Physics Design and Technology. Development of CSNS Accelerator

Transcription

1 Physics Design and Technology Development of CSNS Accelerator Second CSNS International Accelerator Technology Advisory Committee Review Meeting Institute of High Energy Physics, CAS January, 2010, Beijing, China

2 Contents 1 INTRODUCTION TO THE CSNS ACCELERATOR COMPLEX BEAM DYNAMICS OF DTL RCS BEAM DYNAMICS DESIGN THE DESIGN OF BEAM TRANSPORT LINE FRONT END THE R&D STATUS OF 324MHZ DTL FOR CSNS CSNS LINAC RF-SYSTEM STATUS THE PRELIMINARY DESIGN AND PROTOTYPE TEST OF CSNS MAGNETS MAGNET POWER SUPPLY SYSTEM RCS RF SYSTEM VACUUM SYSTEM THE PRELIMINARY DESIGN AND PROTOTYPE TEST OF CSNS/RCS INJECTION AND EXTRACTION MAGNETS INJECTION BUMP MAGNETS PULSED POWER SUPPLY FOR RCS MECHANICAL SYSTEM BEAM DIAGNOSTICS CONTROL SYSTEM CSNS RADIATION SAFETY DESIGN SURVEY AND ALIGNMENT ATTACHMENT:...162

3 1 Introduction to the CSNS Accelerator Complex 1.1 Basic Design Parameters Jingyu The design principles for the CSNS are: 1) Leading role. On the one hand, it will be among the leading spallation neutron sources in the world; on the other hand, it can satisfy the increasing demand of the domestic scientists in different disciplines. 2) Operational availability. As a large platform open for broad users, the facility, especially the accelerator complex must be built to provide beam time almost on schedule. On the one hand, the failure time should be minimized and some installed spares are included in the design; and on the other hand, beam losses should be controlled in a very low level to allow hands-on maintenance for most accelerator components. 3) Cost rationality. The CSNS needs large investment in the construction and also large operation outlay. Therefore, it is a key issue to make the accelerator design rationally in cost. 4) Upgrading potential. Firstly, the CSNS accelerator will be designed in two phases to enable Chinese scientists to take the opportunity to use the advanced neutron scattering tool as early as possible. Secondly, the upgrading will be included in the design from the very beginning to make it realistic and efficient in cost. Thirdly, some reserve will be kept in the design when not confronting significant increase in cost. The construction of a large-scale project such as China Spallation Neutron Source (CSNS) is a part of the national renovation on science and technology. Considering the actual development level of Science and Technology in China and the international development tendency, the proposal to construct a neutron source of 100 kw in beam power with the potential to be upgraded in the future has been approved by the Chinese Central Government. Following the conceptual and feasibility studies, CSNS will be built in two phases: 100 kw at CSNS-I and 200 kw at CSNS-II. The accelerator part has even the reserved potential to be upgraded to 500 kw. The basic parameters are shown in Table Table Basic parameters of the CSNS CSNS-I CSNS-II Beam power (kw) Repetition rate (Hz) Target number 1 1 Average current (µa) Proton energy (GeV) Linac energy (MeV) Accelerator Frame 1

4 High beam power proton accelerator is the key part in the construction of a spallation neutron source. Compared with other accelerator schemes for spallation neutron source, the CSNS accelerator scheme has the following aspects: 1) It is composed of a linac with a modest but upgradable energy and a rapid cycling synchrotron (RCS) of the fixed energy at 1.6 GeV. The space is reserved for adding more DTL tanks to increase the linac energy to 132 MeV at CSNS-II and a superconducting linac to 250 MeV at CSNS-III. This is a compromise between the construction cost for CSNS-I and the upgrading potential. 2) The repetition rate is chosen to be 25 Hz, even during the upgrading phases. This is chosen by the CSNS target-instruments. 3) The total beam loss rate in RCS is designed to be at a relatively low level. In order to achieve this, measures to counteract the space charge effects have been taken and a sophisticated collimation system has been designed, whereas the uncontrolled beam loss rate of 1W/m is maintained in the whole accelerator complex. The CSNS-I accelerator train consists of: an H- ion source of PIG surface type that will produce a beam of 20 ma and low emittance, a low energy beam transfer line (LEBT) using magnetic focusing and electric beam chopping, a four-vane RFQ linac of 3 MeV, a medium energy beam transfer line (MEBT), four DTL tanks of 80 MeV at exit, a linac to RCS beam transfer line (LRBT) of 197 m in length including the reserved space for linac upgrading, an RCS of 1.6 GeV as the extraction energy, an RCS to target beam transfer line (RTBT) of 144 m in length. Whereas the linac and the RCS are cycling in 25 Hz, the two main beam transfer lines (LRBT and RTBT) work in the DC mode. The H- beam is converted to proton beam by a stripping foil at the RCS injection, and this is necessary to accumulate the required large number of protons in the RCS and to paint the linac beam of small emittance into the large acceptance in the RCS with good uniformity in both the transverse and the longitudinal phase planes. Figure Layout of CSNS accelerators 2

5 1.3 Main Technical Challenges and R&D Studies Design and construction of a high beam power proton accelerator are world frontier topics in means of both accelerator technology and accelerator physics. We have also met many challenges during the last years and perhaps will have more in the future. From the viewpoint of accelerator physics, the main focuses are on: linac front-end design, DTL beam dynamics and structure optimization, beam halo collimation in LRBT, RCS lattice optimization, injection and phase space painting, space charge effects in both transverse and longitudinal phase spaces, beam loss budget and collimation in RCS, single turn extraction with slow bumps, and beam manipulation before the target in RTBT etc. From the viewpoint of technical issues, some of the difficulties are pointed out here: high brightness H- ion source, RFQ of high duty factor, DTL linac with high RF frequency, rapid cycling magnets and their power supplies, large scale ceramic vacuum chambers, RCS RF cavities with a fast frequency sweeping and very heavy beam loading, etc. A broad R&D programs have been selected and conducted in the last years. These include: a test-bench for the PIG surface H- ion source; a model section of a DTL cavity including the drift tubes and the focusing magnets; a resonant-type high voltage power supply for linac RF modulator, an RCS main dipole magnet with its resonant power supply to form an one-cell White circuit; an RCS main quadrupole and its resonant power supply; large scale ceramic vacuum chambers for both bending magnets and quadrupole magnets, and RF shielding and TiN coating to the ceramic chambers; a complete RCS RF station including a ferrite cavity, an bias current power supply, an RF amplifier and an LLRF system; a bump magnet and its pulsed power supply with a programmable pulse curve for injection; an extraction kicker and its PFN power supply; some beam diagnostics devices. Although we have gained much experience from the R&D studies, most of the R&D items have met challenges, and some of them are still under way to complete the technical proof. 3

6 2 Beam Dynamics of DTL 2.1 Introduction Jun The CSNS DTL consists of 4 tanks accelerating the beam from 3MeV to 80MeV. In the future upgrade the other 3 tanks will be added, and the output energy will rise to 132MeV in Phase II. In the present design, we take Phase II into account. Table2.1-1 is a summary of the major DTL parameters. Table DTL parameters Phase I Phase II Length[m] Beam energy[mev] Max. repetition rate[hz] Peak current[ma] Chopper beam-on factor 50% Average pulse current[ma] Average current[ua] ua Max. beam pulse length[ms] Max. beam duty cycle 1.05% Particle per pulse Cell geometry The geometries of the DTL cells were optimized by using SUPERFISH with the aim to maximize the effective shunt impedance and to avoid voltage breakdown by keeping the peak surface electric field below 1.3 Kilpatrick. In addition, the drift tube must be large enough to have space for housing the electromagnetic quadrupole magnets inside. The CSNS DTL accelerates beam from 3MeV to 132MeV. Accordingly, the relative velocity βof H- ions increases from 0.08 to 0.48.We divide the βinterval into 9 sections. In different sections, the geometries of drift tubes are different, and in the same section, the cell geometries of drift tubes are kept constant. It has merits both in high effective shunt impedance ZTT and easy fabrication. The main parameters of the cell geometry are shown in Table The effective shunt impedance and the quality factor calculated with SUPERFISH are shown in Fig All the DTL tanks use the same tank diameter. The drift tube diameter and the bore radius are constrained by EMQ size in tank 1 ( β [ 0.08,0.21] impedance increases obviously by increasing the face angle on the drift tubes. ). The shunt 4

7 Beta Table Main parameters of the cell geometry of the DTL Bore Radius (cm) Face angle (degree) Inner nose Radius (cm) Outer nose Radius (cm) Corner Radius (cm) Diamter of drift tube(cm) Cavity Diameter (cm) Flat length (cm) Figure The effective shunt impedance and the quality factor (SUPERFISH values) 2.3 DTL tank parameters The design of DTL tanks are completed by using code PARMILA, and the parameters of each tank are shown in Table The accelerating fields in the DTL are determined based on three conditions: 1. The total RF power consumption with a 30mA beam in a tank is as large as about 2MW so as to leave enough operating margin for a 2.5MW klystron. Each tank is powered by one klystron. 2. The peak surface electric field in the tank is smaller than 1.3 times the Kilpatrick limit, which is considered safe for reliable operation. 3. The length of each tank is shorter than 9m for mechanical requirement. Fig shows the accelerating electric field E 0 and synchronous phaseφs. The E 0 and φs adopted in the 1 st tank are different with those in other tanks. In the 1 st tank, E 0 starts at 2.2MV/m, and then linearly ramps to 3.1MV/m at the middle cell of the tank, and finally is kept constant for the remaining cells. Theφs starts at -30 which provides a large longitudinal acceptance needed in the first tank. And then it gradually ramps up to -25, providing strong longitudinal focusing at low energy. The fields and phase remain constant over the rest tanks, increasing the acceleration efficiency at high energy. 5

8 Table2.3-1 Parameters of the DTL tanks `Tank number total Output energy (MeV) Tank length (m) Number of cells RF driving power (MW) Total RF power (MW) Accelerating field (MV/m) 2.2 to Synchronous phase (degree) -30 to Figure Accelerating electric field and synchronous phase along the DTL 2.4 Focusing design The transverse focusing is arranged in an FD lattice with electro-magnetic quadrupoles. The EMQ uses J-PARC type SAKAE coil, which has no bending radius [1]. Beam stability in high current linac requires k x0, k z0 <90, where k x0 and k z0 are the zero-current wave numbers for transverse and longitudinal oscillations. Besides, the transverse and longitudinal temperatures should be equal according to the equipartitioning theory [2]: 2 2 T k xε nx 2 ε nx zm = = γ 0 = 1 (2.4-1) 2 2 T k ε ε a Here, ε nx and ε nz z nz nz are the normalized RMS emittances of the transverse and longitudinal phase spaces, a and z m are the radii of the bunch in the transverse and longitudinal directions, γ 0 is the relativistic parameter. Since accelerating field E0 and the synchronous phaseφs are usually fixed, the longitudinal focusing strength is fixed. Consequently, the transverse focusing parameters can be calculated from equation (2.4-1) using the initial parameters given at the entrance of the linac. The emittances at the DTL entrance are listed in Table The main focusing parameters along the DTL are plotted in Fig to

9 Table Normalized RMS emittnaces at the DTL injection point Transverse emittance (Norm. rms) Longitudinal emittance (Norm. rms) 0.26 mm mrad MeV deg Figure Zero current phase advance per Figure Required magnetic focusing period field gradient and length 2.5 Quadrupole magnet power supply 223 electromagnetic quadrupole magnets are used in CSNS DTL. In tank 1, two magnets in series are powered with one power supply. In tank 2-7, six magnets are grouped in series for one power supply. Totally 58 power supplies are demanded. Compared with one-to-one power supply scheme, this power supply scheme can save cost. EMQ is adopted for CSNS DTL because it can be adjusted easily. The power supply scheme degrades EMQs flexibility, so a mini power supply is added on each magnet for fine tuning. The tuning range of the mini power supply is about ±5% of the main power. 2.6 Beam simulation The multi-particle beam simulation for the CSNS DTL has been performed with PARMILA code. The simulation starts with 30 ma H beam at the entrance of the DTL with initial uniform distribution. The results have been iterated to obtain acceptable matching in TREACE3D.After this, the match results were input into PARMILA macro-particles per bunch are used. Space charge interaction is calculated via the 2-dimensional PIC method with a mesh. The mesh size is 0.05cm. The simulation results are shown in figure

10 Figure Phase-space distribution at the exit of the DTL Figure Normalized RMS emittance along the DTL Figure RMS and total envelopes along the DTL At the exit of the DTL, the longitudinal RMS emittance increases 2.45% compared with the initial value. The transverse emittance almost has no change. The ratio of the aperture to the RMS beam size is 4 in the first tank and higher than 7 in all rest tanks. 2.7 Error study The goal of error study is to define the manufacturing tolerances of the DTL. Simulations are performed with the PARMILA code. We simulate alignment, focusing and RF errors. Each error is applied on all DTL cells. For each cell, the amplitude of the error is generated randomly and uniformly within a given range [-max, +max]. The results are obtained by averaging 10 calculations. A uniform distribution with macro-particles per bunch is used in these simulations. Space charge 8

11 interaction is calculated via the 2-dimensional PIC method with a mesh. The mesh size is 0.05cm. Individual error is simulated each run, and we haven t do global error simulations. We vary the maximum allowed amplitude of the error to determine the amplitude of each error, with the requirements that beam has no loss and emittance growth is smaller than 20%. The manufacture and RF tolerance for the DTL are summarized as follows for the quadrupoles: 1. Transverse displacements: δ x,y =±0.05mm 2. Ratations : φx,y,z=±3mrad 3. Integrated field : GL/GL=±3% And for the accelerating field: 1. Klystron field: E klys / E klys =±1% 2. Klystron phaseφklys=±1deg 3. Gap field:: Egap/ Egap=±1% 2.8 Tuning of the 1 st tank (a) Tuner and post coupler 12 slug tuners and 30 post couplers are used in the 1 st tank. The parameters are listed in Table The distance from the end of post couplers to the drift tube is 2 cm and the distance from drift tube to tank wall is 0.89λ/4 Table Parameters of the stem, post couplers, slug tuners and vacuum ports Diameter(mm) Number Stem Post coupler Slug tuner Vacuum port The tank diameter is adjusted from 56cm to cm to account for known frequency shifts due to these elements perturbation, including half the 2MHz slug-tuner range and 1.5 MHz frequency shift caused by stems and post couplers. (b) Accelerating field tuning DTL is usually fabricated directly using the cell data given by DTLFISH and PARMILA. However, a general problem appears: the measured E 0 for each cell in a fabricated tank deviate a lot from the design E 0D. Here, we use E 0M to stand for the measured value and E 0D to stand for the accelerating field in PARMILA input file. This deviation will affect the net accelerating voltage and the peak surface field. Engineers used to think the deviation is attributed to the errors in fabrication and assembly. The problem is usually solved by using slug tuners and post couplers. Tuning work often costs a lot of time and the inserted tuners lower the Q value of the tank. We discovered that the deviation, in fact, already exists before the fabrication. It is because that the E 0D in PARMILA are decided by the designer and the designer doesn t know the cell geometries beforehand. Then PARMILA uses the input E 0D to generate cell geometries of a tank, by interpolating the cell data from DTLFISH that 9

12 do not account for E 0D variation information. In a tank, the field distribution in each cell is affected by the field in other cells, and the E 0 of each cell is not only related with the input power, but also with the cell geometries. As the results, the deviation between the E 0M and the E 0D appears. The 1 st tank of CSNS DTL has been simulated with 2D electromagnetic field code MDTLFISH. Here E 0S is used to stand for the simulated E 0 value in MDTLFISH code. It obviously deviates from the E 0D and the difference is illustrated by the ratio of E 0S to E 0D, as plotted in Fig The ratio ranges is rather large: from 0.92 to Taking into account the proved accuracy of MDTLFISH, this result confirms that the E 0 distribution pattern in the tank generated by PARMILA doesn t agree with the designed E 0 distribution pattern. To decrease this deviation, the frequency of the 3 cells at the beginning and the 3 cells in the middle of the 1 st tank of CSNS DTL are tuned by adjusting their gap length in the MDTLFISH simulation. The frequencies of cell 1, 2, 3 are tuned from 324MHz to MHz, and the frequencies of cell 22, 23, 24 in the middle are adjusted from 324MHz to 321.8MHz. In average, there is little resonance frequency shift in the tank. Since the transit time factor T of the tuned cell is changed because we have adjusted the cell gap, we return to PARMILA code to correct the SFDATA tables with new T and other parameters. Running PARMILA again, the code generates a new tank by interpolating from new SFDATA tables. The E 0D and the E 0S for this new tank are shown in Fig The ratio is much less than that in Fig and not difficult to be tuned by tuners and post couplers [3]. Figure The E 0 distribution pattern of DTL-1 for CSNS Figure The E 0 distribution pattern of DTL-1 for CSNS, after tuning [1] H. Ino et al, Advanced Copper Lining for Accelerator Components, LINAC2000. [2] M Reiser. Theory and Design of Charged Particle Beams (John Wiley, New York, 1994). [3] Jun P, Zhiri S, Shinian F. A new tuning procedure for the DTL RF field pattern. Chinese Physics C (HEP & NP) ; (02):

13 3 RCS Beam Dynamics Design 3.1 Lattice scheme Sheng Due to the requirement of the beam collimation for beam loss control in a high intensity proton synchrotron, the lattice with 4-fold structure is preferred for separated-function design: the collimation can be performed in a separate straight section. Also the lattice superperiodicity of 4 is better for reducing the impact of low-order structure resonance than that of superperiodicity of 3. CSNS/RCS design chooses this kind of 4 super-period structure. The dedicated functions are accommodated in the different sections of ring, as shown in the sketch map in Figure Figure 3.1-`1 The sketch map of dedicated functions distributed in the different sections of ring The optimization iteration for lattice design has been done. Compare with the lattice scheme in the feasibility study report (the scheme is also the one presented in the first CSNS international review meeting, called old lattice, here in after), the large change has been made for the present improved lattice scheme (called new lattice, here in after). The old lattice scheme will be reviewed firstly, and then the 11

14 introduction to the new lattice scheme. The comparison between two schemes will also be given. (1) The old lattice scheme The old lattice adopts hybrid structure, which consists of FODO arc and doublet straights. Fig shows the layout of the RCS ring. As a compromising of the magnetic field quality and the volume of the dipole, the length of the dipole is chosen as 2.1 m, and totally 24 dipoles are used for RCS. For accommodating momentum collimator, a gap with large dispersion function is created in the middle of the arc. There are 6 dipoles at each arc. An arc consists of phase advance FODO cells. The dispersion is suppressed by using two groups of 3 half-cells (with 90 horizontal phase advance per cell) located on each side of a missing-dipole half-cell. The long (one 9.3 m and two 6.3 m uninterrupted drifts per straight) dispersion-free straights facilitate injection, extraction, and transverse collimation. Arcs: 3.5 DOFO cells, 315 degrees phase advance Straights: doublet, m long drifts at each straight Dipole : <1.0 T Quadrupole gradient: <5.5T/m Figure The layout of old lattice Figure gives the twiss parameters of one super-period. For there are 3.5 cells at each arc, the lattice function is asymmetric. It contains 24 dipoles and 48 quads. The circumference is 248 m. The base tunes are (5.82, 5.80). The straight section adopts doublet structure, and each straight section consists of two 6.5 m and one 9.3 m long drift space. The total dispersion free long straight section is 89.2 m. In the middle of the arc the missing dipole form a 4.1m straight section for momentum collimation. The peak dispersion is 6 m and the peak beta is less than 25 m in the straight, and less than 16 m in the arc. The gap created by the missing dipole near the maximum dispersion location allows efficient momentum collimation. The FODO arc allows easy lattice optics correction. 12

15 Figure Twiss parameters for RCS lattice in one super-period (2) The new lattice scheme The optimization efforts aiming at reducing the budget of RCS was done. Decided by the characteristics of FODO cell, the gap of dipole is difficult to be further decreased, and also the aperture of quadrupoles located at the position with large dispersion is also hard to decreased. To further decrease the gap of dipoles and the aperture of quadrupoles, and lower the budget of magnets and power supply, the new lattice scheme was proposed and adopted. The new lattice is based on triplet cell, and the whole ring consists of 16 triplet cells, with circumference of m. Also the 4-fold structure is adopted. The number and length of dipole are the same as the old lattice: totally 24 dipoles and 2.1m long for each. In each super period, the structure is mirror symmetry on the middle point of each super period. Figure shows the layout of the new RCS ring. There are 6 dipoles at each arc, located at three triplet cells. In the middle of the arc, a 3m long space is reserved for accommodation of momentum collimator. Momentum collimator take only one of the four arcs, and the other three can be used for long beam instrumentation and other devices. In the two side cells in the arc, a 3.85m space is reserved between dipole and quadruples. In the each super period, an 11m long drift space is left in a triplet cell, and this uninterrupted long space is very good for accommodation of injection, extraction and transverse collimation system. Figure gives the twiss parameters of one super-period. The 48 quadrupoles are powered by 5 families power supply. The maximum beta function is less than 26m, and the maximum dispersion function is less than 4m. Especially in the middle of the arc, the dispersion is large and the horizontal beta function is small, and this is good for high efficiency momentum collimation. Take the advantage of the triplet, the double-waist beta function benefit to not only the decreasing of the gap of dipoles and the aperture of RF cavities, but also the design of injection, extraction and beam collimation. The comparison between main parameters of two lattice schemes is given in the Table

16 Figure The layout of new lattice Figure The twiss parameters of one super-period for new lattice. 14

17 Table The primary parameters of two lattice schemes Scheme New lattice Old lattice Circumference (m) Superperiod 4 4 Number of dipoles Number of quadrupoles Lattice structure Triplet FODO+doublet Number of long drift(dispersion free) Total Length of long drift (m) Nominal Betatron tunes (h/v) 4.82/ /5.80 Chromaticity (h/v) -4.3/ /-7.3 Transition gamma RF harmonics 2 2 RF Freq. (MHz) 1.022~ ~2.25 RF Voltage (kv) Trans. acceptance (µπm.rad) Momentum acceptance 1% 1% Effective length of dipoles(m) Dipole gap (mm) Max. quadrupole aperture (mm) Effective length of quadrupole(mm) 410,450,600, ,700 Number of dipole power supply 1 1 Number of quadrupole power supply Chromaticity correction and dynamic aperture Although the nature chromaticity is not so large, as shown in Table 3.1-1, and it operates under the transition energy, the chromaticity correction is not indispensable, according to the commissioning experience of SNS accelerator, to meet the requirement of commissioning and machine study, the chromaticity correction is designed. Figure shows the distribution of sextupoles. Totally 16 sextupoles are used, and powered with 2 families power supply. The power supply is DC, so the chromaticity can be corrected only at the low energy stage. Figure The distribution of sextupoles in one super period (blue block) 15

18 The triplet based lattice is not good for chromaticity correction design, and with two families power supply, only first order chromaticity can be corrected. Figure shows the horizontal and vertical beta functions for p/p=0, ±1%, with and without chromaticity corrections. The deviation of off-momentum beta function from on-momentum axis is very small, for both correction and no correction. beta-y dp/p=-1%,ξ=-9 0 dp/p=0,ξ= distance/m dp/p=1%,ξ=-9 beta-y dp/p=-1%,ξ= dp/p=0,ξ= dp/p=1%,ξ=-0.5 distance/m Figure Vertical on- and off-momentum beta function for one super-period, without chromaticity correction (left) and with chromaticity correction (right) The tracking was done for checking the dynamic aperture with the nonlinear effect. In two dimensional case (x-y), the tracking results show that the dynamic aperture with sextupoles only for particles of p/p=±1% is around 4.5σx 2.5σy, where σx and σy are horizontal and vertical beam size. With sextupoles and high order field of dipoles and quadrupoles, the dynamic aperture is decreased to 2.5σx 2σy. as shown in the Figure Figure The dynamic aperture with the effect of sextupoles and high order field of dipoles and quadrupoles (left for particles of p/p=1%, and right for particles of 3.3 Closed Orbit Correction p/p=-1%) There are 32 BPM in the whole ring for closed orbit correction, two BPM in each triplet cell. The number of dipole correctors in RCS is 34, in which 17 are for horizontal plane and 17 for vertical plane. The power supply for dipole corrector is programmable, and the dipole corrector should be ramped 10 to 20 steps during one RCS cycle. The maximum correction ability of dipole corrector is 1mrad at 1.6GeV. With these BPM and correctors, the closed orbit distortion can be well corrected. Figure gives the statistics of rms closed orbit for 20 groups of random error. The closed orbit distortion comes from alignment error of dipoles and quadrupoles, the field error of dipoles, and the alignment error of BPMs. 16

19 After correction After correction Events number x_rms(mm) Events number y_rms(mm) Figure The statistics of rms closed orbit for 20 groups of random error Another distortion to closed orbit is the stray field of Lambertson magnet. The Lambertson magnet operates in DC mode, and its stray field generates the distortion to the circulating beam, especially to the beam in low energy stage. According to the calculation results of OPERA-3D, the vertical stray field yields 0.7mrad impact to the beam of 80MeV. A local 3-bump is design to locally correct this distortion. 3.4 Tunes and parameter correction Due to the resonant working mode of main dipoles and quadruples, the field saturation will occur in high energy stage in an RCS cycle. The saturation will lead to the mismatch between dipoles and quadrupoles, and make the deviation of twiss parameters and tunes. When the deviation is large, the correction is necessary. To perform this kind of correction, trim quadrupoles are adopted in the optics design. Figure shows the location of trim quadrupoles in a super period. The total number of trim quadupole is 24. Trim quadrupols are powered with programmable power supply. They also can be used to correct or compensate some of the space charge effect. Figure shows the tune shift due to saturation of dipoles and quadrupoles and the correction results. The average value of saturation error is 1%, and 30 groups of random errors were used for statistics. With the errors due to saturation, the horizontal tune is shifted to 4.78~4.93, and vertical tune is shifted to 4.55~5.14, while the nominal tunes are 4.86/4.78, and it also shows, with corrections by using trim quadrupoles, the tunes are corrected to the nominal value. The trim quadrupole also benefits the response matrix beam measurement. Figure The distribution of trim quadrupoles in one super period (blue block) 17

