BUCKET SELECTOR SYSTEM FOR KEKB
BUCKET SELECTOR SYSTEM FOR KEKB
E.Kikutani,M.Suetake and M.Tobiyama
KEK,Oho1-1Tsukuba,Ibaraki305-0801,Japan
Abstract
A bucket selector system has been developed for the KEK-
B collider.Three VME computers for actual bucket-selecting,bunch-current reading and operator interface are working.The system provides a tool for generating the de-sired?lling patterns in the rings.
1INTRODUCTION
At KEK in Japan,the commissioning of an asymmetric electron-positron collider,KEKB,is now in progress.The collider consists of a3.5GeV positron storage ring(LER) and an8.0GeV electron storage ring(HER).In order to achieve the target luminosity the currents to be stored in the rings are2.6A and1.1A for LER and HER,respective-ly.These beam currents are distributed to many rf buckets and ultimately,the number of buckets?lled will be4607, which is90%of the total number of rf buckets,namely 5120.At the present stage[1],the maximum storable cur-rents are about500mA for both rings and the shortest bunch spacing with which beams are stably stored,with the help of feedback systems,is4ns.
The electron and positron beams are provided by an in-jector linac complex[2].For the synchronized operation of the linac and the rings,a fast timing-manipulator system was constructed,and as a sub-part of this system,the buck-et selector system was developed.With this system,we can inject the beam into any desired rf bucket of the rings.
In the previous paper[3],we described the basic design of the system.In this paper,we will explain mainly the software structure of this system.
2FUNCTIONS OF THE BUCKET
SELECTOR SYSTEM
2.1Generation of desired?lling patterns
In general,it is not easy to store a large number of bunches with a small bunch-spacing.The dif?culty arises from var-ious causes:instabilities(mainly coupled-bunch);heating of the beampipes;degradation of vacuum caused by high beam-current;etc.The strengths of the instability,the heat-ing,etc.are functions not only of the total beam-current but also of the bunch spacing,bunch-currents,etc.Thus,a?ll-ing pattern of bunches,i.e.,which rf buckets are?lled with the beam,is a very important factor in raising the luminos-ity as high as possible.
In principle,the number of possible?lling patterns is very large and we can not explore all these https://www.wendangku.net/doc/c913647675.html,ual-ly the bunches are?lled in trains,and it is convenient to cat-egorize the?lling patterns by the following parameters:(1) the number of the bunch trains,(2)the number of bunches per train and(3)the bunch spacing in a train.By surveying these parameters,we?nd the optimum?lling pattern under a given operation condition.Figure1shows an example ?lling pattern:32bunch-trains,40bunches/train and6ns bunch
spacing.
Figure1:An example of the?lling patterns in LER.
The optimum?lling pattern may change with different various operation conditions,decrease of the vacuum pres-sure,change in the strength of the instability depending on the operating point,etc.Thus,it is important for operators to be able to change the?lling patterns very easily.
2.2Control of?lling order
Next we explain the requirements on the?lling order.If we inject the beam into the same rf bucket continuously at the maximum injection rate(50Hz),the bunches in the buckets surrounding the injection-bucket will be kicked repeatedly. Since the transverse radiation damping time of the rings is about45ms,the kicks will be given to the bunches with-in one-half the damping time.In order to avoid this,the injection-bucket should be changed from pulse to pulse. 2.3Equalizing the bunch-currents in the ring When the beam is lost,either slowly due to vacuum or rapidly due to instabilities,we re-?ll the ring.It is typical that the bunch currents are lost unevenly and the number
of particles to re?ll is different,bucket to bucket.In order to regulate the bunch-currents we have an optional bunch current equalizer (BCE)in our bucket selector system.
3HARDW ARE CONFIGURATION
The rf bucket to be ?lled with the next linac-pulse is de-termined by the trigger timing for the electron gun.This mechanism is essentially common in both electron and positron modes.Five VME-packaged delay modules are used for controlling this timing,and the delay values of these modules are set by a VME computer which is in-stalled in the common VME sub-rack to the delay mod-ules.Hereafter,we call this computer the “linac computer ”because the sub-rack is located near the gun.
An accelerator operator chooses the ?lling pattern (ac-tually,chooses the parameters explained in Section 2.1).These parameters are accepted by a computer installed in the Central Control Room (the “CCR computer ”).The CCR computer and the linac computer are connected by dedicated optical ?ber cables using the interface of the Shared Memory System,which we described in a previous paper[3].
