peak field should be kept m odest, so th a t the perm eability does not drop below a high value (~1000). This will help in tracking, should dipoles and quadrupoles be in series. P rob­ably the peak air gap field should not exceed about 1.3 T. Cost optim ization is secondary to  tracking and field quality.REPORT ON FAST CYCLING MAGNETS WORKING GROUP DISCUSSION1432. Copper losses can be calculated to  ~5% , inductances to  ~ 2%. Iron loss estim ations may be more inaccurate, bu t can be m easured on prototypes. Indirectly cooled con­ductors are apparently available a t current densities of up to  8 A /m m 2. These will considerably reduce losses compared with directly cooled conductors, and should be seriously considered. The radiation resistance of the copper strand  insulation might be a drawback.Grain-oriented steel is expensive and probably not necessary.3. A one-piece lam ination will make the achievement of assembly tolerances easier, but restricts coil design and implies a vacuum chamber which can be inserted.Curved dipoles similar to  the KAON dipoles have been successfully m ade in various laboratories, and these should be assessed to  find a technique best suited to  the m anu­facturer.4. In a combined function m agnet the peak field is significantly higher than  th a t a t the edge of the good field region. In addition, to  achieve the same good field w idth in a CF m agnet probably requires a wider pole than  is necessary in a pure dipole.G radients to  1% or better are achievable.Control of g radient with pole face windings is difficult due to  induced em f’s. Forces on the windings m ust be taken into consideration, and of course they consume aperture.5. The problem  of tracking will be eased by placing quadrupoles in series w ith the dipoles, bu t if significant changes of tune are envisaged then separate circuits are indicated. P u ttin g  taps on the quadrupole coils is probably im practical.The linearity  of quadrupole effective length w ith energization can be optim ized by careful design of the pole ends. Nevertheless a t some level sa turation  will set in and this regime should be avoided. As a very approxim ate guide, pole tip  fields should not exceed 0.9 T , bu t this assumes a well rolled-off pole end.Tracking of gradients to  be tte r than  1% should not be too difficult.For tune variation about one particu lar working point, A v  = ±0.2 should be enough.6 . If possible chokes should be centralized w ith 100% coupling.Design of m agnets should prevent long-range stray fields, but these may still be a problem  with two rings close to  one another. They may, for instance, induce vibration. S tray fields from prototype m agnets should be m easured and their effects assessed.N atural vibration frequencies of both  m agnets and supports should be taken into ac­count.144Klaus Reiniger, TRIU M F W alter Praeg, ANL Hirochi Sasaki, KEK Neil M arks, E SR F/D LThe discussion considered various features of the dual frequency resonant circuits needed for the fast cycling machines. The 50 Hz repetition ra te  system  was m ainly discussed, bu t m any conclusions are relevant to  the 10 Hz system.1. PO SITIO N  AND NUM BER OF DC BIAS SUPPLIESThe present proposal used twelve separate dc supplies, one for each cell, around the d istribu ted  system . This would lead to a significant increase in the cost of this supply, as twelve small units are a factor of between three and four times more expensive than a single supply.A large single unit would be rated  at about 2 kA, 1 kV and would require stabilization to l x l O 4 or b e tte r  (2 or 3 in 105 should be possible). There was doubt whether Canadian industry  could m anufacture the big unit, whereas the m ultiple units, w ith the lower ratings, could certainly be m anufactured. However, it was agreed th a t this may be more a m atter of confidence than  inability. Politically, buying ‘m odular’ com ponents from different m an­ufacturers m ay be seen to  be advantageous, bu t it was pointed out th a t this could lead to technical problems.A m ajor technical advantage of the single system is th a t it allows one of the choke/capac­ito r cells to  be modified to  separate the alternating current in the choke from the dc source. Diagram is given below:REPORT ON MAGNET POWER SUPPLY SYSTEM WORKING GROUP DISCUSSIONLM CM LM CM LM CMAt each choke/capacitor cell there is sufficient capacitance to  resonate the m agnets in series, and the choke in parallel:C m  =  ^  C c h  =  ’At the modified cell, the choke secondary is split, each half being separately resonated. The bias supply, w ith a central earth , is connected between the two half windings; because these145are separately resonated, no reactive current flows in the dc supply. Resistive power, if made up from the choke prim aries, will flow through the supply, bu t this will only require a modest capacitor, and the large 2-10 F capacitors can be dispensed with.This m odification is possible with 12 separate dc supplies, bu t leads to  the complexity being present in all cells.The two capacitors required for dual circuit operation will be necessary in each of the three banks a t the modified cell.The final decision on this problem will clearly involve political and commercial consid­erations as well as technical argum ents.2. CIRCUIT RESONANCE AND CHOKE CO NSTRUCTIONThe m ultiple resonances, possible in a circuit w ith d istributed inductors and capacitors, are elim inated by ensuring th a t there is strong (i.e. ~ 100%) coupling between choke secon­daries. In practice, this involves having a prim ary winding strongly coupled magnetically to each secondary, and then strongly couping the prim aries, electrically in parallel. The circuit then oscillates a t a single resonant frequency, irrespective of inductor/capacito r distributions.Com ponent imbalance then leads to  voltage differences in the circuit. To obtain  the required ‘saw tooth’ voltage distribution to  earth , which is needed to balance capacitative leakage currents and prevent over-voltages to  earth , m agnet inductances m ust have similar values, and the values of the individual capacitor banks are ‘trim m ed’, during commissioning, to  give a balanced distribution.Choke construction should reflect the high coupling required between the prim ary and secondary coils in one cell, and interleaving appears to  be a reasonable solution.A num ber of choke constructions are possible. A single large air gap will give a large choke w ith very linear behaviour. The flux density in the air gap will vary down the axis of the choke, and the num ber of turns in each prim ary/secondary pair m ust be varied to give the same inductance. In this circum stance, the use of an eddy-current shield around the coil would be very advantageous, and should lead to uniform flux densities in the coil stack. Praeg also believed th a t this could lead to  an overall reduction in eddy losses.A m ultiple air gap design, using steel packets between insulated, nonm agnetic spacers in the centre of the coil stack, gives a more com pact, more economical design. This would pos­sibly be less linear (leading to  higher harmonics in the m agnet) and also leads to  mechanical problems of securing the steel