20 Tune-y Tune-x Figure Tune shift with 30 groups of random saturation errors (red point is for 3.5 Longitudinal dynamics design after correction) The waveform of magnetic field variation is sinusoidal in the RCS, as shown in Figure In phase I, the RF cavities with harmonic number of 2 are adopted to reduce the RF voltage demand, and in the future upgrade for higher beam power, dual harmonic cavities will be added. The design of RF voltage and phase curves is an import issue to decrease the beam loss due to the space charge and phase changes. An RF acceleration period consists of three stages: injection, capture and acceleration. There are 7 RF cavities in the phase I to provide total RF voltage of 165 kv, with additional one cavity for redundancy. Figure One cycle of magnetic field of RCS and its RF acceleration period The RF voltage curve and the corresponding RF phase curve are calculated by using the code of RAMA. In the beginning of the acceleration, the bunching factor is about 0.4, and with the increasing of RF voltage, the bunching factor is decreased to The filling factor used in the calculation is 0.8. Fig shows the variation of RF voltage and energy gain during half RCS cycle, as well as the variation of RF 18

21 synchronous phase and bunch center phase. Figure The curves of RF voltage and energy gain during half RCS cycle(left), and the curves of RF synchronous phase (Phi-S) and bunch center phase (Phi-B) (right) To decrease the beam loss during the RF capture, the RF voltage should be increased rapidly in the beginning of the RF acceleration, as shown in the Fig , during the first 0.5 ms, the RF voltage is increased to 70 kv from 21 kv, while the change of the magnetic field is very small. If the initial RF voltage is high, the beam loading factor will be large, so the initial RF voltage is set to 20 kv. The RF voltage is increased to 165 kv during the 5 ms, and the RF voltage is 100 kv at the end of the acceleration. 1-D ORBIT is used to simulate the beam loss during one RCS cycle. The painting procedure is not included in the simulation. With chopping rate of 50%, there are only 3 particles lost in macro-particles during the acceleration. Fig shows the longitudinal phase space at the initial and end of acceleration. Figure The longitudinal phase space at the initial and end of acceleration 3.6 Beam loss and Collimation In the whole accelerator design, beam losses should be controlled in a very low level. Based on the past operational experience, to allow hands-on maintenance for most accelerator components, an average uncontrolled beam loss should be not exceeding about one watt of beam power per tunnel meter. For CSNS case, in the first phase, one watt of beam power per tunnel meter corresponds to a fractional uncontrolled beam loss of To control the beam loss to this level, both longitudinal and transverse collimation systems are required to reduce the uncontrolled beam loss within the acceptable level for hands-on maintenance. By 19

22 using the momentum collimators located at straight section in one middle arc and the transverse collimation located at one long straight section, it is expected to obtain over 95% collimation efficiency. There is one momentum collimator located in the gap at the middle of arc. The type of the momentum collimator is direct absorber made of graphite and copper. The transverse collimation system adopts the two-stage collimation system. It consists of one primary collimator and four secondary collimators. The transverse collimation system takes a separate straight section, just downstream of the momentum collimators. Halo particles are scattered by the primary collimators, and the secondary collimator absorb these scattered particles. It is expected to have a collimation efficiency of over 95% for the whole collimation system. Fig shows the layout of the two-stage transverse collimation system which is located in one straight section of RCS. The acceptance of collimator is a key parameter for high collimation efficiency. Figure shows the designed beam emittance and collimator acceptance. The beam emittance after painting is around 320πmm.mrad, and the primary collimator is movable, the acceptance can be changed, while the acceptance of secondary collimator is set to 420πmm.mrad, and the acceptance in the other part of the ring is 540πmm.mrad. Fig The layout of the transverse two-stage collimation system in the straight section Fig Beam emittance and acceptance of collimators and ring Figure shows the beam loss distribution in the RCS without beam 20

23 collimation. In case of no beam collimation, the total beam loss is small, but the maximum local beam loss is over 5W/m, so the collimation is necessary to keep the uncontrolled beam loss less than 1W/m. Figure shows the beam loss distribution with transverse and momentum beam collimation, in which the total beam loss is about 5%, and the collimation efficiency is more than 96%. Figure Beam loss distribution in the RCS without beam collimation Figure Beam loss distribution in the RCS with transverse and momentum beam collimation 3.7 Space charge effects and emittance growth Due to the high beam density and high repetition rate, the rate of beam loss must be controlled to a very low level. In this kind of high power RCS, especially in the low energy end, the beam is space charge dominated, and the space charge effects are the main source of beam loss. The space charge effects limit the maximum beam density, as well as beam power. Many simulation works have been done for the study of space charge effects by using code ORBIT and SIMPSONS. Various conditions, which may influence the space charge effects and beam loss, are considered, including the effects of different lattice structure, different tune, the combine effect of sextupole field and space charge, different painting beam distribution, etc. Some injection painting optimizations were made for decreasing halo formation and tune spread. The beam loss and emittance growth are compared for different conditions. The simulation results are the foundation of physics design and the choice of some 21

24 design parameters. As examples, some simulation study results are given. Figure shows the emittance growth after painting due to the space charge effects, in which the correlated painting was used. Figure shows the beam distribution in y direction, in which, with the space charge effect, the beam distribution is distorted from the correlated painting result with no space charge effects (left figure), and with the optimization of painting bump orbit, the beam distribution with space charge effect are optimized (right figure). Figure The emittance growth due to the space charge effects, red point represent the particles without space charge effects, and the black points represents the particles with space charge effects. The distribution is obtained at just finishing the correlated painting. Figure The beam distribution in y direction for correlated painting. With the space charge effect, the beam distribution is distorted from the painting result with no space charge effects (left). With the optimization of painting bump orbit, the beam distribution with space charge effects are optimized (right). 3.8 Impedance and instability (1) Impedance budget The impedance of RCS comes from the wall of chamber, bellows, kickers, RF cavities, BPM, collimators, steps and space charge. The impedance of these sources has been calculated both for injection energy and extraction energy. Table and Table give the longitudinal impedance budget and transverse impedance budget respectively. It can be seen from these impedance budget, the space charge impedance dominates in the broad band impedance, and wall impedance mainly comes from the stainless steel chamber. Table RCS longitudinal impedance budget (Ω,n=ω/ω0) Impedance type Kinetic energy Components 0.08 GeV 1.6 GeV 22

25 Space Charge j j96.62 Wall impedance Stainless steel chamber 1.52(1+j)/n 1/2 2.35(1+j)/n 1/2 Ti chamber 0.015(1+j)/n 1/ (1+j)/n 1/2 Copper stripline 0.064(1+j)/n 1/ (1+j)/n 1/2 Collimator 0.083(1+j)/n 1/2 0.13(1+j)/n 1/2 RF cavities (1+j)/n 1/ (1+j)/n 1/2 Total wall impedance 1.68(1+j)/n 1/2 2.61(1+j)/n 1/2 Broad band Ceramic chamber with j n j n TiN coating Step j0.23 j0.55 RF cavities j0.014 j0.033 Collimators j0.17 j0.42 Bellows j0.17 j0.40 Pump ports j0.083 j0.083 Flange shielding j1.12 j2.67 Extraction kicker j13.45 j32.16 Table RCS transverse impedance budget(kω/m,n=ω/ω0) Impedance type Kinetic energy 0.08 GeV 1.6GeV Component Space charge j16.77k j3.30k Wall impedance Stainless steel chamber 28.42(1+j)/n 1/ (1+j)/n 1/2 Ti chamber 0.28(1+j)/n 1/2 0.18(1+j)/n 1/2 Copper stripline 1.19(1+j)/n 1/2 0.77(1+j)/n 1/2 Collimator 6.14(1+j)/n 1/2 3.97(1+j)/n 1/2 RF cavities 0.093(1+j)/n 1/ (1+j)/n 1/2 Total wall impedance 36.12(1+j)/n 1/ (1+j)/n 1/2 Broad band Ceramic chamber with j n j n TiN coating Step j4.74 j4.74 RF cavities j0.23 j0.23 Collimator j12.38 j12.38 Bellows j1.72 j1.72 Pump ports j0.86 j0.36 Flange shielding j16.63 j16.63 The impedance of whole ring can be represented by the following impedance model, and the longitudinal and transverse impedance can be calculated for any frequency: At injection energy : 4 1+ j Z // / n( Ω) = j j n +, 2 n n + j20.87( n 4) Z ( kω / m) = j n j n, j j n + 2 n n At extraction energy: 3 1+ j Z // / n( Ω) = j j n +, 2 n n + j12.28 n 4 ( ) 23

26 Z (kω / m) = j j n j n + j n + 2 n n (2) Collective Instabilities The collective instabilities due to impedance may lead to the beam emittance growth, limitting the beam current, and inducing the beam loss. Based on the impedance budget, it is important to estimate various collective effects, including its rise time and threshold. In a proton synchrotron, these collective effects include longitudinal microwave instability, transverse mode coupling instability, e-p instability, etc. The threshold of longitudinal microwave instability is shown in Table The primary calculation shows that there is no instability occurred. 2 Table The threshold of microwave instability Energy 80MeV 1.6GeV Z // 0 /n (Ω) Z (MΩ/m) The threshold of transverse mode coupling instability is shown in Table The impedance budget is larger than this threshold, and there exists possible instability. But due to the damping effect of space charge, for synchrotron operates under transition energy, the theoretical calculation is always conservative. The details numerical simulation is needed for further study. Table The threshold of mode coupling instability Energy 80MeV 1.6GeV Z (kω/m) The rise time of wall impedance instability is shown in Table The rise time of transverse wall impedance instability is less than a RCS cycle period. Table The threshold of mode coupling instability Energy 80MeV 1.6GeV τ(ms) (n=ω/ω 0 ) On e-p instability, the simulation shows that due to the long bunch space, electron cloud will not accumulate, and e-p instability is not a trouble. 3.9 Injection (1) Component layout In order to depress strong space charge effect, injection into the RCS is by using H- stripping and painting method to match the small emittance beam from linac to 24

27 large emittance beam in RCS. Figure gives the injection scheme. The whole injection chain is arranged in an 11 m long free straight section, consisting of four horizontal painting magnets (BH), four vertical painting magnets (BV), and four fixed orbit bump magnets (BC). The BC magnets are adopted to meet the need of the fixed injection angle, which is required by the injection septum. To avoid using ceramic chamber with a complicated shape in the injection region, the BC magnets will work in DC mode. In general, DC mode will increase the traversal number in the stripping foil, but this problem can be solved by adding an additional offset in the horizontal painting orbit. When the injection is finished, the offset is decreased to 0 quickly, and the additional traversal is avoided. There are a small fraction of H- can not be converted, which depends on the thickness of stripping foil, and bulk of them are H0, a small fraction of them are H-. To control these particles, the second stripping foil is adopted. The bulk of H0 are converted into proton in second foil and sent to injection dump through BC4 and a septum. The small part of H- escaped from main foil stripping, will mostly stripped by the second foil and absorbed by a local absorber (benefits from very small beam power). A small fraction of H0 and H- particles in high excited states may be stripped by the magnetic field of BC3, and will loss in the injection rejoin or become halo beam in the ring that finally be stopped by the ring collimators, this part of beam loss is neglectable. Figure Injection component layout. BC1~BC4:Fixed orbit bump magnets, BH1~BH4:horizontal painting bumpers,bv1~bv4:vertical painting bumpers, ISEP1&2: septum (2) Painting A careful design of the painting scheme is very important to control the emittance blow-up and beam loss. Both correlated and anti-correlated painting schemes are available. In the correlated painting scheme, the beam fills the emittance from inner to outer for both the horizontal and vertical painting, and the beam distribution in the x-y space is nearly rectangular. In the anti-correlated painting scheme, the beam fills the vertical emittance from outer to inner, while fills the horizontal emittance from inner to outer, and the beam distribution in x-y space is elliptical. The painted emittance is around 240 πmm.mrad, and at the end of injection, due to the existence of space 25

28 charge effects, the transverse emittance is around 320 πmm.mrad. Many works have been done to optimise the painting procedure. Fig shows the beam distribution in phase spaces at the end of injection with correlated painting and related emittance evolution. Figure The beam distribution with correlated painting in phase space and x-y real space, and the emittance evolution in painting process Extraction The 1.6GeV proton beam is extracted by one-turn extraction from RCS. In each RCS cycle, two bunches will be extracted at one time. The bunch length is about 70~100ns, and the space between two bunch is about 330~360ns. The beam is vertically kicked by a series of kicker to a horizontal bending Lambertson type septum. The rise time of kicker is required to be less than 250ns and flat top field need to be kept more than 550ns for high efficiency extraction. The vertically kicked orbit should have a separation large enough to put the septa between the acceptance of circulating beam and extracting beam. The auxiliary bump magnets are adopted to provide additional extraction orbit and ease the kicker requirement. The extraction system is performed in one 11 m long straight section. The layout of components is shown in Figure There are 7 kickers, one is put to the position before a defocusing quadrupole in a triplet cell, the other 6 kickers, together with Lambertson and auxiliary bump magnet, are put in the 11 m long straight section. The space between the seventh kicker and auxiliary bump magnet is reserved for installing additional three kickers in case of necessary. Table gives the basic parameters of extraction system, and Table gives the parameters of the 7 kickers. 26

29 Figure The layout of extraction components and vertical orbit displacement Table Basic parameters for extraction system Components Parameter General Acceptance (πmm.mrad) 350 Kicker Lamberson Number 7 Total strength (mrad) Rise time (ns) 250 Pulse width (ns) ~600 Magnetic field (G) 530~620 Number 1 Effective length(m) 2.2 Angle ( ) 13 Field strength (T) Table Parameters of kickers K1 K2-K5 K3-K4 K6 K7 Field strength (T) Angle (mrad) Effective length (m) Flat-top time (ns) >600 >600 >600 >600 >600 27

30 Pole width (mm) Gap (mm) Inductance (uh) Stray inductance (uh) Rise time (ns) Effective current (A) PFN Voltage (kv) Field uniformity (350π) <±1% <±1% <±1% <±1% <±1% Peak field uniformity <±2% <±2% <±2% <±2% <±2% The auxiliary bump magnets work in DC mode. It produces about 30 mm bump orbit at the injection energy, to keep the acceptance of 540πmm.mrad at the location of septa. With the energy increasing, the beam rigidity is increased, and the bump orbit shrinks and approaches to the septa. At the extraction energy, the acceptance is decreased to 250πmm.mrad, and this effectively reduces the requirement to the kicker strength. 28

31 4 The Design of Beam Transport Line Sheng 4.1 Linac to RCS beam transport line (LRBT) 1) Beam line design LRBT transports H- beam from linac to the RCS injection point. It performs following tasks: matching the beam into the ring upon requirements, collimating the transverse and momentum beam halos, decreasing the energy spread and energy jitter by using a debuncher, and reserving enough space and potential for upgrading linac beam energy to 250MeV. Figure shows the layout of LRBT. A branch beam line, named as LDMP2, separates from the LRBT main line at switch bending magnet, transports beam to a beam dump. The length of LRBT is m, and the LDMP2 branch is 33 m. The twiss parameter of LRBT are shown in Figure The rms emittance of linac beam is about 0.6 πmm.mrad. To reduce the beam loss due to halo particles, the designed acceptance of LRBT is 25 πmm.mrad. Figure The layout of LRBT 29

32 Figure LRBT main line twiss parameters LRBT can be divided into several sections according to its function. 1) The match section connected with the exit of linac matches the beam from DTL FODO structure to LRBT Triplet. 2) Long straight section consists of 10 triplet cells, in which 85m space taken by 6 cells is reserved for energy upgrade, and a debuncher and transverse collimators are located here. 3) Matching section between triplet cell and bending section, which consists of two doublet cells. 4) 45 anti-symmetric achromatic section. 5) Match section for injection into ring, in which totally 9 quadrupoles are used to match the transverse phase space. LDMP is used not only for commissioning, but also for stopping protons converted by collimators. The proton beam in LDMP can also be extracted to a branch line for proton application test. 2) Beam collimation The phase advance of triplet cell in LRBT is 60, and this benefits to the collimation design. Three stripping foils are designed in the straight section, once the halo particles hit on foils, they will be converted into protons. The stripping foils are located at the position with double waist envelope function, so the protons will be transported along with the H-minus beam until the switch bending magnet. From there, the proton beam will be separated from the H-minus beam and transported to the beam dump. The triplet has the property of matching both H-minus and proton beams, ensures low beam losses in the straight section. Figure shows the sketch map of transverse collimation. Momentum collimator is installed in the front part of 45 bending section, which has a large dispersion function, and the particles with momentum deviation over 0.5% will be stopped. 30

33 Figure Sketch map of transverse collimation 3) Debuncher The momentum spread of the core beam from the linac is about 0.1%, but the longitudinal beam halo has much larger momentum spread, and the space charge effects also increase the momentum spread in the long transmission in LRBT. The momentum spread is a key issue to induce beam losses in the RCS. A debuncher (a cavity like MEBT buncher) is designed in the LRBT to control the momentum spread less than ±0.1% at the injection point. The debuncher can also correct the momentum jitter of the linac beam when it is less than ±0.2%. By carefully choosing the drift distance, RF voltage and RF phase, the momentum jitter and spread can be well corrected. 4.2 RCS to target beam transport line (RTBT) RTBT transport extracted beam from RCS to target. Compare to the LRBT, the beam power is much high, and beam loss control is much more important. Matching and tailoring the beam to get the required profile and beam density in the target is also an essential design goal. A beam dump is necessary for beam commissioning. The proper space should be reserved for extracting a fraction of beam to the second target station for future upgrade program or to the area for high energy proton beam application. Figure shows the layout of RTBT, and Figure gives the twiss parameter of RTBT main line. The interface between the RCS extraction and RTBT is the exit of the Lambertson. The un-normalized extracted beam emittance is about 70 πmm.mrad (98% particles), but the total particle emittance is more than 100 πmm.mrad. The acceptance of RTBT is 350 πmm.mrad, almost same as the acceptance of RCS beam collimators. The length of RTBT main line is 144.2m, and a 31

34 branch to beam dump is 41.2m. A 13 bending magnet and a group of quadrupoles match the extracted beam to FODO cell in the long straight section, and also eliminate the horizontal dispersion induced by the Lambertson; and in this part, there is enough space to accommodate fast extraction components to extract the beam to the planed second target station or proton application area. When the bending magnet is switched off, the beam is transported to the dump. Before the target, a 15 bending magnet is arranged. By this bending magnet, the back scattering neutrons from the target can not travel far away along the RTBT line, and on the other hand, these fast neutrons can be used for some neutron experiment. The last 4 quadrupoles are used for match the required beam profile to the target. Collimators are designed at RTBT for protection of the modulators and shielding the back scattering neutrons. Figure The layout of RTBT Figure Twiss parameter in RTBT main line The beam profile at the target is required to be 120 mm (h) * 40 mm (v) by the target design, and also the distribution with low peak beam intensity on target is better for target lifetime. To optimize the beam distribution in target, two step field magnets 32

35 are designed in RTBT for horizontal and vertical direction respectively. The field distribution of the magnet is shown in Figure By using this kind of field, it is expected to fold the halo particles into the footprint on the target, and then the footprint of the core beam can be enlarged, resulting in a more uniform intensity profile. Figure The calculated field distribution of the step field magnet 33

36 5 Front End Huafu OUYANG The front end includes the H- ion source, the low energy beam transport line (LEBT), the radio frequency quadrupole accelerator (RFQ) and medium energy beam transport line (MEBT). 5.1 Ion source In Table 5.1-1, the main parameters of the ion source are listed. CSNS ion source adopts the Penning surface plasma H- ion source (same as the ion source of ISIS) due to its current intensity, beam emittance, longevity, reliability, comparatively lower cost, and most important, satisfying the requirement of the CSNS Phase-I on the source. Another factor for choosing Penning H - ion source is that the technical guidance is available from the Rutherford Appleton Laboratory (RAL) for a good cooperation RAL and the Institute of High Energy Physics (IHEP). Table Ion source main parameters Ion type H - Output energy (kev) 50 Repetition rate (Hz) 25 Pulsed current (ma) 20 Norm.rms emittance (π.mm.mrad) 0.2 Pulse width (µs) ~500 Longevity (month) ~1 5.2 LEBT 34

37 Fig The schematic layout of CSNS Fig The beam envelope of LEBT The schematic layout of CSNS LEBT is shown in Fig.1. The length of LEBT is 1680mm. As shown in Fig , three solenoids are used to match the un-symmetric beam form ion source into the symmetric acceptance of RFQ, and an electrostatic deflector as the pre-chopper is chosen to chop the beam for CSNS. The deflector is installed downstream LEBT, i.e., the entrance of RFQ. The expected rising and falling time of the chopped beam is within 20-30ns. Fig shows the beam envelope. In Fig.5.2-3, the schematic layout of the deflector and other relative components is shown. The chopped beam is designed to lose in the RFQ cavity. The sloping deflecting plates instead of parallel deflecting plates are adopted to decrease the deflecting voltage. The gap between the deflecting plates varies with the beam envelope and keeps about 1.2 times the beam envelope size. The needed deflecting voltage is 4.5kV. In order to confine the destruction of charge neutralization caused by the deflector in the local area, a ground potential electrode and an electron-trapping electrode are installed upstream the deflector. In addition, a beam collimator is also installed on the flange of RFQ. The collimator, which is electrically insulated from the flange of RFQ, is split azimuthally into four quadrants that are also electrically insulated from each other. Fig The schematic layout of the deflector and other relative components, for comparison, the beam envelope at a beam current of 40mA is also shown 5.3 RFQ A High-duty factor proton RFQ accelerator has been constructed at IHEP, Beijing for the basic study of Accelerator Driven Sub-critical System (ADS). In the initial commissioning of the 3.5MeV RFQ with an ECR ion source showed a nice performance with a transmission rate about 93% and an output beam current of 35

38 44.5mA. The 352.2MHz RFQ is basically design for CW operation with the RF power source from LEP-II of CERN. Recently, progress in high-duty factor operation from about 7% to 15% is achieved. Fig and Fig show ADS RFQ in the installing process and the commissioning result. Fig ADS RFQ in the installing process Fig mA pulsed current with a transmission about 93% and beam duty factor of 7.15% Based on the success and experience of the former RFQ, the CSNS RFQ structure will keep same as much as possible with the former. The CSNS RFQ still consists of two resonantly coupled segments and each segment consists of two sections that technically connected together. Comparing to the length of 4.75m of the former RFQ, the length of the CSNS RFQ is 3.62m, which is only about 3.9 times long as the wavelength of RF. 3-dimension simulations show that the dipole stabilization rod is not needed any more due to the shorter length of RFQ while keeping enough interval between the operation mode and the neighboring dipole modes. In Table 5.3-1, the main RFQ design parameters are listed. In Fig , some parameter variation versus cell number is shown. In the figure, a stands for the minimum bore radius, m modulation factor, B focusing strength, W synchronous energy and φ s synchronous phase. With this set of parameters, the transmission of the beam got by PARMTEQM is about 97.1% as shown in Fig Parameters Table Main RFQ design parameters Value Frequency (MHz) 324 Injection energy (kev) 50 Output energy (MeV) 3.0 Pulsed beam current (ma) 40 Beam duty factor 1.05% Inter-vane voltage V (kv) 80 Average bore radius r 0 (mm)

39 Vane-tip curvature ρ t ( mm) Maximum surface field (MV/m) (1.78Kilp.) Input norm. rms emittance (π.mm.mrad) 0.2 Vane length (m) Fig Parameter cell variation. Fig Beam dynamics in RFQ for a current of 40mA 5.4 MEBT MEBT has two objectives: (1) to match the beam both in the transversal direction and the longitudinal direction from RFQ into DTL; (2) to further chop the beam into the required time structure asked by RCS. The design of MEBT is carried out with the chopper taken into account although the chopper will not be used for CSNS Phase-I Optical design Figure shows the schematic layout of the CSNS MEBT, which is structurally similar to J-PARC MEBT due to the same exit energy of RFQ and similar DTL structure with J-PARC. Fig The schematic drawing of CSNS MEBT 37

40 The length of MEBT is 3030 mm. It mainly consists of 8 quadupoles, 2 bunchers and 1 chopper. In addition, two vacuum gate valves (GV) and a number of beam diagnostic components are also installed between or in the quadrupoles. 8 sets of steering magnets are incorporated into quadrupoles through wiring on the quadrupole yokes. The first 4 quadrupoles are arranged to serve for the chopper. The last four quadrupoles and two bunchers are adopted to match the beam from RFQ into DTL. Table lists the design value of 8 quadrupoles and two bunchers both for the currents of 20mA and 40 ma. Figure shows the 6 times the rms (6*σ) beam envelope and phase spread in MEBT obtained by code TRACE-3D. Table Parameter value for quadrupoles and bunchers both in the cases of currents of 20mA and 40mA Elements Gradient (T/m)/gap-voltage (MV) Length (mm) 20mA 40mA Q Q Buncher Q Q Q Buncher Q Q Q Fig *σ beam envelope and Fig The rms emittance growth versus phase spread in MEBT the element obtained by code PARMELA Multi-particle simulations Multi-particle simulations are carried out by code PARMELA. As shown in Fig , the rms emittance growth is about 14%, 4.5% and 1.1% in x, y and z 38

