lill JU ILI
Contents
1. NS 30 5 1983 ND FORTRAN Reference Manual ND 60 145 8 EN Norsk Data SINTRAN III Real Time Guide ND 60 133 02 Norsk Data 1984 A SCSI Command Library for ND100 T Ramsey Department of Physics University of Oslo 1991 FYSISK INSTITUTT FORSKNINGS GRUPPER Biofysikk Elektronikk Elementzerpartikkel fysikk Faste stoffers fysikk Kjernefysikk Plasma molekylar og kosmisk fysikk Strukturfysikk Teoretisk fysikk DEPARTMENT OF PHYSICS RESEARCH SECTIONS Biophysics Electronics Experimental Elementary Particle physics Condensed Matter physics Nuclear physics Plasma Molecular and Cosmic physics Structural physics Theoretical physics ISSN 0332 55712. Version 4 CERN 1989 TSVME 204 RAM EPROM memory card Thomson Semiconductors DAISY a VME Based Data Acquisition System for the Oslo Cyclotron Laboratory T Rams y B Bjerke B Skaali and J C Wikne In proceedings of VMEbus in Pesearch Z rich 1988 TPU Trigger Pattern Unit User s manual B Bjerke Department of Physics University of Oslo 1988 NIM Interface User Manual P Hey Christensen The Niels Bohr Institute 1987 CAMAC Branch Driver CBD 8210 User s Manual Creative Electronic Systems S A Geneva SYS68K MOTH A User s Manual FORCE Computers Advanced Systems 1988 SYS68K PWR 20 Power Supply Hardware User s Manual FORCE Computers Advanced Systems 1987 43 13 14 15 16 17 18 19 20 21 22 23 24 VALET Plus User s Guide Macintosh version Version 4 CERN 1989 PILS Reference Manual Version 4 CERN 1990 CAMAC ESONE NIM Standard CAMAC Subroutines Version 4 CERN 1987 DOMINO based MF VME interface S Lied Norsk Data 1987 DOMINO Standard Hardware Description ND 14 001 1A EN Norsk Data 1987 DOMINO and NUCLEUS Software Guide ND 820026 1 EN Norsk Data 1988 VMVbus Vertical Bus System VBR8212 VBE8213 User s Manual Creative Electronic Systems S A Geneva PLANC Reference Manual ND 60 117 5 EN Norsk Data 1986 SHIVA a multitask data acquisition system for the Oslo Cyclotron Lab B Skaali et al IEEE Trans Nucl Sci VOL
3. 2 The ND message box rresrerenrssersonssrarensersvrnvrtnersersserssessrtssssserserssesvenssesverserener REFERENCES isso 43 APPENDIX Volume 2 A Event Builder listing B Move Buffers listing C On line sorting and storage C 1 Data acquisition controller DACQC C 2 The MT transfer program MTDMP C 3 The Exabyte transfer program EXDMP C 4 The dummy task MDUMM C 5 The event monitor task EVMON C 6 The sorting task DS RT C 6 1 DSORT MAIN SYMB C 6 2 DSORT ROUTINES SYMB C 6 3 The sorting application routine D Memory banks E Segments F Exabyte and MT status words 1 Introduction The experimental work at the Oslo Cyclotron Laboratory has concentrated on nuclear structure at high intrinsic excitation energy The group has developed a promising technique based on the measurement of decay after single nucleon transfer reactions with the use of py coincidences A proper interpretation of the experimental results is however often difficult due to low counting rates This fact led to the advent of a multidetector system CACTUS CACTUS is constituted of an array of Nal detectors attached to a frame with the geometry of a truncated icosahedron Fully equipped the CACTUS accommo dates 28 Nal ray detectors In addition 8Si charged particle AE E telescopes are fitted around the target There is also space for 2 Ge high resolution y ray de tectors CAMAC ADCs and TDCs were chosen for the array of Nal detector
4. bit of physical address must match logical address Memory bank 1078 in Mpm5 600008 multiport memory is used Addresse range 21600000 21777777 1600008 1600128 1600248 1777778 Figure 4 2 The link segment 300 containing the data buffer and the Message Box The databuffer is loaded from logical address 60000 on the segment Since the program and the data on link segment cannot overlap in address space the size of the program is limited to 60000 Note that the logical address must match the lower 16 bits of the physical address The physical start address for the databuffer is 2166 0000 which corresponds to page 10730 The corresponding memory bank is 107 Figure 4 3 shows the out line of the memory banks comprising the Mpm5 multiport memory The total size of the memory is 16 MBytes However only the lower 4 MBytes can be accessed from the DOMINO controller ND 100 adr Mpm5 adr 1200 0000 0 1400 0000 200 0000 DOMINO CONTROLLER 1600 0000 400 0000 WINDOW 2000 0000 800 0000 4 MByte 2200 0000 1000 0000 4200 0000 4000 0000 Af acdiesses Me in octal representation One memory bank is 64 KW Le 20 0000 words octal Figure 4 3 Layout of the memory banks of the Mpm5 multiport memory Thedata segment must be fixed contiguosly in memory As the ND 5000 monitor will occupy all free pags in memory after start up the reservation must take place prior to starting the ND 5000 monitor The command FIXC 300 10730