packets.A completely steel-free design would be highly linear, and have less ac losses. However, it would be large, expensive, have high fringe fields, and is possibly not optim um  in term s of dc resistance.It is felt th a t the single large air gap, steel-clad design is the best approach.Separate chokes in each cell had been discussed previously. They would be more ex­pensive (by up to  a factor of 3), but would reduce cable lengths, cable losses, and capacitance to  earth . Discussions with industry  would be useful to  establish the true  situation  relating to  costs.3. AC ENERGY M AKE-UP SYSTEMThe present proposal uses a ‘pulse power supply’ invertor (as a t CEA and NINA) for m aintaining ac excitation. The current pulse enters the network through the choke prim aries,146and is a  half sinewave, w ith a frequency of between three and five tim es the network fun­dam ental. The second integral of this current, i.e. (sinwpf) appears in the m agnetcircuit.This system  is not necessarily the best available, and produces a num ber of problems:(i) The disturbance in the m agnets is quite large, and is a source of tracking inequality between dipole and quadrupole circuits. This is due to  the phase of the pulse varying when a circuit is driven off resonance, leading to  pulse phase differences being present between the free cycling dipole and forced quadrupole circuits.(ii) The pulse will excite the delay line mode in the m agnet network, the capacitor volts having a cosine waveform, w ith sharp edges:- f\ j -(iii) W hen ‘forcing’ a circuit (the quads will be driven a t the dipole frequency), and there is an appreciable difference between the resonant and forced frequency, there is a large increase in the pulse current am plitude.Recently, ESR F had commissioned a study on invertor circuits for ‘W hite C ircu it’ applications from Holec, and the report had recommended the use of a single phase ‘pwm ’ circuit using ‘g to ’s to  generate the square waves from a dc rectifier. Holec had shown th a t by controlling the waveform, lower-order harm onics could be elim inated, leading to b e tte r  tracking between the different circuits. Inform ation on this study was provided to  TR IU M F.The degree of decoupling from the supply was considered. W hilst the pulse power supply appears to  be well decoupled, other circuits also use a series inductor, and the use of a fast voltage servo at the rectifier should produce good isolation from mains transients. However, all the circuits considered will be susceptible to  disturbance by supply fundam ental (due to  phase im balance) and harm onic (from the rectifiers) contam ination. These will appear (heavily a ttenuated ) in the m agnet, and should be examined as a source of servo instability  and tracking error.4. TIM ING AND FREQ U EN CY  STABILITYOne circuit -  probably the network with the highest stored energy -  would be the frequency m aster for the machines. All other supplies will then be frequency locked to this system . As this results in the other networks being driven off frequency, phase differences will result, and as d<j)/du is m aximum  at resonance these can be appreciable for even small frequency shifts. These effects should be minimized by m aintaining a  reasonable tem pera­tu re  control on ring tunnels and capacitor/choke buildings. T ighter control is achieved by switching small capacitance in to /o u t of circuits autom atically. This will m inimize the reac­tive power th a t  is needed from the supplies powering the driven system s, and should result in the ra ting  incrase needed for reactive power being small.Some phase drift will occur, and phase control servos will be essential, holding all driven system s in phase w ith the frequency m aster system.1475. FAULTS AND FU TU R E INVESTIGATIONSThe circuit can clearly have some interesting fault conditions associated w ith switch m alfunctions, capacitor breakdowns, etc. It is not clear what features these will have, and investigation will be necessary. Circuit protection, as a t ANL, where power is dum ped into shorting switches when a fault is detected, should be considered.Reiniger indicated his intention to extend his existing single cell high current tests to a dual cell experim ent. It was also suggested th a t a low voltage/curren t tabletop network, with all twelve cells present, could give valuable inform ation on norm al and fault behaviour in the multicell situation.148R E PO R T ON INDUSTRIAL PARTICIPATION W ORKING GROUP DISCUSSION A to ta l of 15 participants attended these discussion. These broke down into:Questions considered1. Establishing industrial capacity for m agnet production:• M ethods for getting  companies interested in a new product line• Present capability of Canadian industry to  m anufacture m agnets• Are the quantities and schedules feasible for a fabrication period of 3-4 years?2. Production technology• Can the tolerance requirem ents on tooling and lam ination stam ping be met?• The coils will be vacuum im pregnated with insulation of fibreglass, m ica and either epoxy or polyester and a voltage insulation requirem ent of 10 kV. Is this a problem for m ost companies?3. Design assistance• Can companies assist TR IU M F in detailed design studies such as eddy current losses and in providing cost estimates?The discussion was free ranging with a lot of tim e spent discussing current practices in m aking m agnets. The questionnaire was not followed rigidly but was used as a guide. Sum m ary answers to  the questions listed in it are given below:1 . Industrial Capacity for M agnet ProductionThere is always in terest in new product lines, bu t they have to  be sold to company m anagem ent as being beneficial to  the company. A product which does not lead to repeat orders is especially hard to  sell. TR IU M F is planning continued involvement with companies during the PDS year.A t the present, Canadian capacity does not exist to m anufacture all the KAON Factory m agnets. There was a feeling th a t it could be generated if the m anufacturers are convinced th a t it is worth their while to  be involved in the project. It is to  be noted th a t excess transform er m anufacturing capacity which was present 2—3 years ago is now much leduced as a benefit of the improved economy. The response from industry  cannot be predicted for two years tim e which is the earliest we anticipate orders will be placed for the main KAON Factory production. It will depend on the economy at th a t tim e.The size of potential contracts was discussed. M anufacturing facilities will have to  be in the  range of 100,000-200,000 square feet to  handle the larger contracts. Production will have to  be done on an assembly line basis to  meet the schedules. No one m anufacturer could handle the production and we expect to have at least 60 different contracts spread amongst the suppliers. If C anadian companies do not accept the challenge we would have to look toCanadian m anufacturers representatives U.S. m anufacturers representatives French m anufacturers representatives U.S. laboratory  employees TR IU M F employees52143149offshore suppliers. There is also the possibility th a t International Participation  in the project may reduce the