41 direction, respectively The least chopping angle The least deflecting angle calculated at the exit of chopper is mrad based on the beam envelope obtained by TRACE-3D, and the least deflecting angle at the exit of chopper based on the beam envelope obtained by PARMELA is about mrad. The trajectory of the chopped beam center in x direction is shown in Fig Fig The trajectory of the chopped beam center in x direction The buncher Nose-cone CCL type structure and the same geometry is chosen for the two the bunchers. The main parameters of the buncher are listed in Table In Fig , the mechanical drawing of the buncher is given. The buncher is a copper-plated stainless cavity. Table Main parameters of the buncher Beam Kinetic Energy (MeV) RF Frequency (MHz) Beam aperture diameter (mm) 32 Longitudinal length (mm) 162 Inner cavity diameter (mm) 569 Nose-cones separation (mm) 15 Q value (computed) Transit time factor Shunt impedance (MΩ) (linac convention) 2.28 R/Q (Ω) Nominal voltage (kv) 156 Peak dissipated power (kw) Duty cycle 1.30% Peak electric field on nose cones (MV/m) Ratio peak field to Kilpatrick limit 1.47 Fig Fig.12 The The mechanic drawing drawing of of the the buncher 39

42 5.5 R&D of CSNS front end The R&D of the front end includes two parts: the construction of the H- ion source test stand and the experiment of the electrostatic pre-chopper in LEBT R&D of H- ion source Firstly, by virtue of the collaboration with Rutherford Appleton Laboratory (RAL) of England, the major mechanical drawings of the Penning Surface Plasma H- ion source used for ISIS project are introduced. Based on these drawings, several sets of the discharging chamber and the extractor of ion source were designed and machined domestically. Test experiments of the chambers and extractors on the ion source test stand at RAL showed that the chamber performance including the extracting current, the beam duty factor and the longevity etc. are almost same as or better than that of ISIS ion source discharging chamber. It is a critical step towards to the final successful ion source construction. In Fig , the experimental results of the discharging chamber at RAL ion source test stand are shown. Fig Experimental results of the discharging chamber and the extractor at RAL. Left:repetition rate 50Hz, pulse width 200µs, output current 70mA;Right: repetition rate 50Hz, pulse width 200µs, output current 55mA Secondly, as a further R&D of ion source, an H- ion source test stand needs to be constructed in IHEP to check the performance of the ion source. The construction of the H- ion source test stand has been finished, and all subsystem of the ion source has been tested independently. In Fig , the main body of H- ion source is shown. The commissioning of the ion source is now also in process. At present, a stable AC arc current of 40A is got with a about 100V arc voltage as shown in Fig However, we also encounter some trouble on the Freon-chilling system, and the Cesium supplying system. The trouble of the Freon-chilling system is the seal of Freon at the high-voltage insulation section. The problem about the Cesium supplying system will be resolved by replacing the present transport of the Cesium. 40

43 Fig Main body of the H- ion source Fig A Ac arc current, repetition rate 25Hz, Pulse width 1ms Eexperimental plan of the electrostatic pre-chopper In order to examine the reliability (to check the deflecting voltage value, to check the rising time and falling time, to check the confinement area of destruction on the space charge neutralization, etc.) of the pre-chopping design of CSNS, a similar pre-chopping design in the Accelerator Driven Sub-critical Reactor System (ADS) RFQ LEBT is also done. The experiments on the pre-chopper are being also prepared. The design is based on the existent ADS RFQ LEBT layout and the structure of the third vacuum chamber located at the entrance of RFQ. At present, except the power supply system of the pre-chopper, all components are machined and installed. The deflector, the collimator and the electron-trapping electrode are all housed in the existent third vacuum chamber. As shown in Fig , the horizontal length of the deflector is 50mm, the gap between the deflecting plates varies from 20.16mm to 33.84mm, which is 1.2 times the beam envelope size, and the width of the deflecting plate varies from 25.2mm to 42.3mm, which is 1.5 times the beam envelope size. Fig The layout of deflector and relative components in the third vacuum chamber 3-dimensional simulations by MAFIA code are carried out to determine the loaded capacitance of the deflector and the least needed voltage for the 41

44 electron-trapping electrode, as shown in Fig.16 and Fig.17. The loaded capacitance is about C=7.14pF, and the applied voltage on the electron-trapping electrode is -1.25kV. Certainly, a negative electrostatic potential is beneficial to the trapping of the ions and hence the charge neutralization. So, in practice, an absolute voltage higher than 1.25kV will be applied. Fig The cutting view of the simulated 3-dimension MAFIA volume Fig The axial electrostatic potential at the location of collimatorversus the applied voltage on the electron-trapping electrode 42

45 6 The R&D Status of 324Mhz DTL for CSNS 6.1 INTRODUCTION Xuejun In the China Spallation Neutron Source (CSNS) project, a conventional Alvarez DTL accelerating structure will be used to accelerate H - beam from 3MeV to 80.7MeV. The R&D for the 324 MHz DTL linac has been carried out at IHEP in Beijing. A 2.8m module containing 28 drift tubes of the first tank is under fabrication. 6.2 DESIGN OF THE DTL LINAC The CSNS DTL consists of four tanks. The peak current density of the accelerator for the 1 st phase is 15mA and 30mA for the 2 nd phase, and thus 30mA current is assumed in physics design of the 1 st phase DTL. The main parameters of the CSNS DTL are listed in Table Table Design parameters of DTL Tank Number Total Output Energy (MeV) Number of cell Cavity RF power (MW) Total RF power (MW) Accelerating field (MV/m) Synchronous phase(deg.) Tank length (m) We had optimised the tank diameter, the aperture and face geometry of the drift tube using SUPERFISH code to iteratively calculate the shunt impedance and surface field. The average electric fields are from 2.2 to 3.1 MV/m in the first tank and keep 3.1 MV/m in the rest tanks. The peak surface electric field is controlled less than 1.3 times Kilpatrick field Tank As above mentioned, the length of the tanks is too long to process at a time. It should be better to divide into several short technology modules. In our case, each DTL tank is composed of three modules. The tank is made of carbon steel tube with inner diameter 56cm. The internal surface of the tank is coated with OFC of 150µm thick. The mechanical drawing of the first module of the first tank is presented in Fig

46 Figure The mechanical drawing of the first module of the first tank Drift Tubes In CSNS DTL, the focusing periods are designed as the FD lattice and each drift tube (DT) contains a electromagnetic quadrupole (EMQ). For the strong focusing scheme, the EMQs are used with high gradient from 75T/m to 38T/m. In principle, the outer diameter of the DT is decided by the compromise between the size of the EMQ and the RF property requirement. And the diameter of the DT should be designed large enough to house the core, coil and water cooling channel for EMQ and DTs. The DT shell and stem will be made of Oxygen Free Copper (OFC), instead of the commonly used stainless steal to avoid copper plating. Long-term deformation test has been done which convinced us the material selection and design. The total number of drift-tube is 156 and each drift tube has different lengths (53 mm to 165 mm). The shape of the drift tubes is a conventional cylindrical shape and mirror finished to prevent electric discharge. The face angles α of the drift tubes are from 0ºto 60º(as shown in Fig ). The larger face angle both increases the shunt impedance and the surface field. The bore radius increases from 0.6cm in the first tank to 1.3cm in other tanks. These parameters were summarized in Table Fig shows two kinds of the drift tubes. Table Design parameters of Drift Tube Tank number Face angle (degree) Inner radius (cm) Outer radius R o (cm) Corner radius R c (cm) Diameter of drift tube (cm) Flat length (cm)

47 Figure The design model of drift tube For the FD focusing period, every drift tube containing an EMQ, and thus there are no space in tanks to accommodate diagnostic instruments. It brings difficult for beam commissioning and orbit correction. So a high accuracy in fabrication and alignment is strictly demanded in both EMQ in DT and DT in tank. Laser tracker will be used for the assembly of the drift tubes in a tank Quadrupole Magnet In the drift tube, the EMQ is used to supply transverse focusing to the beam. The major parameters of the designed quadrupole for the injection end (3MeV) are given in Table The R&D of the quadrupole for the lower energy section of the DTL is a critical issue for the DTL structure because the size of the drift tube for this section is so small that it is not possible to apply the standard techniques for installation the electromagnetic quadrupole. Figure shows the field design of the electromagnetic quadrupole calculated by using the POISSON code. Figure The field of the EMQ designed by POISSON code The diameter of the magnet is 115 mm and the tube width is 49.89mm in minimum. Hollow-conductor is normally used as a coil. However, in case of such small and high current coil, hollow-conductor is not desirable because a thick 45

48 hollow-conductor enlarges the bending radius around the small magnetic pole and reduces the area of flux. Therefore, the J-PARC type SAKAE coil which has no bending radius has been adopted. The wire cutting and the Periodic Reverse (PR) copper electroforming method are applied to the coil manufacture process. Table Parameters of electromagnetic quadrupole Magnet aperture diameter (mm) 15 Yoke out diameter (mm) 118 Core length (mm) 35 Magnetic field gradient (T/m) 75 Good field region (mm) ±6.5 Effective length (mm) 41.3 Core material Silicon steel 50WW470 Thickness of steel leaf (mm) 0.5 Number of turns per pole 3.5 Integrated field GL (T) 3.1 Water flow rate (l/min) 1 Max. excitation current (A) 528 Conductor area (mm 2 ) Current density (A/mm 2 ) 32 Resistance of coil (mω) 4.25 Inductance of coil (µh) DT aperture diameter (mm) 12 DT outer diameter (mm) THE R&D STATUS OF DTL The R&D of DTL is focused on the cavity manufacture and the drift tube fabrication. Even though IHEP built a 35 MeV DTL about 20 years ago with a RF frequency of, we are still facing some challenges and crucial issues due to the higher RF frequency and high average current in CSNS. A prototype of the first module of the the first DTL tank is under manufacture. The EMQs with SAKAE coil are developed. The PR process is applied to inner surface of the DTL tank. The R&D section of DTL mainly includes a tank, 28DTs, 14 post couplers and 4 slug tuners. Mechanical parameters for this module are listed in Table Table The parameters of first DTL tank Tank length (m) 2.85 Energy range (MeV) Average E 0 (MV/m) 2.2 ~ 3.1 Synchronous phase (deg) -30 ~ -25 Tube face angle (deg) 0 ~ 9 ~ 14 Tank inner diameter(mm) Cells number 29 46

49 No. of Slug tuners 4 Bore radius (mm) 6 DT diameter (mm) Tank manufacture The tank was made of carbon steel with a copper inner surface. It has rather complex structure to be electroformed. It contains 7 large ports for tuners, vacuum, and approximate 60 small ports for drift-tubes, post couplers and pickups. There are twelve straight water cooling channel embedded into tank out-wall. To develop the manufacture technology, we tried various approaches with some short test tanks. An explosive bonding technology was tested. A thin copper plate was tightly bonded with a steel tank in the inner surface. Since the tank had many ports and holes, the bonding condition at the ports and holes was not as good as the inner surface, resulting in vacuum leakage at some ports. Then we tried to apply the Periodic Reverse (PR) copper electroforming technology. It has been successful for both inner surface and all ports/holes. Figure shows the 2.8m tank module after electroforming. The inner copper surface has been polished (as shown in Fig. 4) and the ports have been fine machined for high accuracy and high flatness. Inner surface flatness are between 0.26~0.29um (design 0.6um). The average copper thickness is 0.2mm (design 0.15mm). The outgassing characteristic of the electroformed lining was measured for the samples at room temperature. The conductance modulation method was applied. The measured outgassing rate was about Pa m 3 /s m 2 for 100 hours pumping. For a test tank (process chamber), the total outgassing is Pa m 3 /s. The vacuum test of the first unit tank was also carried out in last May. The vacuum reached Pa only using a turbo molecular pump(0.6 m 3 /s -1 ). Figure Prototype of the DTL tank 47

50 The simulation of thermal deformations caused by heat load dissipated on the tank surface was studied. The temperature of cooling water is 20.The maximum temperature is A little temperature increase is at the side of the tank as shown in Fig Thermal deformation is 4.75µm of diameter. The frequency change is -51.9kHz. Figure Temperature distribution on the DTL tank surface Drift Tubes Fabrication There are 28 drift tubes in the R&D and the specification of them are described in the Table Table DTL Tank Specification of Drift Tube in the R&D DTL-1-1 Face angle α (deg) Magnetic Field Gradient : G (T/m) 75 Integrated Magnetic Field: GL (T) 3 Core length: L [mm] 34( Effective 41.3) Outside diameter of Q-mag d (mm) 148 Bore diameter (mm) Diameter of beam pipe (mm) R c (mm) R o (mm) R i (mm) Number of magnets Beam dynamics requires that the deviation of the quadrupole center from the mechanical center should be within ±50µm. So we decided on ±30µm as a tolerable error of the magnetic field center from the DT center. The drift tube of DTL for CSNS is made of OFC. The merits are no 48

51 electroplating is needed after electron beam welding (EBW) of the two halves of a drift tube, and the surface machining demands only one time to make the mechanical center coincident with the magnetic field center, saving a lot of time in mass production. Figure shows the EMQ and the prototype of the drift tube. However, we realized that OFC is softer than stainless steel and thus we did deformation test before the decision. A mockup drift tube was made and hanged with a heavy load of 77kg for 6 days. The size detection indicated the deformation is less than 2 µm, within the tolerance. The vacuum leakage of the prototype of drift tube was Pa m 3 /s that was much less than the design. Figure The prototype of EMQ and the drift tube The temperature rise of the cooling water was measured. In the coil for water-flow rate of 1 liter/minute and the DC-current of 495 A, the average water temperature rise was 8.6 C. The analysis of thermal deformations caused by heat load dissipated on the drift tube surface was also carried out (Fig ). The maximum temperature was close to the beam pipe, The maximum deformation was only 1.36um. 49

52 Figure Temperature distribution on the tube surface Quadrupole measurement The effective length, magnet field gradient in transverse plane, B-H curve and the excitation property of two magnets were measured by Hall probe. Figure shows the comparison of magnetic field gradient as function of excitation-current for measured and analyzed data. It indicates that the designed gradient of 75T/m can be achieved with an excitation current of 495A, which is lower than that from 3D simulation. G[kG/cm] Field gradient vs.current 7.5kG/cm at ~495A 2 Simulation results 1 Measurement results Current[A] Figure Comparison of magnetic field gradient as function of excitation-current for measured and analyzed data. The magnet installed into drift tube has a concentric tolerance of less than 30µm error. The mechanical centre is fixed by a set of supporting structure on the rotating coil measurement system. Then beam pipe and outer diameter of the drift tube are machined in accordance with the magnetic center which is defined as the position with a minimum dipole component. A small rotating coil measurement system was developed to detect if the deviation of the magnetic field center from the drift tube center is within the required accuracy 30um (Fig ). If not, the drift tube with tolerance redundancy was machine to make it geometrically concentric with the magnetic center. Furthermore, the higher order multipole components in the magnetic field center measured also by the rotating coil were sufficiently small, being less than 0.3% in comparison with the quadrupole component as shown in Fig Table shows the measurement results of an electromagnetic quadrupole. It can be found that the field 50

53 center was deviated only by about 12 um from the mechanical center. The radius of the rotating coil limits the measured area of the good field, so the result is different from the design one. B n /B E-3 1E Multipole Components Number Figure The rotating coil system and the measured higher-order components of EMQ Table Results of the magnetic field design and measurements of EMQ Design Measured Excitation current (A) Magnetic field intensity (T) Field center deviation (µm) ±30 x=-1.77, y= Magnetic field gradient (T/m) Good fields (mm) Φ13 Φ8.8 Good field (mm) Effective field (mm) Core length (mm) Tube outer diameter (mm) 148± Magnet aperture diameter (mm)

54 Beam pipe diameter (mm) 12± Water flow rate (l/min) Water temperature rise ( C) 21.86(@DC600A) 8.6(with tube water 14.86(@DC495A) cooling) 6.4 OUTLOOK The DTL R&D has been carried out in IHEP. We have adopted some new designs and novel process in the R&D. The properties of electromagnetic quadrupole have been measured and studied. A lot of work should be done in the future. The RF tuning issue and the structure stability will be studied in the next step. 52

55 7 CSNS Linac RF-System Status Jian 7.1 OVERVIEW The layout of CSNS linac RF system is shown in Fig The designed output proton energy of linac is 81 MeV. Five klystron power sources are used to power the RFQ and the four DTL tanks. Operating at 324MHz RF frequency and 2.5 MW peak power, the Toshiba E3740A klystron is one of the candidates. Additionally, three solid state RF amplifiers are used to drive two MEBT bunchers and a LRBT debuncher. One-RF-unit-per-cavity independent RF control design is adopted. All the power sources operate at RF frequency 324MHz, repetition rate 25pps, maximal pulse width 650us, and duty cycle 1.625%. The CSNS linac RF frequency is 324 MHz, the same as that of J-PARC, so that the same type of klystron can be used as CSNS high power pulsed RF source for RFQ and DTL. This frequency can give a reasonable room for the electro-magnetic quadrupole inside the drift tube. Basic requirements for CSNS klystron RF power source is given in Table The available klystron like J-PARC type is made by TOSHIBA company, type No. E3740A (Fig , a). Its average output power is 93 kw. At 2.5 MW output power and 25Hz repetition rate, its RF pulse width can be up to 1.5 ms (as shown in Fig , b). Furthermore, another probable candidate is a 324MHz klystron by CPI company (3MW RF pulse power, duty factor up to 3.33%). It is under development. Both klystrons are triode mode, i.e., they are equipped with a modulation anode. Klystron cathode high voltage is kept constant; through applying pulse voltage to modulation anode, pulse output power of klystron can be obtained. 53

56 Compared with diode mode klystron, pulse power supply for the triode mode is relative simple and cheap. Its block diagram can be seen in Fig It consists of three necessary parts: -120kV power supply and energy storage capacitor bank, crowbar and modulator. 2.0 Pulse Width tp [ms] pps 65 pps 50 pps 0.2 Solid line & Point: Measurement value Dotted line : Theoretical value Peak Output Power po [Mw] (a) (b) Figure TOSHIBA E3740A klystron Table Basic requirements for CSNS klystron RF power source Frequency Klystron peak output power at saturation Repetition rate Beam voltage pulse width RF output power pulse width Efficiency Gain 324 MHz 2.5 MW ( < 1 MW for RFQ) 25 Hz 700us 650us (Max.) 55% (Min.) 50 db (Min.) Figure Block diagram of klystron and pulse power supply Now we have already made progress with some key technologies for the linac RF system. The digital low level RF control was already developed and successfully applied in beam commissioning of the ADS(Accelerator Driven Sub-critical) 3.5MeV RFQ accelerator at peak beam 44.5mA, beam duty 7.15%. A proposed new type of 54

57 power supply (100Hz ac series resonance high voltage power supply) passed acceptance test and a satisfactory test results was obtained. R&D of crowbar and modulator are ongoing and we will get final test data soon. 7.2 DIGITAL LOW LEVEL RF CONTROL In the first half year of 2006, the ADS RFQ accelerator was under beam commissioning. We just take advantage of this opportunity to carry out R&D of a digital low level RF (LLRF) prototype for CSNS. Although the operating frequency of ADS RFQ RF power source is MHz, the LLRF key technology is the same as that of CSNS, so that it can be easily transplanted to CSNS digital LLRF in the future. In 2007, LLRF prototype passed acceptance test Hardware and Algorithms A block diagram of LLRF and its actual picture are shown in Fig and Fig An accelerating electric field stability of±1% in amplitude and ±1 degree in phase is required for the RF system. This field control is implemented by a combination of feedback (FB) and feed-forward (FF) algorithms. The two 14bit-ADCs (AD6645) and two 14bit-DACs (AD9764) are installed on FPGA board. The digital FB and FF is carried out by one chip of FPGA (STRATIX II series by ALTERA Co.). The DSP (C6000 series by TI Co.) is in charge of communications and translating values between I/Q and Amplitude/Phase. We adopt AD8345 chip as IQ modulator. The digital algorithms in FPGA can be seen in Fig The maximum RF pulse width is 1.4 ms. Because table length is 1024, the set_point_table_i/q for FB and the FF tables get into the act every µs during the RF pulse. Fig Block diagram of LLRF 55

58 Fig Actual picture of LLRF Set_point_table_I Cavity_read_table_I I FIR - + Kp P I + 开环 / 闭环 FF_table_I + FFbeam_table_I 射频开关 0 饱和限制 I Cavity IF a1 b1 A/D c1 d1 Sample-clk Q FIR - + KI Kp P I + FF_table_Q + a2 b2 c2 d2 射频开关 0 饱和限制 Q Cavity_read_table_Q KI FFbeam_table_Q 开环 / 闭环 Set_point_table_Q Forward IF A/D a1 b1 c1 d1 I Q FIR FIR Forward_read_table_I Forward_read_table_Q Sample-clk Reflected IF A/D a1 b1 c1 d1 I Q FIR FIR Reflected_read_table_I Reflected_read_table_Q Sample-clk Fig Algorithms in FPGA In order to protect the high-power components such as klystron, circulator and RF windows in case of RF arc and over RF reflection, the RF drive and beam should be inhibited within 1µs of the detection and should remain inhibited for the remainder of the pulse. They should be re-enabled for the following pulse. But if detections take place predetermined times in a second, The RF drive and beam should be permanently blocked Performance Prior to beam commissioning, the comparison tests between without and with FB closed loop control were carried out. As shown in the left two waveforms of Fig , in the case of no FB control, due to klystron beam voltage sag, the amplitude and phase decreased by.4% and 19 degrees respectively during the 1.4 ms pulse. When the FB control loop was closed, the flatness of amplitude and phase were changed into the right two waveforms of Fig The fluctuations at flattop were very small. The corresponding errors of the amplitude and phase were ±0.4% and ±0.5, respectively. 56

59 Fig Performance of LLRF feedback control with no beam During beam commissioning, the same comparison tests without and with FB closed loop control were also carried out. The measured waveforms of no FB control are shown in Fig In Fig (a), the upper trace is cavity field RF signal, and the lower one is a RF power reflected from RFQ. It is obvious that the beam loading is very heavy. Fig (b) indicates that the amplitude and phase droop depth are very large (9% and 21 degrees respectively). The two waveforms in Fig (c) were detected from two beam current transformers. The upper trace is for the entrance beam, and the lower one is for the exit beam. It is of bad quality and stability. Transmission efficiency is not so high. Once FB control loop was closed, the cavity field RF became flat (Fig , a), and Fig (b) shows that amplitude and phase were controlled within ±1% and ±1 degree. It led to good beam quality and stability (Fig , c). At that time, the RFQ entrance and exit beam were 48.6mA and 44.5mA respectively. Transmission efficiency was 91.6%. (a) (b) (c) Figure RFQ beam commission without FB closed loop control 57

60 (a) (b) (c) Figure RFQ beam commission with FB closed loop control Also, the function of beam FF control was already examined. It can effectively improve the transient response of the front and trailing edge of the beam loading. In 48 hours continuous operation test, no problem occurred and the stability is very good. 7.3 AC SERIES RESONANCE HIGH VOLTAGE POWER SUPPLY FOR KLYSTRON The proposed -120 kv /50A high voltage power supply for klystron is shown in Fig The design idea of this new type of pulsed power supply mainly utilizes AC series-resonant theory and synchronized discharging property. The power supply mainly consists of six basic units: IGBT frequency converter power supply, exciting transformer B, AC resonant inductance L (1.6H), AC resonant capacitor C (1.585uF), high voltage diodes, DC energy storage capacitor bank Cz. This scheme avoids step-up high voltage transformers and multiphase high voltage rectifiers. The circuit structure is very simple, so it can achieve a good operating reliability and low trip rate, also can be convenient for operation and maintenance. Among these parts, the resonant inductance L and the resonant capacitor C compose a series resonant circuit, 1 f0 = = 100Hz and its natural resonance frequency is 2 π LC. klystron Figure Proposed AC series resonance high voltage power supply for klystron 58

61 By an IGBT frequency converter, the 50 Hz three-phase 380 V power from the mains is converted to f0 (100 Hz) single-phase power. The exciting transformer step up the 100 Hz low voltage input power to about 2.5 kv. The natural resonance frequency of AC series resonance circuit (L and C) is just 100 Hz. So, 120kV peak voltage of sinusoid wave on the resonance capacitor C can be obtained. When the AC high voltage wave becomes negative, behind high voltage diodes, the DC energy storage capacitor bank CZ can be charged. Meantime, discharging rate of modulator plus klystron is 25Hz, which must be strictly synchronized with AC resonance charging pace. That means, every four times the DC energy storage capacitor bank is charged, once discharging pulse occurs. Both of them must be exactly synchronized. A prototype power supply has been built with the design parameters: L = 1.6H, C = 1.585uF, inductance quality factor Q0 >= 250 minimum required. By code simulation in this case, AC to DC efficiency is 87%. In 2008, the prototype passed acceptance test and the results are quite satisfactory. The measured Q value of inductance is no less than 350. At test condition of klystron cathode voltage 66 kv and output RF power 420kW, AC to DC efficiency of the prototype is up to 88%. Fig to Fig are IGBT frequency converter power supply, AC resonant inductance and AC resonant capacitor respectively. In Fig.7.3-5, the sinusoid wave is AC resonance charging voltage on AC resonant capacitor, and the pulse trace is discharging wave of modulator and klystron. Fig is klystron output RF power. Figure IGBT frequency converter power supply 59

62 Figure AC resonant inductance Figure AC resonant capacitor Figure AC resonance charging voltage vs. klystron modulation anode voltage (discharging) 60

63 Figure Klystron output RF power Next step, we are going to modified resonance frequency to be 400Hz, so as to reduce noise and volume of the components. The parameters of AC resonance inductance and capacitor are selected as 0.799H and 0.198uF. Especially as for inductance, oil-immersed air-core inductance with magnetic shielding screen is optimal selection (Fig ). The newly-designed scheme is ongoing. Figure Oil-immersed air-core inductance with magnetic shielding screen 7.4 CROWBAR AND MODULATOR As shown in Fig. 3, high voltage DC power supply feeds the constant high voltage (Vk) to the klystron cathode. M-anode modulator generates the anode pulse voltage (Va) to control the electric power for the klystron. The crowbar is a necessary high voltage fast protection device. In the case of an arc occurring inside the klystron, the crowbar can remove the high voltage energy from the klystron and DC capacitor bank within a few microseconds in order to avoid the damage of the klystron M-Anode Modulator Originally, we had designed a scheme of the hard tube floating deck modulator, in which the HV tetrode TH5188, made by THALES company, was employed. We thought that the modulator of this type could flexibly adjust klystron m-anode voltage so as to get a good beam perveance. But afterwards, we heard from J-PARC that the 61