5. box through the Move_Buffers task 5 2 The ND message box A similar message box is located in the Mpm5 multiport memory of the ND5800 computer Figure 5 4 shows the layout The field Event Rate is defined by the 15 0 Address message box 1 3F C002 massage box 2 ma SF C004 message box 3 Event Length 3F C006 message box 4 3F Coos message box 5 IF COOA om won message box 8 3F C010 Figure 5 4 The ND message box data rate monitor task EVMON The field named Command is used to transfer commands between the data acquisition control programand the on line sorting system MT flag is the semaphore field Event Length is calculated by the on line sorting task The field named SO_flag is a flag used to abort the sorting task when a new databuffer is waiting to be transferred to the Nord memory The next two fields are the same as in the VME message box The three last parameters are all defined by the on line sorting task 42 REFERENCES n 2 3 4 5 f6 U 8 9 10 11 12 Mod 4418 V CAMAC 8 Input ADC User s Guide SILENA Societa Industriale per l Elettronica Avanzata Mod 4418 T CAMAC 8 Input 1OC User s Guide SILENA Societa Industriale per l Elettronica Avanzata Fast Intelligent Controller FIC8230 User s Manual Creative Electronic Systems S A Geneva VALET Plus User s Guide Version 4 CERN 1989 VALET Plus Hardware Guide
6. communications bus Octobus on the ND side The VME device looks like a DOMINO to the MF system and the MF system looks like a VME master to the VME system The controller is designed as a VME master only using the SCB68172 bus controller chip 3 2 2 Internal VMEbus crate The DOMINO controller may either be used mounting one VME card on the top of the motherboard or by mounting a separate 3 slot VME backplane to the rear of the MFbus crate In our case we needed 2 slots to accommodate the VBR8212 VBE8213 modules Unfortunately our computer is not equipped with the proper cabinet the MAXON cabinet and there is not sufficient space on the rear of the crate The problem was solved connecting the VME backplane to the MFbus backplane with a flat ribbon cable It is obvious that this violates the VMEbus specifications However we have experienced no problems so far An ancillary 5V power supply for the VME modules is located inside the 3 slot VME crate Note that one of the VME slots is defect at our installation 3 2 3 VME VME link The VME VME connection is realised using the VMVbus concept of CES 19 which extends the VMEbus to a multiple crate environment The transfer band width of this system is in excess of 8 MBytes s branch lt 10 meters Transfer on the bus is mapped through a 1 MByte window 24 The VMVbus is a multiplexed bus for the full 32 bits address and data of the VMEbus The electrical signals are transmitted differen
7. computer is called the I O server running a utility called the Bridge The link between the VALET and the I O server is made via a RS232 V24 serial link Both systems request services from each other across this link using a RPC Remote Procedure Call technique 2 3 1 Event Builder The Event Builder task is coded in PILS 14 The actual coding is done on the Apple Macintosh using the program editor QUED M When designing the pro gram attention was put on speed Coding elegance was sacrificed for the sake of maximum throughput Asa result the main read out loop contains no procedure calls Test has shown that calls to subroutines give rise to a considerable overhead in PILS Figure 2 4 shows a simplified float diagram of the task A double buffering technique is applied in order to achieve high throughput The two data buffers are 32kW long a word length of 32 bits is used The event loopstarts with the detection ofan event a master gate provided by the TPU master module Before any read out is performed a pile up flag is checked Ifa pile up condition is found the event is discarded and all digitalization devices reset The hit pattern words are fetched from the TPUs Each pattern word is divided into four hit pattern nibbles again to gain speed in checking which ADCs TDCs to read out A nibble is a four bit entity Read out of CAMAC devices is done using calls to the standard CAN AC library 15 Finally the ADCs and TDCs