num ber of m agnet orders to  be placed in Canada.2. Production TechnologyTooling can be built to  meet the tolerances on our m agnets, obviously the tighter the tolerances the higher the cost. It was pointed out th a t some tolerances are relative and not absolute, bu t our aim  will be to  produce m agnets identical to  a few parts in 104.The fact th a t we will require vacuum im pregnation and voltage levels of 10 kV to ground did not appear to  be a problem. The large transform er m anufacturers work with higher levels than  this and all are familiar with vacuum im pregnation.3. Design assistanceCompanies were very interested in becoming involved with the design of our m agnets. AC machine designers are much more experienced than  TR IU M F staff in calculating eddy current and core losses especially for non-sinusoidal excitation.There was a feeling th a t participation by industry engineers would be m utually bene­ficial. However, there might be a problem in finding engineers who would be willing to come to  Vancouver for a period of tim e. This aspect will have to  be followed up individually with the companies.Finally, we gave photographs of typical TR IU M F m agnets and photographs of DESY m agnets plus a copy of a typical specification to  Canadian m anufacturers. We will also send them  a copy of the Economic Im pact Study by Coopers & Lybrand. Three of the representa­tives are having m eetings w ith their company m anagement to  report on the workshop in the week following the workshop. We told them th a t if they need further help or special visits we will do all th a t we can to  assist them.150A SHIELDED ENERGY - STORAGE CHOKE FOR RAPID CYCLING SYNCHROTRONSW.F. PraegArgonne National Laboratory, 9700 S. Cass Ave., Argonne, IL 60439, USAABSTRACTThe conventional design of energy storage chokes with a picture frame iron enclosure is briefly reviewed. Eddy current losses in the coil wind­ings caused by transverse leakage flux are calculated. Eddy current shielding is proposed for the coil assembly to protect it from leakage flux and to eliminate the need for stranded conductors. The eddy current shield forces nearly 10 0% coupling between the coils, permitting designs with an H-frame iron enclosure and multiple air gaps.INTRODUCTIONLarge rapid cycling synchrotrons (RCS) comprise many resonant cells. Each cell has an energy storage choke that carries the ring magnet dc current and an ac current. These chokes may be constructed as separate units or they may use one iron structure common to all of them. In either case, the design usually consists of pancake coils with an air core and a "picture frame" iron enclosure to confine the return path of the magnetic flux. For such a structure the main flux is transverse to the axis of the coil windings and rises linearly from zero, at the outside of the coil to a maximum on the inside of the coil as illustrated in Fig. 1. The flux dis­tribution is not the same for each of the pancake coils since the coil turns are at different distances from the top and bottom of the enclosure. Effective uniformity of flux linkage of the different choke coils is enhanced by coupling all coils with an auxiliary winding. This is more effective if all chokes are in one common iron frame as compared to widely space individual chokes. To reduce eddy current losses, the coils are wound from conductors comprising many small insulated wires in parallel. These wires are cooled either by inserting the choke assembly in oil or by providing a water cooled heat sink at the center of the wires.The major criteria for the choke design are:a. uniform flux linkage in the choke coils of all resonant cells,b. small eddy current losses in the coil assembly,c. cost and simplicity of construction.The purpose of this paper is to stimulate interest in exploring whether these design criteria can be satisfied more efficiently by enclos­ing the choke coils with eddy current shields. Of special interest are designs that have the chokes of the resonant cells in one structure.151CONVENTIONAL DESIGNSEnergy - storage chokes with a common magnetic structure have been analyzed. Figure 1 illustrates the design of a recently developed prototype choke. ^Eddy Current Losses in the Coil WindingEach of the twelve pancakes of the choke in Fig. 1 has seven turns of 58 stranded wires as shown in Fig. 2. With a 50 Hz current of 880 A per turn, the field at the center of the choke is * 2140 G. The average flux density in the seven conductors of the pancake coil rises in steps of 305.6 G from 153 G for the turn on the outside to 1987 G for turn No. 7 on the inside. Eddy current losses per unit length of a wire with radius r in a field Hq transverse to the wire axis are^wherep = resistivity= permeability of free space<5 _  f— P _ ] V 2  _  s k j_n  d e p t h''ir u f J of = frequencyThe function Ff^l is given by aJs ■----  -jr -j '“ O'F I’t ) = -Re \j2j -r-------- —  t where J and J are Bessel Functions* (2)0 o ( \ r r r ri o i ' •J oFigure 3 shows e (/^-) with its approximations.With 2r = 0.26 cm and a skin depth of 0.93 cm for copper, we have r < 6 and the power losses per unit length can be calculated by combining eqs. (1 ) and (3) which givesIT ,B I 2 rr \42 p K  i U )  * (4)oThe losses per cm length in a single wire and in the 58 parallel strands of a conductor are given below for the seven turns of a pancake.1  ( L ) 44 for r < 6rfor r >(3)152Turn 1 2 3 4 5 6 7VerticalField 1 53 459 764 1071 1 376 1682 1 988 GSingleWire0.0151 0.1 36 0.3755 0.738 1 . 2 2 1 .82 2.54 mW/cm58 wires 0.874 7.86 21 . 8 42.8 70.7 1 05.5 1 47.8________mW/cmThere is also a horizontal field component in each of the pancake coils due to the current flowing in all other pancakes as illustrated in Fig. 4 by the dashed bars. With 7 x 880A = 6160 amperturns per pancake, the norizontal fields and eddy current losses for the conditions as shown in Fig. 1 are shown below.PancakeNo.IXIIIIXIIIIXIVIXVVIIIVIVIIhorizontalfield1442 1051 709 405 1 28 1 28 Gsingle wire 1 .34 0.711 0.324 0.1 06 0.011 0.011 mW/cm58 wires 77.7 41 .22 1 8 . 8 6.11 0.611 0.611 mW/cmThese losses must be added to the losses caused by the vertical field. Thelosses per turn depend on the wire length which is a function of the twist of the stranded wires.Eddy Current Losses in the Water Cooled Heat SinkThe time average eddy current power losses in a conducting cylinder due to a transverse sinusoidal field have been calculated asIA I 2 r dr d0 (5)1 z1j = current densityr, Q, z = cylindrical coordinates (x axis at 0 = 0 )A = magnetic vector potential.The losses in the 1.5 cm O.D. copper heat sink with a wall thickness of0.15 cm are given below for the design shown in Figs. 1 and 2.p = Re —  / / 2 |J |2 p r dr dO = Re ~  f  j 2 0 r , 0 r lwhere153Turn 1 2 3 4 5 6 7verticalfield 1 53 459 764 1071 1 376 1682 1 988 Glosses/cm 0.0 11 0.096 0.265 0.52 0 . 8 6 1 .29 1 .80 W/cmlosses/turn 3.75 31 . 