64 modulator often caused tetrode breakdown at 110kV in the modulator oil tank. It resulted from corona discharge that occurred at TH5188 socket. Moreover, at that time, THALES stopped producing TH5188 tetrode. Therefore, we started to R&D another type of J-PARC modulator a 120kV high voltage semiconductor switch is substituted for the tetrode. The schematic diagram is shown in Fig The m-anode pulse voltage is generated by switching the cathode voltage through dividing resistors in the m-anode modulator. The anode voltage can be adjusted by the dividing ratio, and the pulse duration is controlled by switching device connected with the resistors in series. The semiconductor switching device is purchased from PEEC company in Japan, the same as that of J-PARC modulator as can be seen in Fig It mainly consists of 150 FETs (Field Effect Transistor) in series configuration, so that it can withstand -120kV movement voltage. At present, fabrication and assembly have already finished (Fig ). Before filling tank with oil, -40kV test in air was carried out. At 700us pulse width as shown in Fig , its rise and fall time are 9us and 22us respectively. High voltage test will be expected in January X3 1V/1A 油温 油位 / 或气压开关 X1 3 相 380V 2KW 调压器 T1 调制信号入 电源异常 脉冲检测 Ig 保护 Ug 保护 110% 100% 90% T BNC-C-CK3 X7 X5 X8 BNC-C-CK4 X4 AC220V X6 CX2-23J8M/K8Q X2 4 5 CX2-23J8M/K8Q R10 1/5w 4 5 R1 84/6/200W M AC220V T2 AC220V T3 R7 10K/25W R9 5/2W Ig Ug R5 20K/500W C2 1U AC220V R2 (18.6K/2)*10/200W Figure Schematic diagram of CSNS m-anode modulator 62

65 Figure The semiconductor switching device made by PEEC company Figure CSNS m-anode modulator internal construction (a) 700us pulse width (b) 9us rise time (c) 22us fall time Figure Modulator test result at 40kV in air Crowbar In the original CSNS crowbar design scheme, thyratron CX2098B, made by EEV company in England, was preferred as crowbar tube, because the peak anode voltage of a single tube can be up to 180kV. It is helpful to simplify trigger curcit, which leads to avoid miss-fire malfunction and achieve good reliability. But in the 63

66 second half year of 2008, EEV declared to stop producing CX2098B thyratron. So we have to do R&D on ignitron type of crowbar. In CSNS crowbar prototype, 7703EHVNP ignitron, made by Richardson in USA, serves as crowbar tube. Its peak anode voltage is 50kV and peak anode current is 100kA. Therefore in CSNS crowbar circuit as can be seen in Fig , the 4-series ignitron configuration is adopted. Each ignitron is controlled by one ignitor trigger module. The control signals of the four modules are fed from a 4-output driver module via glass optical fibers. The internal construction of crowbar is shown in Fig We set up a test stand (as shown in Fig ) to measure its trigger-to-discharging speed. Without being immersed in oil, at high voltage -40kV in air, the delay time from switch to be closed to crowbar discharging is less than 6us of design specification. The measured time is 4us as shown in Fig Full voltage acceptance test of overall crowbar system will be finished in January V 稳压电源 BNC-C-CK5 X 度温度开关油位下限开关 CX2-23J8M/K8Q X 撬棒放电电流监测 X4 PEARSON 101 R1 R2 20/100W AC220V IG5F M/50W C1 4000P/60KV R1 R2 20/100W AC220V IG5F M/50W C1 4000P/60KV AC220V R1 R2 20/100W AC220V IG5F M/50W C1 4000P/60KV R1 R2 20/100W AC220V IG5F M/50W C1 4000P/60KV -120KV 输出 -120KV 输入 X6 PEARSON 110 光缆光缆光缆光缆 X5 TTL 宽度大于 100nS BNC-C-CK6 检测触发电路 IG5F-T4 脉冲电流过流, 时间过宽保护 外触发 ( 光缆 ) X2 X3 输出监测 ( 光缆 ) Figure Schematic diagram of CSNS crowbar Figure CSNS crowbar internal construction 64

67 Figure Crowbar test stand 4us CT2 CT1 Figure Crowbar test result of 40kV in air 65

68 8 The Preliminary Design and Prototype Test of CSNS Magnets Wen CSNS accelerator consists of an 80 MeV linac, a rapid cycling synchrotron (RCS) of 1.6 GeV, which mainly has 24 AC dipole magnets and 48 AC quadrupole magnets, beam transport line from Linac to RCS (LRBT), which has four DC bending magnets and 47 DC quadrupole magnets, and beam transport line from RCS to target (RTBT), which has 6 DC bending magnets and 36 DC quadrupole magnets. The RCS has a fourfold symmetric lattice. Each super-period consists of 6 bending magnets and 12 quradrupole magnets. The parameters of the main magnets are listed in Table 8-1. Since the RCS is used to produce pulsed spallation neutrons, it should provide an extremely intense beam to the production target with very small loss. The multi-turn injection has to be used to get the required beam current. To accommodate such a high intense beam, the acceptance of the synchrotron and the aspect ratio of the magnets (inner diameter over magnet length) should be very large in comparison with a storage ring, especially for quadrupole and sextupole magnets. Table 8-1 Parameters of the RCS magnets (check red number with power supply) Parameter RCS160B RCS206Q RCS265Q RCS222Q RCS253Q RCS230S Number of magnets Field strength B (T) B (T/m) B (T/m) B (T/m) B (T/m) B (T/m 2 ) Max. field Min. field Gap or radius 160 mm 103mm 132.5mm 111mm 126.5mm 115mm Effective field length 2.1m 0.41m 0.9m 0.45m 0.62m 0.2 Turns per coil DC current A 813.3A 866.7A 715.6A 828.7A 168.1A AC current 872A 579.7A 617.7A 510A 590.6A Field quality BL/BL BL/BL BL/BL BL/BL BL/BL BL/BL Good field region ±106mm ±90 mm ±120mm ±98mm ±114mm ±109mm Resistance 21.5mΩ 11.3 mω 21.8 mω 12.1mΩ 17.3 mω 78.8 mω Inductance 38.8mH 8.69mH 24.9mH 9.33mH 17.3mH 34.9mH Design and prototype test of the RCS magnets The main magnets of the RCS are dipole magnets and quadrupole magnets, which are operated synchronously with each other. These magnets are characterized as follows: 1) The magnets are excited with a repetition frequency of 25 Hz. 2) The magnets have an aperture larger than usual ones. 3) A saturation of the magnets has to be small in order to assure a close tracking 66

69 between dipole and quadrupole magnets. The high repetition rate of the magnetic field would lead to following issues. 1) A large eddy current induced in iron core, end plates and coil conductors of the magnets. 2) Eddy current also induced in the metal components nearby the magnets. 3) Vibration of the magnet coil that could damage the insulation layers if it is not treated properly. 4) High voltage induced in the coils that requires much stronger insulation level of coil than the usual. There are two effects caused by eddy current. One is eddy current loss and overheat in the magnet components, the other is disturbance to the guiding field due to the eddy current field. To reduce iron loss in the cores of the magnets, the iron cores are stacked and adhered with non-oriented silicon steel sheets of J23G-50, which has properties of high saturation flux density and low iron loss as shown in Table 8-2. The thickness of the sheet is 0.5mm. The maximum magnetic field in the iron core is designed less than 1.6T. Table 8-2 The properties of the J23-50 silicon steel sheet Parameter Value Magnetic Inductance B T Iron Loss W 1.72/50 4.5W/kg Coercive Force H c Os To reduce the eddy current in the coils, the aluminum conductors stranded on the stainless steel duct with cooling water have been developed and used to wind the coils for dipole magnets, while hollow copper conductors with small cross section area have been used for the coil winding of quadrupole magnets by means of special winding method. And to reduce the eddy current loss in the end parts and end plates of the magnets, the end shape of the magnet cores have been carefully designed, so that the magnetic field component perpendicular to the end plates is not too large. Therefore, the Rogowski shape and some slits cutting the path of eddy current are made in the end parts and end plates of the magnet cores. In order to investigate such effects, one prototype dipole magnet and one prototype quadrupole magnet have been designed and fabricated. The dipole magnet has a curved H type core with two parallel end plates. Its drawings are shown in Figure 8-1, and pictures of the prototype magnet and one coil winded with stranded aluminum conductor are shown in Figure 8-2. The treatment of the end part areas is shown in Figure 8-3. To improve the uniformity of the integrated field from 0.25% to 0.05% and to decrease the saturation of field, the end areas of the magnet have been carefully chamfered. To decrease the temperature rise of the end plates from more than 400 o C to less than 100 o C, 17 deep slits have been cut. The measured temperature rise is about 100 o C after 9 hours operation with the full AC current. The measurements of the magnetic field have been done at DC and AC current condition respectively. The results are shown in Figure 8-4 and Figure

70 Figure 8-1 Cross section and drawing of the dipole magnet Figure 8-2 The prototype dipole magnet and the coil winded with aluminium stranded conductor Figure 8-3 The treatment of the end plates and measured temperature rise uniformity distribution DC field uniformity at different current uniformity 2.00E E E E E E E E E-03 DC E-03 DC1145+AC816 25Hz amplitude -1.80E-03 x[mm] AC 25Hz field uniformity compared to DC one Figure 8-4 Field uniformity measurement 68

71 Under DC current excitement, the measured field uniformity is better than either in injection energy or in extraction energy of RCS. However, with AC current excitement, the field uniformity is worse than DC one, probably due to eddy current effect. Measured DC field linearity Measured AC field linearity Figure 8-5 Field linearity measurement The linearity of integrated field is about 1% under DC measurement, while that is about 3.4% under AC measurement. At present,the problems appeared in prototype dipole magnet are following, 1 There appeared some obvious broken cracks on the coil surfaces after long time AC current operation. The possible causes are vibration of the coils, non-uniform magnetic field force in the coils and too thick epoxy resin layers in some areas of the coils. 2 From the view of the power supply, the linearity of the magnet inductance must be much larger than the measured field linearity, which is estimated to be 14%. This is not understood very well now. We don t know what reasons make the linearity of inductance much larger than the calculated one, while the measured linearity of the field does agree with the calculated one. The prototype quadrupole magnet has an aperture of 308 mm. The core consists of four parts, which are laminated and adhered by J23G-50 silicon steel sheets. The coils have a saddle shape, which are winded by hollow copper conductors with cross-section dimension of 9 9mm2 and the inner diameter of 6mm. To reduce the voltage drop on the RCS quadrupoles, we use a special way to wind the coils to reduce the inductance. Each pole has four coils that are arranged in parallel rather than in series. That is to say, power supplies will see one fourth of coil inductance while magnet sees the total turns of the coils. To keep the same length of four pole s coils, the lead of each coil will be exchanged relative position before connected with the coil of next pole. The prototype quadrupole magnet and its field measurement system are shown in Figure

72 Figure 8-6 The prototype quadrupole magnet and rotating coil measurement system We adopt end chamfer to suppress the high harmonic contents of the field and cut slits on the pole region to reduce temperature rise. The magnet has a large aperture, so it is necessary to confirm the non-linearity of inductance and field by 3D field analysis. We used TOSCA program for this purpose. The simulated non-linearity of excitation function and inductance are shown in Figure 8-7. The DC measurement results can be seen in Figure 8-8 and 8-9. The non-linearity of the field is about 1.6%. The largest measured harmonic field is b6, which is However, after end chamfering, it is suppressed below the required value of For the prototype quadrupole magnet, the AC measurement will be done in the next months. Figure 8-7 The simulated non-linearity of excitation function and inductance 70

73 散裂四极磁铁削斜前励磁曲线 ( 霍尔探头 ) BL/I Current(A) Figure 8-8 The measured excitation function by hall probe B6/B2((@r0=141mm) 7.00E E E E E E E E Current Figure 8-9 Measured harmonic element of b6 before and after end chamfer The magnets in LRBT are all conventional DC magnets, and solid iron cores are adopted. The iron cores are divided into small sections that will be welded to form a whole magnet. For RTBT, the magnet sizes and field intensities are much larger which increases the design difficulty. Most magnets are also conventional DC magnets except RT-136B with fast pulse excitation. Special 8 iron structure is used for RT-220Q because its horizontal width is limited to 900 mm. Laminated cores are adopted for RT-180Q and RT-220Q to meet the high field gradient requirement and enhance the field quality. For other magnets, including the correctors, solid iron cores are used. In the preliminary design, most magnet designs meet the physical requirements with an exception that integrated field qualities of some dipole magnets need to be improved by further end chamfer. 71

74 9 Magnet Power Supply System Jing 9.1 Introduction The 1.6GeV proton synchrotron proposed in the CSNS project is a 25Hz rapid-cycling synchrotron (RCS) with injection energy of 80MeV. Beam power is aimed to 100kW at 1.6GeV. The power supply system consists of four sub-systems: one for Linac, one for LRBT (Linac to RCS transport Line), one for RTBE (RCS to Target transport Line) and one for RCS. The power supplies are DC mode output except RCS power supplies. RCS power supply system includes a dipole power supply, 5 quadrupole power supplies, 4 sixtupole power supplies, 36 trim B power supplies and 6 trim Q power supplies. The power supply s specifications are listed in Table Fig and Eq illustrate the typical magnet current waveforms. I m =I dc -I ac *cosω 0 t, (9.1-1) where ω 0 =2πf 0, f 0 =25Hz. In order to avoid drawing a large reactive power from the magnet to a.c. lines, the White Circuit type resonant network was adopted as the structure of the B&Q magnet power supply system [1]. In Fig B magnet current of the RCS the CSNS project, the total number of 24 dipole magnets and 48 quadrupole magnets are divided into 6 families according to the lattice design. Each family is powered independently and ensures the accuracy of the running through the tracking and control system. Table Specifications of the power supply set Peak Current Peak Voltage Power Ratio Tracking accuracy (A) (V) (kw) RCS-B % RCS-Q % RCS-Q % RCS-Q % RCS-Q % RCS-Q % RCS-S % 72

75 RCS-Trim BH % RCS-Trim BV % RCS-Trim Q % 9.2 Configuration of Resonant Networks: The total number of dipole magnets is 24, which excited by one power supply in serial. The total number of Quadrupole magnet is 48, which divided to 5fimalies and excited by 5 power supplies independently. The parameters of the resonant network are listed in Table 9.2-1, and Table Table Dipole resonant network parameters Magnet parameters Resonant network parameters Number of Dipole Magnet 24 Gap(mm) 160 Effective length (mm) 2100 Frequency (Hz) 25 field-strength(t)(injection ~ extraction) ~ AC Peak Current(A) 872 DC Current(A) Total power loss(kw) 44.8 Inductance(mH) 38.8 Power Supply for Dipole Magnet 1 Magnet No./Cell 2 Resonant cell 12 Magnet inductance (Lm)/cell (mh) 77.6 Capacitance /cell (µf) Choke inductance/cell (mh) 77.6 Peak Voltage/cell (V) Table Quadrupole resonant network parameters QPS Q265 Q253 Q222 Q206 Q206 Magnet Parameters Total No Radius(mm) Effective length (mm) DC Current(A) AC Current (A) Peak Voltage (V) Total Loss (KW)

76 Inductance(mH) Magnet No./cell Resonant network parameters Resonant cell No Capacitance(µF) L ch (mh) Peak voltage/cell(v) To ensure the 25Hz resonant frequency, the parameters of per cell are strictly identical, and can be adjusted to compensate the error of the resonant components. After comparing the characteristic of the parallel type of white circuit and the serial type white circuit, serial type white circuit is adopted for the BM and QM, which is easy to implement the harmonic injection to compensation the magnet field saturation, and in same time the choke of the serial white circuit is easy to manufacture. But there is a disadvantage, the power supply is difficult to design and manufacture due to only one power supply is used to feed the whole resonant network. Configuration of resonant networks is shown in Fig (AC+DC) PS Fig The series resonant network The inductance of choke (L ch ) equals the total inductance of magnet per cell, and the choke is parallel connection for dc bypass. The current uniformity in the magnet groups is guaranteed owing to the series connection of the circuit. The output current is directly applied to the magnet; it is more flexible to perform current regulation. The number of resonant cell is based on the voltage across the capacitor. The parameter errors of resonant components bring the resonant frequency drifting. Power line and fixing techniques need strictly uniformity. The temperature drifting of resonant components bring out the magnet current drifting. All of resonant power components 74

77 have water-cooling to minimize the temperature drifting. 9.3 Resonant components Capacitor The parameters for BM and QM resonant capacitor are listed in Table The full film capacitor is selected, which has low loss, long life, and low temperature effect, and used in the prototype of the dipole resonant network. Every capacitor bank has the regularly capacitor used for compensating the error of resonant frequency, which is the result of the errors of the magnet inductance, choke inductance, the circuit distributing parameters, and so on. The variation of capacitance as a Fig Capacitor bank function of temperature is about -0.04%, and the tangent of the loss angle is less than 0.02%. Fig shows the capacitor bank Choke On the premise that the repetition frequency and the magnet inductance are fixed, the inductance of the choke should be determined from the economical viewpoint [1]. According to experience, the ratio of the choke inductance to the magnet inductance is determined to 1. Considering batch production and cost reduction, we adopt a multi-gap, iron-clad, oil-cooled choke. Fig shows the prototype of choke. Non-linearity of the inductance is specified 1%, and it is better than 1.4% in the choke prototype. It seems good enough for the resonant networks. The total weight of the choke is 32T. Fig shows the result of the inductance non-linearity test. Fig Choke 75

78 inductance (mh) current (A) Fig The choke inductance linearity measurement 9.4 Power Supply Design IGBT (Insulated Gate Bipolar Transistor) power supplies will be used for both BM and QM networks. Lower harmonic components and fast response can be obtained using IGBT power supplies. The specification is listed in Table and Fig shows the topology of the power supply. In order to limit the magnet AC voltages to the ground, the power supply is connected to the ground in the middle of the output stage. The input transformer is used to match the voltage from the 10kV AC line, and a simple six phase diode rectifier is used to get a DC voltage, the voltage across the C1 is adjusted by the booster type DC-DC converter. A 2Q type bridge is used to control the output current. A full digital controller module is used to control the IGBT. Fig shows the scheme of the power supply control strategy. The reference of the waveform is designed in a look-up table, which can avoid the large data transfer. Fig The topology of the power supply Fig Digital PS controller 9.5 Tracking system of magnet power supply Aim and task: The RCS power supply system has 6 independent WHITE CIRCUIT, The expression of the magnet current is: I m =I dc -I ac *cosω 0 t 76

79 Aim:Ensure the phase control accuracy and constant ratio in amplitude between the magnet currents (magnetic field) in 6 networks during accelerator repetition cycle. Task: (1). current accuracy better than 0.1%; (2). phase error less than 10uS The Scheme for Control Control Policy:A 25 Hz common signal is used to synchronize for BM and QM, and also a 1MHz clock signal is used the synchronize the digital Power Supply Control Module (DPSCM). The 25Hz common signal is used to trigger the start address of the look-up table, which storage the wave form of the power supply set reference. Inside the DPSCM, the phase of the output current will be calculated. The error of the phase will be used to change the start address in the next cycle. But maybe it can be simple to control the phase of the output current, because we found it is stable enough in the prototype, if the tracking error is better than 0.1%. Thanks to DPSCM, it is easier to change the control strategies without changing the hardware. References [1] White, M.G., et. al, CERN-Symposium(1956) [2] J. A. Fox, Pro. IEE 112(1965)

80 10 RCS RF System Hong The purpose of the Rapid Cycling Synchrotron (RCS) RF system is to capture and then accelerate the proton beam from 80MeV to 1.6GeV with a repetition rate of 25Hz. The ring RF system is designed to provide the maximum RF voltage of 165kV and operate on the harmonic number h=2. The RF frequency range is from 1.02MHz at injection to 2.44MHz at extraction. Table 10-1 summarizes the main machine parameters related to RF system. Table 10-1 RF machine parameters for CSNS RCS Parameters Value Beam power 100kW Circumference 228 m Injection energy 80 MeV Extraction energy 1.6GeV Number of protons per pulse 1.87E13 Circulating dc current 1.5/ 3.6A Repetition rate 25Hz Harmonic number 2 RF frequency 1.02~2.44MHz Max. RF voltage 165kV (h=2) In order to minimize the beam loss during the RF capture and acceleration the RF voltage and synchronous phase are optimized with the RAMA code in accordance with the magnet field ramping in the RCS. Fig shows the curves calculated by RAMA code of the RCS RF voltage, synchronous phase, and RF frequency, within an accelerating period of 20ms RF frequency (MHz) RF voltage (kv) Synchronous phase (deg.) Time (ms) Time (ms) Time (ms) (a) RCS/RF frequency pattern (b). RCS RF voltage patterns (c) RCS synchronous phase patterns Fig RCS RF system operating curves in a cycle 78

81 The RCS RF system is composed of 8 ferrite-loaded RF cavities (h=2), including a spare one. Each cavity has own RF high power tube amplifier and LLRF control loops. The space for installation of additional three second harmonic (h=4) cavities is reserved for the phase 2 upgrade in the future to increase the beam bunching factor RF cavity A classical and conservative NiZn ferrite-based design was chosen. The ferrite-loaded coaxial resonant cavity has 2 accelerating gaps with single ended. Ferrite loaded material is Ferroxcube 4M2. There are 56 ferrite cores installed in a cavity, each section holds 28 pieces. The gap inductance can be shifted from 8.1µH to 1.4µH as a bias current varies from 200A to 3000A. The gap capacitance is 3nF. The resonant frequency of the cavity can sweep within 1.02~2.44MHz to meat the need of the CSNS RCS ramping operation. A nominal peak RF gap voltage of 12kV is required. The maximum RF magnetic flux density reaches 225Gs around 1.2MHz, as show in Figure The power density in ferrite would be less than 0.25W/cm 3, which is calculated according to the Q value measured from a small sample ring. Table lists the RCS RF system parameters RF magnetic flux density (Gs) RF frequency (MHz) Fig The function of RF magnetic flux density vs. RF frequency Fig The cavity mechanical layout Table RCS RF system parameters RF frequency range MHz (h=2) Number of cavities(h=2) 8 (1 spare) Cavity length 2.7m Diameter of beam pipe 170mm Number of accelerating gaps 2 Ferrite type Ferroxcube 4M2 Number of ferrite cores 56 pieces Size of the ferrite rings Ø500 Ø mm 3 Max. cavity voltage (h=2) 24 kv 79

82 Peak cavity power loss < 150 kw Power density in ferrite < 0.25 W/cm 3 Cooling water flow per cavity 210 L/min Gap capacitance 3 nf Gap inductance 8.1 ~1.4 µh Bias current supply A The cavity length is about 2.7m, and ferrite length is 1.4m. Two accelerating gaps are interconnected in parallel via two coaxial links and driven by a tube power amplifier located adjacent to the cavity. A magnetic coupler configuration between the amplifier and the cavity is used. Fig shows the cavity mechanical layout. Because the bias current goes through the inner conductor, i.e. beam pipe, the bias power loss depends on the resistance of beam pipe. In order to reduce the loss and to simplify the cavity mechanical configuration, the beam pipe is made of oxygen-free copper with stainless steel flanges. The ceramic in the gap is 95% Al 2 O 3. Fig shows the prototype of cavity inner conductor. After baking and evacuating the vacuum inside beam pipe is better than mbar. According to the test result of the small ferrite sample ring, the cavity dissipation is inferred to be about 140kW, and the average loss is less than 50kW. The ferrite power loss is taken out by cooling water in copper plates. The efficient heat transfer between ferrites and the cooling plates is a key issue. The flatness of cooling plates can be controlled in machining less than 0.1mm on the dish surface. The cooling plate is made from dual rectangular copper conductor winding in a spiral shape. The cross section of water duct is 9 4mm 2, and the water velocity is less than 1.8m/s. The prototype of full size cavity has been completed, as shown in Fig Fig The cavity inner conductor Fig The prototype of RF cavity Cold test of cavity has been done. The curve of the RF frequency vs. bias current and S 11 parameters of cavity are shown in Fig and Fig respectively. There are some high order modes in the cavity. Obviously the mode of 6.9MHz is introduced by a parasitical inductance of the links. It must be removed or damped in order to avoid exciting an extra field by a third harmonic component of the cavity voltage. Fig is the spectrum of cavity voltage. When the RF frequency is 2.34MHz, the third harmonic frequency is just 7MHz overlapping with a cavity high order mode, so the amplitude of this third harmonic component is too high 80

83 Resonance frequency (MHz) Bias current (A) Fig The cavity resonant frequency vs. bias current Fig S 11 parameter of the cavity, some high order parasitical modes can be seen Fig Spectrum of cavity voltage 10.2 Cavity dynamic tuning system During the RF frequency sweeping and a cavity detuning happen due to the beam loading in an accelerating cycle, RF cavity should operate on resonance over the period of 20ms to minimize reflected power reversing to the tube amplifier. Cavity dynamic tuning system can tune the resonant frequency of cavity automatically by adjusting the bias current, according to the phase between anode and grid voltage of the tube. Fig shows the scheme of cavity dynamic tuning system. 81

