COUPLER DEVELOPMENTS AT CERN

Transcription

1 COUPLER DEVELOPMENTS AT CERN G. Cavallari. E. Chiaveri, E. Haebel. Ph. ~egendre(*) and W. Weingarten CERN, Geneva, Switzerland 1. SOME RECENT HISTORY Since the second workshop on RF superconductivity three years have passed and we are here to monitor and discuss the developments which took place during this time interval. Let us start with a recapitulation of the state of coupler construction at CERN in We had then just closed one chapter, on cell coupling [l] and started to study beam tube coupling [2,3-41, having a close look on the question of "trapped modes" which, if concentrated in the centre of a multicell cavity, cannot interact with a beam tube coupler. We then found that in restricting the number of cells to four with the correct amount of intercell coupling, and endcells compensated simultaneously for several modes [5], "trapped modes" can be avoided at least up to three and a half times the fundamental mode frequency. We regarded this result as a sufficiently safe basis to switch over to beam tube coupling with two higher order mode (hom) couplers, one on each side and with 65O in between them, and in addition one beam tube power coupler. 2. BASIC COUPLER SPECIFICATIONS The characteristics of the cavity and the machine, in our case LEP, determine the basic coupler specifications. The LEP beam current of 2 X 3 ma and the cavity voltage of 8.6 MV determine the 4 kw power rating of the fundamental mode coupler and its external Q of according to the two equations: Pb = V. Ib. sin@ acc S and Here R/Q is the specific shunt impedance where U is the stored cavity energy and 9, the synchronous phase angle. Presented at the third Workshop on RF Superconductivity October 1987, Argonne, Ill.. USA (*) CERN Fellow, now at MC Dornell-Douglas, Geneva. Switzerland. 565

2 A definition of hom coupler requirements is more involved. The probably simplest but reasonable method for big machines with important time intervals Tb between bunch passages is to require a decay time of hom voltages no longer than Tb l61 For LEP this criterion leads to This value compares well to the Q, of copper cavities and in fact special external hom damping is not presently foreseen for them. Since in LEP a feedback system damping transversal beam instabilities will be installed dipole modes of the cavity may even have higher Qex up to 15. We will see that actual hom coupler constructions substantially surpass these requirements. It has also technical sense in the LEP environment: the LEP beam current is made up of very short bunches and has therefore a comblike spectrum with lines of intensity 2 Ib = 12 ma in a distance of 44 khz. If such a line falls on a weakly damped resonance with high R/Q very big powers may be deposited in couplers and finally in their terminating resistances. One has therefore a strong technical interest to damp such modes, here especially the TM,,, mode, to the aperiodic case with 2Q /o<<tb ex where the deposited power reaches its minimum value of For LEP and all modes below 12 MHz: Higher frequency modes will no longer be confined to cavities and adjacent beam tube regions. Since all cavities in a chain are then coupled together, meaningful field distribution calculations are no longer feasible. All what can be said is that fields will be in the form of standing waves and hence couplers must be simultaneously sensitive to both kinds of fields, electric and magnetic ones. On the other hand they can be mounted far from the cavities so that fundamental mode suppression is not a problem. The sketch below shows a possible coupling concept based on a two-conductor screened transmission line which has two different transmission modes, the symmetric and the unsymmetric one. As sketched they will couple to the electric and magnetic beam tube fields respectively.

3 load load We will equip the connections between cavities with ports, ready to receive high frequency couplers of this kind. if necessary. ago? But what was the state of beam tube coupler development at CERN three years THE TYPE I COUPLER Based on the earlier designed and successfully tested on-cell couplers and the in-parallel acquired theoretical understanding [7] the geometry of a hom beam tube coupler had been defined. Fig. 1, giving its outline. is a reproduction from our paper at the last workshop. The connection to the load resistor is lateral and a demounting flange is foreseen. Coupler type I, first model of 1984

