SCAN -9909024 OCR Output

7|\1|1|!||\EINXIIIIHIIIIIWlfllIINIIUIJIIIIIIMIIIIIIIIIHIIICERN LIBRARIES, GENEVA

Geneva - May 1977

P0O030187

September 14-17, l976

Conference, Chalk River, Ontario, Canada.Paper presented at the l976 Proton Linear Accelerator

practical solutions are described.the design of such a structure are explained, andenergy gain around O.1 MeV/m. The guide·lines foroperated almost CW with a travelling wave giving anquency swing during the acceleration cycle. It ishigh-energy proton linac, except for a small fre

The SPS accelerating structure is essentially a

SUM ARY

G. Dome

TRAVELLING WAVE DRIFT—TUBE STRUCTURE FOR THE CERN SPS

THE SPS ACCELERATION SYSTEM

CERN-SPS/ARF/77-11

'cf p OXXCERN - SPSO A f. "

EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH+E 7Q U 6

©¢Snr¤—svs-·RRF Q4-)!6 J 5 Jllll.1977

FIGURES OCR Output

REFE RENCE S 1 9

ACKNOWLEDGEMENTS 18

MANUFACTURERSll. 18

POWER AMPLIFIERS10. 16

FEEDER LINES I5

TERMINATING LOAD I5

13INPUT AND OUTPUT COUPLERS

IlBEAM LOADING AND RF POWER

HIGHER MODES OF THE ACCELERATING STRUCTURE Il

TECHNOLOGY

CHOICE OF THE ACCELERATING STRUCTURE

REQUIREMNTS FOR THE ACCELERATING STRUCTURE

INTRODUCTION

.*252

Table 0f contents

in the SPS ring. OCR Outputb) This value corresponds to 1 X 10protons circulating13

a) At the present time, a 3rd cavity is being ordered.

tion easily provides the required bandwidth without any variable tuning

at a reasonable power level. On the other hand, travelling wave opera

necessitate variable tuning for a cavity operated with a standing wave

Although the relative frequency swing is only 4.4 X 10"°, it would

power is needed during most of the acceleration cycle.

The last requirement in Table I stems from the fact that the RF

close to 1Duty factor

UID = 89 kWBeam power per cavity (at ¢_ = 45°)

ID = 70 mAAverage be m current

eU = 1.27 MeV(at ¢_ = 45U)Energy gain per cavity for one traversal

Synchronous phase angle (from crest of the wave) ]¢S| Z 45

Peak accelerating voltage per cavity VI = 1.8 MV

Number of cavities

to 200.396 MHzto top energy (400 GeV)

199.526 MHzFrequency swing from injection energy (10 Ggv)

200.222 MHzFrequency at transition energy (24 GeV)

Specifications for the SPS Acceleration System

Table I

specificationsl’ shown in Table I.

June 17, 1976. The RF accelerating system was designed to meet the

ground near Geneva. It reached its design energy for the first time on

The SPS is the European proton synchrotron of 400 GeV built under—

velocity is v, the electric field induced by the beam along the axis of OCR Output

If the RF component of the be m current is IL and the particle

El.

peak value of the effective accelerating voltage per cavity is essentially

(several hundreds m); therefore, in the absence of beam loading, the

attenuation constant. At 200 MHz, the attenuation length lk! is large

where vn is the group velocity of the travelling wave, and u is the field

v_ Q c3>R = AL EL = 2uR 2 1

is related to the series impedance by the equation

of the structure

(2)R1 (in E-) = EZ/power loss per unit length

defined as

The effective shunt impedance R1 per unit length of the structure,

travelling wave.

(including the transit time factor), and P is the power flux of the

where E is the peak value of the effective accelerating field on the axis

(1). R2 (ln 55) — En E2

as

parameter for RF power is the series impedance R2 of the structure, defined

terminated in a matched load for travelling wave operation. The important

Each accelerating cavity is essentially a loaded waveguide of length 2,

2. REQUIREMENTS FOR THE_ACCELERATING STRUCTURE

the cavities are in a tunnel, 60 m underground.