84 Bus bar 3kA Bias Supply B+ RF Set Value P D + PI Controller Digital LLRF Control Fig The scheme diagram of cavity dynamic tuning system The performance of tuning loop depends on the capability of a bias current supply. A switching bias supply has been developed using, due to IGBT modules with its merits of high efficiency. Fig is the main circuit diagram of the bias supply. Four rectification modules of 750A are connected in parallel to provide a DC bias current of 3000A. A multiple PWM control is used to suppress the output ripple. Table is the specification of bias current supply in design and measurement respectively. The tuning loop response to small perturbations extends to 2 khz, which is dominated by the response of the bias cuurent supply. Fig is the picture of prototype of the bias current supply. Fig and Fig show the measurement results of the response to small signals and tracing accuracy. The spectrum of cavity gap voltage and the ripple of bias current and voltage are shown in Fig and Fig The peak ripple voltage and current are less than 10V p and 2A p respectively, but the ripple effect on the RF voltage seems not too much. The ratio of RF component and sideband component resulted from ripple is more than 50dB. As the bias current goes higher the sideband component becomes lower. IGBT module: FF600R06ME3 600A/600V Fig Main circuit diagram of switching bias supply 82

85 Table The specification of bias current supply Specification Design value Measurement Nominal output current A A sweeping Nominal output voltage ±40V/3000A Current Stability 0.2% 8hours Ripple voltage 0.2%+10mV Much higher Band width ( to small A signal) Bias current slewing rate >400A/ms 400A/ms Tracing accuracy < 0.1% 0.25% for sin waveform Fig Prototype of switching bias supply for the ferrite-loaded RF cavity Fig The measurement of tracing accuracy at sin waveform 83

86 20Log(Vout/Vin) A+30sinωt f(hz) Log(Vout/Vin) A+30sinωt f(hz) Fig The measurement of band width of the bias current supply I b =500A V gap =500V p-p fo=1.27mhz Vrf / Vripple >50dB Fig Spectrum of the gap voltage and the sideband frequency (switch frequency of 20 khz) component caused by the ripple of bias current Voltage ripple Current ripple Fig Ripple of output voltage and current of the bias supply 84

87 10.3 RF power amplifier A tube power amplifier is developed using the tetrode 4CM500000G (CPI product), with the same configuration and performance as TH558. The final tube amplifier with a configuration of grounded-cathode is operated in class AB1. The maximum plate dissipation of the tube is 500kW. It is sufficient for offering the cavity power dissipation and beam power, also remaining a lot of power margin for the cavity s dynamic tuning, RF feedback control and protracting the life time of tube to enhance the reliability of RF system. Between the preamplifier(ssa) and the final tube amplifier there is a wide-band tetrode amplifier of 500W, as a feedback amplifier, plus a power combiner to serve both as driver and fast feedback amplifier. Feedback of the gap RF signal provides reduction of the cavity impedance to the beam. A particular design is the use of a tuned low-q resonant grid circuit, to reduce the affect of grid capacitive reactance as RF frequency changing. Advantages are higher gain, smaller drive power and the possibility to increase the feedback loop stability. An adjustable ferrite loaded inductance is employed in the resonant grid circuit, and the inductance varies as a bias current to keep the input and output of the feedback amplifier in phase. Layout of the ring RF system is sketched in Fig Table is the specification of tube power amplifier. Table Frequency range Repetition rate Duty factor Maximum peak power Cavity shunt impedance Operating mode Maximum input signal The specification of tube power amplifier MHz 25 Hz 1/2(RF on), 1/3 (power output) 500 kw (4CM500000G tetrode) 2 kω 400 Ω AB1 +5 dbm 85

88 3kA Bias Supply Bus bar HV 4CM500000G FB SSA RF in Grid Tuning Host computer Ethernet CPU PowerPC PMC Flash memory SDRAM 100MHz Ethernet FPGA (EP2S180) Nios II DAC 14bit CLK C L K Generator ADC 14bit Insulator Insulator RF D rive r Bias driver Vgap Vgrid WCM FCT BPM VxWorks PCI Bridge Digital LLRF CPCI Bus Fig Layout of ring RF system The anode DC voltage of 15kV is provided by a solid-state high voltage supply using PSM (Pulse Step Modulation) technology. It comprises 24 IGBT modules in series, and each module produces 700V. This kind of high voltage supply is benefit of easily operating and maintaining, and also no demanding of a crow bar protection. The whole module is water-cooled. The transformer is 300kVA. Fig shows the picture of a R&D prototype of RF power amplifier. The high power test on a RLC resonant dummy load of 50kW, in a duty factor of 5% and a repetition rate of 25Hz, is ongoing now. IGBT modules Tetrode Control unit RF tube power amplifier Transformer Anode DC voltage supply Fig Prototype of RF power amplifier 10.4 Ring LLRF control system LLRF control system comprises two main parts of digital control loops based on FPGA technology and system interlock protection based on PLC controllers. A 86

89 generic, modular LLRF control architecture will be adopted. Whole control function is implemented by FPGA programming. Digital control system is based on cpci bus with high speed of 132M bit/s and high reliability. cpci is a popular industrial control bus, so that the cost of the commercial PCI module and development should be low. In order to simplify and compact the system configuration, the FPGA boards, which implement same functions in different cavity system, for example the amplitude loops for the 8 cavities, will be put into one cpci crate to build up a control module, together with a CPU board with a PowerPC processor running the VxWorks operating system in the module. It communicates with the central control system via the 100/1000 Mbytes/s Ethernet ports. The layout of the ring LLRF control system is sketched in Fig RF Local Control Central Control Ethernet WCM FCT BPM FPGA amplitude & phase based on cpci C P U FPGA amplitude & phase based on cpci C P U FPGA feedforward & orbit based on cpci C P U PLC Controller Interlock protection 8 ch RF signal combiners V gap Beam Signal Beam Line Fig Scheme of ring LLRF control system A major component of FPGA control board prototype will be developed. It is composed of FPGA Altera Stritax II EP2S180 with embedded DSP blocks and a soft CPU core of Nios II, a 100Mbytes/s Ethernet port, 4 channels 14bit ADC and 4 channels 14bit DAC on the board (Fig ). The sample clock is 100MHz. Because of the large FPGA size and capability to run virtual processor cores within the FPGA logic, the control functions and data processing can be completed only in FPGA, which enables the digital control system working faster. A DDS (Direct Digital Synthesis) core made-up by FPGA programming will be used to generate a sweeping RF signal. Its frequency, amplitude and phase can be regulated easily via programming. The required accuracy of the RCS LLRF system is amplitude ±1% and phase ±1 The block diagram of RCS LLRF is shown in Fig

90 PMC Flash memory SDRAM SDRAM DDRAM 100MHz Ethernet FPGA (EP2S180) Nios II Internal Clock 4 ch DAC DAC 14bit 14bit CLK Generator 4 ch ADC ADC 14bit 14bit Insulator Insulator External Clock PCI Bridge FPGA board Fig Configuration of FPGA board Frequency Set Frequency Correct DDS D irect D igital Synthesizer Phase M odulation Synchronous Phase Control Loop Cavity Voltage Control Loop RF Voltage Amplitude M odulation Adaptive Feedforw ard Orbit Feedback RF Phase IQ D em odulation Direct Feedback Sum Am p. Beam Frequency Beam Phase Cavity Tune Control Loop Grid Voltage Phase Tube Amp. Beam Amplitude RF voltage Detector Bias Supply Anode V oltage Phase Cavity BPM FCT WCM Fig Block diagram of RCS LLRF system for one cavity Digital LLRF R&D is based on an Altera FPGA development board of DK-DSP-EP2S60N. After the control loop programming, simulation and table experiment, the test of amplitude and automatic tuning loops has been done under low power using a solid-state amplifier of 500W together with a ferrite-loaded RF cavity and a 3000A bias supply. Fig shows the set up of digital LLRF system test. OSC DPO4034 FPGA Development board Diagnostic & Command Signals V rf Cavity C gap A Bias External Clock ADC ADC Clock Internal generator Clock DAC FPGA EP2S60 PMC Flash memory SDRAM DAC Nios II 100MHz Ethernet RF current transducer I rf Digital LLRF SSA 500W Amp Fig Set up of digital LLRF test The amplitude and tuning loops implemented in FPGA is shown in Fig

91 DDS I Digital Demodulation Q AMP Pattern Filter I R Filter Q Phi A A-B B A PI Controller Digital Modulation Bias pattern I Digital Demodulation Q Filter Filter I Q R Phi A-B B Tune Phase Pattern B A-B A PI Controller + A/D Bias Supply D/A D/A A/D RF In FPGA RF in Driver Cavity Fig Digital control loops in FPGA The high power test of RF power tube amplifier is ongoing now in the factory. After acceptance test it will be transported to IHEP recently. Next step is the ring RF system prototype installation and adjustment, then starting the high power experiment and measurement of RF system. 89

92 11 Vacuum System Haiyi 11.1 Introduction The CSNS vacuum system consists of linac, beam transport lines of linac to ring and ring to target, and Rapid Cycling Synchrotron (RCS). The linac vacuum system is described in the separated equipment such as H - ion source, RFQ and DTL. A conventional design is applied on vacuum system of transport lines. We focus on the design and fabrication of the RCS vacuum system in this report. The goal of the RCS vacuum system is to provide a friendly environment for the circulating beam. The basic requirements to the RCS vacuum system are: The operating pressure of RCS will be in 10-8 Torr range to maintain sufficient beam lifetimes and to minimize ion-induced pressure instability. Ceramic chambers will be chosen in the dipole and quadrupole magnets to reduce the eddy-current effect which will produce the perturbation to the magnetic fields and the large ohmic losses. RF shielding is necessary to lower the impedance of the circulating current. The shielding will be designed as a high-frequency pass filter, where the eddy current effect can be avoided. The inner surface of the ceramic chambers will be coated with TiN to reduce the secondary electron yields. All the vacuum components under the radiation environment in the tunnel should be fabricated with high reliability and long lifetime in order to minimize maintenance. The vacuum pressure must be below the threshold at which there is appreciable loss of beam due to nuclear scattering, coulomb scattering and residual gas ionization. The latter is the limiting factor. The interaction between the proton beam and the residual gas produces the ions; the newly created ions are repelled by the positive space-charge potential and are accelerated towards the wall. The ions bombard the chamber wall and desorb molecules from it, and will result in possible pressure instability. To prevent the instability, the minimum effective linear pumping speed of CSNS RCS vacuum system should be more than 22 L/s m. The main parameters of the RCS vacuum system parameters are listed in Table Table Main parameters of the CSNS RCS vacuum system Circumference of ring [m] Average circulating current [A] 3.6 Required pressure [Torr] Dipole & quadrupole chamber material 97 99% Alumina ceramic 90

93 Ion desorption coefficient [molecules/ion] 5-10 Shape Race-track cross-section Dipole ceramic chamber Number 24 Bending radius[m] Length [m] 2.8 Inner aperture size Exterior aperture size 216(H) 134(V)[mm mm] 236(H) 152(V)[mm mm] Quadrupole ceramic chamber Type-A (16) Type-B (16) Type-C (8) Type-D (8) Exterior diameter [mm] 202 Inner diameter [mm] 182 Exterior diameter [mm] 261 Inner diameter [mm] 241 Exterior diameter [mm] 218 Inner diameter [mm] 198 Exterior diameter [mm] 249 Inner diameter [mm] 229 Total volume [m 3 ] Total inner surface [m 2 ] Ceramic vacuum chambers The vacuum chambers within the dipole and quadrupole magnets of RCS must be designed to limit the eddy currents induced in them. It is not possible construct these chambers of stainless steel, aluminum or copper, which are normally used in other accelerator vacuum systems, because the magnetic field cycles in 25 Hz at the RCS. Therefore, an alumina ceramic material constructed of isostatically pressed methods with high strength, good radiation resistance and good vacuum properties, is a better choice for the dipole and quadrupole vacuum chambers. Several approaches have been developed to fabricate long ceramic chambers in other laboratories. At Rutherford Appleton Laboratory, the ISIS chambers were fabricated by joining short segments of ceramic pipes together by using glass glazing rather than the more conventional method of metallizing the ends of the ceramic pipes and brazing metal ring between them. This provides a strong, vacuum tight and radiation resistant join. Ceramic flanges are attached to chambers in the same way. Wire cages following the beam aperture were fitted inside the ISIS ceramic chambers to form RF shielding and conduct the high frequency image current. A different approach was used by J-PARC, the ceramic pipes were joined by metallizing the ends and brazing them together. Titanium flanges were welded to Ti sleeves by TIG welding, while the Ti sleeves were attached to the ceramic pipes by metallizing and 91

94 brazing. The electroformed copper strips are covered around the ceramic pipes for RF shielding and capacitors are connected between the each strip and the Ti flanges in order to cut off the eddy current loop. The inner surface of the ceramic chamber is coated with TiN to preclude any charge buildup and reduce secondary electron yield, and this is important to avoid the instability of potential electron cloud Dipole ceramic chambers The ISIS approach will preferably be adopted for the CSNS RCS dipole ceramic chambers since this method is considerably less expensive and relatively simple than that of metallizing the ceramic pipe ends and brazing them together. The design scheme of the dipole ceramic chamber by glass joining is shown in Fig The dipole chamber with a race-track cross section is divided into 7 sections with 15 bending in the dipole magnet. The ends of each ceramic pipe are ground flat and pre-glazed on each face with a thin glass layer of about 0.4 mm. The inner surfaces are not ground, but must be within the allowable errors. The outer walls are partly ground for the installation. The ceramic pipes are then placed one above the other in an air furnace and heated to about 1100 C to melt the glass at a rate of 50 C/h. The ceramic flange is also joined to the ceramic sections by glazing. According to the ISIS experience it is necessary to hold the ceramic pipes in place using ceramic dowels in the end faces to avoid sideways slip. Since the glass joint replace the thin metal rings for brazing, the whole structure is resistant to corrosion by acid which can be formed in the air at high radiation levels. The nominal clearance between the chamber wall and the magnet pole is 3 mm and the wall thickness of ceramic chamber is 8 mm and 15 mm in the horizontal and vertical directions, respectively. The calculated maximum tensile stress in the walls is 11 MPa under vacuum load. Length:2782mm Bending pipes:5 426mm Straight pipes:2 320mm Ceramic flanges:2 27mm Fig Scheme of the dipole ceramic chamber by glass jointing 92

95 To check the feasibility of the glass joint, a testing ceramic chamber with the length of 1 m and the bending of 5 has been designed and fabricated. The testing dipole ceramic chamber is shown in Fig It consists of the four segments of ceramic pipes and two ceramic flanges. Even if the ceramic surfaces are highly polished, the micro-cracks on the surface of the ceramic flanges will occur. That will be easy to result in a leakage for metal gaskets. Therefore, a thin ring-shaped glass layer of about 0.2 mm has been glazed to the ceramic flanges, and this provides an excellent sealing surface with a leak rate of less than Torr L/s. The ultimate pressure of the testing ceramic chamber has reached Torr, and the outgassing rate is less than Torr.l/s.cm 2, which meets the requirement of the vacuum system. Metal seal of Helicoflex Delta type and demountable V-type chain clamp will be used to join the ceramic flange and the flange of the adjacent stainless steel chamber for easy assembly and good reliability. Fig The testing dipole ceramic chamber Quadrupole ceramic chambers A prototype of the quadrupole ceramic chamber has been produced by metallizing of the ends of ceramic pipes and brazing them together, since it is relatively easy to be fabricated in a vacuum furnace. The one-meter long quadrupole chamber has a circular cross section (282 mm inner diameter) and a wall thickness of 10mm. Ceramic pipes were constructed of isostatically pressed high purity alumina, and joined by brazing with Cu-Ag alloy, and the Ti flanges were welded to the Ti pipes which were joined to the ceramic pipes by brazing. Mo-Mn metallization layers were coated by a thin layer of nickel to protect them from oxidation. Fig shows the prototype of the quadrupole ceramic chamber. 93

96 The one-meter long ceramic chamber consists of two segments of the ceramic pipes. Because of the requirements of the installation accuracy, the total allowance tolerance of circularity and straightness is less than 3 mm. The exterior surfaces of the sintering ceramic pipes were ground to reach the required sizes. The ultimate pressure of the unbaked ceramic chamber reached Torr, and the outgassing rate is less than Torr l/s cm 2, which are in the good vacuum condition. Both the approaches of the glazing and brazing for the dipole and quadrupole ceramic chambers have been tested, and we will finally decide which is more suitable for the CSNS vacuum system according to the ratio of performance and cost. Fig Prototype of the quadrupole ceramic chamber RF shielding of the ceramic chambers The RF shielding of the CSNS ceramic chambers will be assured by wrapping Cu strips of 0.4 mm in thickness on the external surface of the ceramic chambers. The impedance of the shielding is affected by the skin depth. A thickness of 0.4mm is acceptable, considering the skin depth in copper is about 0.1mm. The RF copper strips are fixed in the kapton insulation sheets with a width of 5 mm and an interval of 5 mm. One end of each Cu strip is connected with a copper screen across the flange by a capacitor in order to cut off the eddy current loop. The geometrical design of the RF shielding in the ceramic flanges junction should be smooth to reduce the impedance. The exterior RF shielding offers a large aperture for the beam and is easy to check the damage of the copper strips. However, Kapton film is a thin organic polymer which supplies high the resistance to radiation, but the performance of the adhesive between Cu strips and kapton films need to be tested under high radiation. Fig shows the RF shielding construction of the ceramic chamber. 94

97 Fig RF shielding method of the ceramic chamber TiN coating of the ceramic chambers The inner surface of the ceramic vacuum chambers will be coated with Titanium Nitride (TiN) of about 100 nm in thickness in order to minimize the secondary electron yields (SEY), thus avoiding the so-called e-p instability caused by electron multipacting. DC magnetron sputtering will be adopted to coat TiN inside the ceramic chambers. An experimental system of TiN coating has been set up, as shown in Fig This system consists of a Ti cathode with magnets inside, a turbomolecular pump group, a capacitance manometer and a cooling device etc. The magnets are stacked with opposing poles, resulting in a looping magnetic field of sufficient strength on the cathode surface. Argon and nitrogen gases of Very high purity are fed into the vacuum chamber. The flow rates are adjusted using mass flow controllers. The dense plasma generated and confined by the electro-magnetic field increase the ion density and the sputtering rate by a factor of ten or more as compared with that of DC sputtering, allowing the coating of 100 nm TiN in an hour. The vacuum chamber is baked at 150 C for 48 hours to degas before TiN coating. Fig Experimental system of TiN coating 95

98 An anode screen made of thin copper sheet of 1mm in thickness sheet is wrapped outside the ceramic chamber in order to produce an electric field with the Ti cathode. This will result in a uniform thickness of TiN along the ceramic chamber. The uniformity of TiN coating thickness is better than 10%, and the ratio of Ti and N is about The secondary electron yield of the ceramic sample is expected to be measured The Pumping System The RCS ring with a circumference of m will be divided into 8 vacuum sectors by all metal gate valves, which are four arc sectors and four straight sectors. The gate valves are interlocked by vacuum gauges to protect the ring from the catastrophic vacuum failure. The valves also allow the repairing and modification of components without venting other sectors to ambient air. Fast-closing valves are installed in the extracted beam line. These valves protect the RCS from contamination by the spallation target when the proton beam window fails. Roughing down to approximately 10-7 Torr will be achieved by the oil-free turbo-molecular pump group with a pumping speed of 400L/s. The main pumping is preferably achieved with ion pumps duo to its high reliability, no moving parts, long life and high radiation resistance. In addition, the ion pump current is proportional to vacuum pressure and will give more detailed pressure profile along the ring. The power supplies of ion pumps will trip to protect the ion pumps from damage if the ion current rises above a pre-set value. Ion pump currents can be stored in a databank, enabling the operators conveniently to find problems. Lumped ion pumps with a pumping speed of 800L/s in an interval of about 10 m will be used to reach the operation pressure of Torr on average. Three sets of Pirani and cold cathode gauges will be installed at each vacuum sector to measure pressure. One residual gas analyzer will be installed at each vacuum sector and provide a quick analysis for the partial pressure of residual gases. Duo to the high radiation levels in the tunnel, all the vacuum electronic devices will be located at the service building. References [1] H.C. Hseuh, Design of the SNS Accumulator Ring Vacuum system, Proceedings of PAC 99, New York, 1999 [2] J R J Bennett, The performance of the vacuum system of the 800 MeV proton synchrotron, ISIS, Vacuum/volume 44/number 5-7/page /1993 [3]M. Kinsho, Development of alumina ceramic vacuum duct for 3 GeV-RCS of the J-PARC project, Vacuum 73 (2004)

99 12 The Preliminary Design and Prototype Test of CSNS/RCS Injection and Extraction Magnets Wen The RCS accumulates protons via H - stripping injection with 25 Hz rate. The injection of the RCS takes place in a 9 meter long straight section, which includes four shift magnets (BC) forming a horizontal orbit bump, eight symmetrically placed dynamic bump magnets (BH&BV) for the phase space painting in both the horizontal and the vertical planes, two septum magnets and two strippers as shown in Figure Fig Outline of the H- painting injection system The extraction system of the RCS consists of seven fast kicker magnets and a Lambertson magnet. When the beam needs to be extracted, the seven kicker magnets deflect the beam vertically in such an angle that the beam could leave the RCS centric orbit and enter the downstream Lambertson magnet. Then the Lambertson magnet will deflect the beam horizontally for an angle of 13 degree and make the beam leave the RCS thoroughly. The injection dynamic bump magnets are pulsed magnets. To decrease the eddy current loss, the magnet core must be laminated with 0.15 mm thin silicon steel sheets. In addition, the 25 Hz pulsed field induces eddy currents in the magnet components, especially in the end plates, and heat generation is also of great concern. The extraction kicker magnets are also pulsed magnets, but the field pulse is much shorter than that of the injection bump magnet. Instead of using silicon steel sheets, the core of the kicker magnets must be formed with ferrite blocks, and the magnets should be placed inside the vacuum tank. To prove these design considerations, one prototype injection bump magnet and one prototype extraction kicker magnet have been developed and tested. The parameters of the two prototype magnets are listed in table Table 12-1 Parameters of the prototype magnets Bump magnet Kicker magnet Peak field 2313Gs 552Gs Core length 220mm 220mm Beam H140mm V170 H133mm V206 97

100 Aperture mm mm Field rise time Field top time 1ms 50µs 250ns 600ns Peak current 18000A 6000A Repetitive rate 25Hz 12.1 DESIGN OF THE BUMP MAGNET 25Hz As shown in Figure 12-1, the narrow space between the bump magnets and injected beam ducts requires that the core of the magnet must be designed as a C-type or an asymmetric W-frame structure. Considering the low stray field on the injecting beam and the good accuracy of repetitious assembly, the asymmetric W-frame structure is preferred. The drawings of the bump magnet are shown in Figure Fig Drawing view of the bump magnet The core is laminated with 0.15 mm thin silicon steel sheets, which have the property of low eddy current loss. The coil has two turns and can be split into upper and lower parts. When the BH magnet is powered to full current, the large Lorentz force will cause the coil to vibrate with 25 Hz. To reduce the vibration of the coil, the holes are made in the coil and core, and by using 12 bolts through these holes, the coil are tightly pulled on the core. The eddy current induced in the end plates of the core is expected to be large, so some slit cuts are produced in the end plates to cut the eddy current loops and therefore to decrease the eddy current losses. The insulation of the coil requires a high tension and a high radiation resistance, so Kapton films are applied for the coil insulation from the core TESTS OF THE PROTOTYPE BUMP MAGNET After the assembly of the bump magnet, the performance of the coil has been tested. The insulation test passed under a DC voltage of 4.5 kv, which is 1.5 times higher than the full operating voltage. The welding quality of the coil conductor and ceramic insulators used to insulate the power supply from the cooling water pipes have also passed through the hydraulic pressure test of 20 kg/cm 2. The measurements of the magnetic field are shown in Figure The field waveform shows that the rise time is 1ms and falling time 0.45ms, and the peak field 98

101 is 2018 Gauss while exciting peak current is 1.8kA. The field uniformity is better than 1.5% within 60% of magnet aperture. 场均匀性 ( 长线圈 ) 4.00% 2.00% 0.00% -2.00% -4.00% -6.00% Y=0 Y=20 Y=40 Y=60 Y=80 Y=-20 Y=-40 Y=-60 Y= % X(mm) Fig Testing of the magnetic field The measured temperature rise of the magnet is shown in Figure The temperature on the surfaces of the coil and core is not higher than 42 o C, but the temperature on the electrically connecting copper bar is higher than 100 o C, which needs to be modified wider to decrease the temperature in the future design. Figure Measured temperature of the magnet The vibration of the magnet has also been measured. The results are listed in the Table Figure Vibration measurement of the magnet 99

102 Table Vibration measurement results of the magnet X direction Y direction Z direction Main frequency component(hz) 25,75 125,150 25,75 Amplitude (µm) DESIGN OF THE KICKER MAGNET The designed kicker magnet has a single turn structure. The magnet consists of a vacuum tank, a ferrite core, a copper current bus, two ceramic insulating supporters, a sliding support plate and a high voltage feedthrough. The vacuum tank is made of cylindrical stainless steel tubes with a diameter of 600 mm. It not only provides vacuum for the kicker magnet, but also supports the whole magnet. The window frame core consists of eight C-shaped Ni-Zn type ferrite blocks, which are developed in a Chinese ferrite company. The performances of the ferrite are as good as that of CMD5005, which are listed in Table The ferrite has the properties of high frequency response, low loss and low outgassing rate, which are important for fast pulsed magnets in high vacuum application. The current bus is made of several pieces of copper plates, which are soldered together. To strengthen the insulation between the high voltage current bus and the ground, two large pieces of ceramic plates were produced and added on the top and bottom of the magnet. They were also used to support and protect the ferrite core. The picture of prototype kicker magnet is shown in Figure Table Properties of the ferrite CSNS ferrite CMD5005 Initial permeability Saturation flux density 280mT 320mT Remanent flux density 138mT 170mT Coercive force 16.1A/m 13.5A/m Curie temperature 250oC 130 oc DC volume resistivity 107Ω*m 107Ω*m Mass density 5.2kg/cm3 5.1kg/cm3 Fig The prototype kicker magnet 100