8 82.8 152 234 324 416 W12 pancakes 45 383 994 1828 2809 3887 4991 WThe losses in the 84 copper heat sinks due to the vertical 14.9 kW. Additional losses are caused by the horizontal fields dashed in Fig. 4. They are tabulated below and amount to 3.6 khfieldsshown.arePancakeNo.IXIIIIXIIIIXIVIXVVIIIVIVIIhorizontalfield1 442 1 051 709 405 1 28 1 28 Glosses/cm 0.94 0.502 0 228 0.0745 0.007 0.007 W/cmeach pancake 191 3 1 022 464 1 .52 1 4.2 14.2 WThe total eddy current losses in the heat sink are - 18.5 kw. They could be reduced by replacing the copper tube with one made from stainless steel. However, this would also reduce the efficiency of the heat sink.PROPOSED DESIGNS WITH THE COILS ENCLOSED BY AN EDDY CURRENT SHIELD Picture Frame Iron CoreBy enclosing the pancake coils with copper, except for a small verti­cal gap to avoid that the enclosure becomes a shorted turn, the vertical transverse flux is prevented from entering the coil area by means of eddy current shielding. The vertical flux distribution is now as shown in Fig.5 for one quarter of the choke assembly of Fig. 1. with the transverse flux excluded from the coil conductors, there is no longer a need for stranded wires. The pancakes can now be wound from conventional water cooled copper conductors suitable for the operating frequency. With a cop­per shield 0.9 cm thick and for a 2140 G, 50 Hz field at the center of the choke the eddy current losses in the shield were calculated with PE2D as 96.6 W/cm. With a median circumference of 290 cm, the losses in the water cooled shield would be - 28 kw. These losses can be reduced with a thicker shield wall, a more compact pancake design, or by using an H-Frame core as described below. More importantly, the shield forces a more uniform flux throughout the coil assembly; this improves the mutual coupling between the154chokes of different resonant cells without large equalizing currents in the1 2auxiliary windings. 'The eddy current pattern of a simple shield as shown in Fig. 6 , can be improved with thin contact making copper sheets between the pancakes at the air gap location as shown in Fig. 7 (i.e., sheet 1/16" x 2" x width of shield).Flux Within Shielded volumeThe magnitude of the horizontal fields to which each of the 12- pancakes is exposed due to the 6160 amperturns in each of the other 11 pancakes is illustrated by the solid bars of Fig. 4. These fields can be eliminated with copper shield plates between each pancake as shown in Fig.8 . These plates make electrical contact with the walls and have one gap to prevent a shorted turn for the main choke flux.H-Frame Iron CoreWith the shield excluding external transverse flux from entering the space occupied by the choke coils, it is no longer necessary to have an air core design. A multi-air-gap iron-core choke can now be considered as illustrated by Fig. 9 which has the same inductance as the choke of Fig. 1. The losses in the eddy current shield should be much smaller than for a shielded picture frame because the iron guides the flux around the shielded area and the field strength around the multiple air gaps is small. Each of the twelve pancake coils has 10 turns. The core comprises twelve 2.12 cm thick G-10 plates separated by thirteen 4.27 cm high sections of grain oriented silicon steel. This design will be analyzed at a later date. Of course, the eddy current shields could also be made from aluminum.AcknowledgmentI am grateful to R. Lari of Vector Fields, Inc., who calculated the losses in the eddy current shield with PE2D.References1 . j .a . Fox, "Resonant Magnet Network and Power Supply for the 4-GeVElectron Synchrotron NINA," Proceedings IEEE, Vol. 112, No. 6 , June 1965.2. H. Sasaki, et al, "A DC Biased Rapid Cycling Magnet System Operating ina Dual Frequency Mode," this workshop.3. H. Kaden, "Wirbelstrome und Schirmung in der Nachrichtentecknik," Springer Verlag, 1959.4. A. Ivanyi, I. Bardi, and U. Biro, "Analytical Solution to Problem 2 ofthe International Electromagnetic Workshop," Elelctromagnetic Workshop, Rutherford Appleton Laboratory, Chilton, Didcot, Oxon, England, March1 986.155VER TIC A L F I E L DSTOFig, 3 F (g) for calculating the losses in solid round conductors.Fig. 1 Picture-frame choke assembly and flux distribution.Fig. 2 Stranded conductors of pancake coil.I — j2T+a r-mFig. 4 Horizontal fields in pancake coils,10.0001-Fig, 5 Flux distribution in one quadrant of choke with shielded coils.156Fig. 6 Eddy current pattern of a simple shield.Fig. 7 Eddy current pattern of shield with thin contact making sheets between shield walls.Fig. 8 Shield plates between pancake coil' to eliminate horizontal field.Fig. 9 H-frame choke with multi-air-gapiron-core and shielded pancake coils. Choke has same inductance as choke in Fig. 1.PANCAKE COIL OUTER S H IE L D(VERTICAL FIELDS- * - 0 )p a n c a k e  c o il  s h ie l d(h o r iz o n t a l  f i e l d s - ^ 0 )157DUAL VERSUS SINGLE FREQUENCY RING MAGNET POWER SUPPLIES FOR RAPID CYCLING SYNCHROTRONSW.F. PraegArgonne National Laboratory, 9700 S. Cass Ave., Argonne, IL, 60439, USAABSTRACTFeatures of dual and of single frequency ring magnet power supplies for rapid-cycling synchrotrons are compared.INTRODUCTIONRing magnets of rapid-cycling synchrotrons (RCS) are usually excited in a resonant circuit by a dc-biased cosine wave current. Particle acceleration takes place during the rise time of the current. After beam extraction, the magnets are reset during the fall time of the cosine wave. Only 50% of the cycle is used for acceleration. A significant reduction in peak radio frequency (rf) power can be achieved by making the magnetic field rise slowly and fall rapidly. This can be done by adding harmonics to the fundamental cosine wave1' or by using a dual frequency circuit.1'3'4' 1 The B and B waveshapes associated with the various cir­cuits are shown in Fig. 1. Adding a 2nd harmonic (f , 2f ,---) reduces thepeak value of b by 33% as compared to a single frequency wave (f , -..-).By adding a 2nd and a 3rd harmonic (fQ, 2fQ , 3fQ , ...) b can be made toremain practically constant during most of the acceleration time. Notethat at the beginning (injection) and end (extraction) of acceleration,B is about the same as with a single frequency circuit. Analysis of a 2nd harmonic circuit shows that the voltages on the additional components are larger than the magnet voltages and the control of the amplitude and phase of the various circuit currents is difficult. This may be the reason why this circuit is not in use.These difficulties can be avoided with a circuit that has the magnet and choke in one core assembly. 2 A second harmonic (2fQ ) and a dc bias field can be added to a first harmonic (fQ ) as shown in Fig. 2. Figure 3 illustrates the phase-relation and magnitudes of the magnetic fields. Desirable features of this circuit, which is especially suitable for com­bined function magnets, are:1. No dc flows through the first harmonic ring magnet coils.2. No 1st harmonic magnet current flows through the dc coil or through the 2nd harmonic coil.3. With the inductances of the magnet and choke connected in series, the circuit needs only a fraction of the tuning capacitance required for a conventional circuit. For LM = LCH, the circuit requires 1/4 of the capacitance of a conventional circuit.In 1980, the writer developed, for an upgrade of a 30 Hz, 500 MeV synchrotron, a dual frequency circuit. 