8. is inserted in the LOAD MODE MODE file which is run automatically after a computer restart 35 4 2 2 The local control program The local control program DACQG accepts the basic acquisition control com mands from the SHIVA system and notifies the relevant tasks Communication withSHIVA isdoneusing theintemaldevice 212 Communication with the sorting and the MT dump tasks is achieved using monitor calls AT and ABORT and through the Message Box Exabyte operations are performed using the SCSI Command Library 24 Terminal I O is on terminal 44 SHIVA terminal A listing of the program source code is given in Appendix C 1 4 2 3 Magnetic tape transfer task The transfer of raw data to magnetic tape of 6250 bpi STC type is achieved by the ND 120 task MTDMP The task utilizes the monitor call ABSTR MON 131 for the transfer This call implies a DMA transfer to the datafield of the MT unit it is thus a very fast way of moving the data The task is responsible for the handshaking with the Move Bufiers task of the DOMINO controller Location 2 of the Message Box is used for this purpose listing of the program source code is given in Appendix C 2 4 2 4 Exabyte transfer task The transfer of raw data to the Exabyte EXB 8500 is done by the ND 120 task EXDMP The task utilizes the SCSI Command Library for the transfer This task is responsible for the handshaking with the Move Buffers task of the DOMINO controller Location
9. mode Command Note that command prorapt in MoniCa debugger is gt and in PILS command mode Command The message shown above Welcome to is printed by a script named STARTUP PIL which is automatically run at PILS start up After a warm start you will enter PILS command mode immediately when an application is running the screen will be blank Issue CTRL C to return to PILS command mode It may be usefull to rur the STARTUP PIL script manually to set up the correct path select USE from the PILSfile menu According to the documentation VALET Plus is not multifinder compatible However itis possibleto run under multifinderif VALET Plus is quitbefore another Macintosh application is started The quickest way to quit the BRIDGE is to click on the close icon gt Bridge d r7 STARTUP PIL 2 1 1 Preparing for a run Before the acquisition is started it is highly recomended to perform some basic test functions and run through the set up for CAMAC devices Two applications are available for this purpose GAMPARI for CAMAC devices set up and TEST The corresponding compilation and loading scripts are named CAMPARI CMD and TEST CMD They are located in the folder CMD Select the USE command from the PILSfile menu to activate them TEST Issue the command USE Test cmd Test_prog or use the menus as described previously The following menu will appear i Oslo Cyclotron Laboratory i Test Programs Test TPU Test S
10. stricktly apply it is not a semaphore as defined by Djikstra The flags and other communication parameters are stored in message boxes in memory Two massage boxes have been implemented One is located in the VME processor board memory and one in the ND5800 multiport memory The first is accessible by the Event_Builder and the Move_Buffers task the latter by all ND tasks and the Move_Buffers task 5 1 The VMEbus message box A sketch of the message box located in the FIC 8230 on board memory is shown in figure 5 3 The message box is found in the upper part of the FIC 8230 memory slave port The message box is constituted of 10 words each of 32 bits length Address message_box 0 buffer address 201F FFDO message box 1 buffer length 201F FFD4 massage box 2 semaphore 1 201F FFO8 message box 3 semaphore 2 201F FFDC message box 4 201F FFEO message box 5 V vmestatus 201F FFE4 message box 6 201F FFE8 message box 7 201F FFEC message box 8 201F FFFO message box 9 201F FFF4 N U Not Used Figure 5 3 The VME message box 41 The buffer address is computed by the Event Builder task The value is read by the Move Buffers task The two databuffers have one associated semaphore each The field named V vmestatus is defined by the Event Builder task It is set to 1 when it is running 0 otherwise The field V_ndstatus is defined by the ND tasks and takes the same values These fields are exchanged by the ND message
11. to compilefoad the ND 120 tasks Data acquisition control program MT data transfer program Exabyte data transfer program Simulates handshaking without MT dump Event monitor and alarm program Block data containing the data buffer Shadow process for the on line sorting task Mode file to compile and load sorting system On line sorting main program ND 5800 Subroutines Data buffer and message box declarations Declaration of sorting spectra Range curve Gain and shift values for on line sorting Sorted spectra histograms are stored as disk files on user KJF SWAP Table 4 2 The ND 120 5000 source code files 39 5 Handshaking The acquisition system is constituted by several task running on loosely coupled processors It is evident that these tasks must synchronize their activities by some sort of handshaking mechanism Figure 5 1 shows a simplified sketch of the three major parts of the acquisition system the front end the DOMINO controllerand the Nord computer Figure 5 1 The tasks of the acquisition system The handshaking is implemented using flags or semaphores in memory Inter rupts are not used in any part of the system Figure 5 2 displays the use of such flags for task synchronization Event Builder i shared memory VME tront end DOMINO conkroller Mem Move Buhara Local DOMI controller MPbui s MON pot memory Figure 5 2 The use of semaphores in the DAISY system The term semaphore does not