Additionally,we have one more VME computer to con-trol two bunch-current monitors for the LER and HER.They output the 5120bytes of data per ring synchronized with a signal which triggers the injection kicker.This com-puter,the “bunch-current computer ”is also connected to the linac computer with the optical cables and the bunch-current information is transferred to the linac computer.Figure 2shows the hardware con?guration of the sys-tem.The distance between the bunch-current monitor and the CCR is about 300m and that between the CCR and the linac is about 1000m.By using dedicated optical cables,the transfer speed of the bunch-current and the ?lling pat-tern information is not dependent on the control network under which all the control components are supported.
4SOFTW ARE
4.1
General
The control system of the KEKB accelerators,including the injector,has been constructed based on the EPICS environment[5].In this environment,each piece of hard-ware is controlled by a VME computer,while the opera-tor interface is performed by UNIX computers.The VME computer is Motorola Power PC 750(266MHz clock)with memory of 64Mbytes.The operating system is VxWorks 5.3.1.
The program running in the linac computer plays a main role in executing the functions explained in Section 2.In designing the software,the most important point is the reg-ulation of the execution time of the program.The maxi-mum injection rate of KEKB is 50Hz,that is,all the pro-cedure should complete within 20ms.In order to keep the execution time constant,we adopted a scheme as follows:
Shared Memory (VME board)VME computer optical cable for shared memory control network
delay modules
Figure 2:Hardware con ?guration of the bucket selector system.
1.The number of EPICS records is limited in the linac computer to make it approximately dedicated to the bucket selecting,
https://www.wendangku.net/doc/c913647675.html,munication with the operators,which is asyn-chronous with the beam-injection,is done by the CCR computer to keep the linac computer free from opera-tors ’interrupts.
4.2Software on the bunch-current computer
The program in the bunch-current computer acquires the 10240bytes (two rings)of bunch-current data from the monitors.Its execution speed is fast enough to synchro-nize to the injection kicker,whose rate is 50Hz at the high-est.It is coded as a device-support of the EPICS system and the data is available through an EPICS record.Besides this record,the bunch-current information is written into the Shared Memory.After writing all of the bunch cur-rents,an interrupt signal is also sent to the linac computer through the Shared Memory System.
4.3Software on the CCR computer
As described above,the parameters specifying the ?lling pattern are accepted by the CCR computer through a num-ber of EPICS records.They are written into the Shared Memory and copies are available on the Memory Board under control of the linac computer.The device support for
the Shared Memory loaded in the CCR computer functions as the device support of the bucket selector system.
The operator interface was developed with MEDM,which is the standard user interface in the EPICS system.Figure 3shows the operator-interface panel.Through this panel,an operator can specify the ?lling pattern and sever-al options,such as turning the bunch current equalizer on and off.In addition,the operator can read some useful in-formation:the last-?lled bucket;total number of bunches;and so
forth.
Figure 3:Control panel of the bucket selector system.
4.4Software on the linac computer
A ?ow-chart of the program in the linac computer is shown in Fig.4.This is coded not as device support but as a sim-ple C-language program under VxWorks.The hardware accessed by the code are the Shared Memory Board and the delay modules,both of which are VME modules.
The code consists of some ?ow-control part and the cal-culation/setting of the delays including the 12access times of the Shared Memory and 5access times of the delay mod-ules.Execution time for one trigger was measured to be about 40micro seconds with the bunch-current equalizer off.A separate measurement showed us that one access of
a VME module takes approximately 1micro seconds.If the bunch-current equalizer is switched on,reading of the bunch-current information (4.6kbytes at most,5ms)will be added.Adding the processing time of the interrupt under VxWorks,21micro seconds,the total time should be some 5ms,which is much shorter than 20ms.
semaphore
Figure 4:Flow chart of the code in the linac computer.The execution of the program in the linac computer is initiated by an interrupt from the bunch-current computer.The in-terrupt management is done with code prepared under the VxWorks System.
5
EXPERIENCE
With the bucket selector system,the beams are smooth-ly injected to the rings.When the operation scheme is changed,changing of the optics,for example,the optimum ?lling patterns are easily surveyed with this system.
6REFERENCES
[1]K.Oide et al.,in Proceedings of the 1999Particle Accelerator
Conference,p.288.[2] A.Enomoto,in Proceedings of the 1-st Asian Particle Accel-erator Conference (KEK Proceedings 98-10),p.292.[3] E.Kikutani et al .,in Proceedings of the 1-st Asian Particle
Accelerator Conference (KEK Proceedings 98-10),p.525.[4]M.Tobiyama et al.,,in preparation.
[5] A.Akiyama et al.,in Proceedings of the 1999Particle Accel-erator Conference,p.343.