1 For a reduction of B by 1/3 as158compared to a single frequency circuit, the frequencies during acceleration and during magnet reset must have a ratio of 1:3 as shown in Fig. 1. Note also the reduction in b during injection and extraction as compared to all other circuits. The resonant frequency is changed by adding or removing a capacitor bank C2 in parallel to a capacitor bank C1 with a solid state switch. Switching takes place when all the circuit energy is in the magnet and choke of the circuit, thereby keeping switching transients to a mini­mum. The design curves of Fig. 4 show a ratio of “ 1.5 as optimal.This ratio corresponds to values of f ^ / f Q = 0.667, o = 2' ~ 2fCl/Co = 0 , 2 6 and C1 + C2 = 2 , 2 6 Co* further increase in fQ/f1 willdecrease B during acceleration by a much smaller percentage (f1/f0 ) than it will increase the magnet voltage during reset (f2/fQ ). For example with fQ/f1 = 1-6, B during acceleration will decrease by 6.2% and the magnet reset voltage will increase by 25% as compared to the corresponding values for f0 /f-j = 1 »5.In the early 1980's, there were no suitable gate-turn-off (GTO) thy­ristors available and, therefore, silicon controlled rectifier (SCR) turn­off circuits were required for practical circuit applications.1. Recently large power GTO's have become available and the solid state switching circuit can now be simplified by using a combination of GTO's, SCR's, and diodes. Dual frequency circuits with flat bottoms and with flat tops utilizing GTO's have been described; as well as circuits with only dual frequency operation. 5 , 7 , 8  For a comparison of dual frequency and single frequency operation, the circuits of Fig. 5 and 6 will be referred to.BRIEF COMPARISON OF FEATURES OF DUAL FREQUENCY VERSUS SINGLE FREQUENCY RING MAGNET POWER SUPPLIESEconomicsFor a high rate of pulses per second (pps), the reduction in peak rf power with a dual frequency ring magnet circuit is 56% and the savings far outweigh the increase in cost over a single frequency ring magnet power supply. Cost comparisons have not yet been made to establish the lower limit for the pps, which may be near 10 pps (6 . 6  Hz and 20 Hz).Circuit ReliabilityThe addition of the dual frequency switching circuit adds relatively few thyristors to the overall power supply and the switching stress on these components is relatively low. The dv/dt and di/dt stress on these thyristors is much smaller than the stress on conventional power supply thyristors. For example, the thyristors of the 24-phase power supply of the 500 MeV, 30 Hz RCS at Argonne National Laboratory (ANL) are continually phase controlled to generate a unidirectional current i = 2300 A - 1300 A cos 188 t. Every 1.39 ms a thyristor is switched, it must commutate half the magnet current, and must hold off voltages that rise at 100 v/^s to generate a biased 30 Hz wave from the 60 Hz commercial power source. This circuit has operated satisfactorily for over 10 years. The addition of a capacitor switch as shown in Fig. 5 for dual frequency operation would add only one high voltage (HV) switch to the existing 24 thyristor switches. This would not degrade the overall circuit reliability by much because the HV switch operates when its voltage is zero.159With reference to Fig. 6 , the switching modes are as follows:1. At time tQ (the switch closes, connecting C„ in parallel with C ).^ IAll the circuit energy is in the choke and in the magnet. The capacitor voltage is zero and the current in C1 is at its negative peak. As the voltage on capacitor increases, the diode begins conducting and capacitors C1 and C2 share the current while the choke discharges its energy into the magnet and the capacitors. The current transfer into C2 is determined by the inductance of the interconnec­tions; transmission lines rather than bus bars should be used for this purpose to keep the inductance and the duration of transients low. The reverse voltage on the thyristor assembly is limited to the diode voltage drop.2. At time t1 (the capacitor current goes through zero).At time t1, the capacitor current goes through zero; the capacitors are at their peak energy (voltage). Diode D turns off as the magnet vol­tage begins to fall and the thyristor assembly must be turned on. After t1, the choke and capacitors discharge into the magnet. As the capaci­tors discharge, the thyristor current increases at a rate of di/dt= ^  IC2 C°S ,J>t* w-‘-th IC2 —  the mating is < 2n 33 1 / 3 s~ 1x 6kA < 1 .25A/ys which is small compared to typical thyristor ratings of < 800 A/ys. The slow current rise necessitates gate drives for the thyristors long enough to go past the thyristors holding currents.3. At time t2 (the switch opens, disconnecting C2)The capacitors are discharged (zero voltage), the capacitor current is at its positive peak. All the circuit energy is stored in the magnet and in the choke. A hard gate drive to the GTO turns it off and the current in the series connected chain of one GTO and several SCR's goes to zero with the aid of parallel connected snubber circuits (not shownin Fig. 5). The forward voltage on the thyristor assembly rises at arate of < 18.8 V/ys which is negligible when compared to the assembly rating. The peak forward voltage at time t3 is three times larger (< 30 kV per cell) than the peak voltage during acceleration. The design rating of the switch assembly for blocking a peak forwardvoltage of 30 kV would be about 60 kV peak.Power SupplyThe number of resonant cells and power supply feed points depends on what peak voltage to ground can be used. The major considerations are capacitive currents through the coil insulation and resistive currents through the cooling water circuits. The capacitive currents are determined by the magnet voltage and by the coil insulation. During acceleration, the magnitude of the magnet voltage in a dual frequency circuit is only 66% of the magnitude that a single frequency accelerator with the same repetition160rate would have. During magnet reset, the voltage is twice as large as for a single frequency accelerator. However, by then the beam has been extracted and capacitive currents are of little consequence. Coil insulation make up less than 10% of the magnet cost. A peak magnet cell voltage to ground of ±15 kV should cause no problems for an insulation system based on mica tape and a coil and core design that prevent corona discharges. In principle, power can be supplied continuously ' or pulsed. ' 10 The continuous 24-phase power supply at ANL mentioned earlier provides for the ac and dc losses of around 500 kW with a peak power demand of about 1.4 MVA. This relatively inexpensive power supply operates from the 480 V, 3-phase, 60 Hz power line without causing noticeable interfe­rence with other circuits.  ^ Power supplies for much larger loads should be connected to a 13 kv or 60 kV power line. A more expensive but most likely better solution is a pulsed power supply as shown in Fig. 5. It provides complete isolation from the power line during the make-up pulse.In this circuit, power losses are made up by a current pulse on the primarywinding of the choke. This pulse is usually applied during magnetreset. ' 10 However, applying the pulse during acceleration with its peak occurring at time t is better. The peak power supply and magnet voltages are smaller and the pulse can be made longer. Tests in 1984 at ANL showednegligible disturbance in the magnet current due to a make-up pulse duringacceleration.CONCLUSIONTable 1 is a summary of some of the features of a single versus a dual frequency power supply. For a synchrotron operating at 50 pps, I recommend a dual frequency power supply.Table 1: Comparison of Single and Dual Frequency Ring MagnetPower Supplies _____ _______________ _Circuit Function Compared fo fi " T7s; f 2 = 2fopeak rf power 1 . 