12. two ADC TDC channels Figure 2 8 shows how the patterns from the three TPUs are allocated Note that the channel number equals bit number 1 TPU1 has been allocated to NIM devices Only the odd numbered channels are used for data read out Channel 1 and 3 are used for the particle telescopes The 8 telescopes are multiplexed into two groups aand b The first data word contains the AE energy the second the energy deposited in the end counter E Channels 5 7 9 11 13and 15 may be used for other NIM interfaced detectors as for example Ge counters Here the first data word gives the energy while the next contains time information The even numbered channels may be used as logical bits The actual meaning of sucha bit must bedefined in thesorting routine They are typically used to denote asingles or a coincidence event TPU 2 and TPU 3 are allocated to CAMAC ADCs and TDCs Again each pattern bit corresponds to two data words The first is an energy that is an ADC is read out The second is the time hence a TDC is read out This allocation scheme may be changed to suit the experimental conditions The only part which will be affected is the sorting routine lt bens lt NE Men m lt fes lt toc lt fos 7 lt tos a lt NS lt 18 iocos Toca lt 1 ADC2 2 lt Ta EY Mer lt Ne Toer CET ances E T E T E T E T E 4T E T Li T E T E T E T E T T E T E T E T E T m am am Am 4am am a
13. 2 of the Message Box is used for this purpose A listing of the program source code is given in Appendix C 3 4 2 5 Dummy handshake task A dummy task which takes care of the handshaking when no storage medium is supplied listing of the program source code is given in Appendix C 4 4 2 6 Data rate monitor task task EVMON calculates the data rate and gives an alarm signal when the datarate drops to zero The calculation is performed with a variable time interval The start value is 30 s If this interval is to short to give a reliable value it will be increased by 15 s The maximum interval is 10 minutes The alarm will go off when no databuffer has been received within the interval The tone is constituted of groups of 5 beeps separated by 2 s After 50 beeps the alarm will go off for timebase seconds and then start again In cases of extremely low event rate it may be necessary to disable the alarm This is done by typing ABORT EVMON in SINTRAN command level 4 2 7 On line sorting The purpose for the on line sorting task named DSORT is to sort event data into histograms spectra These spectra are mainly used for monitoring the experi ment The sorting is rather CPU demanding thus the ND 5000 processor is used for this purpose Creating a real time task on the ND 5000 requires a ND 120 shadow process named DSORT DRIVER This very simple program basically consists of one single statement the NDSOOF call to P
14. F LOAD PLANC PL 68020 100 LI E D LI E U EX Downloading of the code to the DOMINO controller can unfortunately not be done through a mode file The reason for this is that there is no synchror ization mechanism between the DOMINO controller and SINTRAN The following commands must be given nd domino domi monitor open path server 50 soft place kjf daisy ces run t m CR CR e Please note that the domino monitor is accessible for user SYSTEM only 4 Sorting and data storage On line sorting and data storage is done on the ND 5800 mini computer This computer has two CPUs a 32 bit ND 5000 CPU and a 16 bit ND 120 CPU The latter is responsible for I O operations This part of the acquisition system consists of several real time tasks some run on the ND 5000 part some on the ND 120 The storage devices available is Exabyte EXB 8500 and STC magnetic tape The storage capacity of the Exabyte is 5 GByte in 8500 mode and 2 5 GByte in 8200 mode The capacity of the magnetic tape is about 180 MByte 4 1 Controlling the data acquisition The operator control of the data acquisition takes place on theSHIVA 21 console terminal Terminal 44 The acquisition system has been interfaced to the SHIVA command input task A sub menu choice L in the top level menu implements the commands The commands are DAISY LOAD UNLOAD DAISY START DAISY PAUSE DAISY STOP DAISY STATUS DAISY DUMP DAISY CLEAR DAISY SNAPSHOT roummoom