important when the power amplifiers are located at the surface, whereas

matched load even with varying frequency and beam loading. This is very

a long feeder line to the power amplifier, and continues to appear as a

It also presents another major advantage: the cavity may be connected by

operating mode approaches the edge of a passband, where vg vanishes. OCR Outputperfect synchronism with the bea . It is no longer valid when thespace—harmonic of the beam induced field, which is closest to

a) It should be mentioned that Eq. (4) represents only that particular

Vrf

_ i b T ji ·7T — sin T ·1;·— I (6). S].n_ 2 S]..II_ _ Jo 2 _ RZR I ER. Q ······I·····. T2 _ 2 i· 2traversal of the cavity is

With Eq. (4), the effective voltage V seen by the beam upon one

the particles leave the cavity.

the particles enter; in the second case, z is taken from the end where

first case, the distance z in Eq. (4) is taken from the cavity end where

acceleration is of the forward wave or of the backward wave type. In the

as positive or negative according to whether the passband used for

as positive. The particle velocity v, on the other hand, should be taken

In all formulae, the group velocity vc should be taken essentially

finite velocity (or energy) Of the particles.

synchronism between the wave and the particles; it corresponds to a de

In Equation (5), wo is the angular RF frequency at which there is perfect

(5)T=;"*—(w·wg)

travelling wave and the proton bunches along the cavity:

the middle of the cavity. The angle T is the total phase slip between the

where z is the distance from one cavity end. All phases are referred to

b b 4onE:-I?-%s‘2}—?£

_.w _&—]T—· JT(z)2

the cavity is given by

limit for the filling time 2/vg of the cavity: OCR Output

i.e. at injection. According to Eq. (5), this condition sets an upper

it was decided that Irl should not exceed w radians in the worst case,

In order to limit the voltage reduction due to T at other energies,

the acceleration cycle.

transition. In this case, T and ¢_ have always the same sign throughout

RF amplitude |E2| but only the RF phase ¢, then r must be zero at

while ]V| is kept constant. If this is to be done without changing the

¢h must abruptly change sign (going from a negative to a positive value)

Fig. l shows the phasor diagram for Eq. (6). At transition energy,

than in normal proton linacs.

to work with a rather low accelerating field E, about l0 times lower

"linac". On the other hand, this recirculation allows the latter linac

in order to recirculate the beam some l00 000 times through the SPS

which does not apply in a normal linac, is the small price to be paid

accelerating voltage is given by Eqs. (5) and (6). Such a reduction,

travelling wave and the particles; the corresponding reduction in

cannot be maintained throughout the acceleration cycle between the

Because of this slight change in RF frequency, perfect synchronism

frequency of the particles.

order to maintain an exact harmonic relationship with the revolution

cycle; this is the reason why the RF frequency must also increase, in

cumulative increase in the revolution frequency during the acceleration

hand, the small increase of velocity from turn to turn produces a

structure to be exactly periodic, as for an electron linac. On the other

traversal of the cavity is very small, which allows the loaded waveguide

For relativistic particles, the velocity increase during one

very close to 2IO.

sents the beam—induced voltage VL. For tightly bunched beams, IL isrf

Eq. (6) is the RF impressed voltage V, while the last term repre

axial electric field in the middle of the cavity. The first term in

where ¢ is the phase of the bunches with respect to the crest of the

value of E2 and of the RF power. OCR Output

Fig. 1 that T = 0 corresponds to a larger |Vrf|, i.e. to a maximuma) Because T and ¢S have always the same sign? it is clear from

<12>P = KVI2 + |V.|2 + 2IVI |vI¤¤s ¢1 Q [ bS]

or, from Fig. 1 when T = O:

_ "rf

The RF power is given by Eq. (1) as

is the beam—cavity coupling impedance.

c 8_ r (11)

Rzmzwhere

c b(10)V = E2·ej¢— r I

to

For simplicity“”, we now assume r = O in Eq. (6), which then reduces

Optimization for power efficienc

straight section of the SPS ring.

which is the maximum length available for a single cavity in a long

R 5 22.5 m (9)

To this condition should be added

(8)2 5 (215 m) v_/c

which may be rewritten as

(7)R/vn 5 0.718 ps or E4 903£

identical cells. OCR Output

1. The structure should be a wave—guide periodically loaded with

Summary

the relatively small nominal beam current I0 in Eq. (16).

which is an extremely large value at 200 MHz. This large value comes from

(17)R1/Q 2. 5000 S2/m

(9):