0 0.44/NB, 6 at injection and extraction 1 .0 0 . 6 6capacitive and resistive ground currents during acceleration1 . 0 0 . 6 6phase shift between I & B in cores small smallerpeak magnet and choke voltages 1 . 0 2 . 0circuit losses 1 . 0 > 1 . 0capacitor switching none takes place at zero voltage. dv/dt < 20 V, di/dt < 3 A/ys.161REFERENCESM.H. Foss, W.F. Praeg, "Shaped Excitation Current for Synchrotron Magnets," IEEE Transactions on Nuclear Science, vol. NS-28, No. 3, June 1981.W.F. praeg, "Dual Aperture Dipole with 2nd Harmonic," IEEE Transactions on Nuclear Science, Vol. NS-30, No. 4, August 1983.W.F. Praeg, "Dual Frequency Ring Magnet Power Supply with Flat-Bottom," IEEE Transactions on Nuclear Science, Vol. NS-30, No. 4, August 1983.W.F. Praeg, "A Multi-Function Ring Magnet Power Supply for Rapid- Cycling Synchrotrons," IEEE Transactions on Nuclear Science, Vol. NS- 32, No. 5, October 1985.H. Someya, et. al, "Application of a GTO Thyristor for a Dual Resonant Frequency Circuit for the Magnet of a Rapid-Cycling Synchrotron," IEEE Transactions on Nuclear Science, vol. NS-32, No. 5, October 1985.W. Praeg and D. McGhee, "Ring Magnet Power Supply for a 500 MeV Synchrotron," IEEE, IAS, 1978 Annual Meeting, Conference Record,Catalog No. 78CH1 346-61 A.T. Adachi, H. Someya, and H. Sasaki, "Magnet Exciting System with Dual Resonant Frequency Circuit," European Particle Conference, Rome, Italy, June 7-11, 1988, (KEK Preprint 88-17).H. Sasaki, et al, "A DC Biased Rapid Cycling Magnet System Operating ina Dual Frequency Mode," this workshop.J.A. Fox, "Resonant Magnet Network and Power Supply for the 4-GeV Electron Synchrotron NINA," Proceedings IEEE Vol. 112, No. 6 , June 1965.K. Takikawa and H. Sasaki, "Design Studies for the KEK Booster Magnet Power Supply," National Laboratory for High Energy Physics, Tsukuba- Gun, Ibaraki, Japan, KEK-74-7, August 1974.W. Praeg, "Tolerable Limits of Voltage Fluctuations Produced by MagnetsPulsed Directly from Alternating Current Power Lines," IEEE Transactions on Nuclear Science, NS-16, No. 3, June 1969^162Fig. 1:B AND B WAVESHAPES FOR VARIOIARK RESONANT RING MAGNET POWER SUPPLY CIRCUITS.c [01{ 3 ]-utr-l - 2 f0 -F IL T £ R  ^o ) EQUIVALENT CIRCUITFig. 2:TWO RINGS WITH COMMON WINDINGS, MAGNET CORE, AND POWER SUPPLIES.Fig. 3:FLUX DENSITIES IN TWO-RING CORE WITH FIRST (fG) AND SECOND (2 fD) HARMONIC AND DC EXCITATION.b ) MAGNET S FLUX PATHS-LAMINATIONSRING I RING z ( c h o k a )163Fig. 4: RESPONSE OF DUAL FREQUENCY CIRCUITVERSUS f /f,.o 1Fig. 6: WAVESHAPES OF THE CIRCUIT OF FIG. 5.Fig. 5: DUAL FREQUENCY RING MAGNET CIRCUIT AND PULSEDPOWER SUPPLY.7164CALCULATING THE FREQUENCY RESPONSE OF MAGNETS WITH LAMINATED CORESW.F. PraegArgonne National Laboratory, 9700 S. Cass Ave., Argonne, IL 60465, USAABSTRACTA method is described for calculating the frequency response of mag­nets with laminated ferromagnetic cores energized from alternating current (ac) with a direct current (dc) bias. This is useful during the design of magnets and their power supplies for regulation and ripple studies and to estimate magnet losses and the circuit response to transients.INTRODUCTIONThe ring magnets fo rapid cycling synchrotrons are often energized^ with a unidirectional current that comprises a single or dual frequency cosine wave superimposed on a direct current bias. During the design stage, knowledge of power losses and the frequency response of the magnets is essential for studies of transient response, power supply regulation, etc., of the overall ring magnet circuit.This paper shows how the inductance and the resistance due to core losses can be calculated for any magnet based on measurements on a small core sample. The capacitance of the magnet, transformed to the coil termi­nals, is calculated from the coil and core geometry. Finally, from this data the frequency response of the magnet is calculated.MAGNET RESISTANCE DUE TO CORE LOSSES AND MAGNET INDUCTANCEComplex PermeabilityNeglecting capacitive effects, the impedance of a coil containing a ferromagnetic core is Z = R + ju>L. After subtracting the coil resistance RCu from R, we haveZ = R + ju>L Fe(1 )Resistance Rpe is due to core losses (hysteresis and eddy currents). Both the core losses and the inductance are dependent on the properties and the geometry of the core, and Eq. (1) can be written as  o aZ = jo u n Jwhere o = 2irf,f = frequency,"jj = n - ju = complex permeability of core material, L R(2)n = number of coil turns,165A = effective core area, i  = length of core .From Eqs. (1) and (2), the real and imaginary parts of the complex permeability areR—  L .  Fe!J -)|JR " 2 A " 1 2 A ' ^3)n j  con jFor a defined core and coil geometry as, for example, a toroidal core shown in Fig. 1, and for a controlled core magnetization, the complex per­meability can be calculated from an impedance measurement.The toroidal test core has a B-coil to measure the dc and the ac mag­netic fields. This one layer coil is made from thin (i.e., flat) wire and wound close to the core to reduce errors due to flux in the space between the B-coil and the core. The ac excitation winding is wound over the B-coil and has a relatively large cross-section to keep the copper losses small in comparison to the iron losses. The dc magnetization coil can be a few turns through the center of the hole of the toroid with the return far enough away in order not to effect the core material. 2Before taking measurements on core samples with the appropriate ac and dc magnetization, it is essential that the core laminations be demagne­tized. The dc magnetization is than slowly increased until the desired dc bias field (i.e., 6.75 kG) is reached; thereafter the ac magnetization is increased from zero to its maximum value (i.e., 3.85 kG). For some core material, including silicon steel, it takes time before a final impedance value is obtained for a given ac and dc magnetization. This effect (NACHWIRKUNG), which causes changes in the permeability of several percent, can be eliminated by operating the sample at a given magnetization for about half an hour before the impedance measurement is made.Figures 2 and 3 show the complex permeability, y , and its components and u of 0.35 mm thick silicon steel (96% Fe, 4% Si) measured at 50 Hz without dc magnetization. These values were calculated with Eq. (3) from impedance measurements with an inductance bridge on a coil containing core laminations. A ^ iriusoidal^ excitation current was forced by connecting a relatively large resistor in series with the bridge and by making the coil impedance and the impedance of the comparison arm of the bridge small as compared to the other two bridge arms.The reason that, in the above figures, the maximum permeability is only u = 3200 y as compared to the usually quoted values of y = 7000 for 2% silicon steel is the sinusoidal excitation. When the sinusoidal excitation drives the core into saturation, the corresponding flux contains large harmonics; the amplitude of the fundamental frequency of the flux is considerably smaller than its peak value. The curves in Figs. 2 and 3 illustrate the effect of hysteresis because at 50 Hz eddy current effects are small. At higher frequencies both hysteresis and eddy currents affect the complex permeability. In eddy current problems, it is convenient to make use of the equivalent skin depth, 6 , where the field166strength has a value e-1 = 0.37 of the field on the surface of the lamination.