15. ILE Ed RDC s Choose A E Test SILENA 4418 7 TDS s D Test CAMAC ete Initialize CACAC Crate EXIT from testprogram TEST gt A The TPU test option launches a sub menu The results of the test are observed by inspecting the front panel LEDs B Test the CAMAC ADCs Two functions are provided an internal test function and a test using external input signals The internal test is intended to show whether all channels are working The external test requires some input signals from the detectors Data are read without using the hit pattern consequently most channels will bee zero This option may be used to see if all channels receive data Another useful application is to check the offsets Zero energy i e no signal should give zero as a result Frequently one will read numbers like 4094 6 and so on This implies that the offset should be adjusted This is done by the program CAMPARI C The same test for the CAMAC TDCs D Test the CAMAC Branch Highway bus The Branch Highway connectors are unfortunately rather fragile It is therefore highly recommended to check it before starting the acquisition The test function writes bit patterns to a register in the Borer Display Unit located in slot 1 and reads them back Any bit errors will be reported The test runs 10000 cycles before terminating E Reset the CAMAC This option issues a Z cycle on CAMAC dataway CAMPARI The purpose of this application is to set the parame
16. NO9100105 UNIVERSITY OF OSLO DAISY the Oslo Cyclotron Data Acquisition Systern T Rams y Department of Physics University of Oslo Norway OvP Received 1991 08 30 Report 91 21 ISSN 0332 5571 DEPARTMENT OF PHYSICS REPORT SERIES EL a h i ud OE U d ee dU Le n ZEN ma TTL a n Ju LW N gt EE an E Meg TETTE Preface This document is intended for anyone that will use the Oslo cyclotron data acquisition system or wants to get some deeper understanding of the structure of the system Each chapter is opened with a more general description The contents should be known to all users More technical information is found in the last part of the chapter The source code listing is found in a separate volume DAISY Volume 2 Appendix available at the cyclotron laboratory Oslo 2s Teu d Tore Ramsey Contents 1 Introduction SPUR e Ceo PENE E NT EA 3 2 VME front end system esreersrereesesreneernnsanssvessensevssearesanerersennnsevasennensanvensaversessensern 2 1 Getting started oes 2 1 1 Preparing for a run 2 1 2 Starting the data acquisition sse tenter 9 2 1 3 Resetting the VME system 9 22 Hardware ss 11 2 2 1 Processor unit 11 2 2 2 Trigger pattern unit 12 2 2 3 NIM interface system 13 2 2 4 CAMAC interface 14 2 2 5 Power supply a
17. O controller address is then given as 1F6000 2 Set bit 31 803E C0900 The multiplication by 2 is due to that MC68k processors use a byte as the basic addressable unit while ND uses a word 16 bits The DOMINO controller can access only 24 bit 16 MByte on the VMEbus directly When a standard address is used the address map for VME will be from 100 0000 to 1FF FFFF Bit 24 is set to 1 to route to the VMEbus address space The VMV system performs a mapping through a 1 MByte window s an example consider the first location of the message box which is located at 201F FFD0 In slave mode the address is 1F FFD0 The VMV link maps the address issued from the DOMINO controller as follows c f 19 EF FFDO DOMINO address OF FFFF VMV mask 10 VMV sets MSB Which gives 1F FFDO the address in slave mode A full listing of the Move Buffers task source code is found in Appendix B 3 3 2 Compilation and loading Compilation and linking of the Move Buffers task is done running this mode file on the ND 5800 computer end BE Rae ses Sena nanan eee ee ME Er Mode file to compile and load the Move_Buffers task File kjf daisy move buffers mode del fi kjf daisy ces nrf End planc pl mc68 i com kjf daisy move buffers kjf daisy ces LINK LOAD REL DOM ces DEL DOM ces SET DOM ces SET SEGMENT NUMBER OB OPEN SEG ces LOW ADDRESS 600000B D LOW ADDRESS 400000B P LOAD ces NR