103 In the middle of the back legs, two copper strips are used to carry the beam induced image current and are helpful to reduce the coupling between the beam and the ferrite. High voltage current is fed from one end of the ferrite core. The inner surfaces of the magnet will be coated with TiN films to reduce the secondary electron yield. Before assembly, the vacuum tank was cleaned first, and then the ferrite blocks and all of other machined parts were cleaned and baked at 200 o C. Firstly, the kicker magnet was assembled, aligned and fixed on the sliding plate. Secondly, the sliding plate carrying the magnet was slid into the vacuum tank. By adjusting the sliding plate in the vacuum tank, the magnet was moved to its final position, which was determined by beam dynamics. Thirdly, the tank was closed as shown on right hand sided Figure TESTS OF THE PROTOTYPE KICKER MAGNET The kicker magnet is powered by a pulsed power supply. The 40 kv pulse from the power supply will be sent into the magnet in the vacuum tank by a coaxial HV feedthrough. The feedthrough has a tri-polar structure, which can isolate the ground of the pulsed power supply from that of the vacuum tank. After assembly, the vacuum test was done firstly. The required vacuum pressure for the kicker magnet is Torr. Since large quantities of ferrite and ceramic were used in the system, to overcome the outgassing of the ferrite and ceramic blocks, one 500L ion pump was installed under the kicker assembly. The whole system was baked at 180 o C under vacuum and kept for 48 hours. After cooling down, the vacuum pressure reached Torr., which met the vacuum requirement. The distribution of the integral magnetic field was measured using a long search coil, which has the loop size of 3mm in width and 1100 mm in length. The magnetic field waveform and the exciting current waveforms are shown in Figure The rise time of the waveform is about 250ns, and the top time is about 600ns. Fig Waveforms of the field and current Fig Measured field error distribution The distribution of the integrated magnetic field has been measured within the 60% of the magnet aperture. The uniformity of the field is about 1% as shown in Figure , which has agreement with the simulated results DESIGN OF THE LAMBERTSON MAGNET 101

104 The Lambertson magnet has been chosen as a septum magnet for the beam extraction of the RCS because it has many advantages, such as anti-radiation structure, higher reliability and lower power consummation, etc. The only disadvantage of it is that the stray field is supposed to be large, so some effective measures will be adopted to decrease the stray field to the allowed value. At this point, two dual-layer shielding plates were employed at ends of Lambertson magnet respectively, the effect of magnetic shielding is very good. Figure shows the sketch structure of Lambertson magnet. Figure shows the simulated results of distribution of stray fields in circulating-beam pipe, and the ratio of stray field and the field at the extracting center is only 4.7E-4, which much lower than that of SNS, the latter arrives at 2.1E-3. And the ratios of stray fields at the different positions were also calculated with 3D simulations with a result that the differences between ratios at center of circulating-beam pipe and closing to septum is smaller than 1E-4. Fig Structure of Lambertson magnet Fig Distribution of stray field in circulating- pipe 102

105 13 Injection Bump Magnets Pulsed Power Supply for RCS Yunlong Bump magnets power supply for RCS injection system consists of BCPS BHPS and BVPS, which provides power to four constant bump magnets in series, four horizontal painting bump magnets in series and four vertical painting bump magnets in series respectively, as shown in Figure Figure 13-1 Injection bump magnet power supply system The falling edge of the pulsed power supply for RCS injection painting bump magnets is determined by the given physical function and can be changed as needed in operation, as shown in Figure Table 13-1 lists the design specifications of the pulsed power supply for injection bump magnets. Pulse current of horizontal painting Pulse current of vertical painting bump magnets bump magnets Figure 13-2 Output pulse current waveform requirements to the pulsed power supply for RCS injection painting bump magnets 103

106 Table 13-1 Specifications of injection bump magnets pulsed power supply for RCS BCPS BHPS BVPS Pulse recurrence frequency (Hz) Pulse rise time (us) > 550 > 550 > 550 Pulse Flat Top Time (us) Pulse fall time (us) (Programmable) (Programmable) Pulse current(a) Load inductance (uh) 8.24(2.06 4) 6.16(1.54 4) 8.52(2.13 4) Inductor resistance (mω) Inductor maximum voltage (V) Maximum rate of descent(a/us) Stability (%) <0.5 <0.5 <0.5 Tracking error (%) 2% 2% 2% According to the design specifications, the descent rate of pulse current of bump magnet pulsed power supply must reach 260A/µs and the current tracking error is less than 2% when the maximum pulse current of pulsed power supply is about 18000A. The main circuit of the bump magnets pulsed power supply is composed of many IGBT power conversion units in the form of 5 in series and then 10 in parallel,which can produce high-power pulse current, as shown in Figure Because of its special structure and double control strategy in each H bridge, the equivalent switching frequency of power supply can reach 100 times of the actual IGBT switching frequency. For example, if the actual IGBT switching frequency is 10kHz, the equivalent switching frequency is up to 1MHz. Thereby it can improve the tracking accuracy of output pulse current to meet the tracking accuracy requirements( 2%). 功率变换单元 L 1 P 功率变换单元 L 2 R 0 L 0 C 1 C 0 C 2 10 个并联 功率变换单元 L 10 N Figure 13-3 Main circuit of bump magnet pulsed power supply Each power conversion unit of the bump magnets pulsed power supply consists of five power modules in series, as shown in Figure Each power module includes rectifier, boost chopper and four-quadrant chopper. DC-side bus voltage is stable at 750V, which can be realized by the control of boost chopper. Each arm of the 104

107 bridge in a four-quadrant chopper is formed by four 450A/1200V IGBTs in parallel. Each power module can output high pulse current above 1800A. Due to the modular structure, we just need to replace the fault module according to the fault instructions of power supply control system when the power failure occurs. In this case, it has no use for online troubleshooting, which can be convenient for power maintenance and improve operational efficiency. Phase-shifting transformer is used to provide separate three-phase voltage to the five power modules. 隔离变压器 Figure 13-4 Schematic diagram of power conversion unit Bump magnets pulsed power supply is the key equipment of RCS injection system. There are no mature products and technologies in domestic company because it is a high challenge in power supply technologies. The development of power supply prototype began in 2007, and was completed more than two years later. It has passed the acceptance test and met all the design requirements. Test results are shown in Table Figure 13-5 is the photo of the pulsed bump magnets power supply. Figure 13-6 and Figure 13-7 represent the output pulse current and output pulse voltage waveform of bump magnets pulsed power supply respectively. Table 13-2 Injection bump magnet pulsed power test results Design Test Results specifications Pulse current 18000A 18000A Pulse rise time 1000µs 1000µs Pulse Flat Top Time 50µs 50µs Pulse fall time 300~550µs 400µs Maximum rate of descent 260A/µs 260A/µs tracking accuracy of pulse current falling <±2% ±1.9% edge equivalent switching frequency >800KHz MHz 105

108 Pulse current stability <±0.5% 0.38% (10000A 4hours continuous running) Load parameters 12µH 12µH Peak output voltage 3120V V Frequency 25Hz 25Hz Figure 13-5 Injection painting bump magnets pulsed power supply (yellow- Given current,blue- Output pulse voltage) Figure A Output pulse current (yellow-given current,blue- Output pulse current) Figure V Output pulse voltage 106

109 Pulse power supply for RCS extraction kicker magnet The extraction system in RCS consists of seven sets of kicker magnet pulse power supply, which provide high-power exciting current to seven fast kicker magnets respectively, as shown in Figure13-8. Figure 13-8 The sketch of extraction kicker magnet pulse power supply system The characteristics of the extraction pulse power supply for the kicker magnet needs a very rapid rising time for the output pulse current (~250ns). The simulated waveform is shown in Figure13-9. The designing parameters of extraction pulse power supply are listed in Table ms ms ms ms Time Figure 13-9 The output pulse current wave of the extraction pulse power supply Table 13-3 The designing parameters of extraction pulse power supply for kicker Bunch energy 1.6 GeV Bunch cycle 830 ns (at 1.6 GeV) Extraction time gap 330 ns Extraction Bunch Length 500 ns 107

110 Extraction frequency 25 Hz Extraction kicker inductance 0.58 nh Extraction kicker magnet excitation current ~5800 A Transmission cable characteristic impedance 12.5 Ω PFN charge voltage ~40 kv Excitation current pulse rise time ~250 ns (1% - 95%) Excitation current pulse flat top time ~ 600 ns Excitation current pulse flat-topped degrees < ± 2 % PFN Voltage stability < ± 1 % Design requirements for the extraction kicker magnet pulse power supply are as follows: 1)beam deflected by the kicker magnet should not hit the following quadrupole magnet; 2) furthermore, this beam should not hit the components of the septum magnet and transport line; 3) the pulse power supply should be located outside the RCS tunnel, escaping from high radiation extraction region for keeping the system of a high maintainability and reliability; 4) the operating voltage of PFN should be no more than 40kV considering the lifetime of equipment, the stability and the fault tolerance as well as the long-term operation of the system. The Blumlein-type Pulse Forming Network (PFN) and the high-voltage high-frequency transmission cable are used as the scheme of the RCS extraction kicker magnet pulse power, as shown in Fig The PNF of 6.25Ω is composed of four-group PFNs with a characteristic impedance of 6.25Ω joined as two in parallel and then two in series. The charging voltage can be reduced by the Blumlein-type PFN, resulting in an operating voltage less than 40 kv as well as an equivalency between the charging voltage of PFN and the discharging voltage of the magnet. Four high-voltage cables with a 50Ω characteristic impedance are set in parallel for achieving a transmission cable with a 12.5Ω characteristic impedance. There is a matching resistor with a 12.5Ω characteristic impedance set in parallels and on the right of the PFN, which is used for reducing the effect of the induced current resulted from the accelerated beam as well as reducing the transmission impedance of the beam. The excitation current of magnet is doubled by using the terminal-shorted circuit. The hollow-anode-type heavy hydrogen thyratron can withstand large reverse current. The fast recovery diodes and absorption resistance are in series and then paralleled with PFN for reducing the effect of the reflected current. When the beam is accelerated, the saturate reactor is used for isolating the PFN structure as well as the absorption resistance of 12.5Ω, and for sharpening the pulse leading edge. 108

111 Figure The sketch map for extraction kicker magnet pulse power supply system Because of the strict demands on the difficult technique of the power supply, there is not mature product or technology in domestic company. A prototype of the pulse power supply was fabricated in 2007.It satisfies all the design requirements, as listed in Table A photo of the prototype is shown in Figure13-11 while the measured pulse current waveform is shown in Figure Table 13-4 The testing results of the extraction kicker magnet pulse power supply Design specification Testing results Charge voltage(kv) Excitation current amplitude(ka) Rise time(ns)(2%-97%) < 250 < 250 Flat-top width(ns) > 600 > 600 Flat-topped degrees of current pulse(%) < ± 2 ± 1.56 Pulse amplitude stability < ± 1 < ± 1 Excitation current pulse jitter(ns) <±10 <±10 PFN Voltage stability(%) < ± 1 < ± 1 Extraction frequency(hz)

112 The tank of pulse power supply PFN magnet link of Transmission cable and kicker Figure RCS extraction kicker magnet pulse power supply Figure Pulse current waveform measured from the load on the magnet 110

113 14 Mechanical System Huamin 14.1 Collimator A classical two-stage collimator system is applied for RCS. The transverse collimation system is arranged in a dedicated straight section as shown in Fig To adapt to the existing lattice, we put the primary collimator in front of the secondary drift, and the following four secondary collimators which occupy suitable phase advances from the primary one. Fig Location of the transverse collimators in CSNS/RCS Primary collimator The primary collimator is a kind of jaw-type one, the chief components are thin scrapers which have adjustable distance to the beam center. Once the primary collimator is placed inside the machine, it becomes the main aperture restriction. For easier operation, a special design consisting of a movable inner surface for the primary collimator was proposed. Fig shows the structural figure of the primary collimator. Fig The structural figure of the primary collimator The heating of primary collimator is severe. On the one hand, interaction between halo particles and scraper leads to a large amount of thermal deposition; on the other hand, the scraper is too thin to be cooled directly with cooling water. The thermal analysis is shown in Fig The scraper adhered with a copper sheet in which the cooling water cycles. The maximum temperature of the scraper is about

114 . Fig Thermal analysis of the primary collimator Secondary collimator The secondary collimators are self-shielding structures capable of absorbing the scattering halo particles and containing the shower of secondary particles. In the first step of design, the absorber must stop the scattering protons. The lost particles kinetic energy is lower than 200MeV in the first phase of CSNS and lower than 400MeV in the second phase. So the absorber should stop protons with 400MeV kinetic energy at least and absorb most of the energy of the probably beam loss in 1.6GeV. Fig shows the structural figure of the secondary collimator. Fig The structural figure of the secondary collimator Momentum collimation Careful designing of CSNS, in particular chopping in the linac, mean general momentum losses should be very small. However, it is likely that there will be some longitudinal loss, including1loss due to RF trapping inefficient; 2loss due to error conditions; and3loss due to leakage from the betatron collimation. Therefore, the momentum collimation is included. The length of the momentum collimator is about 3.5 m and the copper jaws are 2 112

115 cm long. To prevent the copper jaws from arriving their melting point and decreasing the lifetime of the momentum collimator, we should use water-cooling in the jaws design. According to the simulation results, the momentum collimation efficiency is about 90.6% which meets the needs stripping foil In CSNS, we design two kinds of stripping foils respectively located in the LRBT and the injection area of the ring. At LRBT, we use the foil scraper for scraping the beam halo. We intend to set three scrapers in LRBT, as shown in Fig , in order to control the emittance. Every device has two pairs of scraper head respectively in the vertical and horizontal direction. Fig Location of the stripping foil scraper in LRBT At the injection area, injection stripping foils are designed. We intend to set a primary foil and a secondary foil there, as shown in Fig The primary foil is used for stripping the main injection beam which is injected into the ring; the secondary foil is used for stripping the left particles which are then throw to the beam dump. Fig The stripping foil at the injection area Foil design 113

116 As the carbon has a lot of excellent performance, we choose the carbon-carbon foil for both the two kinds of stripping foil. Comprehensive considering the lifetime and stripping efficiency, we choose the thickness of 120µg/cm 2, 80µg/cm 2 and 10mg/cm 2 for the scraper s foil, primary and secondary injection foil. For the relatively thin foil, we decide to adhere it to frame by epoxy, and supported it with carbon fiber. For the thicker foil (secondary injection foil), we intend to mount it on to the frame by a clamp, as shown in Fig Fig Left: the scraper s foil; Meddle: primary injection foil; Right: secondary injection foil Mechanical structure of the foil scraper device In our design, the distance of the scraper head s movement is about 20 mm and the precision of the movement is expected to be less than 0.1mm. Considering the interference of the motion, the vertical and horizontal scrapers are in different working plans. The scraper head insert into the vacuum chamber through a seal bellow. We adopt a cantilever structure to mount the scraper head to a bracket. For the drive mechanism, we plan to use a trapezoidal screw friction. The two scraper heads at the same direction use opposite rotary direction screw-nuts which connect to the same screw, shown in Fig Fig The structure of the foil scraper Mechanical structure of the primary injection stripping device The structure of the primary injection stripping device is mostly referenced to SNS. We d like to use a chain structure to store the reserved foils and change them. 114

117 The number of reserved foil is expected to be 20. We intend to be hung those foils on the pins which connected to the chain, asshown in Fig The chain structure is installed on a plant which could horizontally move through a rack-gear structure. The distance of the horizontal movement is about 40mm and the precision of the horizontal movement is expected to be less than 0.1mm. The driver sprocket is driven by two pairs of bevel gears and a spline shaft. The precision of the sprocket rotary motion is expected to be less than 0.05rad. Fig The structure of the primary injection stripping device Mechanical structure of the secondary stripping device Comparing with the primary stripping device, the secondary stripping device s structure is much simpler. It is also referenced to SNS. The foil is directly connected on an arm. The arm is driven by a rack-gear structure. The distance of the arm s movement is about 60mm and the precision of the movement is expected to be less than 0.1mm, as shown in Fig Fig The structure of the secondary injection stripping foil Control system design of stripping devices We preliminarily choose step motor as the drive power, and want to use the encoder to establish a closed-loop feedback control system in order to control the movement. After re-equipping the foils every year, the control system needs to be re-calibrated The Girder System of the RCS Magnet The design requirement for the RCS ring magnet girder: dipole magnet is about 115

118 25 ton, quadruple magnet is about 10ton. The magnets have exciting force at 25Hz. So the magnets girders are need of vibration isolation. Dipole magnet and quadruple magnet girder have same structure, method of installation and method of adjustment, as shown in Fig Fig The girder of dipole magnet The Structure and Shape of the Magnet Girder The magnets girder is designed in divided structure, so that the magnet girder has no coupling between horizontal movement and vertical movement. The girder is divided into two-layer. After adjusting the magnet, the movable two-layer are connected and locked by bolt. Because of the vibration and distributed asymmetry of the weight, the girder has less deformation after installing the magnet. The main part of girder is welded into case texture, as shown in Fig The main part of magnets girder is made of structural iron. Fig The main part of girder The Adjustment Mechanism of the Magnet Girder a) Vertical Adjustment Equipment The magnet girder is supported by four worm and gear lifter, as shown in Fig , using stepping motor to drive the lifter for adjusting. According to the calculation, the worm and gear lifter supports 150kN of the driving force to raise the dipole magnet on the girder. The centre of worm wheel is feed screw nut. The supporting axle is made of feed screw. Through a pair of worm and gear transmission, worm is driving by a stepping motor. 116

119 Fig The worm and gear lifter b) Horizontal Adjustment Equipment Girder horizontal adjustment is carried on between up and down layers. The horizontal thrust is composed of force of sliding friction between up and down layers and gravity caused by girder s unevenness. According to the calculation, the dipole of magnet horizontal load is 57kN. Horizontal adjustment equipment uses differential screw gearing, as shown in Fig , which is made of a differential screw and two nuts. In order to get an angle of rotation at the time of adjusting, a rolling bearing and a stationary shaft can be used to fasten every nut and girder. A range of horizontal adjustment is: X direction ±30mm,Y direction ±30mm,the maximal rotation angle:±1.55degree. Fig The differential screw gearing Vibration Isolation and Fastening Magnet RCS magnet is working with 25Hz AC plus DC power supply. Therefore, vibrating frequency of magnet is 25Hz with a range of 0.1mm. Through investigation, vibration isolator must be applied when magnet is fixed, so that the vibration can not be transferred to the ground and then to other devices as much as possible. It is made of steel wire being pressed in mould. The vibration isolator has the capability of radio-resistance and small volume. Magnet fix: On the basis of fixing four vibration isolators which undertake most of the weight of magnet, magnet is fixed by four bolts. The vibration isolator is also used between the girder and blots. It makes the fixing not only has the function of fastening, but also vibration will not influence the girder, as shown in Fig

120 Fig The installation of the girder 14.4 Pillow Seal and Proton Beam Window Pillow Seal In the Target-Instruments Hall (TIH) of CSNS project, the proton beam line will be in a high-radiation area where hands-on maintenance cannot be performed. In the event where a vacuum connection must be broken along the beam line for the purpose of maintenance, it will be impossible to do so manually. Therefore, in collaboration with the Chinese Factory, we have started to study a remote controlled vacuum seal flange (the CSNS Pillow seal). In a pillow seal, the vacuum seal is achieved through the expansion of diaphragms and bellows, and the misalignment and the tilting of the mating flanges are less severe than in the original design. Care must be taken such that the diaphragms of the pillow seals are not damaged during insertion. The diaphragms are therefore designed to be attached at the center of a plug shield, and are contracted to within the thickness of the plug shield before insertion. This ensures that the diaphragms are not damaged as long as the clearance defined by the two mating flanges is large enough for the plug shield to pass through. Vacuum connections, which will be used for the beam ducts of the primary proton beam line and the front-end secondary extraction beam lines in the vicinity before the target, are expected to suffer both from tremendously high radiation and from corrosion induced by NOx which is produced by air activation. Non-radiation hard components such as rubber O rings therefore can not be utilized. Also, it is essential to adopt a vacuum flange which can be operated by remote handling. Although a pillow-seal was originally developed at PSI for this purpose, we have recently started design work for developing a dedicated CSNS pillow seal in order to match the following requirements. 1) An aperture of 250 mm in diameter, 2) It can be expanded more than 20 mm by pneumatic air (70-90 mm in stroke) while remaining parallel to the beam duct flange, 3) A leak rate in the order of 10 7 Pa m 3 /s. Recently a test prototype of CSNS pillow-seal was manufactured successfully and its leak rate was experimentally measured to be Pa m 3 /s. A picture of it is shown in Fig

121 Fig A seal theory experiment prototype of CSNS pillow-seal At present, we are working on related R&D projects including how to improve the leak rate of the pillow seal, how to develop the pillow seal prototype, how to install the pillow seal safely by using a guide rail, how to extend it smoothly by compressed air using a pantograph support mechanism and how to evacuate the groove between the two pillow rings for differential pumping Proton Beam Window The proton beam window with a sound structure is essential, which successfully divides RTBT from the target area. Since the generated beam should not loose significant amount of energy while propagating through the proton beam window, it is necessary to design the proton beam window with a minimal thickness. Fig shows a conceptual diagram of the configuration. Fig A conceptual diagram of the configuration A proton beam deposits substantial energy onto the proton beam window. Therefore, the development of a reliable cooling method to maintain the window temperature below structural damaging point is a key to successful operation of the accelerator. Since water is selected as cooling medium, the design issue is basically solving a convective heat transfer problem for multiple cooling configurations and 119

122 providing sufficient cooling to maintain the structural integrity of the window. At present, computational analyses were preliminarily performed to investigate the thermal hydraulic characteristics of water cooling, and to provide most effective water cooling configuration of the beam window structure. Furthermore, based on the selected cooling method, structural analyses will be performed to examine if the structural integrity can be maintained during several operating conditions. Width and height of PBW is given by the proton beam size. At present, PBW consists of curved surface similar to a cut-away shape of a cylinder. It consists of two plates. Between the gap of plates, H 2 O flows. Fig shows PBW design with curved surface. Fig PBW design with curved surface 120

123 15 Beam Diagnostics 15.1 Beam diagnostics layout TaoGuang Beam diagnostics in Front-end In LEBT,there are three beam instruments. Two beam current transformers (BCT) monitor the beam intensity and the transport efficiency, and one double scanned slit system measures the emittance. The measurement of the emittance at the very early beginning of the whole accelerator is considered important. Fig Beam diagnostics layout in MEBT In MEBT line, there are eight beam position monitors (BPM) installed in the quadrupole magnets, four wire scanners (WS) for beam profile measurement that is important for verifying the beam transverse dynamic design. Three fast current transformers as phase detector will be used to measure the beam energy by the time of flight method. Two current transformers will monitor the beam intensity and the transport efficiency in MEBT. In order to perform the transverse matching to the DTL, a double-slit system for both the horizontal and vertical planes is installed to measure the beam emittance Beam diagnostics in DTL Linac In the DTL cavities, there is no space for installing beam instrumentation devices. Thus, one BCT and one FCT are placed between each junction of the DTL cavities. In addition, twelve beam loss monitors (BLM) are distributed evenly along the DTL linac. For the commissioning of each DTL cavity, a temporary beam line including two FCTs, a double-slit emittance measurement system, two BPMs, a BCT and a WS will be assembled. This beam instrumentation line can supply the measurements of beam energy, emittance, current, position, phase, profile, and so on Beam diagnostics in LRBT 121

124 Fig Beam diagnostics layout in LRBT In LRBT, the basic beam optics has been designed with triplet focusing cell. The distance between triplets is about 12 m. At each triplet, a BPM and a BLM detector are installed. Because the beam energy at LRBT is too high for emittance measurement using a double-slit system, four beam profile monitors are used to measure the emittance. After the bending magnet, a profile monitor can measure the beam energy spread. In front of the beam dumps and the injection foil at RCS, beam profile monitors will be used to match the required beam spot. There is a debuncher cavity to reduce the momentum spread in the LRBT line, thus a wall-current monitor (WCM) is installed to monitor the bunch length Beam diagnostics in RCS Fig Beam diagnostics layout in RCS The beam diagnostics layout in RCS is shown in Fig Corresponding to 122

125 the triplet lattice structure, thirty-two BPMs will be installed in RCS ring for the COD correction. Furthermore, those BPMs share the space with corrector magnets. There are additional three BPMs for the RF feedback system, one BPM for tune measurement and three RF BPMs responding to the linac RF frequency to monitor the painting procedure. There are three FCTs and two WCMs, except one WCM for monitoring the bunch length, others for the feedforward and feedback control of the ring RF system. By taking the reference of the J-PARC IPM (ionization profile monitor), we reserve the same space for two IPMs that will not be built in the first phase. At the injection section, five multi-wire profile monitors will be inserted just for the low current commissioning and the first beam set-up of each run of the RCS Beam diagnostics in RTBT Fig Beam diagnostics layout in RTBT In RTBT, according to the beam dynamic design, one BPM just needs to monitor the beam position in one direction. For the beam diagnostics near the target area in RTBT, three movable profile monitors are used to tune the beam optics at the two step-like field magnets and the target, and an inline wire halo monitor to monitor the beam position at target. Combined with the proton beam window, a thermocouple outside the beam is used to make sure the beam position on the window. A list of the total CSNS beam diagnostics is shown in Table Table Beam diagnostic devices used in the CSNS BCT DCCT BPM BLMFBLM WS Harp/ MWPM EM FCT WCM Tune IPM LEBT 2 1 MEBT DTL LRBT RCS ? 123