Inserting |V| = 1.8 MV and I0 = 70 mA, this yields with the condition

Q 903 903 I0(16)R 2. 11 mz 4 .L 2, .Z.... = ..... i

or with Eq. (15):

R22.2 = -5 .,5 903 RN

But condition (7) yields, with Eq. (3):

I(15)V Rom? = aLl: 4i

power efficiency at the nominal values of]V\ and I0:

hence with Eq. (ll), there is an optimum R25} which provides maximum

cb b(14)vl = ris lv

s bFor IV\Ȣand Ifixed, P is minimum when

r lv c b(13)V r I P=l|Vll —-L-l-+—Q+2cos¢» 8 b I

This may be rewritten as

currents, the cavities should be shortened. OCR Outputyielding maximum power efficiency; while for still higher beambeam current, the final cavities described in Table II come close tothe SPS by a factor 5, when multipulsing its CPS-injector. For such a

a) Already at present it is contemplated to increase the beam current in

increased"’.

power efficiency, it could do so in the future, if the beam current is

and energy gain per turn; but according to condition (15) for maximum

choice do) optimum power efficiency at the design values of beam current

this choice of accelerating structure does not provide (nor could any other

parallel, and to optimize it for maximum series impedance R2. Obviously,

Therefore, it was decided to use the simpler structure with all bars

times the nominal value.

while condition (16) can still only be met with a beam current several

practically unchanged. Conditions (8) and (9) are then easily fulfilled,g g

parallel to the first one, reduces vto v/c = 0.1 while R1/Q remains

other hand, going to a simpler structure where the second set of bars is

violates condition (9) and still fails to verify condition (16). On the

n/2 mode between adjacent bars. The maximum length 2 allowed by Eq. (8)

Orders of magnitude are R1/Q = 500 Q/m at 200 MHz and vn/c = 0.2 at

The bar passband is of the backward wave type.

the structure is operated in w mode (i.e. H/2 between adjacent bars).

and to provide a resonant coupling between each set of parallel bars when

envelope. The second set of bars is used to produce a large bandwidth,

alternately by horizontal and vertical bars placed across a circular

bandwidth. This structure contains two sets of drift tubes, supported

the cross—bar structure appears to have the largest R1/Q and the largest

Among all structures suitable for accelerating relativistic protons,a)

CHOICE OF THE ACCELERATING STRUCTURE

and (16).

turn and beam current, 2 and R1/Q should verify conditions (8), (9)

3. For optimum power efficiency at the design values of energy gain per

the transition frequency (200.222 MHz).

The travelling wave should be exactly synchronous with the beam at

7.

flowing. The surface finish of the copper is better than 5 um. OCR Output

line is turned to the top of the cavity, where no electric current is

a cylinder and welded longitudinally. In the final assembly, the weld

clad with copper by hot rolling. Afterwards, the sheets were rolled to

copper: 5 mm of copper on 20 mm of steel. The flat steel sheets were

The cylindrical envelope is made of ordinary steel clad with OFHC

Manufacturing of the cavities

4. TECHNOLOGY

the optimum is very flat.

has led to the set of parameters shown in Table II (p. l0). As usual,

0.067& and condition (8) was not fulfilled. For n/2 mode, optimization

and 2v/3 modes were investigated; in the latter case v_/c was at most

Once 2 is determined, R2 is optimized under condition (8). Both n/2

therefore the matching of the couplers is less critical.

is no impedance transformation by the cavity between these drift tubes, and

between the two extreme drift tubes be a multiple of H; in that case there

total number of drift tubes is chosen in such a way that the phase shift

at least five identical sections made of an integral number of cells. The

than Q.5 m. With a maximum allowed length of 22.5 m, this corresponds to

For manufacturing, the cavity had to be divided into sections shorter

tunnel.

to accommodate the cavity and all the ancillary equipment in the SPS

impedance, the inner cavity diameter had to be limited to 750 mm in order

a somewhat larger cavity diameter would yield a slightly better series

aperture), drift tube length, stem dia eter, cavity diameter. Although

have been varied: drift tube diameter (determined by the necessary beam

using a classical bead perturbation technique. All geometrical dimensions

Many reduced scale (1:5) models of the structure have been measured,

drift tube assembly (Fig. 3) can easily be taken out and replaced.