4 The coupler which evolved from this first design is illustrated in fig. 2. Not much had obviously to be changed except for the method of suspension, now by a small diameter Nb tube instead of a ceramic cylinder. This tube also admits liquid He to the interior parts of the coupler. The exterior parts between the beam tube and the dismounting flange are within the liquid He container which surrounds cavity and beam tubes and is welded to the dismounting flanges, such that the seals and the ceramic window on the output are between the cavity and the insulation vacuum. All coupler parts beyond these flanges are cooled only by solid conduction but the design is such that these parts only see very moderate magnetic fields, not higher than 1 G at 5 MV/m accelerating field in the cavity. The highest magnetic fields are on the 2 mm antenna tube, 15 C due to the injected fundamental mode current. The demounting flanges are of stainless steel (CF type) with copper plated contact areas. They are brazed [8] to the 1 mm Nb tubes carrying them. The other assembly work of the Nb parts making up the coupler has been done by EB welding. After manufacture the fundamental mode suppression filter is tuned by successive machining steps of the filter condenser. A broadband transfer measurement method as described by us at the last workshop and in successive papers [7,12] serves to control the progress versus the correct frequency. A chemical polishing taking off 5 pm with a subsequent thorough rinsing finishes the preparation of a coupler. With the load resistors outside of the He bath a fundamental mode Qex of at least (Pex - 3 W) is required. This corresponds to a tolerance of * 3 MHz in filter tuning as can be read from a plot of Qex against filter frequency in fig. 3. Up to now four couplers of this type have been built and are now used on the 4-cell cavities LEPl and LEP2. The first two of them have been examined in a test program based in the beginning on a single cell cavity with two coupling ports, one on each beam tube, and mounted in a vertical cryostat with the couplers fully immersed in liquid He. In a first test run EgCc = 5.8 MV/m was reached. This had been also the maximal field of the cavity in a preceding reference measurement without couplers. At 5 MV/m one coupler "B" had a fundamental mode leakage of 4 mw and the other "A" of 2 W but this coupler had a filter tuned approximately 3 MHz too high. After dismounting this coupler (i.e. with only coupler "B" left) after 18 h of He processing 7.1 MV/m were reached, again limited by the cavity. The plot of Q, against Eacc was as in the reference measurement without couplers, proving the absence of supplementary coupler losses. To check the repeatability of such results then both couplers were remounted but this time each in the previous position of the other. In fact the tubes which receive the couplers (coupling ports) have diameter tolerances which influence the filter frequency. In the subsequent test again Eacc = 5.8 MV/m was reached and fundamental mode leakages of 3 W and 1 W for coupler "A" and "B" respectively. Finally, to check the hom power handling capability of vacuum feedthroughs. N type connectors and cables, 2 W at 56 MHz (frequency of TM, mode) were transferred through couplers and cavity.

5 Fiq. 2 Coupler type I as present -

6 X 2 > ro 3 U 4 + a, I U C U a, s., 3 V) m a, E V) m s., a, 4 Q 3 CI a, C1 X JA V- X c a, 3 U Z V- s., a,.u.* 4 V- V) 3 V) s., a, > L) X G' a.* LL

