The TRISTAN control system

29

Nuclear Instruments and Methods m Physics Research A247 (1986) 29-36 North-Holland, Amsterdam

THE TRISTAN CONTROL SYSTEM Shin-ichi KUROKAWA, Atsuyoshi AKIYAMA, Kazuhiro ISHII, Eiichi KADOKURA, Tadahiko KATOH, Takashi KAWAMOTO, Eiji KIKUTANI, Yoshitaka KIMURA, Haruyo KOISO, Ichitaka KOMADA, Kikuo KUDO, Takashi NAITO, Katsunobu OIDE, Shigeru TAKEDA, Kenji UCHINO and Junji URAKAWA National Laboratory for High Energy Physics (KEK) Oho-mache, Tsukuba-gun, Ibarakr-ken, 305 Japan

Manabu SHINOMOTO, Michio KURIHARA and Ken-ichi ABE Omrka Works, Hitachi, Ltd, Hitachi-stir, Ibarakr-ken, 319-12 Japan

The 8 GeV accumulation ring and the 30 GeV main ring of TRISTAN, an accelerator-storage ring complex at KEK, are controlled by a highly computerized control system. Twenty-four minicomputers are linked by optical fiber cables to form an N-to-N token ring network. The transmission speed on the cables is 10 Mbps . From each minicomputer, a CAMAC serial highway extends to the controlled equipment. At present, twenty minicomputers are connected to the network and are used to control the accumulation ring . The software system is based on the NODAL language devised at the CERN SPS. The KEK NODAL system retains main features of the original NODAL: the interpretive scheme, the multi-computer programming facility, and the data-module concept. In addition, it has the following features : (1) fast execution due to the compiler-interpreter method, (2) a multi-computer file system (3), a full-screen editing facility, and (4) a dynamic linkage scheme for data modules and NODAL functions. The accelerators are operated through five operator consoles, each of which is managed by one minicomputer in the network. An operator console contains two 20-inch high-resolution color graphic displays, a pair of touch-panels, and ten small TV monitors . One touch-panel is used to select a program and a piece of equipment to be controlled ; the other is used mainly to perform the console actions. 1. Introduction An electron-positron colliding beam facility, TRISTAN, is now under construction at KEK [1]. It consists of three accelerators : a 2.5 GeV linac, an 8 GeV accumulation ring (AR) and a 30 GeV main ring (MR) . AR and MR are controlled by a single computer control system. The diameter of AR is 120 m, and that of MR is 960 m. The equipment is distributed around the accelerators and the controllers are installed in sixteen site buildings. Fig. 1 shows the layout of TRISTAN and the location of the main control room. The TRISTAN control system [2,3] must fulfill the following requirements : (1) It must control hardware equipment distributed over a circumference of 3 km from the main control room . The number of points to be controlled exceeds 30000. (2) It must adapt flexibly to various changes in machine design and operation style. Because large-scale accelerators are very complex machines, thorough investigation of their characteristics is necessary before they can show their full potential. Moreover, accelerators are the tools for physics experiments ; therefore the style of 0168-9002/86/$03 .50 © Elsevier Science Publishers B.V . (North-Holland Physics Publishing Division)

operation and the equipment may require changes in order to cope with the needs of physics. (3) To assure the required flexibility, the software system must make programming easy for machine operators, accelerator engineers and accelerator physicists who may not be expert programmers. (4) To do complicated calculations, such as correction for closed orbit distortions, it is necessary to have a general-purpose computer with a large computing power and a large memory . We have solved these problems by adopting the following: - A set of distributed minicomputers connected by an optical local-area network (LAN). - The use of NODAL, an interpretive language with a multi-computer facility. - The connection of the LAN to the central laboratory computer system in order to use general-purpose computers for complicated tasks. The architecture of the TRISTAN control system is explained in section 2. NODAL and its KEK version, KEK NODAL, are described in section 3 . The man-machine interface system is explained in section 4. I. OVERVIEW OF EXISTING SYSTEMS