18. OFF TPU2 0B lt gt sw4 2 1 OFF TPU3 0C lt gt sw4 3 OFF others ON 15 14 13 12 B 7 6 2 1 0 3 Write Z3 AKT e e tst LeCr Enable Test w Lecroy discriminators Etst Test the event line opt Optional LED opt Optional LED Int En Interrupt enable Figure 2 3 The TPU master status register Note that the TPU modules must be placed in special slots i e slot 7 8 and 9 as the ECL supply and termination voltages are supported here only Currently one master and two slave modules are installed in the VME crate ne extra module configured as master is available There are no extra printed circuit boards Figure2 3showsthe TPU status register in read and write mode respectively The current settings are indicated 2 2 3 NIM interface system An interface from VME to NIM ADC s is provided using the NBI design A comprehensive description of this system is found in 91 The system consists of three parts The NIM Controller is implemented on a double height VME card The interface to the VMEbus is found on this module The VME base address is defined by switches BA8 BA31 Currently the address is set to EFFF80 This is done by setting BA20 i e switch 5 on component U43 to ON while all other switches are OFF We utilize only the standard VME address a 24 bits address Note that the address lines 24 31 are not correctly connected on the PCB board This must be corrected if a full 32 bits ad
19. TUS D Enab l e Disable TPUS o EXIT Ctri C STOP data acquisition DAISY gt After initialization all three TPU modules will be enabled If one or more modules are not to be used i e they will not have any cable connected to the ECL input connector they must be disabled An unconnected module will make the system hang Starting the acquisition is simply done by typing A The time and date of the start up is printed out Stopping the data acquisition is done by CTRL C The program is restarted by typing RUN 2 1 3 Resetting the VME system If problem should occur during loading of the VALET Plus system or if the system otherwise should get stuck a reset cycle must be issued Press the upper of the two buttons on the FIC 8230 front panel extended console The button is indicated by an arrow on figure 2 1 Figure 2 1 The VME front end system On top the ADC interface crate is visible 10 2 2 Hardware block diagram of the components comprising the front end part of the acqui sition system is shown if Figure 2 2 Figure 2 2 Block diagram of the VMEbus front end system 2 2 1 Processor unit The processor module used is the FIC 8230 Fast Intelligent Controller from CES 3 It is based on the MC68020 processor running at 16 67 MHz The FIC 8230 has a dual bus architecture which permits simultaneous local computation and direct access transfer from VMEbus to the global on board memory The unit is equippe
20. are reset and the Event Builder is ready for next event t The pile up flag is set by the input signal from the Quad Pile Up Reject NIM module A pile up condition occurs when two pulses overlap thus producing a spurious energy signal 15 Resat ADC s TDC s Reset butter poinler Read TPU1 PATT Y Read ADC s 8 TDC according lo Paitem Store data in butter M Repeat for TPU2 TPUS e Rese ADC s amp TDC s Figure 2 4 Float diagram of the Event_Builder Measurements of the processing speed has been undertaken The measurements are carried out monitoring the INHIBIT signal from the master TPU on an oscil loscope By inserting jumps to the reset statement on various spots in the code the time elapsed for each operation can be found The read out time for an event with two parameters only is about 215 jus a three parameter event will take approximately 240 us Based on these numbers the maximum read out speed using a typical event length is about 180 kBytes s 16 On basis of these measurements we also observe that at least 701s will elapse before any read out of ADCs TDCs can take place After this period all ADCs TDCs will have finished their conversion Consequently it is not necessary to test this condition 2 3 2 Memory allocation Data buffers The data buffers are freely allocated in PILS workspace The actual start address defined at compilation time is found using the PILS statement addres
21. d to terminate when the DOMINO controller signals that a new buffer is waiting to be transferred to the ND computer In this way the sorting will not slow down the data acquisition Since the sorted histograms are intended to be used to monitor the experiment it is not vital to have all data sorted A listing of the sorting task is found in Appendix C 5 1 and Appendix C 5 2 An example of thecurrently used userdefined subroutine is found in Appendix C 5 3 The source file is named SPROG SYMB and is found on the user KJF SORT Theon line sorting system is loaded running the mode file DSORT MODE found on user KJF DAISY Go to the user KJF DAISY and type MODE DSORT MODE Give the name of the application routine file name in reply to the question It is recommended to test compile the application routine in advance Use the compiler switched UNASSIGNED VARIABLE CHECK ON and ARRAY IN OEX CHECK ON to minimalize coding flaws 4 2 8 Survey of the ND 120 5000 files The Nd part of the acquisition system consists of several files All files are located on the user KJF DAISY the password is VME Table 4 2 provides a list of the source code files Binary files and segment files are not included DAISY 100 LOAD MODE DACQC SYMB MTDMP SYMB EXDMO SYMB MDUMM SYMB EVMON SYMB DAISY BLOCK SYMB DSORT DRIVER SYMB DSORT MODE DSORT MAIN SYMB DSORT ROUTINES SYMB BUFFER DECL DECL SPEC DECL DECL RANGE DATA GAINSHIFT DATA Mode file