7 Subsequently these couplers were mounted on the 4-cell cavity LEP2 for a first test in a vertical cryostat. The couplers were then in a horizontal position and He gas pockets could form in the internal cooling channel. The fundamental mode leakages were now 12 W and 1.2 W at 5 MV/m which was also the highest field reached in this test. The field limit was due to non-resonant electron loading of the cavity which could be traced back to a weakness of coupler construction: Its structure has to be well centered within the coupling port. This was accomplished by three small A12, rods distributed evenly around the central disk where the output line is also connected and where the fundamental mode electric field is zero. It turned then out that chips came off these rods during coupler mounting and contaminated the cavity if produced with a length to assure very exact centering. In response better materials were looked for and among other ceramic materials quartz glass and sapphire examined and retained as suitable. Mounted with sapphire centering rods on cavity LEP2 in its regular horizontal cryostat [9] (with the restricted cooling mentioned earlier) end-cell fields corresponding to Eacc = 5 MV/m have already been applied successfully. For this last test also a new connecting scheme to the external loads has been introduced: Each coupler has now two external 5 Q load resistors linked by separate 5 Q cables to a connector T. The connection between this T and the vacuum feedthrough on the coupler output is via a 9 cm long rigid line of 75 Q wave impedance. This connecting scheme handles higher hom powers and produces better hom dampings than a simple 5 Q termination of the coupler. Fig. 4 gives broadband transfer curves for both kinds of termination. Hom damping figures measured on LEP2 in the superconducting state will be given later in a comparative table. They exceed the requirements in LEP. We therefore regard this type I coupler as a serious candidate for application in LEP. 4. HOM COUPLERS WITH CUT-OFF CHARACTERISTIC 4.1 The first model There is nevertheless a whole series of competitors under study. They all have one common ingredient, shunt inductors between the inner and outer conductors of a coaxial line and in consequence a low frequency cut-off like waveguides. The first representative of this species was a model examined two years ago [l1 at CERN. Its outline is given in fig. 5. We can interpret this device in two ways: Either as an excentric coaxial line with shunt inductors added or as a length of a special type of ridqed waveguide, the so-called lunar quide. As such its wave impedance Z and propagation constant are frequency dependent

8 Fiq. 4 Broadband transfer of type I coupler 1.2 GHz.To load Fis. 5

9 By choosing a cut-off frequency oc above the fundamental and making the coupler long enough any required fundamental mode suppression can be obtained. But it turned out that a coupler fitting into the space between beam tube and cavity equator was too short for a sufficient suppression and a simple filter, here a series LC resonator parallel to the load was nevertheless necessary. The other alternative, a parallel LC resonator in series with the load was not yet examined since more difficult to realize at this model stage. On the other hand condensers, in lumped element highpass filters the series element, complementing the shunt inductors, were soon introduced in a simplest form: as capacitive couplinq between load resistance and coupler structure (fig. 6). window I I to Load Fig. 6 At this stage J. Sekutowicz at DESY started to work also on this principle with special emphasis on structures with a reduced number of posts [l l]. A close and intensive collaboration - one of the authors of this report spent half a year at DESY - led then to the development of a hom coupler concept for the 5 MHz HERA cavity [l21 with a post supported structure welded to the beam tube (more on the realization and the results of this concept in [l 31). 4.2 The distributed filter hom coupler (type 11) At CERN we care much about the demountability of couplers, the motif being that such couplers could equip not only standard niobium cavities but also niobium sputtered copper cavities as presently developed at CERN [14]. Looked upon under this angle of view the post supported structure of fig. 5 seems not very suitable. But finally a solution could be found which even fits into the coupling port of the earlier described type I hom coupler. Key element of this coupler is a sequence of posts with condenser plates added (fig. 7). Fiq. 7