30 0

S. Kurokawa et al. / The TRISTAN control system 500 M

2 . System architecture 2 .1 . Network and minicomputers The complexity and size of TRISTAN compels us to adopt a distributed computer control system . Twentyfour 16-bit minicomputers (Hitachi HIDIC 80-Es and HIDIC 80-Ms) are distributed around the accelerators . These computers are linked together by optical-fiber cables to form an N-to-N token ring network (Hitachi Data Freeway) . Fig . 2 depicts the overall configuration of the TRISTAN control system . The details of the network are given elsewhere [4,5] . The HIDIC 80-E and HIDIC 80-M are 16-bit minicomputers with 1-Mips computing power . Each minicomputer is equipped with a 256-Kword memory, a 17-Mbyte magnetic disk drive, a console typewriter, a serial printer, and one or two CRT terminals (DEC VT100 or equivalents) . On the HIDIC 80-M a 4-Mbyte RAM is installed and is used as an additional "disk" device ; this has an access latency of 2 ms which is a substantial improvement over the 40 ms of a disk . The minicomputers are classified into two groups : the system computers and the device-control computers . The nine system computers are located in the central control room . Each of them supports one of the central-control functions of TRISTAN such as servicing an operator console (OPO-OP4), alarm-processing (ALO and ALI), library (LBO), and program development (DVO) . DVO is equipped with two magnetic-tape drives,

Fig . 1 . Layout of TRISTAN .

operator's console

*control station (CST) omaster station (MST)

optical fiber ring network (10 Mbps)

Accumulation ring

Main ring

Fig. 2 . Overall configuration of the TRISTAN control system .

S. Kurokawa et al. / The TRISTAN control system

31

which are used to backup the contents of magnetic disks. The fifteen device-control computers control hardware equipment such as magnets and power-supplies (MGO-MG4), radiofrequncy equipment (RFO-RF2), beam transport equipment (BTO and BT1), vacuum equipment (VAO and VAI), beam monitor equipment (BMO and BMl) and miscellaneous equipment (GPO). VAO, VAI, BMl and GPO are placed in the main control room while the others are located in the site buildings around the accelerators . Each device-control computer has two CRT terminals attached : one located near the computer and the other located near the controlled devices, connected by modems, for tests and diagnosis.

been used there for more than ten years. Today, NODAL or NODAL-like languages are also used at the CERN CPS [9], at DESY [10], at Rutherford Appleton Laboratory [11], amongst others . We have also adopted NODAL as the control language for TRISTAN [12] . One of the most important features of NODAL is its multi-computer facility . The following example shows how the multi-computer facility of NODAL works :

2.2 . CA MAC serial highways

When the interpreter in the computer executing this program encounters the line 1 .20, it sends the lines of group 2 and array A to MGO and waits until MGO sends back the answer ; the interpreter on MGO starts the interpretation of these lines and remits the array A by the REMIT command. The interpreter task in the originating computer then types the contents of the array A in the line 1.30. As this example shows, one NODAL program can be composed of several sub-programs, some of which are sent to other computers and interpreted there. The programmer, therefore, can write a multi-computer program on a single computer, leaving the complicated networking process to the NODAL system ; thus enhancing the efficiency of programming under the multi-computer environment. There are two variants of the EXEC commands, EXEC-P and IMEX, operating at different priority levels . Another important feature of the NODAL system is the idea of the data module . This is a special device handler in the NODAL system and is an external procedure referenced in NODAL. In the above program MAG(l, `CUR') is an example of a data module call, where MAG is the data-module name, I is the unit number, and `CUR' is the property. The unit number identifies the piece of the equipment to be controlled, and the property distinguishes operations on the equipment. Each data module has a corresponding two-dimensional array called a data table which contains the parameters necessary for the data module to work. Data modules and data tables are loaded into the computers where the equipment serviced by the data modules is connected. There exists in NODAL another type of external procedure, a function, which is a subroutine referenced in NODAL and which performs some particular service. Data modules and functions may be coded either in assembly language or in a high-level language and compiled and linked to the NODAL system .

From each computer, except LBO, a 2.5 Mbps bitserial CAMAC serial highway extends to the equipment . There are 40 CAMAC crates for the AR and 140 for the MR . The reasons for adopting the CAMAC serial highway are the following: (1) Process interfaces in large control systems must be based on a well-established standard from the point of view of maintenance and extensibility. CAMAC is a widely used standard in high-energy physics. (2) The CAMAC serial highway has many advantages for a long-distance process-control data-highway [6]. The bypass and loop-collapse functions are useful tools for maintenance and diagnosis. Many kinds of CAMAC modules are commercially available from many vendors. (4) A test system can easily be made with a low-cost personal computer, because CAMAC is a computer independent standard . In order to guarantee reliable data communication on the serial highway where the maximum distance between crates is 800 m, U-port adapters (UPAs), with the bypass and loop-collapse functions, are used where necessary. Two-wire coaxial cables connect the adapters. 2 .3 . Linkage to the KEK central computers