22. d with a 2 MByte DRAM plug on module and the MC68881 floating point co processor The FIC 8230 has master and slave VMEbus ports and a master only VSB port OurCPU board has been configured to run the VALET Plus system 4 developed at CERN The VALET Plus firmware is delivered on 16 EPROMS Two 128k x 8 11 bit EPROMS holding the MoniCa code is mounted on the CPU board using a piggy back module 5 The 14 remaining EPROMS are mounted on a separate EPROM socket card the TSVME 204 6 The FIC 8230 is connected to the Macintosh SE through a RS232 link 2 22 Trigger pattern unit The Trigger Pattern Unit TPU constitutes an interface to the trigger logic of the experimental set up 7 The TPU module supplies information on which detec tors actually giving signals The list of active parameters is called the pattern word The pattern word is used by the read out program to set up an event structure and to read out the digitalization devices full description of the module is found in 8 The documentation available at the cyclotron laboratory contains the latest updates and fixes The TPUs were originally designed with interrupt capabilities This has however neither been needed nor implemented STATUS REGISTER FFFFX0 16bit read write 15 14 13 12 B 7 0 lem elector Current setting cw control word strap 29 1 lt gt OFF ADC code Strap27 28 21 0 lt gt OFF TPU base address U43 TPU1 0A lt gt sw4 2
23. dress is applied The PCB layout was performed on our CAD CAE system using the Visula software irom RACAL REDAC The cards were manufactured by EB Elplex One NIM Controller module is installed in the VME crate and one spare module is available In addition there are several printed circuit boards available The NIM Controller is connected to the ADC Controller througha flat ribbon cable The ADC Controller mainly contains timing logic which is not used in our ap plication The ADC Controller is located in a separate single height Euro crate PCB layout of the ADC Controller was performed on the institute s CAD CAE system using the Visula software from RACAL REDAC The cards were manufactured by EB Elplex One module is in operation another is available Several PCB cards exist The ADC Interface modules are located in this crate one module for each NIM ADC There are 16 modules available additional PCB cards may be obtained from NBI The single height crate has a power supply of its own a VERO PK 25 monovolt delivering 5A on 5V The crate backplane is a home brew a flat ribbon cable with 64 pins A C row euro connectors fitted 2 2 4 CAMAC interface Interface to CAMAC devices is accomplished through the CBD 8210 CAMAC branch driver from CES 10 It is connected to the CCA2 2110 crate controller through Branch Highway The branch number isset to 1 on the CBD 8210 The Branch Highway cable suffers from rather fragile connect
24. ed into parameter memory when the CAMPARI application is started As the parameter memory is cleared after a power off of the CAMAC crate the program should always be run to load the contents of SETUP DAT So far the ADCs and TDCs have proven to be verv stable as far as the offset is concerned The values currently saved to SETUP DAT should probably be correct Tabel 2 1 presents the current values from SETUP DAT The numbers in paranthesis are the actual register contents It would of course be more convenient to combine all test and set up function into one single application program The small amount of memory 2 MBytes available on the FIC 8230 makes this impossible ADC1 ADC2 ADC3 ADC4 TDCi TDC2 TDC3 TDC4 Common mV 20 4 20 4 20 44 204 ULD mV O 255 0 255 0 255 0 255 0 255 0 255 0 255 0 255 LLD mV 0 0 o 00 og o0 o0 O0 0 Offset ch 4 132 6 134 4 132 6 134 0 128 0 128 0 128 O 128 0200V 255 12V 10096 of full scale 0 of full scale 128 O channels Table 2 1 The current settings of the parameter memory SETUP DAT 2 1 2 Starting the data acquisition The front end system part of the data acquisition is started by typing the command Es Event_builder USE Event cmd in PILS command mode The following menu appears on the screen i slo Cyclotron i i Data Reais tion System I A START data acquisition B DUMP SE data buffer buffer one C Print STR