10 The impedance of these "posts" becomes zero at a frequency a, ) and in moving this frequency from zero towards the cut-off frequency any requlred cut-off damping can be obtained. An alternative description would be to speak of a distributed filter which must only be tuned near to the fundamental mode frequency to produce sufficient damping. And since the fundamental made current is now distributed on many elements magnetic fields can be kept moderate. A hom coupler construction [l51 making use of this idea is sketched in fig. 8. The distributed filter, with posts and condenser plates joined together into metal sheets, is adjacent to the antenna tip and eliminates the injected fundamental mode current. Therefore behind this element a standard 1 mm CF flange can be placed for dismounting. A support section in form of a normal lunar guide follows. Here also, by a small tube, liquid He is admitted to the interior of the antenna. Coupling via the demountable vacuum window to the 5 R load is capacitive. Note the minimal expenditure for cooling, the filter being only cooled by solid conduction! A prototype constructed from & RRR niobium and with an antenna penetrating 45 mm into the beam tube has been recently tested on a single cell cavity. The hom damping was then four times higher than required for LEP and the coupler worked perfectly up to an Eacc of 5 MV/m. Here (at a fundamental mode output power of 1 W) internal heating started. We are confident that higher fields can be easily reached without additional cooling channels if high RRR niobium is used or simply coupling is somewhat reduced in using a shorter antenna. 4.3 The one post hom coupler (type 111) The design of this coupler profited in many respects from experience with the HERA hom coupler. The concept of using a small number of posts is here at an extreme: only one support post is employed. Other design goals were however somewhat different. In view of applications with sputtered cavities it was decided to make at least the filter and the window (i.e. the most demanding parts for manufacture and preparation) demountable. Further it was tried to get excellent coupling especially to the dipole modes which govern beam stability in linacs, this evidently in view of possible applications of LEP type cavities in such machines. In this context dimensions were also looked for which allow scaling to higher frequencies without arriving at dimensions which are difficult to realize. Fig. 9 gives the geometry which evolved [16]. As in all our coupler designs the outer diameter is near to 1 mm allowing to use a standard CF flange. The antenna diameter in contrast has been increased from 2 mm to 45 mm corresponding to a wave impedance of 5 R in this front part of the coupler. The front part continues with the support post of 35 mm diameter and ends in the flange which due to the use of a stop filter in series with the load has to support only the small hom currents (of the order 5 A in LEP). The pieces of this front part which would be welded to the beam tube can be manufactured and assembled with low precision except of a good centering of the antenna. This centering is necessary to allow a connection between antenna and 5 Q load which exerts a minimum mechanical stress on the vacuum window. As link serves a 6 mm long line of 27 R wave impedance and 16 mm outer diameter whose flexible center conductor ends in a sliding contact into which the center piece of the 5 R coaxial vacuum window is inserted.

11

12

13 The filter, housed around this connection, is a capacitively loaded A/4 resonator. Among the dipole modes of the LEP 4-cell cavity the TM,, mode at 685 MHz emerges by its R/Q which is 2.5 times bigger than that of any other. Special care has therefore been taken to enhance coupling to this mode. The microwave circuit of this one post coupler allows to create two resonance peaks. As the broadband response of fig. 1 shows the bigger peak has been placed on the TM, mode and a smaller one between the TE, and TM,,, mode families. Fiq. 1 Broadband response of type I11 coupler Dipole mode external Q values of a first copper model of the LEP cavity carrying two such couplers, one on each side and with 65O between their mounting planes, are in the table below together with corresponding results from the LEP2 cavity equipped with two type I couplers.

14 Mode f (MHz) R/Q (R) Qex (type I) Qex (type 111) TExll TElll 12 3 TE1ll TE 111 l 18 T" T" T" T" W l8 68 T"1ll As we read from this table two type I couplers result in highest shunt impedances of 69 KC2 at 514 MHz (TMlo mode) and 54 KQ at 688 MHz (TM, mode). The mounting of type I11 couplers produces 3 to 4 time smaller impedances with the highest values: and 187 K(2 at 688 MHz (TM,) 155 KQ at 513 MHz (TM,). The type I11 (or one post) coupler is actually manufactured in niobium and will be tested after the workshop. Since the magnetic fields are everywhere low in the coupler, no more than 2 C at Eacc = 5 MV/m. no difficulties are expected. 4.4 The type IV coupler Before we close this chapter on hom coupling we want to represent a fourth design idea which had less priority and is therefore still on the working bench. Nevertheless because of its simplicity it could become the favourite if we decide to weld complete couplers to the cavity. The underlying idea was to go in the direction of lumped element highpasses i.e. to introduce additional capacitive couplings. Investigating this possibility P. Legendre then found that a capacitive gap in a lunar guide together with a "window" in the supporting metal sheet forms a stop filter which can be tuned onto the fundamental mode. Fig. 11 gives a sketch of the geometry worked out so far together with the broadband transfer and, Q, values. The results are already very good indeed and entirely statisfactory for LEP. Nevertheless we will "nurse" the design a bit more and try to obtain a response peak at the TM, mode.