To manage tasks which overload minicomputers, LBO is connected to the KEK central computer system, which consists of three loosely-coupled mainframe computers (Hitachi M-280D, M-280H and S-810/10) . The connection is made by KEKNET [7], an in-house highspeed network at KEK. For the details see ref. [5]. 3. NODAL 3.1 . Generalfeatures of NODAL

NODAL [8] was devised at the CERN SPS as an interpretive language for accelerator control and has

1 .10 1 .20 1 .30 1 .40

DIM A(10) EXEC (MGO) 2 A ; WAIT (MGO) FOR I= 1, 10 ; TYPE I A(I) ! END

2.10 FOR I=1, 10 ; SET A(I) =MAG(I, `CUR') 2.20 REMIT A

I . OVERVIEW OF EXISTING SYSTEMS

32

S. Kurokawa et at / The TRISTAN control system

3.2. KEK NODAL

0 1

The KEK version of NODAL [131 is enhanced over the original NODAL by the following : (1) The addition of a full-screen editing facility, (2) The speedup of execution by the compiler-interpreter method, (17 (3) Provision of a dynamic linkage scheme for data modules and functions, (4) Provision of a multi-computer file system . The details of the structure of the KEK NODAL are given in ref. [13] ; therefore only subsequent modifications and items not previously covered are discussed here . 3.2.1 . Full-screen editor

The full-screen editing facility of KEK NODAL uses the cursor-control capability of DEC VT100. A user can edit his program by moving a cursor on the screen, inserting, deleting, or changing characters . We have also made an editor for programming in PCL, a FORTRAN-like compiler language for the HIDIC 80 . PCL is used to code data modules and functions in the KEK NODAL system . The mode of operation of this editor has been made nearly the same as that of the NODAL editor in order to maintain the same environment for the programmers. 3.2 .2 . NODAL interpreter tasks

A NODAL program is data for the interpreter which works as a task under the multitasking real-time operating system PMS . There exist three types of such tasks: (1) the terminal interpreter task for interactive NODAL programs, (2) the remote execution interpreter task for NODAL programs transmitted by EXEC, EXEC-P or IMEX commands, and (3) the real-time interpreter task activated either by the computer timer, the LAMS from CAMAC modules, or the accelerator timing system . 3.2 .3 . Organization

of memory

Since the HIDIC 80-Es and HIDIC 80-Ms are 16-bit computers, each task must work in a logical space whose maximum size is 64 Kwords ; therefore the 256Kword physical memory on the HIDIC 80 is organized into 13 logical spaces (LS) (see fig. 3) . The organization of these logical spaces for the device-control computers is as follows. LS#0 is the space for tasks that support the network, and LS#1 is that for driver tasks for the console typewriters, serial printers, etc. LS#8 is the space on which compilers, linkers and loaders work . LS # 9, LS # 10, LS # 11 and LS # 12 are the spaces for the NODAL interpreter tasks, with LS#4, LS#5, LS#6 and LS#7 as corresponding working areas. Working areas contain NODAL lines (the word "line" means the intermediate codes for one line of a NODAL source

LS#0

64kw shared

PMS , etc (211

t

1 8

t

2

NR task (10 )

R

sub )

user

R sub (161

9 10 11 12 3

11 DT (12)

4 5

7

Fig. 3. Organization of memory into logical spaces . Arrows mean that the same physical memory is shared by several logical spaces . The letters WR, NR, R and DT stand for the words, "working area", "non-resident", "resident", and "data table", respectively . Numbers between parentheses mean the memory size in Kwords. program), variables, etc. LS#3 is used for data tables . The remote execution interpreter task for the EXECP command runs on LS#9, those for the EXEC and IMEX commands run on LS#10, the interpreter task for the first NODAL terminal runs on LS #11, and that for the second terminal runs on LS#12. The real-time interpreter tasks run either on LS#11 or LS#12, sharing the same logical spaces with the terminal interpreter tasks. Four NODAL interpreter tasks reside simultaneously in the memory, sharing the central processing unit . Small tasks coded in PCL can run on LS#2 ; the maximum size of these tasks is limited to 3 Kwords. Since the execution speed is fast and the overhead of loading is small in these PCL tasks, they are suitable for programs which are run repeatedly for a short period of time . 3.2 .4 . Increasing the speed

of NODAL

We have increased the speed of execution of KEK NODAL by using the following methods: (1) the adoption of intermediate codes, (2) the use of hashing for variable search, and (3) the use of cache tables for lines and functions. The following example illustrates how a source program is converted to the intermediate codes:

33

S. Kurokawa et al / The TRISTAN control system

Source : FOR I =1, 100, 2; SET AA = I + 1 This line is converted to the following 23-word long intermediate codes in hexadecimal form : F301 4900 : F104 4180 0000 : F104 47C8 0000 F104 4280 0000 :1518 : F302 4141 : FB01 4900 F104 4180 0000 :0200 :0900 :0600 :0007 where " :" denotes the division between units of codes that have definite meanings . The meanings of the above codes are: (variable I) (real 1.0) (real 100.0) (real 2.0) (FOR) (variable AA) (read variable 1) (real 1.0) (add) (store) (FOR end) (end of line) This shows that constants represented in ASCII string forms (for example, "100") are converted to the floating-point format (47C8 0000), and that commands are represented by the number codes and arranged in the reverse-Polish order. The penalty we must pay for the adoption of the intermediate-code scheme is that some commands take a long time to execute, because these commands require the compilation to the intermediate codes at run time . In the KEK NODAL system, programs for compilation at run time are non-resident and need time-consuming disk accesses when they are called . An example of such commands is $DO, which is followed by a string containing a NODAL source program as the command body. However, the use of the commands that need run-time compilation is not frequent ; the overall speed of KEK NODAL for the average program is increased by this modification . The search time for variables is reduced by the hashing method . Variables which appear in a NODAL program are connected to one of sixteen lists according to the hash value calculated from the variable name . The average number of variables in a list thus becomes one sixteenth of the total number of variables. This speeds up the search, since the time needed to search for a variable is roughly proportional to the number of variables in a list . The cache tables in the working area contain the addresses and other necessary parameters of the most recently referenced lines and functions . When the interpreter encounters a line number or a function, it first searches for it in the cache tables ; if the search fails, it then proceeds to search in the working area or in the EFUN table (see the next section). Since the probability of finding frequently referenced lines and functions in the cache tables is high, the search time for these items is reduced. The results of the benchmark test of KEK NODAL on the HIDIC 80E is given in table 1. In this benchmark test, a command was executed N times (N is a large integer, such as 10000) by the FOR loop, FOR I = 1, N; (command) The overhead of the FOR loop was measured by the FOR loop without a command, and the result was

Table 1 Results of benchmark test FOR LOOP SET A =1 SET A = B SET A = A+ 1 SET A = A -1 SET A = A+0.9999 SET A = A/0.9999 SET A = A^0.9999 SET D1(4)=1 SET D2(4,4) =1 SET D3(4, 4, 4) =1 IF B > 1 WHILE B( =1 ; DO 33 .20; 33 .20 RET DO 54 ; 54 .10 GOTO 54 .20; RET $SET C ="A" $SET C ="A" "B" $SET DS(4) ="A" $IF C ="X" ; $SET C =1 $DO "S A =1" SET A = SIN(1.001) SET A = SQRT(l .001) SET A = DEF(1.001)

0.319 ms 0.463 ms 0.668 ms 1.066 ms 1.063 ms 1.066 ms 1.059 ms 1.453 ms 1.150 ms 1.414 ms 1.673 ms 0.771 ms 0.769 ms 0.612 ms 1.152 ms 0.879 ms 1.683 ms 1.134 ms 1.113 ms 3.664 ms 265.306 ms 1.569 ms 1 522 ms 2.805 ms

subtracted from the result of the former measurement. The penalty paid in having to compile the $DO command is clearly evident. 3 .2 .5. Dynamic linkage of data modules and functions

In the KEK NODAL system, data modules and functions are coded in PCL. They are compiled and loaded as core images on the resident user subroutine area or on the non-resident subroutine areas on the disk . The names, the addresses of the load modules, the numbers and types of the arguments of functions and data modules are contained in the EFUN table. When the NODAL interpreter calls a function or a data module, it gets the necessary information for linking from the EFUN table. Since the linking of data modules and functions are buffered through the EFUN table, we can compile, link and load a new function or a new data module independently of the rest of the NODAL system . The only requirement for entering a function or a data module into the NODAL system is to set the relevant parameters into this table. This is done using a NODAL function EFUN . In KEK NODAL, functions are usually subroutines or functions coded in PCL, with a free choice of the number and the type of arguments. Data modules are also subroutines coded in PCL; however, the number, the type and the order of the arguments are strictly determined by the calling protocol from NODAL. I. OVERVIEW OF EXISTING SYSTEMS