by horizontal bars (Fig. 2). It was designed in such a way that each

taining a set of identical and equally spaced drift tubes, all supported

Consequently, the structure is made of a cylindrical envelope con

the regulation of the cooling water temperature to within i7.5°C. OCR Output

assemblies at CERN. An overall, fine tuning of f25 kHz is provided by

the spacing between pedestals before brazing of the final drift tube

was measured. The frequency error was corrected by carefully adjusting

ll drift tubes especially used for this purpose, and the w/2 mode frequency

each section individually. The section envelope was fitted with a set of

tuned to the transition frequency at r/2 mode. This was achieved by tuning

to avoid any variable tuning in the cavities. Nevertheless, they must be

As mentioned, the original purpose of travelling wave operation was

Tuning of the cavity sections

than 2 X l0-G Torr.

Under normal working conditions, the vacuum in the cavities is better

resonator.

tested successfully under a peak current of 30 A/cm in a A/4 coaxial

holes: they normally carry a peak current of 3 A/cm, but they have been

case, for example, of the contacts between the pedestals and the envelope

inserted in grooves which are machined in the OFHC copper. This is the

All RF contacts are provided by silver-plated copper-beryllium springs,

system is made of copper or stainless-steel parts.

to it by an epoxy resin with good thermal conductivity. The whole cooling

waterpipes (with l20 mm! bore) running all along the structure and glued

O.&°C along a stem. The envelope itself is cooled sufficiently by 8

the temperature difference along a pedestal radius is l.1°C; it is only

the matched load at the end of the cavity. At the nominal field level,

circuits being connected in parallel; the same waterflow then cools

The total waterflow per cavity is 550 litres/minute, all cooling

by stainless steel flanges.

electron-beam welded. On their outside ends, the stems are terminated

perfect tightness against water, both ends of the stems have been

stems, these are cooled by a high speed waterflow. In order to ensure

together under vacuum. Since half of the power dissipation occurs on the

The drift tube assemblies are made of OFHC copper pieces brazed

9.

bea loading. OCR Outputb) The power dissipated in the cavities increases only slightly with

the stem diameter has been increased from 85 um to 86.5 nm.

a) For practical reasons during brazing of the drift tube assemblies,

13 kWPower dissipation per cavity

293 kwPower flux without beam loading PO = E5/R2

89 kV/mAxial electric field (without beam loading) E0 = lv/zlPower attenuation Zol = m2/(vgQ) 0.0é56 neper

1.h MQBea cavity coupling impedance TP = RZZZ/8

0.712 usFilling time 2/v

20.196 mInteraction length 2 = SQ cells

20.570 m2 half-cells for the input and output couplersCavity length: 5 sections of 11 cells, including

27.1 kQ/mSeries impedance R2

0.0946vg/c

19650Q (measured on the final, full scale cavity)

611 QmR1/Q

85 mmStem diameter

150 mmDrift tube length

170 mmDrift tube outer diameter

130 mmDrift tube inner diameter

750 mmCavity inner diameter

376 nmCell length (BA/4)

to 191.2 MHzto w mode)

221.6 MHzLower bar passband (0 mode

w/2Operating mode at transition frequency (200.222 MHz)

Parameters of the SPS travellinz wave structure

Table II

10.

being fitted with a 200 MHZ rejecting filter. OCR Outputa number of cells; they couple as strongly as possible to 629 MHz whileprobes with resistive loads have been designed to be put at the top ofprovide insufficient coupling to the 629 MHz mode. Therefore, damping

a) Since the time of the conference, these end-plate loops have proven to

maximum value for the RF power which is required in the presence of beam loading:

cAssuming again T = 0, we get from Eq. (l2) with |Vb| = rlb, the following

the same cavity voltage V against beam loading, as shown clearly by Fig. 1.

Nevertheless, the amplifiers must deliver more power in order to maintain

would otherwise go into the terminating load.

of the beam, and the beam power is merely subtracted from the power which

the bea towards the feeder line: the power amplifiers ignore the presence

sign as the particle velocity. Therefore, there is no RF power induced by

(of the synchronous space—harmonic) of the travelling wave be of the same

cavity end which must be fed by RF power in order that the phase velocity

From Eq. (4), the beam induced field vanishes at z = O, i.e. at the

6. BEAM LOADING AND RF POWER

once the da ping loops are installed“’.