15 C: X d - N I - + a - N r X - LL V) W O, X + g y B E gs mm mm *" 7-,-4d*m - m mm \* so o- g4 r C X h 8 m" "lg lg mmg a- 8 o g z X =.- 8 m CV S,-+E -

16 To allow a comparison of these different designs table 2 gives Qex values for the most significant modes of our 4-cell copper model cavity all measured in the same way: with one coupler mounted and dipole modes polarized into its plane. An inspection of this table shows that with regard to low shunt impedances Qex = R/Q a type I11 coupler gives the best overall results. TABLE 2 External Q values with one coupler on the 4-cell cavity. Dipole modes polarized in plane of coupler. f (MHz) R/Q (Q) Type I; 5 Q Type I, 25 Q Type I1 Type I11 Type IV TE T"iio TM TM TM Q ex Q ex Q ex Q ex Q ex & R/Q = Vgcc/(oU); dipole modes: Vac, Type I = Antenna 25 mm in beam tube. 5 cm from axis. Type I11 = Antenna 28 mm in beam tube. Type 11 = Antenna 45 mm in beam tube. Type IV = Antenna 35 mm in beam tube. 5. THE POWER COUPLER Let us again look back to the time of the last workshop. We then had already clear ideas about a future beam tube coupler. It had to be an antenna coupler using as much as possible the technology of a proven design: the 125 kw power coupler for the LEP copper cavities [17]. A similar line of development was followed [l81 also at DESY, for even higher demands for the r.f. power. 5.1 Construction Fig. 12 reproduces a drawing already shown 3 years ago: The coupler uses a warm cylindrical window integrated into the waveguide-coaxial transition to a 1 mm diameter 5 Q line running down to the beam tube. The antenna is normal conducting. For cooling cold He gas is injected into the antenna from the warm side. through the doorknob. The outer conductor is SC from the beam tube up to the level of the cavity equator. There a CF flange forms the transition to copper plated stainless steel and there also 4.2 K cold He gas is injected to intercept, in flowing up to the warm side, heat from solid conduction and RF dissipation. To create a good heat exchange the gas flows are confined to thin annular channels with obstacles in form of wire rings to create turbulence. For the thermal design methods created for magnet current leads had been applied [19]. A second CF flange on the warm side allows to demount antenna and window without opening the cryostat. This was the

17 state 3 years ago. In the following time not much had to be added to the design. Special copper joints which assure a good RF contact in the CF flanges have been worked out and it has been decided to use an antenna tip made from niobium instead of copper to avoid a possible contamination of the cavity from this source. For this purpose the welding of niobium to copper had to be mastered. A drawing of the coupler is in fig. 13. Fig. 12 Power coupler. ideas of

18

19 5.2 Pretests and conditioninq Such a coupler is now mounted on LEP2 but only after a series of pretests and processing steps. In fact we worried in the beginning especially about two points. First about losses in the connecting flanges and more generally about losses which cause additional evaporation of helium; secondly about the use of a warm window. Using a warm window makes certainly the construction and maintenance of a power coupler simpler but it may cause surface contamination by desorbed gases. Taking this hypothesis serious we condition couplers thoroughly at room temperature. Couplers are normally manufactured in pairs. They are then mounted, after careful rinsing, face to face on a special connecting piece with water cooled contacts sliding onto the antenna tips (fig. 14). The couplers are then pumped and baked and power is started to be transferred from a tetrode power amplifier to a matched load. Typically 24 to 3 h are needed to increase the power to 4 kw. During this forming process bursts of reflected power and pressure occur and serve to switch off power rapidly. Several hundred bursts rising the pressure to about 1-4 Torr are counted before reaching 4 kw (i.e. 2 V and 4 A). The rate of progress is particularly slow around 5 kw and between 25 and 3 kw. The same is observed [2] when forming LEP copper cavities and could indicate that the window is the principal site of gas desorption. A second observation pointing in the same direction is that the coupler length has practically no influence on the forming time. Since during mounting on a cavity internal coupler surfaces are inevitably in contact with air also gas exposure experiments were made after the first forming. Their parameters are in the next table. Species of gas (and duration of exp.) Forming time up to 4 kw Number of gas bursts ) 1-~ Torr Remarks N (very short); air (6 h) 2 Air (.2 h).8 h.6 h 7 1 Slow progress between 2 and 28 kw Air (36 h).8 h 4 Special efforts were necessary to prepare the coupler for the LEP2 cavity experiment in the SPS [Zl]. To facilitate the potential use of the power coupler as a passive damper of the fundamental mode [22], for this experiment coupling has been increased so much that Eacc = 5 MV/m corresponds to 4 V in the coupler. This means that a beam power of 16 kw would produce a match i.e. a pure travelling wave in the coupler. But the maximum beam power in the SPS experiment is only 1 kw and hence in the coupler a nearly standinq wave pattern is formed with a voltage minimum at the antenna tip (the risk to double the voltage has been taken because the coupler length chosen places then the window in a voltage minimum too). We have simulated these conditions in placing a coaxial short at the antenna tip and have succeeded to produce the required 4 V in the coupler in driving this hiqhly mismatched load via a 1:4 impedance transformer and a trombone from the tetrode amplifier(*.c). (*) This was also a pretest of the power circuit constructed by D. Boussard and H.P Kindermann for the SPS experiment [21].