126 RTBT Total ? 1. LRBT includes the dump line and the injection dump line. 2. RTBT includes Ring dump line Beam monitors According to the beam characteristics, the beam diagnostic devices are divided into two categories. One includes those used in the FE, DTL and LRBT, or together named as BR (before ring) section. The other includes those used in the RCS and RTBT, or together named as RT (ring and/to target) section. The beam monitors in the BR section deal with beams of small transverse emittance and short bunch length, whereas those in the RT section deal with beams of large transverse emittance and very long bunch length. For the ring commissioning, the minimum peak current for the injection should be 5 ma Beam Current Transformers For observing beams of different characteristics, the BCT is divided into slow current transformers (SCT) and medium current transformers (MCT). In the BR section, SCT observes the macro pulse current that length is about 500 µs, and MCT observes the chopped pulse current which length is about 500 ns. The rise time of SCT is less than 10 µs; the current measurement range is from 5 ma to 50 ma; the measurement accuracy is 1% of full scale. These parameters have been thought about the future need. The rise time of MCT is less than 10 ns. We plan to buy some ACCTs from Bergoz Company as SCT and fabricate some SCT and MCTs by ourselves. As for the MCT, the design of the electronics is considered to a delicate task. In RCS, The function of SCT is similar to DCCT. We are hesitating if the SCT for RCS is indispensable. The MCT is used for observing the progress of multi-turn injection. For the design of the MCT sensors, we will follow the one used in J-PARC RCS. In RTBT, the transverse beam size and the beam current are similar to those in RCS, thus MCT can be used as the current monitor in RTBT. The BCT readout system is designed in a same way and consists of a multiplexer and an oscilloscope. One can control the multiplexer to switch among different BCT waveforms DCCT DCCT observes the circulating beam current in RCS. The whole system will be bought from Bergoz Company. Because the detector is sensitive to the outside magnetic field, a good magnetic shield shell is needed. In order to prevent the magnetic core damaged from the heat induced by beam, a water-cooling system is designed for the DCCT Linac Beam Position Monitors 124

127 In the whole BR section, the bunch structure is similar, thus we choose the same type BPM, the one of shorted-end strip-line type. This type of BPM has a good mechanical structure and just needs four feedthroughs. The measurement accuracy of the strip-line BPM is 1% of aperture radius. The Bergoz LR-BPM module is chosen as its electronics system. The sum signal of a BPM is proportional to the beam current. Thus we need to employ the SUM function of LR-BPM. The ADC of bit for reading out the BPM data has a sampling rate of khz. In order to save space, the BPM shape is designed carefully so that the BPM can be inserted in quadrupole magnets. As the sum of four signals from the four electrodes can provide the information for the bunch phase, the cable length of each electrode should be phase matched, and they will have same electrical length Wire Scanners We adopt the crawling wire scanner system to measure the beam profile in the BR section. For the measurement errors, the WS system can give the beam size in about 10% accuracy. The maximum peak beam current for measurement is 50 ma; the pulse length is µs; the beam repetition rate for measurement is limited to no larger than 1 Hz. The system has three wires to measure the beam profile in X, Y and 45 directions. The heat deposit on wires by the beam can break the signal wire, and should be designed with care. For the low energy beam in the MEBT, we chose carbon wires of 35 µm in thickness, although the mechanical property of carbon Fig Wire scanner wires is bad. For the beam energy larger than 3MeV, the heat deposit is no longer a critical problem, then we use tungsten wires that have better mechanical property Emittance Measurement For the low energy beam in LEBT and MEBT, we chose the scanning double-slit system to measure the emittance. The driving system of the EM is the same as the WS. Because the first sampling slit is made of a metal plate of high melting temperature and cooled by water, the system can work with a higher duty factor than the WS system. But it is still needed to short the beam pulse length when the system is used Fast Current Transformers In the BR section, the bunch length is less than 3 ns, thus the rise time of the 125

128 phase detectors should be about hundred picoseconds. We plan to buy FCTs from Bergoz Company as the linac phase detectors to guarantee their use during the CSNS commissioning. And more, FCT has a merit that its longitudinal dimension is short, and thus can be installed at space-tight locations. In the RT section, the minimum bunch length is about 80 ns; we will follow the J-PARC design to fabricate the FCTs by ourselves. We plan to use the down converter and digital I/Q demodulator technology to extract the phase parameter RCS Beam Position Monitors There are three types of BPMs in RCS. One is for the COD measurement, and its number is 32. They are installed in the corrector magnets for saving space. One BPM of the same type are used for the tune measurement. Those BPMs can provide the beam positions in both the horizontal and vertical directions. The second type of BPMs provides only the beam positions in the horizontal direction for the RF feedback control. All the above BPMs use diagonal cut electrodes. The third type of BPMs is a strip-line type for measuring the bunch structure in 324 MHz. According to J-PARC design, these BPMs can monitor the painting process during beam injection. For the COD BPMs, we have adopted the same design as the ones used in J-PARC. It aims 1% of aperture radius accuracy. They have a good linear response to the beam position. One BPM set consists of one pair of horizontal electrodes and one pair of vertical electrodes. The BPM outside shell is made of Titanium, which is a light material and has low residual radioactivity. In order to reduce eddy current loss, the shell has to be thin (1.5 mm) but with multiple ribs to keep enough strength against pressure. The electrode is made of Ti alloy whose conductivity is small to reduce eddy current. For the electronics of RCS BPM, we want to develop it by ourselves. The system uses 6U VME structure to tackle the complicated signal processing for the turn-by-turn measurements. It has step attenuators and a variable gain amplifier or their combination. Then it follows with or without a filter process. A 14-bit ADC with Msps performs the sampling. At last, the data is processed in an FPGA circuit and sent to IOC through VME bus Tune measurement At beginning, we just focus on coherent Tune measurement. The system consists of an exciter and an individual BPM. The exciter has a pair of stripline electrodes and gives either horizontal or vertical transverse kick to the beam. Using white noise with limited bandwidth, only resonated frequency power excites the beam. Signal from an arbitrary signal generator is amplified by a 1 kw amplifier and transferred to the power divider near the exciter. The opposite phase signals are fed to the two electrodes from the down- to the upstream of the beam. The signal from BPM is fed to a real-time spectrum analyzer and betatron sidebands will appear around the harmonics of the revolution frequency. 126

129 Beam Loss Monitors We will use three types of BLMs. One is the ionization chamber, the second is FBLM (fast BLM), and the last is a neutron detector. We use BLMs for machine commissioning and machine protecting (MPS). The ionization chamber system is the main BLMs and will be connected to the MPS. Different system has different response time. For ion chamber, FBLM, and neutron detector, the response time is about 10 µs, 1 ns, and 50 µs, respectively. When beam energy is very low, the beam loss situation can be easily detected by the neutron detector system. So the neutron detector system will be installed just in the linac part with the beam energy less than 80 MeV. We choose BF 3 as the detector working gas. In the future, the gas can be changed to 3 He. The electronics of ion chamber system, we plan to follow the SNS design. For FBLM, we will use an oscilloscope of 500 MHz in bandwidth to observe the output from the photomultiplier Wall Current Monitors WCM is used to measure the bunch length. For the different beams it deals with, in the BR section, the WCM needs wider bandwidth but the beam current is relatively low; in RCS, the bandwidth of the detector can be narrower, but the dynamic range of bunch intensity is quite large. In the BR section, the bunch length is about 1 ns, so the bandwidth of the WCM should be about 1 GHz. We will adopt the WCM type developed by Fermi lab, see Fig Various ferrite cores are used to force the image current through the resistive gap rather than allowing it to flow through other conducting paths. The voltage across the gap is Fig The inside of Fermi WCM measured at four equally spaced positions and combined with a shunt resistive combiner to eliminate the effect of beam position. In RCS, we will choose the ferrite core that can endure higher bunch current Harps The harp or MWPM (Multi Wire Profile Monitor) is usually installed near beam dumps and beam windows. It is also used in the areas where the beam profile measurement should be fast but can be less accurate. In the initial RCS commissioning, it is important to perform a good transverse matching at the stripping foil, so Harps have been selected for the beam profile monitoring at the injection region. If the space is enough, we will adopt the SNS design where both signal wires and 127

130 shield wires are designed. If the longitudinal space is short, we will adopt the J-PARC where only signal wires are present but the distance between the signal wires is slightly larger. To increase the signal voltage for proton beam at injection area, the harp wires at the injection are replaced by a Ti foil of 1.0 mm in width and 10 µm in thickness. As for the harp near the proton beam window, it needs remote handling. We plan to follow the SNS design. This is described in detail in the mechanical design part of this report Ionization Profile Monitors Although IPM is very useful in commissioning RCS, the components such as magnet, MCP, and EGA are relatively expensive, and the technical issues look to be relatively more complicated than other diagnostic devices. Therefore, we decided to reserve the space for adding them in the future. For the IPM design, the one used at J-PARC is a good example. For example, the magnetic field is about G, and the electric field is about kv/m R&D studies and summary Taking the advantage of carrying out beam halo formation studies in an ADS RFQ accelerator, which is supported by a different resource, we are carrying out several R&D items on the CSNS beam diagnostic devices, especially those for the use in the linac Prototype of Linac BPM We have already made two BPM prototypes. The diameter is 75 mm. One of them has separators between the electrodes. Poisson and CST MicroStudio codes have been used in the design. The BPM transverse size is calculated following the wideband match mode. Fig Mapping of the BPM The sensors mapping is measured by using antenna method. The Bergoz LR-BPM module was bought as the BPM electronics. The output is observed using an oscilloscope. 128

131 Fig the BPM output signal Those two BPMs were tested using the ADS RFQ beam. However, we have met a problem of the RF noise. As shown in Figure , the beam signal is submerged by the RF noise. Further effort is needed to solve the problem Prototyping BLM Following the SNS ion chamber design, we made three ion chambers. The third one can work at a higher voltage of 2.8 kv. Until the radiation dose of rad/h, as shown in Figure , it has a linear response. Its sensitivity is about 12 pa/rad/h, with the working gas composed of N 2 and Ar with 1:1 ratio. If we increase the portion of Ar, the maximum sustainable voltage decreases. We will try to increase the portion of Ar but with a purer Ar gas Fig Ionization chamber response and maintain a high working voltage with radioactive dose at the same time, and hope to improve the sensitivity in this way Prototyping FBLM An FBLM composed of a plastic scintillator and a photomultiplier has been finished. The signal is observed using an oscilloscope of 1 GHz in bandwidth. The prototype was also tested with the RFQ beam. The system is very sensitive, but it detects also the cosmic ray. Thus we need to find a method to eliminate the influence of cosmic ray. On the other hand, we plan to develop a liquid scintillator as FBLM for it can sustain longer time under the hard radiation condition than a plastic scintillator Electronics development for bunch phase measurement By collaborating with Fast Electronics Laboratory of University of Science Technology of China, we have developed a set of phase measurement electronics associated with a FCT. 129

132 The system has four input channels. It can provide phase difference between arbitrary two channels. Then with the measured bunch phases we can obtain the beam energy using the TOF method. The laboratory test of the electronics has been finished. The system can work well for an input signal varying from 50 dbm to 1.5 dbm. The resolution is better than 0.05 degree. However, we found that the cable assembly can affect the measurement result. The test with the RFQ beam will start soon Prototyping WS We are constructing a wire scanner with halo scraper. The driving system, wire fixing method and signal method are all the same as the CSNS WS design. The sensor is driven through a stepper motor. The motor controller has been chosen to be OMS VX-2 as it can communicate with the EPICS. Acme thread lead screw is chosen for it owns self-locking. This helps eliminating the brake noise to the electronics. We use an optical grating ruler of 0.01 mm in accuracy as the position sensor. Until now, we can control the stepper motor through the EPICS system. As mentioned before, for the measurement of the proton beam with the energy below 3 MeV, carbon wires are used to reduce heat deposit. We Fig Prototype of solder the carbon wires to copper bars, and use wire fixed structure springs to form the required tension on the wire. The wire signal electronics converts the current of the signal to voltage to eliminate noises, similar to other diagnostic devices RCS BPM prototyping For the RCS BPM, the electronics is considered the key technology. According to the RCS commissioning plan, the BPM output will have a large dynamic range from the minimum voltage of 5.4 mv corresponding to the initial injection commissioning to the maximum voltage of 50 V in the normal operation. We have designed a test circuit based on FPGA technique. In the FPGA, we can use the FFT processing to each bunch response, so we want to use /Σ in frequency domain to get the position data. Another aim is to store the turn-by-turn position data with a period of 30 s in the electronics memory using FIFO mode. We can download those data by off-line method. Because we worry about the uniformity of impedance matching transformer, we did not install the transformer between the sensor and the electronics. Through the studies of about one year, now the electronics can measure the minimum level is 10 mv, but more effort is needed to eliminate or reduce the noise that is about hundred mv Manpower 130

133 By now the beam diagnostics group have six staff members, nine students, and three short-term engineers. We plan to add five more persons as the staff per year during the period of from 2010 to The problem is that the fresh persons are absent from experience. They need be trained during the R&D studies. Therefore, we will look for some experienced people in part time mode to give us powerful help. 131

134 16 Control System Chunhong The overall task of the CSNS control system is to collect all the valuable information of the devices and the beams from the accelerator and the target as well as the conventional utilities, and control the working parameters of the devices. According to the requirement of accelerator control, the operator in the central control room can control and monitor all the devices distributed in the accelerator as so to accomplish H- beam generation and beam acceleration in the linac and beam transport in LRBT, beam injection into RCS, beam accumulation and acceleration in RCS, and beam extraction from RCS and transport to the target where neutrons are generated for experiment researches. As a high current intensity proton accelerator, the beam loss must be strictly controlled so that the residual radiation will not pose serious problems for the hands-on maintenance. Thus, the control system should have high reliability and real time response, not only provide friendly man-machine operation interface and have good ability of the information processing, but also have a fast device protection system. Figure CSNS Control Scope 16.1 Control Scope According to the physical distribution of the facility, the control system consists 132

135 of the consoles, computers/servers, network, database, physics application software, device control system, machine protection system and timing system as shown in the Figure The accelerator control system will not include the control or data acquisition for the target and the experimental stations. It does not include the personnel protection system, neither, which is a separate system but will be monitored in the central control room. According to the types of the devices, the device control system is divided into the ion source control, linac LLRF control, magnet power supply control, RCS LLRF control, vacuum control, injection/extraction control, beam diagnostic control and collimator control. Here, the linac LLRF and the RCS LLRF control will be accomplished by the two RF groups, the control system will be responsible for high level interface. The task of the control system is to control and monitor thousands of devices covering the linac, LRBT, RCS and RTBT. The detailed devices include: (1) Approximately 250 magnet power supplies in linac, LRBT, RCS and RTBT. The control precision and stability are (2) About 140 ion pump power supplies, 20 gauges and 30 vacuum valves. The statuses of the valve and gauge need to be monitored. The valves need to be controlled and interlocked with the vacuum gauge. (3) 2 sets of injection pulsed power supplies and 7 sets of extraction power supplies (4) For the 8 RCS RF stations, the cavity pressure, the RF phase and the RF frequency need to be controlled and monitored. (5) Over 300 beam diagnostic devices need to be monitored, and some movable devices need to be controlled. Some key parameters such as the beam current and profile, BPM, emittance, beam loss need to be displayed on the consoles. (6) The H- Ion source, linac RF power sources and water cooling need to be controlled. (7) Timing system will provide trigger signals to the ion source, linac/rcs LLRF, magnet power supplies, injection/extraction pulsed power supplies, target rotors and beam diagnostic devices. (8) MPS is responsible for the safety interlock of devices and displaying the running statuses of the protected devices at the central control room. The fast protection system is responsible for monitoring the beam loss of the beam line. Once the dose rate by the beam loss is over the preset threshold, the fast protection system is responsible for turning off the beam in the front-end or damping the beam into the beam dump. (9) The accelerator control system should be able to exchange information with those for the target and the conventional utilities, and display the relevant information in the central control room. In addition, the operation status of the CSNS project will be displayed through the web for people viewing it outside the network. Moreover, the control system will provide a friendly man-machine interface and a device alarm handler system for the operators to monitor the devices. It also will provide high lever application platform for the beam commissioning. In addition, considering the front-end devices in a strong magnetic field, the control system will adopt isolation, shield and grounding technologies against the 133

136 noise from the devices. Considering the long-term stable operation of the control system, the control system should have high reliability and availability, and should be upgradable, expandable and easily maintainable Design philosophy and construction management The design philosophy of the control system is to adopt a distributed architecture, to use EPICS toolkits and standardized software and matured hardware. The hardware selection should consider both the price and the performance. The control system will establish a good collaboration with the control groups of the other labs both in the world and in China for technical support and information exchanges. There will be a quality control and software project management during the development and construction of the control system. It will be proceeded in the following phases: user requirement investigation and analysis, conceptual design, feasibility analysis, R&D prototype, preliminary design, system development, system testing off-line and on-line, system installation and integration, commissioning and operation. At the beginning and the end of each phase above, the control system need to ask advices and comments from the responsible persons of the controlled devices so that some problems can be solved in time to make sure the quality of the system and the development on time. When the key phase ends, the control system should have a review meeting for the testing results and set up the documentation on the developed hardware and software to make sure the quality of the system Feasibility study and R&D The conceptual design of the CSNS control system started in September of It took about four years for the system to go over investigation, user requirement investigation and analysis, conceptual design, feasibility analysis, R&D prototype and preliminary technical design. From September of 2005 to March of 2007, the control system investigated the status of the control systems for proton accelerators in the world including the system architecture, system integration tool, front-end control technology, software development platform and high level application software. Then, we studied some new control technologies such as RTEMS and a high level application software XAL used at SNS. At the same time, we accomplished the market investigation of the selected hardware. After that, we accomplished the conceptual design report and the budget estimation. The conceptual design was reviewed in the CSNS international review meeting held in May of Since May of 2007, the control system has started feasibility studies and R&D prototype with a very limited budget. There were totally five prototypes executed: a WE7000 for the control of a pulsed power supply for the injection, an ion source control system using a PLC, a fast waveform monitor for the ring magnet power supplies, an IOC using embedded ColdFire for the digital power supply control module, and an EVG/EVR for timing system. During this period, we carried out again 134

137 the user requirement investigation and obtained the new specifications of the controlled devices following the design evolution and accomplished the feasibility study report and budget estimate. The goal of the R&D and prototyping is to solve the key technologies of the system construction and select hardware including hardware interface for the quantity ordering during the system construction. Thus, the R&D and prototyping are very important for the development and construction of the control system. The results of the R&D and prototyping will affect the design scheme and the construction plan. So far, the WE7000 has been tested with the injection PS and it is proven that it could meet the requirement. The item on the waveform monitor and verifier was found to be difficult to continue, and was stopped. The ion source control system is being on-site integrated and tested now. It turns out that the strong noise from the high voltage platform is a serious problem and needs to be solved. The prototype using ColdFire µc5282 was set up on an Altera FPGA evaluation board with µc5282 development Kit in May of The goal of the prototype is to put the ColdFire chip onto a board which is called digital power supply control module (DPSCM). The DPSCM has been developing by the power supply group since This R&D item is going slow, still under way. At the same time, we will prepare a backup scheme for the interface to the DPSCM System Architecture Since 1990 s, distributed processing system has been widely adopted in the world. Thus, the CSNS control system will be also a distributed system which will adopt the standard two-layer architecture with PC/Linux workstations as the client and VME IOC/PC IOC/Embedded IOC as the server. There will be no field bus to a third layer, but extensive use will be made of a field bus or serial interfaces from the lower layer, to the equipment. At the early development stage, EPICS and VxWorks will be used for EPICS development environment. The new version VxWorks 6.5 is under consideration. Early IOCs will use MVME5100(PPC604) already applied in the BEPCII control system. The feasibility research of MVME5500/MVME6100 and Coldire µc5282 running RTEMS has started since September of The architecture of the control system is shown in Figure

138 Figure The architecture of the control system 16.5 Equipment Interface The interface from the control system to the equipment will be through VME Power PC processors (Input-Output Controllers, or IOCs). For technical guarantee, IOCs will use MVME5100(PPC604) that were already applied in the BEPCII control system. However, the feasibility study of MVME5500/MVME6100 running on RTEMS has been started since IOCs will be installed for some subsystems. The preferred interfaces to the equipment will be analogue, digital and serial (RS232, RS485 etc). There will be many PLC-based subsystems for the systems such as the vacuum, the equipment interlock and the conventional facilities. CSNS intends to use a same PLC manufacture for all those subsystems. Mainly for the reason of cost, YOKOGAWA PLCs will be adopted. Soft IOC (PC-based) will be used for interfacing to the PLCs via the Ethernet/IP. The new type of YOKOGAWA PLCs F3RP62-2L/L1 with EPICS IOC will be used in Vacuum control and monitor. ZTEC scopes with a built-in EPICS will be used for waveform acquisition and display of the extraction PS and the beam diagnostic. EVG/EVR timing system already applied in the BEPCII control system will provide 25Hz synchronous operation and accurate time stamping of data Front-end control Fron-end includes the H- ion source and the LEBT. However, the ion source control system is typically a single component of the CSNS control system. It includes 5 parts: power supplies, vacuum, temperature measurement, water cooling and timing. All are required to be controlled locally in the tunnel and remotely at the central control room. The power supplies will work in the high voltage environment. So, they require the control system with high reliability and availability. The prototype of H- ion source control system has been built up and put into 136

139 commissioning at lab as shown in the Figures and APC/Linux running soft IOC exchanges the data with PLC/CPU via Ethernet. The PLC/CPU communicates with the PLC/IO (all inputs and outputs) modules through the FA-Bus optical fiber. Thus, there is no need for extra electrical isolation between the ground and high voltage platform. Figure Control structure of H- ion source Figure OPI of the ion source control system Some experience and lessons have been learnt by prototyping of the ion source control system, for example: Ground isolation, temperature modulation (from relay switching on/off to PID control), more detailed specifications of device interface, processing control of high voltage ramping and so on. In the next step, the embedded PLC/IOC will be used to replace the combination of the soft IOC and the PLC/CPU Power supply control There will be about 250 various magnet power supplies distributed in the linac, LRBT, RCS and RTBT. Some magnet power supplies are DC power supplies, the others are pulsed or dynamic AC power supplies. Although these PSs are different, the control interface to the different PS will be standardized. Since the ring is a Rapid Cycling Synchrotron (RCS), the power supplies for the main dipoles and quadrupoles will provide sinuous waveform outputs of 25 Hz to the magnets. The power supplies of RCS are divided into three families. The requirements (except timing) from the physics to the controls system are not strict as following. For the power supplies of main dipole/quadrupole magnet, the requirements are as following: 1) Setting AC and DC amplitude values remotely; 2) Monitoring remotely AC/DC amplitude read back values with updating rate of 1-2 second; 3) Collecting the amplitude values and phases of the AC and store AC sinuous waveform within 5-10 minutes (10 khz), and checking or comparing the output waveform from 9 sets AC powers with a sampling rate of 10 khz. 137

140 When a fault occurs and a stop command issues, waveform acquisition will be stopped and previous waveform data within 5 or 10 minuets will be kept. 4) Collecting DC amplitude value and save data within 5 or 10 minutes with a sampling rate of 1 khz. When a fault occurs and a stop command issues, data acquisition will be stopped and previous DC amplitude data within 5 or 10 minuets will be kept. For the power supplies of the correction quadrupole magnets and correction bending magnets, they are programmable and required to do table ramping up during the beam rising half cycle (20ms) with steps of 1 ms and do ramping down to the initialization. The requirements are as following: 1) The ramping table can be downloaded from the control system. 2) Collecting the current values with the triggering time within 5-10 minutes, and checking/comparing the current historic data with the triggering time remotely (1 khz). When a fault occurs and a stop command issues, previous data with 5-10 minutes will be kept. For the power supplies of sextupole magnets (DC power supplies) the requirements are as following: 1) Setting the current and monitoring the current readback with an updating rate of 1-2 seconds. 2) All the sextupole magnet PSs can perform slow ramping when changing operation modes. A Digital Power Supply Control Module (DPSM) has been developed by the Power Supply Group since The DPSCM is an intelligent controller with a serial port (RS232) and a so-called Ethernet port. The DPSCM is a PWM module which is based on Altera FPGA with an embedded µc/noisii RTOS. It can accomplish the control function of a PS. The size is 3U board. It has three evolutional versions. The first version is with a serial port (RS232) and a so-called Ethernet port (Modbus/TCP). From the control side, we tested the Ethernet port and found out no any Modbus/TCP protocol support on the board. The version 2 is only with a serial port (RS232). The version 3 is under going with imbedded IOC to be explained in some details in the next. Figure Interface structure of PS controller Figure Design flow of FPGA and IOC 138

141 A prototype using ColdFire µc5282 was set up on Altera FPGA evaluation board with µc5282 development Kit in May of Following the manual how to create a simple ColdFire and Alteral FPGA IOC by W. Eric Norum, this prototype took almost one and a half year to set up an example IOC on ColdFire µc5282 and can read 8 LED lights of FPGA built in the development evaluation board. Figure is interface structure of PS controller. Figure is the design flow of FPGA and IOC, which makes FPGA and EPICS developed in parallel mode. The goal of this prototype is to put the ColdFire µc5282 chip onto the DPSCM board. So, the version 3 of the DPSCM was built with a serial port (RS232) and a 64 pin connector to connect ColdFire µc5282 in October of Currently, this prototype is still in the development Vacuum control Figure Architecture of Vacuum Control and Monitor and Interlock Vacuum control includes about 140 ion-pump power supplies and about 20 gauges and 30 gate valves. The preferred interface will be using YOKOGAWA PLCs for the ion-pumps, the ON/OFF control and status monitoring of gate valves and as well as the interlock. The Digi serial port server will be chosen for the gauge monitoring via RS232/RS Injection/Extraction control The RCS injection system is to inject the H- coming from linac into RCS by stripping method. It includes 8 painting bumper magnets, 4 of which are used for horizontal and others are used for vertical. Each kind of magnets works in series, which requires two power supplies, BHPS and BVPS. These power supplies output pulsed currents to the bump magnets. The output 139