(4)wherep = resistivitypQ = permeability of free spacejj = relative permeabilityIt is common to distinguish between low frequencies where the field distri­bution in the laminations is approximately uniform and the lamination thickness d < 2 6 , and high frequencies where skin effects are pronounced, d > 26. The boundary between these two frequency ranges has been defined so thatThe cut-off frequency defined by (5) isFor 0.35 mm thick, 4% silicon steel with an initial permeability of 330 p and a peak permeability of 3200 p the cut off frequency is 13 kHz and 1?3 kHz respectively; well above 50 8z.The effects of hysteresis and eddy currents on 4% silicon steel are shown in Fig. 4 for sinusoidal excitation. 3 For f « fc only hysteresis effects are present; the 50 Hz curve shows a change in permeability from|p"| = 330 p for Hac = 0 tojpj= 1150 for Hac = 100 mA/cm. For f » fQ and H -*■ 0 on?y eddy current effects are present as shown on the lower curveafor the range from 50 Hz to 9 kHz with a corresponding change in |p | from -- 330 to ~ 265. All other curves show the combined effect of hyste­resis and eddy currents.Figure 5 shows how a dc magnetization of 1 A/cm reduces the permeabi­lity values of Fig. 4.Equivalent Magnet CircuitIn order to calculate the eddy current and hysteresis losses of an ac magnet from measured values of the complex permeability of the core material, we need an equivalent magnetic circuit. With reference to the magnet cross-section shown in Fig. 6 , the ac excitation current i, the ac field strength H, various lengths of the flux path i ,  the number of turns of the magnet coil n and the ac magnet flux <j) are related by:d = 26 (5)167/ H 0d 4 = i n = H Z + H ?. + . . . + H I (7 )J I  g g 1 1 N N K ' 'B d>H = 7 =^  (8)(7) into (8 ) gives1 1 ,  I( g 1 N1 n = ♦ + r —  + . • • + - — J ( 9 )g g y A  y A1 1  N Nmultiplying both sides by — -— —-and rearranging results inj go n. £ A. I1 ______ g 1 Nj 10 n  ^ " 2 + 2 + . . . +  - (1 0)j 0) n ygAg j to n y ^  j u> n y ^2. ., _ 7 a) n y A . ,with Z = ^-------    , equation (10) can be written_ i  1 1 1j toncj) ~ Z _ Z + Z + * * ' + z g 1 NEquation (11) illustrates that the total magnet impedance Z, neglect­ing the ohmic resistance and the capacitance of the coil, can be thought of as being the parallel connection of N impedances, each having n turns and the electromagnetic properties of its path length. Such an equivalent cir­cuit allows one to compute separately the impedance of the various magnet sections as a function of frequency. The total magnet response being the parallel connection of the various sections. The complex permeability "y of each section is measured on the one core sample by magnetizing it with the ac and dc fields corresponding to this section. The equivalent circuit for the magnet of Fig. 6 , neglecting coil resistance and capacitance, is shown in Fig. 7.EQUIVALENT CAPACITANCE OF MAGNET COILThe magnet coil turns have turn-to-turn and turn-to-ground capaci­tances. An equivalent capacitance, connected across the coil terminals can be found by transforming the individual capacitors to the coil terminals and adding them up. For example, Fig. 8 shows the coil conductor arrange­ment of the magnet coils of the former 12 GeV Zero Gradient Synchrotron (ZGS). From the geometry of the coil and the electrical properties of the insulation an equivalent circuit of the coil turns and associated capaci­tances can be drawn as shown in Fig. 9. For this circuit, the terminal capacitance was computed as 0.075 yF.FREQUENCY RESPONSE OF MAGNETThe equivalent magnet circuits of Fig. 10 are obtained by connecting in parallel with the terminal capacitance, C, the Rpe and L values calculated from the equivalent magnet circuit of Fig. 7 and adding the coil168resistance RCu. Figure 11 illustrates the Rpe and L values calculated for an octant of the ZGS. Figure 12 shows the frequency response of the octant during injection and with a dc magnetization of 21.5 kG. After the ZGS was built, the actual measured frequency response of an octant agreed very well with the calculated values.For the ZGS, response to power supply ripple was of interest. For the 50 Hz synchrotron of the KAON-factory, the response to much larger fields must be known. The field values should start from rated values at 33 and 100 Hz and decrease to a few gauss at < 10 kHz. A family of curves with different rates of field decay with frequency may be desirable.REFERENCES1. W.F. Praeg, "Dual Frequency Ring Magnet Power Supply with a Flat- Bottom, IEEE Transactions on Nuclear Science Vol. NS-30, No. 4, 8/83.2. W.F. Praeg, "Impedance, Time Constant and Ripple Flux of a Ring Magnet Octant of the ZGS," Particle Accelerator Division, Internal Report WFP- _3, 11/27/63.3. R. Feldtkeller, "Spulen und ijbertrager mit Eisenblechkernen," S. Hirzel Verlag, Stuttgart, 1949.1afTT| b L- |A = a x b x stacking factorFig. 1: Toroidal Test CoreCross-Section3000p.20001000In 7^/1- ii—T / —ir  — ' - r It~ •— riT / |/ ir / i Ii0eI/00Fig.0,5 1.0 15 20r|M3: Components of p for 4%Silicon Steel at 50 HzFig. 2: Complex Permeability y ofSilicon Steel at 50 Hz1691100 1000 900 800 700 600 500 A-00 300 200 100 00 100 200 300 400 500.1VFig. 4: Complex Permeability of 4%Silicon Steel without dc Magnetization1 2 0 0 jj 1100 1000 900 800 700 600 500 400 300 200 100 00 100 200 300 <.00 500p„Fig. 5: Permeability of 4% SilicoSteel with 1 A/cm dc MagnetizationFig. 6: Illustration of Flux Pathand Excitation in a MagnetFig. 7: Equivalent Circuit ofMagnet of Fig. 6170Fig. 8: Coil Conductor Arrangement of ZGSMagnet Coil^ --------Rcuy r f 1 v ^/y ^ ----------- (z ^ :7 (biFig. 10: Equivalent Circuits oof MagnetAnchor PlateFig. 