dinal mode (with E_ # 0 on the cavity axis) and will hopefully be suppressed

ing in the cavities. This instability seems to be driven by a longitu

observed at high energy in the SPS, together with a 629 MHz signal grow

Bunch-to-bunch dipole oscillations of the beam have already been

rating mode, and as much as possible to all other modes.

are designed in order to couple as little as possible to the main accele

end-plate; they are terminated by 600 W, air—co0led loads. These loops

Therefore, two special coupling loops have been foreseen on each cavity

to the higher modes, they do not provide sufficient damping for them.

verse instabilities. Since the main couplers are not intended to couple

a mode is large enough,it can drive the beam into longitudinal or trans

synchronous with the beam. If the beam-cavity coupling impedance for such

In the higher passbands of the accelerating structure, many modes are

5. HIGHER MODES OF THE ACCELERATING STRUCTURE

ll.

1.06 MV. OCR Output

From Eq. (21), the peak accelerating voltage per cavity would then be

value; from Eq. (20), the corresponding power efficiency would be 0.828.

with the actual cavities. This current is 5.4 times the present nominal

I0 = 0.377 A for the beam current which would yield maximum power efficiency

The nominal P = 340 kW inserted in Eq. (21) gives IL = 0.754 A or

factor, it was decided to specify a power level of 500 kW per amplifier.

the possible increase of beam current in the future, and for some safety

I0 = 70 mA. With provision made for the losses in the feeder lines, for

power required at the input of the cavity, for an average beam current

(19) gives P = 340 kW with a power efficiency PL/P = 0.261. This is theSInserting now the nominal values of lvl, ¢and Ib in Eq. (18) and

b rn(1 + COS ¢n) ° = 2 with |V| rclb . ( l)4P I2 =-—-—-——-—

c Sto maximum power efficiency for given values of P, rand ¢

condition inserted in Eq. (18) determines the beam current which corresponds

cbwhen |V| = rl, which is exactly the sa e condition as Eq. (14). This

(Z0)(Pb) 2 cos ¢ r max

For a given ¢_, the power efficiency reaches its maximum value

E? = 4 cos ¢S ;rE-+ ——~—- + 2 cos (19)vl rllcb c b lvl

deduced from Eq. (13), which yields for the power efficiency

SThe power Pk = % ]V|Ib cos ¢delivered to the beam is most easily

8 r cb b s+x "+2|V|I ¤¤s¢ l (18)P=—¥|E-E

I2!

the usual ceramic disc, resulting in more uniform dielectric losses. OCR Output

field is roughly parallel to the cera ic, but it is more uniform than in

radially in order to form a big A/4 stub on the air side. The electric

formed into a short ceramic cylinder, by expanding the coaxial line

breakdown. Instead of the usual cera ic disc, the window has been trans

The basic design principle was to have it easily replaceable in case of

The most remarkable part of the couplers is probably the RF window.

this is very important to prevent beam instabilities.

adjacent modes, thus reducing their coupling impedance with the beam:

edges. Consequently, the terminating load of the cavity also damps the

the w/2 operating mode, just because the latter is far from the passband

It was possible to make the matching of the couplers broadband around

up to outlets at the bottom of the big stubs (Fig. 5).

This air is forced down and flows along the coaxial lines of the couplers,

four high·speed fans (working with 400 Hz A.C.) blow air into the line.

the window into 100 Q at the T—junction. Just above the T—junction,

cascade of two A/4 transformers which transforms the 50 Q impedance at

Each piece of line from the windows to the T—junction is essentially a

practically loaded with their 50 Q characteristic impedance by the structure.

The two short coaxial lines from the windows to the coupling loops are

measurements on the first two sections of the first cavity hith 22 cells).

reduced scale model with 16 cells; the final adjustments were made by

The shape of the couplers was determined by measurements on a l:5

drift tube assembly (Fig. 4).

end up into two coupling loops connected to the pedestals of the first

T-junction into two smaller ones, the central conductors of which eventually

matched load at the other end. The big 50 Q coaxial line splits at a

which connect the cavity to the power amplifier at one end, and to a

provide a matched transition between the cavity and the big coaxial lines

Both ends of the cavity are equipped with identical couplers. They

7. INPUT AND OUTPUT COUPLERS

I3.

at the end of each acceleration cycle. OCR Outputreconditioning pulse of 600 kW during 300 us is now being inserted

a) In order to avoid this periodic reconditioning of the cavities, a

to the 550 kW level

of hard conditioning are then necessary in order to bring the cavity back

the maximum level steadily goes down to 350 kW after one week. Two hours

level up to 550 kW; but even when the cavity is being used for acceleration,

tube assembly. Progressive conditioning of the cavity pushes this maximum

to be limited by multipactoring between the end plate and the first drift

On the other hand, the maximum power which can be fed into a cavity seems

very fast during the acceleration cycle, with no noticeable RF reflection.