34

S. Kurokawa et al / The TRISTAN control system

3.2.6. The CA MAC handler A CAMAC handler, based on the IEEE standard subroutines for CAMAC [14], is implemented in the KEK NODAL system . This handler works under the multitasking operating system, PMS. Simultaneous access to CAMAC by more than one task is arbitrated by the handler; the task which issues a CAMAC command while the computer is processing the CAMAC command issued by another task is made to wait until the end of the latter command. Since the subroutines of the CAMAC handler are callable not only from PCL but also from NODAL, it is easy to make test programs for CAMAC in NODAL. The combination of the multi-computer facility of NODAL and the NODAL callable CAMAC handler, allows the interactive checking of every CAMAC module installed in any computers. 4. Operator console There are five equivalent operator console (OPC) units arranged in a line [15] (see fig. 4) . Each OPC unit

--TV _20"

monitor graphic display

touch-panel

Fig. 5. Schematic structure of one unit of the operator console.

is connected to one of the computers OPO-OP4 and contains two 20" color graphic display monitors (GDM), a pair of touch-panels (TP) and ten 10" TV monitors (TVM). Fig. 5 shows a schematic diagram of one unit .

Fig. 4. Photograph of operator consoles in the TRISTAN control room .

35

S. Kurokawa et al / The TRISTAN control system

CLOSED ORBIT DISTORTION 10

8J-07--03 13 :57 :23 -(MEAS) ''

8

3:Ci=

:H

H Hiffill IS Boom Me Mamma 4 :~ "M = 3 M om n o . non 1 81 1 os a ama= 1so . no a :::?:o 2 : ; ::: ::: HOME :-", ::l:f i~ i :"::"i "s::=":~âi:;i" ~::::: ~~::~: 0 g;9 -:?:::aL:U::: r ::asi9 r . i . . .21.I 4,1111 ` == sz:5::; a I a:::ia Was ~11 s .'. "E s:q i: ;s éa==HHHH asB ~~i 1: -4 E ___ _ _ M _" =s;=:sâ He ag-i s" Ia111 1 :: Cl l" -6 :imi 1 :==:: : :I:a:~"i :uu:=:EE:=::==E _::::::: ::=:=à': oo :::1111:=zs:a:E::as"::::=s== .:3" _8 :iesa::::::::sommom=s:":: :: :"~i B~l : as "IN M :o::E:::0nmauo -10 40 45 50 55 60 65 70 75 10 15 Su rz+ SE MIP NE H IP IP 10 ._.,,_~.........:Eï;; 1o ;~ HIM 2_: _ _:~: ;= =:a= ai:=. 1 sé= :iE == :" ? M:: @I 8 :. ._."....".. " -"._..:::- ""C"""""""-".1. ""-e. "":"": 21Hés a: :°s:: = :aa:" a: s::s::a ~é = : : ...._.._:. ...... == 1É 6 :iâ:"-_:E::::::::" :::::=:_::::_::::_ ::a::::::: _a::i=:a::a::::=s::::?::::::::::ss:::::::e:= : : " 4 :Ea ::~ : 3"M : 11112 :::= ==C~~:::::::E:E_:Z :~~=C~:~C1Ç="=l=:BC:~:Cl:a 2 '_~ ." :.i~ :~;s...: """:sü` : : '::: =::1':::: : "::" ttt..ss;.s. ._.s:r.....sa aa: ! , 0 ï: :a::lasso:: :EE3s: :_:'till~i ""::~a â==:: Es ::é£ : ::E :==:~_ . " :::::::::E:=111111111 eü:. : -2 HE-M . . .-M " i:.0 ~_::::.::::::::::::11-: I :" -4 .. ~;...;..a . mm _. o o :: " " . CC:ZZCC : a, 0 s=sms: :sl ss:1 1 N1 aagsii':_-~ ':âa===s s:=:=" !Z'I 111 -6 :a11:a~ :CCM~ü : :: ::'C ;~~ ! :iC:11:1I:" :'! _ 8 .i-. ::E ~:~=EE~: :9iC s -10 i:Es:: 5 " :10 15 20 : 25 usui. 30 35 . 40.~a: 45,::::.:::::.am 5b 60:1s"_""-"""9""":" 65 70 75 11 so Fig. 6. Hardcopy of a display of closed-orbits on a color graphic display. 6

r-~

u

IN

411011111111S

-a-0'

E E u

4.1. Color graphic display monitor The GDMs (Japan Radio Company Ltd. model NWX 235) are used to display high-resolution graphic data such as the closed-orbit of the accelerator (see fig. 6), etc . A GDM has a resolution of 1024(x) x 980(y) pixels and has the functions of windowing, segmentation of images, movement, rotation and zooming of a segment, etc. One can draw pictures in 16 colors at a time, out of a palette of 4096 colors. The GDM is connected to the HIDIC 80 by an RS232C link. Sixty graphic functions are implemented in the KEK NODAL system to support interactive graphics.