20

21 5.3 Helium boil-off To measure the He boil-off due to the power coupler two conditioned equal couplers have been mounted on a single cell cavity with well-known losses (fig. 15). The second coupler had been terminated by a matched load. Then the input to the first coupler is also matched and in very good approximation the voltage in the cavity proportional to the square root of the transferred power P: Fig power couplers on single cell

22 To accentuate possible coupler losses in comparison to the cavity losses coupling (i.e. ), Q, had been chosen so as to transfer 4 kw at Eacc = 3 MV/m. Dissipation was then measured in adjusting the power of a heater resistor in the He bath such that boil-off became equal to the boil-off with 4 kw transferred. In deducing from this power the calculated cavity dissipation at 3 MV/m a coupler loss of 2.5 W was determined. As a control for possible contaminations the cavity was finally remeasured with the power couplers demounted. It was found that maximal field and Q, had not changed. Acknowledgements The construction and testing of these couplers involved the collaboration of the RF and cryogenics groups of the EF Division and the ST workshops for metal forming, electron beam welding, brazing and surface treatments. We gratefully acknowledge all contributions.

23 REFERENCES [l] E, Haebel, Part. Acc. Conf. Santa Fee (1983) [2] R. Sundelin et al., CLNS 33/561, March [3] D. Proch et al., Part. Acc. Conf., Santa Fee (1983) [4] H. Piel, Proc. of 1984 Linear Accelerator Conference (Seeheim) 26. [5] E. Haebeii, P. Marchand and J. Tuckmantel, Proc. of 2. Workshop on RF Superconductivity, CERN (1984) 281. [6] M. Tigner, Proc. of the Workshop on RF superconductivity, Karlsruhe (198) 289. [7] E. Haebel, Proc. of the 2. Workshop on RF superconductivity, CERN (1984) 299. [8] E. Chiaveri, report in preparation. [9] R.G. Stierlin, this workshop. [l1 E. Haebel, CERN/EF/RF 85-2, June [l l] J. Sekutowicz, private communication. [l21 E. Haebel and J. Sekutowicz, DESY, M-86-6, July [l 31 J. Sekutowicz, this workshop. C. Benvenuti et al., this workshop. E. Haebel, P. Legendre, CERN/EF/RF 87-1, April E. Haebel, CERN/EF/RF 87-4, 25 November J.P. Boiteux and G. Geschonke, LEP note 57, November B. Dwersteg, DESY M D. Giisewell and E. Haebel, 3. Int. Cryog. Eng. Conf., ICEC 3 (197) 187. G. Geschonke, private communication. D. Boussard et al., to be published. E. Haebel, CERN/EF/RF 87-3, 2 November 1987.

24