Once the cavity has been conditioned, this power range is being traversed

still occurs on the windows, at a power level ranging from 10 to 50 kW.

As shown by the temperature rise at the window rings, multipactoring

disappeared.

yield. with the titanium coated window assemblies, sputtering has totally

chosen for its low secondary emission coefficient and its low sputtering

resistance being at least 500 MQ measured in vacuum). Titanium has been

the inner side of the ceramic was vacuum flashed with titanium (the D.C.

layer of 0.5 um which was also applied to the inner conductor of the line;

Afterwards, the gold coating was replaced by a vacuum—deposited titanium

resulted in a nicely conducting window and a large reflection of RF power.

stainless steel, and progressively reaching the other end. This process

line and onto the ceramic, starting from the end close to the coated

RF power, this gold coating was sputtered onto the inner conductor of the

deposited gold layer of 2 um, and the ceramic was left uncoated. Under

the early windows, the first copper layer was coated with an electro

layer of 10 pm, then with a vacuum-deposited copper layer of 2 um. In

exposed to RF fields are coated, first with an electro—deposited copper

cooled (Fig. 6). The caps and that stainless steel cylinder which is

a Fe—Ni-Co alloy, which are brazed to the ceramic; these caps are water—

metallic pieces of stainless steel are covered at their ends by caps of

ductivity (roughly 10 times higher than alumina). The neighbouring

The ceramic itself is beryllium oxide, because of its good thermal con

lh.

Rosenthal-Stemag). OCR Output

This ceramic is basically a magnesium silicate ("Frequenta", made by

moulded with a uniform hydrostatic pressure applied from all sides.

losses; but in order to achieve this performance, the ceramic must be

power level of 3 MW (travelling wave) before cracking under dielectric

which is radiation resistant. With this shape, the ceramic can stand a

both ends also, the inner conductor is supported by a triangular ceramic,

contacts on the inner conductor at both ends of each line section. At

expansions between inner and outer conductor are compensated by sliding

The line is made of sections 5.55 m long. The different thermal

under natural cooling by air convection and by heat radiation.

aluminium. It can transport a CW travelling wave of 750 kW at 200 MHz,

normal air, The inner conductor is copper, the outer conductor is

The feeder lines are big 3&5 mm, 50 0 coaxial lines, containing

9. FEEDER LINES

A/A transformer is water cooled; in the second case, it is air—cooled.

dielectric of which can be either alumina or air. In the first case, the

30 mm thick, while the matching element is a A/é coaxial transformer, the

separated: the window which stands the water pressure is an alumina disk,

50 Q coupler on the other side. Finally, the two problems have been

15 atm), and which provides a transition from the 5.8 Q water line to the

on one side (the water column is 60 m high, and the test pressure is

problem is to produce an input window which stands the static water pressure

is increased by addition of a small amount of sodium nitrite). The main

line; it is heated by dielectric and conduction losses (the conductivity

(Fig. 7). Water is both the dissipative and the cooling element of the

with a uniform 6Ug" coaxial line, 6 m long, filled with water as dielectric

200 MHz, and capable of dissipating 500 kW of RF power. This is achieved

The output coupler must be terminated by a load which is matched at

8. TERMINATING LOAD

l5.

a) In fact, the amplifier interlock switches it off at a VSWR of 1.4. OCR Output

ting voltage.

be stepped up in power so as to compensate for the reduced accelera

interruption, provided the remaining tubes of all amplifiers can

faulty acceleration cycle, and acceleration can continue without

Switching to the new operation mode may be done at the end of the

cavity is reduced in proportionality to the number of tubes left.

three other tubes. ln such a case, the accelerating voltage in the

2. lf one out of the four tubes fails, operation may continue with the

matched load when looking towards the amplifier.