142 waveforms of the currents are required to be programmable and can be modified easily according the theoretical design and the commissioning experience. The YOGOKAWA WE7000 measurement system is used to generate arbitrary waveform to the power supply through a module WE7121 and read the waveform from the power supply as the feedback through a module WE7116. The prototype demonstrate that it can accomplish effective control of the pulsed power supplies. WE7121 is a function generator module of 10 MHz with DDS (direct digital synthesis). The different waveform data can be stored in the memory in advance and sent to the PS through the built-in D/A converter. This waveform can also be read out using a clock signal with a selectable frequency. Then, the waveform is generated by passing the data through the D/A converter. The selectable waveform includes sinuous, square, ramp, triangular, and arbitrary waveforms. WE7116 is a digitizer module of 2-ch and 20 MS/s with a built-in A/D converter. This module can measure non-isolated voltage signals with respect to the ground potential. The digital data from the A/D converter cab stored in memory on board according to the sampling interval and the trigger condition settings. The testing result of the control system using WE7000 with a injection painting bump magnet pulse power supply is shown Figures and Figure is a user interface of WE7121 module, and Figure is a user interface of WE7116 module. Figure User interface of WE

143 Figure User interface of WE7116 module The extraction system is to extract the proton beam of 1.6 GeV from the RCS by a group of kickers that kick the beam vertically in the extraction Lambertson magnet in the single turn mode. Then it needs 7 kicker magnets and 7 pulsed power supplies that uses PFN network for discharging. The repetition rate of extraction is 25 Hz, and two bunches will be extracted in every cycles. The control system is requested to control and monitor the PFN charge voltages and monitor the pulse currents. The ZTEC oscilloscope (1GS/S, 300MHz, 12 bits, 2 Ch) with a built-in EPICS system is selected to obtain the charge voltages and pulse currents. The ON/OFF control of the high voltage charge power supplies will be implemented by YOKOGAWA PLC. The oscilloscope and PLC communicate with front-end IOC through Ethernet. The architecture of extraction kicker pulse power supply control system is shown in the figure

144 PC Linux IOC Etherne PLC V oscilloscopes I oscilloscopes High voltage PFN Kicker Analog Set Analog ReadBack Kicker power supply Kicker magnet Figure Extraction kicker pulse power supply control system RF control system The Linac LLRF and RCS LLRF control will be accomplished by the two RF groups, the control system will be responsible for high level interface. TheLinac LLRF consists of a VXI crate and customized I/O module and a PC which comminicate through RS422. The Control System will use Portable Channel Access Server(PCAS)to communicate with the API of the LLRF The RCS LLRF will be implemented using customized CPCI I/O modules by RCS RF group. The Control system needs to develop EPICS driver and database Beam diagnostics control The beam diagnosis system will include more than 300 beam diagnostic devices. The requirements of the beam diagnostic system are: (1) the beam detectors and the other relevant devices can be controlled and monitored remotely and locally. (2) the beam parameters can be displayed in 3-D or 2-D image, and the data can be saved into the historical database. There are two methods of data acquisition. One is wiring-scan using sampling card in sampling-hold mode. Another is waveform sampling. For the waveform sampling, it is impossible to pick up all channel signals to the IOC database at the same time since the data size is too big. If needed, any channel s waveform can be measured. The waveform sampling has two methods. One is using waveform sampling card based on VME bus. Another is using oscilloscopes with the Ethernet port to measure waveform. The ZTEC s oscilloscope with a built-in EPICS system will be chosen for the waveform measurement. Table shows the different 142

145 types of data acquisition for the beam diagnostics. Table Different type of data acquisition for the beam diagnostic devices SCT 15 EM 4 BPM_box 72 MCT 10 WS 28 FCT 10 WCM 4 Harp 2 Foil Video 1 FBLM 20 BPM_Strip 39 DCCT 1 Tune 2 BLM 169 Fcup 1 Beam measurement system as shown in Figure will be made up of the front sensors, the electronics process circuit and the interface to a PC. The beam position measurement in RCS has different modes: first turn, bunch by bunch and turn by turn. It will adopt a high-speed A/D and DSP technology or digital modulation technology to send out the beam signal to a VME IOC. The other beam parameters will be measured using an oscilloscope and frequency spectrograph connected to a local PC. EPICS/PCAS software will be installed on the PC and used to share the data through the Ethernet. Figure Beam measurements system MPS The CSNS Machine Protection System (MPS) must protect beam tubes and the devices including the magnets and magnet power supplies, the vacuum devices and the RF system from damage, when an abnormal beam status or device status happens. It is also important to minimize the radiation dose produced by the abnormal beam for hands-on maintenance. It will also cut off the beam when the beam-on-target parameters drift outside the specifications. The MPS diagram is shown in Figure

146 Figure MPS Architecture The MPS will be designed to allow a carrier signal to propagate to the front-end equipment when the machine is in a safe state. There are two machine protection networks: the auto reset network and the latched network. Each of the networks connects to the machine protection hardware consisting of input interlock signals and hardware logic that transmits a carrier signal or inhibits the signal from being transmitted in a fault condition. The lack of a carrier indicates a fault condition that will inhibit the ion source from producing beam. Once RCS starts acceleration after accepting beam from the linac, there will be no way to dump the beam except completing RCS cycle. Then the beam will be extracted from RCS and dumped into the beam dump. MPS will use Yokogawa PLCs (FA-M3) for the input interlock signals and output off signals to the devices. The device local interlock will be done within the devices and send the interlock signals to MPS. MPS will have several PLC stations in the linac, RCS, the target station to collect the interlock signals and propagate the state signals through the FLNet to the Master PLC. The FLNet can accomplish data scanning within 200 µs. The program flow chart is depicted in Figure

147 Figure Program flow chart Timing The Timing system will use EVG/EVG which has been successfully applied in the BEPCII. It will provide t the trigger signals to the Ion source, Linac/RCS LLRF, magnet power supplies, injection/extraction pulsed power supplies, target rotors and beam diagnostic instruments. It will also provide 1MHz Clock synchrotron to DTL-Q RCS-B RCS-Q RCS- Trim and a 324MHz Clock to the Linac RF as well as accurate time stamping of data. The structure of the Timing system is as shown in the following figure. Figure The Timing system 145

148 16.6 Central computers The central computer system is a large and complex infrastructure consisting of a lot of servers, computers, disk arrays, client terminals and so on. They will provide development, applications, relational database, network management, data archiver and viewer, alarm management, error logging, logging services and IOC booting. Thus, the central computers must be reliable. At the early development stage, we selected an IBM Blade server running Linux (12P/48C CPU, 48GB memory, 12T disk storage) as shown in the following table as the physics application server and an EPICS server. Table BladeCenter Blade Sever Hardware summary Processor Two Intel Xeon 5500 series processors, 2.93 GHz Memory 8 GB of memory with 12 VLP DDR-3 memory DIMMs Network Broadcom 5709S onboard NIC with dual Gigabit Ethernet ports with TOE Adapter Hard Disk 2*300G 2.5" SAS hard disk drive,raid-0-1 Table BladeCenter blade server chassis summary six blade server on the bays 12 SAS (12.0 TB) hard disk drive Hot-swap and redundant switch modules supporting SAS, Gigabit Ethernet, and Fibre Channel options four hot-swap and redundant load-balancing power supply modules An IBM PC server and IBM storage will be used for archiving server and Oracle server as shown in the table Table IBM System x3850 M2 Server overview processors Memory internal storage Network interface 8*Intel Xeon Series 7400 up to 2.66 GHz (6 core)/1066 MHz front-side bus 32 GB PC DDR II 2*146GB 2.5" Serial Attached SCSI (SAS) Integrated dual Gigabit Ethernet w/tcp-ip off-load engine Table IBM System Storage DS4700 Express Host interface 8 host ports model 72, 4 host ports model 70 Fibre Channel (FC) Switched and FC Arbitrated Loop (FC-AL) standard, Auto-sensing 1 Gbps/2 Gbps/4 Gbps Drive interface 4 drive ports Fibre Channel (FC) Switched and FC Arbitrated Loop (FC-AL) standard, Auto-sensing 1Gbps/2 Gbps/4 Gbps Supported drives 16*1 TB 4 Gbps SATA E-DDM Disk drives RAID levels 0, 1, 3, 5, 6, 10 A lot of PCs running MS Windows will be used for client terminals. The LED 146

149 monitor will be used for display running status of the key systems including the target and the conventional facilities Database VDCT will be used for IOC DB configuration during the device control testing since it s convenient for developers to create IOC DB. The RDB is planned to manage conventional control system information together with the control system configuration. The RDB will store equipment information (magnet measurement and survey/alignment), device configuration parameters (Channel/device name, constants, calibration, I/O address, etc), the state (all settings) of the machine and the historical real-time data. All IOC DBs will be generated out of the RDB so ensuring that there is a central repository for all control parameters. CSNS will choose Oracle as RDB. The Oracle DB design will refer to SNS global database schemes. The CSNS RDB schemes are currently being defined Application Software Most CSNS device application requirements can be met through the standard EPICS tools, control panels through EDM, alarm management through Alarm Handler, archiving through Channel Archiver together with Oracle database. Since web browsers have become an easy way to view and manipulate the control data, the CSNS control group is also planning to developed web-based control panel. Considering the successful application of XAL package at SNS, CSNS also plans to adopt this high level application framework for the physics application. Recently, the EPICS community is planning to merge XAL into the Eclipse framework. Eclipse framework is also used for future EPICS version called Control System Studio (CSS). The CSNS control team is undertaking to separate SNS specific code from the present XAL release and develop the specific beam line devices at CSNS by extending XAL devices on the Eclipse framework. Since RCS is different from the accumulator ring at SNS, the online model for the RCS should be developed. As the first step to establish XAL for the CSNS accelerator, we have created the database framework of the CSNS accelerator by using standardized rules and interfaces, and imported most of the equipments data of the CSNS linac accelerator and two beam transfer lines. This provides a data platform for testing the XAL accelerator physics application before the construction of the accelerator. Secondly, we have developed more generic software to replace the original interface software from database to XAL, so that XAL can be applied to CSNS or other accelerators more easily. Thirdly, we have generated the XAL initialization configuration file by using the interface software automatically. Most of the XAL software written for the SNS accelerator can be introduced into the CSNS project, and they also support the development of the CSNS virtual accelerator. The CSNS virtual accelerator is an important part of the high level application. We have successfully developed the framework of the CSNS virtual accelerator by having improved the functions and appearance of the SNS virtual accelerator. The 147

150 CSNS virtual accelerator now can simulate the work conditions or status of the equipments in the linac and the two beam transport lines, and communicate with clients by the EPICS channel access. Thus, it achieves the purposes of debugging both the physics application software and the device application software. In the future, it can be easily expanded to include all the CSNS accelerator sections Network A preliminary design of the control network is based on 100 Mbit switched Ethernet with a Gigabit switched Ethernet backbone. The network infrastructure will connect the control system computer room to each of the local control and instrumentation areas with single- and multi-mode fibers. There will be a firewall between the control network and the campus network. The control network will use a central core switch in the central control room and edge switches at each local control and instrumentation areas. The control network will be separated into several subnets. EPICS CA gateway will be used for the different IOC PV access and the effective management of traffic and security. Figure The topology of CSNS control system network Consoles and Infrastructure The consoles will be workstations and for the cost reason they are likely to be PCs running Linux. There is also a requirement to provide some PCs running windows for Win32-based applications. All the systems the accelerator, the target station and the conventional facilities - will be operated and monitored from the central control room, although there will be local control rooms available for commissioning and troubleshooting. According to the equipment locations, there are one central control room and 148

151 four local control stations near the controlled equipments as shown in Figure The central control room has an area of about 500 m 2, which is on the second floor in the facility building. The functions of the central control room are to operate the accelerator and display the status of the machine including the parameters of the target station. The central control room will be divided into several separate areas to satisfy different demands and ensure the operators working in a good environment far away from the noise as shown in Figure Figure The distribution of the central control room and local stations There are some of facilities demanded for the control room and local stations: Air condition and ventilation facilities. Electrical power supply including insulation transformer and UPS. Grounding system (independent grounding for central control room). Frames and grooves for cables. Racks and its equipped devices, such as server and PCs. Consoles and display wall Safety and fire alarm system 149

152 Figure Central Control Room Layout (the size in meter) 150

153 17 CSNS Radiation Safety Design 17.1 Outline of CSNS Qingbin The China Spallation Neutron Source (CSNS) will be constructed in the Dongguan city, GuangDong Province for the material structure researches. The CSNS accelerators consist of a 68m long a high-energy (130MeV) linear accelerator, an Rapid Cycling Synchrotron (RCS) with a circumference of 231m, and a low energy transport line (LRBT) in connection with the Linac and RCS, the proton beam extracted from RCS through high energy transport line (RTBT) at final will hit tungsten target at target station. The neutron beam produced from target station will be thermalized by the moderator, and then will be extracted to three general Spectrometers, which will be installed in the neutron scatter hall Shielding Design Objectives The radiation shielding design for CSNS is based on the standard of the People s Republic of China GB [1] and GB5172 [2], and also refers to ICRP publication 60 [3]. Conform with these standards, the effective dose limit for occupational exposure should be less than 50mSv per year. However, in applying the ALARA ( as low as reasonably achievable ) philosophy, the design goal is to maintain exposures well below this limit. Especially, the shielding design objectives used in the shielding calculations of the CSNS are listed as follows: 1) the occupational exposure design objective of dose rate in non-controlled areas on accessible outside surfaces of the shield is 2.5µSv/hour at normal operation. 2) Shielding considerations with respect to general public, the annual dose equivalent contributions to an individual at the site boundary should be less than 0.1mSv. 3) An additional consideration used in the Hand-on maintenance: residual dose rate of 1mSv at 30 cm from the component surface, after 100 day irradiation at 4 hours after shutdown. 4) In the computation of shielding thickness, the safety factor 2 should be considered. In particular, for the sake of convenience, a radiation area is classified into several areas as follows: 1) Radiation monitored area: registered radiation workers can enter freely any time. Such as Linac, RCS facilities hall, and the surface of concrete shielding. 2) Radiation controlled area: access to this area is limited and permissions and the entering procedures are required. Such as Linac second tunnel, RCS second tunnel. 3) Forbidden area, access to this area is in principle forbidden. Such as Linac tunnel, RCS tunnel, area inside target station shielding, and hot cell located at the target station. Table Design value for radiation shielding Area Design Value Chinese Law Site Boundary 0.1mSv/year 1mSv/year 151

154 User Area(General Area) 0.5 µ Sv/hour 1mSv/year(2000hour) Monitored Area 2.5 µ Sv/hour 50mSv/year controlled Area 25 µ Sv/hour Forbidden Area >25 µ Sv/hour Ground Water Activation 5mSv/hour 11mSv/hour Since beam line tunnels are going to be constructed underground, the shielding design value is based on ground water activation and the soil activation near shielding walls. For CSNS, 5mSv/h is adopted in the shielding design Shielding Calculation Methods In the high energy proton accelerator, during protons are accelerated, transported and stored, proton loss will occur. The high energy neutrons will be produced through nuclear reactions. To calculate the dose outside the shielding of CSNS, The Tesch s equation and Moyer s model are used for the bulk shielding design below and above 1GeV proton energy, respectively. Parameters determined experimentally by KEK are employed for the Moyer s model. The DUCT-III code is going to be used for duct streaming calculations. The accuracy of the code has been examined by various benchmark experiments and the detailed calculations using Monte Carlo simulation. In case of detailed evaluation for bulk shielding and duct streaming, Monte Carlo codes, FLUKA and MCNPX are used. The Stapleton s equation was used for Skyshine calculation. Its accuracy has been examined by comparing with Monte Carlo calculations Radiation safety design for CSNS (1) Safety design for the bulk shielding Since beam line tunnels are going to be constructed underground, the shielding design value is based on ground water activation and the soil activation near the shielding walls. For the CSNS, 5mSv/h is adopted in the lateral and bottom shielding design. For the overhead shielding, 2.5µSv/hour is adopted in the overhead shielding surface. According to the operation parameters and the beam loss parameters of CSNS shown from table We calculated the maximum dose equivalent rate at the top shielding surface by Tesch s equation and Moyer s model. As can be seen from table , the results show that the annual dose equivalent will be within the guideline without adding any shielding. Table Assumed beam loss for the Linac Position Energy Average Beam power Annual operation beam current loss time [ MeV ] [ A ] [ W/m, W ] [h] LEBT RFQ W

155 MEBT-chopper W 6000 DTL W/m 6000 DTL W/m 6000 DTL W/m 6000 DTL W/m 6000 DTL W/m 6000 DTL W/m 6000 DTL W/m 6000 LRBT W/m 6000 LRBT collimator W 6000 Injection region W 6000 RCS W/m 6000 RCS collimator x-y W RCS collimator z W 6000 Extraction region W 6000 RTBT W/m 6000 Table Required shield thickness for the Accelerator Position Distance from beam line Shielding thickness(cm) Dose Rate (cm) Concrete Soil (micro-sv/h) LEBT E-10 RFQ E-08 MEBT-chopper E-07 DTL E-09 DTL E-06 DTL E-04 DTL E-03 DTL E-02 DTL E-01 DTL E-01 LRBT E-01 LRBT collimator E+00 Injection region E-03 RCS E-01 Extraction region E+00 REBT E

156 Table Annual dose rate caused by Skyshine 01# 09# 19# 27# Distance Dose rate Distance Dose rate Distance Dose rate Distance Dose rate Source term m micro-sv/y m micro-sv/y m micro-sv/y m micro-sv/y LEBT E E E E+00 RFQ E E E E-09 MEBT-chopper E E E E-08 DTL E E E E-10 DTL E E E E-07 DTL E E E E-05 DTL E E E E-04 DTL E E E E-03 DTL E E E E-02 DTL E E E E-02 LRBT E E E E-03 Injection region E E E E-04 RCS E E E E-02 Extraction region E E E E-01 RTBT E E E E-02 Sum 3.06E E E E-01 (2) Skyshine calculation Skyshine: Calculation by using Stapleton semi-empirical formula. During CSNS operation, the dose rate at CSNS boundary site 01 #, 09 #, 19 #, and 27 # caused by Skyshine are shown in table From table , we found that neutron and γ through the shield to the general public at the site boundary may result in the maximum dose of 0.675µSv / a, compared with the value of design objective 0.1mSv/a low-more than two orders of magnitude Environmental Impact Assessment June 17, 2008, in the Dongguan City, Guangdong Province, the State Ministry of Environmental Protection, organized the China Spallation Neutron Source project environmental impact assessment expert review. Review concluded that: "The report, made a rather clear and systematic description for the necessity of the construction, the project features and the associated environmental impacts. Report format and content are in line with national requirements of the relevant provisions, prepared on the basis is sufficient, Purpose of the evaluation is clear, the description for the identification and evaluation of environmental impact factors is clear, the content of a more comprehensive and adequate to meet the project environmental impact assessment requirements. The environmental impact of the project is acceptable, the conclusion is credible. " 154

157 17.6 Pre-evaluation of occupational hazards National Institute for Radiological Protection, China CDC on December 23, 2007 in Beijing, organized the "China Spallation Neutron Source (CSNS) project pre-assessment report for occupational hazards". The expert review was formed the following views, 1) The laws, regulations and standards used for the "Pre-assessment report" are accurate, the content of the assessment is comprehensive, the methods of the assessment is correct, are in line with the Ministry of Health requirements. 2) The main sources for the project items and occupational hazards identification are accurate, Hazard level of analysis is reasonable. "Pre-assessment report" for the protection facilities and measures to be established make an objective assessment. 3) Pre-assessment report" conclusion is correct, proposal is feasible Summary (1) CSNS of the bulk shielding design adopted with mature international formula and the method of combining Monte Carlo simulation, shielding thickness to meet the safety requirements. (2) CSNS is a large-scale radiation facilities, the resulting radiation is mainly instantaneous. As long as the accelerator a shutdown, radiation field then disappears, but also no longer cause air, cooling water and soil activation. In other words, as long as the accelerator can cause a shutdown of the main sources of environmental impact that is disappearing. (3) CSNS serious accidents may occur in staff being diverted to high radiation areas, then result in personal injuries. In the CSNS protective design, through the establishment of the safety interlock system and strict operating procedures to prevent such accidents from occurring. And that such incidents would not in itself cause any impact on the environment. (4) CSNS in operation and under accident will be not caused harmful effects to the staff, to the environment, and to the general public. To sum up, CSNS is a high-energy high-current proton accelerator-driven strong neutron source of a research unit. there are many factors that potentially affect the environment. But as long as to take effective protective measures and establishing strict rules and regulations, its impact on the environment can control the state's standards. 155

158 18 Survey and Alignment Lan The general layout of the CSNS is shown in Figure It consists of a linac of about 197 m in length, the linac to RCS transport beam line (LRBT) of about 40 m in length, an RCS ring of 228 m in circumference, and the RCS to target transport beam line (RTBT) of about 144 m in length. Figure 18-1 General layout of CSNS 18.1 Alignment network design We plan to set an alignment network along the accelerator tunnel. The alignment network is used as a reference for installing,locating or adjusting the accelerator devices. It can also be used to monitor the deformation of the accelerator alignment along with time. According to the major structure of the accelerator complex, the network can be divided into the linac-lrbt network, the RCS network, the RTBT network and the surface network. According to the ways of the measurement and the method of data process, alignment network can be divided into horizontal network and vertical network Layout of surface network Surface network is used to control the locations of buildings and devices and to provide high accuracy reference for lower networks. Figure shows the CSNS s surface network. It is composed of 7 permanent monuments based on bed rock, Z01, Z34, S01, S21, TR21, TR25 and TRS18. Using this surface network we can locate the linac, the transport lines, the RCS ring and the target. 156

159 Figure Layout of the CSNS surface network Figure Survey scheme of the CSNS surface network Surface network survey scheme The surface network survey is carried out by using a total station. We set measuring station on every permanent monument. In each station, the total station apparatus is centered on the local permanent monument, then measures the directions and distances of other monuments as shown in Figure Simulation results of the surface network survey According to the precision of the total station apparatus, the accuracy of distance measurement is 1.5 mm, and the accuracy of direction measurement is 5. With these conditions, we have simulated the measurement error of the CSNS surface network. It turns out: the global point error is 1.9 mm and the relative point error is 1.2 mm Layout of the tunnel alignment network Considering the capability of laser tracker and the accuracy requirement of survey network, we plan to set monument sections with the interval of 6 m along the tunnel. In each section, there are four monuments as shown in Figure , two on the floor, one on the inner wall and one on the outer wall. outer wall monument inner wall monument Floor monuments Figure Monument section layout of the tunnel alignment network 157

160 There are 40 monument sections along the RCS tunnel, 34 monument sections along the linac tunnel, 7 monument sections in the LRBT tunnel and 25 monument sections in the RTBT tunnel Tunnel alignment network survey scheme Tunnel horizontal network survey scheme We use a laser tracker to carry out the horizontal network survey by move station method. The measuring station is set between every two neighboring sections. At each station, the laser tracker measures 3 backward sections and 3 forward sections as shown in Figure There are 4 monuments in each section, so the laser tracker should measure 24 monuments at one station. The number of common monuments between neighboring stations is 20. In order to obtain the horizontal coordinates, at each station we need to establish a horizontal frame for the laser tracker. Four monuments in one section Laser tracker Figure Survey scheme of the horizontal network Simulation results of the RCS horizontal network survey According to the precision of the laser tracker apparatus, the accuracy of distance measurement is 0.08 mm, and the accuracy of direction measurement is 5. With these conditions, we have simulated the measurement error of the RCS horizontal network. It turns out: the transverse point error is 0.08 mm and the longitudinal point error is 0.15 mm Tunnel vertical network survey scheme We use a Leica DNA03 to carry out the vertical network survey and set the measuring stations between every two neighboring monument sections. At each station, we measure one backward section and one forward section. Together 8 158

161 monuments will be measured Simulation result of the RCS vertical network survey Simulation result: the vertical point error is 0.06 mm Magnet fiducialization scheme We use a laser tracker to carry out magnet fiducialization. Through measuring the fiducial planes and fiducial points of the magnet, we can establish a frame for the magnet. In this frame we can obtain the position relation between the fiducial points and the central line of the magnet. Fiducial points Fiducial planes Figure Fiducialization scheme of a magnet The errors of the magnets fiducialization in X, Y, Z directions are about mm Alignment scheme for accelerator devices in tunnel We use the alignment network and laser tracker to perform the alignment of accelerator devices in tunnel. By setting a laser tracker station near the device to be aligned, we can measure the monuments in the 3 backward sections and in the 3 forward sections. Through the best-fit method, we can determine the position of the laser tracker relative to the alignment network. Then using the laser tracker to measure the fiducial points on the device, we can obtain the position of the device relative to the alignment network and adjust the offset of the device to the required tolerance. We can perform the alignment for all the devices within the measuring range of this station at a time. Then we repeat the alignment in other monuments section to complete the alignment of the whole accelerator. 159

162 Fiducial point Laser Tracker Figure Alignment scheme of the accelerator devices 18.4 Error analysis of the final positions for some major accelerator devices The error of the final position of a device is the statistical sum of the fiducialization error, the installation error, the measurement error and the alignment network error. In the error analysis below, the coordinates definitions are defined as: X orientation and Z orientation are in the horizontal plane; X orientation is perpendicular to the beam, and Z orientation is the same as the moving direction of the beam; Y orientation is perpendicular to the horizontal plane. Final position error in X of a dipole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.06 mm, 0.15 mm, 0.05 mm and 0.08 mm: mm Final position error in Y of a dipole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.06 mm, 0.15 mm, 0.05 mm and 0.06 mm: mm Final position error in Z of a dipole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.08 mm, 0.15 mm, 0.05 mm and 0.15 mm: mm Final position error in X of a quadrupole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.05 mm, 0.15 mm, 0.05 mm and 0.08 mm: mm Final position error in Y of a quadrupole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.05 mm, 0.15 mm, 0.05 mm and 0.06 mm: mm Final position error in Z of a quadrupole, with the fiducialization error, the installation error, the measurement error and the alignment network error being 0.08 mm, 0.17 mm, 0.05 mm and 0.15 mm: mm 18.5 Motion monitoring scheme for the accelerator devices 160