4.2. Touch-panels One touch-panel system is composed of a character video-RAM CAMAC module (CVRAM), a mediumresolution RGB video monitor, a touch-panel, a touchpanel controller and a touch-panel interface CAMAC module. Fig. 7 shows the block diagram of the system. The video monitor is 14 in. 8-color TV monitor. The blue is used as the background color of a whole screen. The E260-13SM type touch-panel made by Elographics Corp. is used, which gives an overall resolution of 4 mm in both axes. I. OVERVIEW OF EXISTING SYSTEMS

36

S. Kurokawa et al. / The TRISTAN control system touch-panel interface module

await for the commissioning of the MR scheduled for the fall of 1986 . For each console computer, a library of 500 NODAL programs has been built up and 200 of them are used for daily operation and study of the AR . The total number of programs is 1500 at present. The majority of these programs were written by ten people in a year from the summer of 1983 ; subsequently, the production of new programs and the modification of the existing programs have been continuous . This rapid accumulation of programs proves the productivity of the NODAL system.

console computer CAMAC serial driver

E 3~ ôEE ôô C d 4 v

âm UÔ

color video display unit

touch-panel c r Il er

touch-panel

touch-panel controller

Fig. 7. Block diagram of the touch-panel system . An operator can select a program or a piece of equipment using the left-hand touch-panel . Forty-two buttons can be displayed on this menu-selection panel. In each button outline three lines of messages, each with up to nine characters per line, can be written. These messages show program names, equipment names, measured values, etc. Since the number of buttons on this panel can be as large as 42, the number of touches to reach the desired action can be kept small. The right-hand touch panel is used to set a parameter of a piece of equipment or to control switches . On this panel, keys such as numeric pads, alpha-numeric keys, on-off switches, or up-down switches are displayed according to the requirements . 4.3 . TV monitor

The TV monitors display various signals from screen monitors in the beam, from cameras in the accelerator tunnels, etc. They are also used to display various information about the accelerators generated by CAMAC CVRAM modules.

Acknowledgments We wish to thank Professor T. Kamei and Professor G. Horikoshi for their support during the work . References

[2] [3] [4] [51 [6] [71 [8]

[10] [111 [12]

5. Summary

[131

At present, twenty minicomputers have been connected to the network and are used to control the AR which was commissioned in October 1983 . In the spring of 1986, the whole control system is to be completed to

[14] [15]

T. Nishikawa and G. Horikoshi, IEEE Trans. Nucl . Sci. NS-30 (1983) 1983 . H. Ikeda et al., IEEE Trans. Nucl . Sci. NS-28 (1981) 2359 . A. Akiyama et al., Proc. Europhysics Conf. on Computing in Accelerator Design and Operation, Lecture Notes in Physics 215 (Springer-Verlag, Berlin, 1984) p. 367. H. Koiso et al ., IEEE Trans. Nucl . Sci. NS-32 (1985) 2068 . S. Kurokawa et al., these Proceedings (1985 Accelerator Controls Workshop) Nucl . Instr. and Meth . A247 (1986) 202. IEEE Standard Serial Highway Interface System (CAMAC) ANSI/IEEE Std 595-1982 . Y.Asano et al., Nucl . Instr . and Meth . 159 (1979) 7. M.C. Crowley-Milling and G.C . Shering, The NODAL System at the SPS, CERN 78-08. B.E . Carpenter et al ., System Software of the CERN Proton Synchroton Control System, CERN 84-16. H. Frese et al ., IEEE Trans . Nucl . Sci. NS-26 (1979) 3379 . P. Haskell, GRACES Reference Manual Version-2. RL84-043 . A. Akiyama et al ., KEK NODAL User's Guide, KEK Report 84-5 (1984) in Japanese. S. Kurokawa et al., IEEE Trans. Nucl . Sci. NS-32 (1985) 2071 . IEEE Standard Subroutines for CAMAC, ANSI/IEEE Std 758-1979. S. Takeda et al ., IEEE Trans. Nucl . Sci. NS-32 (1985) 2062 .