1. If the four tubes have the s me output impedance, the cavity sees a

ones:

Among other advantages, the use of hybrids presents the following

shown in Fig. 8 and Fig. 9.

by using three coaxial hybrids (which are 3 dB directional couplers) as

the 500 kW level. The final solution combines the output of four tubes

necessary to combine the output power of several tubes in order to reach

on the market, the choice fell on high—power tetrodes. Still it was

power of 500 kW at 200 MHz. Because 200 MHz klystrons were not available

Each cavity is fed by an amplifier capable of delivering a CW output

l0. POWER AMPLIFIERS

the VSWR as measured at the a plifier output is always less than 1.25

For the complete system, including feeder line, couplers, cavity and load,

during acceleration, the VSWR of both lines is always smaller than 1.05.

are about 95 m and 110 m, respectively). In the frequency range used

0.037 neper for one line and 0.043 neper for the other (the line lengths

The total attenuation between the power amplifier and the cavity is

1 MHz. OCR Output

bandwidth of the whole amplifier chain, from the low level to 500 kW, is

The bandwidth of one unit, measured at the 3 dB points, is 3 MHz. The

deliver a CW power of 140 kW, that is 10% more than the nominal 125 kW.

Finally, it should be mentioned that the amplifier units can safely

50 Q loads.

by 8 small coupling loops resonating at 840 MHz and terminated by small

840 MHz, which killed the first tubes. This mode was selectively damped

this circuit also showed a most dangerous oscillation on a H1] mode at

ferrite in order to prevent breakdown at the output coupling loop. But

value for all frequencies. The anode circuit was fitted with U-17

line, as seen between the control grid and the screen grid, at a low

by the very lossy Eccosorb ZN, which keeps the input impedance of this

U—l7. The screen—grid circuit is a very low impedance line tenninated

range from 30 MHz to 50 GHz. The control grid circuit is fitted with

and Cuming, which has large electric and magnetic losses in the full

range from 400 to 2000 MHz; the other is the Eccosorb ZN from Emerson

losses at 200 MHz but presents a large peak of magnetic losses in the

respect: one of them is the Siferrit U—l7 from Siemens, which has low

vent higher mode oscillations. Two kinds of ferrite were used in this

usual in high power amplifiers, they had to be damped in order to pre

tubes (Fig. l0). They use a grounded screen-grid configuration. As

The a plifier units themselves allow a quick replacement of the

of six tubes.

time as guaranteed by the manufacturer is a total of 30,000 hours per set

type RS 2004 J, in order to meet the SPS specifications. The tube life

hybrids. The tubes were especially developed by Siemens as tetrode

the 50 kW output of which is divided into four parts by using smaller

The four tubes of 125 kW each are driven by another identical tube,

17.

a) On leave from Brookhaven National Laboratory. OCR Output

charge, G. Rogner.

a visitor in the SPS for one year in 1972-73, and to our engineer-in

The mechanical design of the cavities is due to P. Grandu”, who was

amplifiers, and the author was responsible for the accelerating cavities.

touched upon here). H.P. Kindermann was responsible for the power

D. Boussard was responsible for the low—level system (which was not

RF group of the CERN SPS Division, under the leadership of C. Zettler.

The present paper describes the work which has been done by the

Acknowledgements

Colorado), but which are now delivered by Quartex (Paris).

which were first ordered directly from the U.S.A. (Coors, Golden,

The couplers were made at CERN except for the beryllium oxide windows,

Germany). The terminating load was made by Varian (Palo Alto, California).

(Munich). The accelerating cavities were made by Leybold—Heraeus (Hanan,

and Munich). The hybrids and the feeder lines were made by Spinner

The power amplifier plant was made by Siemens (Berlin, Erlangen

ll. MANUFACTURERS

18.

North*Holland Publishing Company, Amsterdam, 1970, p. 707-709. OCR Output"Linear Accelerators" (Edited by P.M. Lapostolle and A.L. Septier),

3. G. D6me: "Review and survey of accelerating structures", in

Energy Accelerators, Frascati, September 1965, p. 38-41.300 GeV proton synchrotron", V International Conference on High

2. W. Schnell: "A wide band travelling wave accelerating method for a

cram/1050, Geneva, 14 January 1972, p. 65-79.1. C. Zettler: "The Acceleration System", in "The 300 GeV Program e",

REFERENCES

19.

Fig. 1 Phaser diagram for Eq. (6)

9; \€P

Vrf OCR Output