9: Capacitance Network of ZGS Magnet Coil171Fig. 11: ZGS Ring Magnet Induc­tance and Resistance as Function of Frequency for 1 Gauss ac- Magnetization with and without dc-Magnetizations172A PROTOTYPE KICKER MAGNET FOR THE KAON FACTORYVolker Rodel CERN, Geneva, SwitzerlandSeveral kicker magnet systems are required for the KAON Factory. The most critical requirements with respect to rise time and kick strength are for the kickers for ejection of the Booster ring and the Driver ring. The performance requirements are listed in Table 1. In the following we look at a possible prototype magnet for either case.Magnet Design - Driver Extraction Driver Extraction Kicker For protons the magnetic deflection is:a = 0.3 I|dLwhere a is the deflection angle in mrad, j B d i  the kick strength in T*m, p is the momentum in GeV/c. For 30 GeV protons p = 30.9 GeV/c and for ot — 1 mrad we get/ Bd£ = 0.103 Tm which is the required kick strength for the ejection kicker.Table 1 Performance requirementsBooster ejection Driver ejectionMomentum (GeV/c) 3.82 30.9Deflection angle (mrad) 2.5 1Kick strength (Tm) 0.032 0.103Kick flat top duration (us) 0.66 3.5Kick rise time (1 to 99%) (ms) 80 80Kick fall time (99 to 1%) (ms) 5 25Aperture (width x height) (cm) 15 x 8 15 x 7Repetition rate (Hz) 50 „ . Q 10r -  i  r \ —QGas pressure in vacuum tank (Torr) 5 • 1 0 ^ 5 • 10 3Rise TimeThe kick rise time depends on the current rise of the pulse generator and the travelling time of the current wave in the delay line-type magnet. We have173where t r is the kick rise time, i.e. the rise time of / B*d£, Tg is the current rise time of the pulse generator, and Tfi is the travelling time of the magnet.To get the specified valuet j- 30 n sand assumingwe getTM = 56 nsNumber of Magnet Module NHigh kick strength and short rise time are conflicting requirements. Therefore the magnet has to be split in N modules such that the total magnetic length I of the magnet iswhere is the magnetic length of a magnet module, andL = N • L[.jwhere L is the total magnet inductance and LM is the inductance of a magnet module.From (1) we getAssuming a single turn excitation coil and a high relative permeability of the yoke ferrite, the flux density B in the gap is:where w is the horizontal aperture and I is the magnet current on the kick flat top. We assume a horizontal beam deflection.The magnet current I is given by the generator voltage and thecharacteristic impedance of the magnet module ZM . We get( 1 )/B • d£ = N • B • ( 2 )( 3 )174since the wave travelling along the magnet module has half the voltage of the generator voltage. Withwhere is the matching capacitance of the magnet module, andtM = /LM‘ CMusing (2) and (3) we get:V/BdA = N • 1  . t m  . ^For a maximum generator voltage of 60 kV (see below) and w = 0.15 m we findthe number of module must be:N > 9.2which means: N = 10Maximum Generator VoltageTo limit the maximum generator voltage to 60 kV has the advantage that standard coaxial cable RG 220/U can be used as a transmission cable between the pulse generator and the magnet module. This cable has a characteristicimpedance of 50 ft. Several cables can be connected in parallel which givesimpedances ofZM = 50 a/m (4)with m = 1, 2, ...Magnetic Length of a Module From (2) we getI  =M N* 6With (3) and the module inductancew iI Ml m  = UO • —where h is the gap height, we getWe choose a high magnet impedance:Zp[ — 25 ai.e. 2 coaxial cables of 50 a in parallel. A higher impedance would result in an unnecessarily long magnet.175The operating generator voltage isVg = • 60 kV = 55.2 kVwhich gives a magnet currentand a flux density in the gap ofB = 19.7 mTfor a gap height h = 0.07 m.The (magnetic) length of the magnet module is thenThe magnet module as well as the pulse forming network (PFN) is split in several cells. The optimum number of cells can be determined by computer simulation.Choosing 20 cells with a magnet module we get a ferrite thickness of 26 mm. The capacitance per cell is thus 112.5 pF. They could be built by ground plates interleaved with the high voltage plates at a distance of 6 mm. The effective area of the plates is 19.5 cm2 .A Ni-Zn ferrite (like the Philips type 8C11) having a high saturation induction (<0.3 T) and a low coercitive force should be used. The flux density in the yoke should be calculated by a magnet program (like P01SS0N).£[>1 = 0.52 mModule DesignThe module inductance is:Thus the total matching capacitance isL.iCM = ~ n t~  -  2.24 nFThe main parameters are summarized in Table 2.176Table 2 Main parametersBooster ejection Driver ejectionMaximum generator voltage (kV) 60 60Characteristic impedance 25 25Maximum magnet current (kA) 1 .2 1 .2Number of magnet modules 3 10Number of vacuum tanks 1 2Magnetic length of magnet module (m) 0.59 0.52Module inductance (uH) 1.40 1.4Module filling time (ns) 56 56Booster Extraction KickerThe main parameters have been determined also for a maximum generator voltage of 60 kV and for a module travelling time of 56 ns. The results are summarized in Table 2.ConclusionA possible design for a prototype kicker magnet module for the Booster and Driver ejection has been studied. It is well within the limits of today's technology. Further investigation of the electrical circuit in conjunction with a pulse generator is necessary along with a more detailed study of the mechanical engineering.177LIST OF PARTICIPANTSA. ASTBURY, TRIUMF, Vancouver, B.C., Canada R. BAARTMAN, TRIUMF, Vancouver, B.C., C anadaB. BERKES, L B L /Paul Scherrer Institu te, Villigen, Switzerland E. BLACKMORE, TRIUMF, Vancouver, B.C., CanadaG. BOWDEN, SLAC, Stanford, CA, U.S.A.R. CASSEL, SLAC, Stanford, CA, U.S.A.G. CLARK, TRIUMF, Vancouver, B.C., Canada M. CRADDOCK, TRIUMF, Vancouver, B.C., Canada J. DOORNBOS, TRIUMF, Vancouver, B.C., C anada G. DUTTO, TRIUMF, Vancouver, B.C., Canada L. ELLSTROM, TRIUMF, Vancouver, B.C., C anada M. FEATHERBY, SAIC, San Diego, CA, U.S.A.D. FIANDER, CERN, Geneva, SwitzerlandJ-P. GARNIER, SGN Company, St. Quentin en Yvelines, FranceC. HADDOCK, TRIUMF, Vancouver, B.C., CanadaK. HALBACH, Lawrence Berkeley Laboratory, Berkeley, CA, U.S.A.E. HALLIN, University of Saskatchewan, Sask., CanadaM. HAROLD, Rutherford Appleton Laboratory, Didcot, England A. HARVEY, SLAC, Stanford, CA, U.S.A.G. IIEMMIE, DESY, Ham burg, West Germany K. IIENRICHSEN, CERN, Geneva, SwitzerlandE. HOYER, Lawrence Berkeley Laboratory, Berkeley, CA, U.S.A.J. HUMBERT, Fermilab, Batavia, IL, U.S.A.R. HURKENS, M inistry of Advanced Education & Job Training, Victoria, B.C., C anadaG. KARADY, Arizona S tate University, Tempe, AZ, U.S.A.H. KARMAKER, GE Canada, Peterborough, O nt., CanadaD. LOBB, TRIUM F/University of Victoria, Victoria, B.C., C anadaG. MACKENZIE, TRIUMF, Vancouver, B.C., C anadaF. MAMMARELLA, TRIUMF, Vancouver, B.C., C anada N. MARKS, ESRF, Grenoble, FranceM. MAY, Fermilab, B atavia, IL, U.S.A.D. NELSON, Lawrence Berkeley Laboratory, Berkeley, CA, U.S.A.W. NEVES, University of British Columbia, Vancouver, B.C., C anadaA. OTTER, TRIUMF, Vancouver, B.C., Canada D. PATEL, Hammond M anufacturing Co., Guelph, O nt., C anada D. PERCO, Transelectrix Tech., Guelph, O nt., C anada L. PICKUP, Pacific Levitation Systems, Vancouver, B.C., C anada W. PRAEG, Argonne National Laboratory, Argonne, IL, U.S.A.H. QUINN, Magnacoil Corp., Danvers, MA, U.S.A.P. REEVE, TRIUM F/University of Victoria, V ictoria, B.C., C anada K. REINIGER, TRIUMF, Vancouver, B.C., C anada V. RODEL, CERN, Geneva, Switzerland J. RUMMLER, DESY, Ham burg, West Germany M. SABADO, SAIC, Princeton, NJ, U.S.A.H. SASAKI, KEK, National Laboratory for High Energy Physics, Ibaraki, Japan178V. SOUKUP, Pacific Levitation Systems, Vancouver, B.C., C anadaG. STINSON, TRIUMF, Vancouver, B.C., C anadaB. SUCHY, Beckland Systems Ltd., Burnaby, B.C., CanadaT. SUZUKI, TRIUMF/KEK, National Laboratory for High Energy Physics, Ibaraki, Japan O. SZAVITS, Paul Scherrer Institu te, Villigen, SwitzerlandH.A. THIESSEN, Los Alamos National Laboratory, Los Alamos, NM, U.S.A.V. VERMA, TRIUMF, Vancouver, B.C., C anadaG. WAIT, TRIUMF, Vancouver, B.C., C anada U. WIENANDS, TRIUMF, Vancouver, B.C., C anada R. WINJE, SAIC, Princeton, N J, U.S.A.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items