25. in coincidence with as NA1SP Nals in coincidence with as TGE1SP Summed TAC with gate on as TNAISP Summed TAC with gate on as 1 General purpose singles spectra S8 Table 4 1 The on line sorting spectra 32 4 2 Program architecture In this section the tasks comprising the ND part of the data acquisition system is described Both the ND 120 CPU and the ND 5000 CPU are utilized The ND 120 CPU takes care of the I O intensive tasks while the ND 5000 CPU is used for heavy calculations i e the on line sorting All programs are coded in ND Fortran 77 22 Figure 4 1 displays the overall architecture of the various real time programs A local control program DACQC takes care of the interface to SHIVA The com munication between the SHIVA command handler task and DACQC is imple mented using the internal device mechanism in SINTRAN SHIVA MAIN User Command input 2 3 d d Ii 7 CE A iM H Figure 4 1 The program architecture of the ND 120 ND 5000 tasks The common data namely the databuffer and the Message Box is accessible to the tasks using the link segment mechanism 4 2 1 The link segment The link segment mechanism permits several real time tasks to have access to a common data set 23 Figure 4 2 shows the outline of the link segment The segment used is segment number 300 which is named DAISY An inspection of the link segment status is obtained giving the SINTRAN command LIST SEGMENT DAISY Lower 16
26. k 32 bit words This gives a transfer rate of about 2 MBytes s in good agreement with the valuecalculated using the numbre of processor cycles and clock frequency 16 67 MHz The choice of programming language for the DOMINO Controller was an easy one PLANC is the only cross compiler excisting The source code is compiled on the ND 5800 MC68020 PLANC compiler linked with the corresponding libraries and then downloaded to the DOMINO Controller Figure 3 2 shows a flow diagram for the main loop of the task running in the DOMINO Controller named Move Buffers The Move Buffers task acts like an intelligent DMA controller 3 3 1 Memory mapping The DOMINO controller has direct access to the physical memory of both the VMEbus and to the ND Mpm5 multiport memory Addresses with bit 31 set to 1 are taken as MFbus addresses Note that a task running on the ND 120 processor must use an address offset when accessing the Mpm5 memory due to the local ND 120 memory The offset is 1200 0000 ND 120 Mpms DOMINO controller TOS mm 476000H BO3E CO00H 200 1774 0000B MESSAGE BOX 2176 00008 7400 47E000H 803F CO00H Figure 3 3 The address mapping in the DOMINO controller B denotes octal H hexadecimal Figure 3 3 displays the addresses of the data buffer and the Message Box viewed from the DOMINO controller The data bufferstart address is 21660000 seen from a ND 120 task which gives the physical Mpm5 address 766 0000 1F6000 The DOMIN
27. lace Library This call will start the ND 5000 task The Place Library must be loaded together with the driver task Fortran is chosen as the sorting language The user part is implemented as a subroutine which is linked to the main sorting program Another way of implementing a user defined sorting algorithm is to use a dedicated sorting language Examples of this approach is the previously used TONE sorting language the EVAL language used at NBI and LINDA a sorting language for ND 500 5000 computers currently being developed by the Bergen group The advantages using a Fortran subroutine is that coding is done in a language well known to the physicists The compiler is well tested and supposedly flawless The disadvantages with this approach is that thesorting probably is less effective It also lacks specific mechanisms as for exainple gate to ease the coding process Another problem is that Fortran gives the user to many possibilities Remember that this subroutine Is called for every event Using for example I O statements will slow down the sorting dramatically A certain degree of self imposed discipline is thus necessary 37 The implentation of the sorting tasks is done using an event matrix to isolate the user subroutine from the details of the buffer structure and handshaking The layout of this matrix is shown in Figure 4 4 COMMON event matrix 0 8 0 15 Figure 4 4 The event matrix The sorting is designe
28. m m am Am am am Am sim m am He lt Toc27 Figure 2 8 Pattern map A full listing of the Event Builder task source code is found in Appendix A 20 2 3 5 Compilation and loading STARTUP PIL CLEAR PAGE SET INDENT 3 SET PATH disk DAISY cmd PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT PRINT EVENT CMD CLEAR ALL CLEAR PAGE PRINT loading event builder n HHAH HHHHHH gt Welcome to the DAISY front end system The following programs are available USE Event cmd Data Acquisition USE Campari cmd CAMAC ADC TDC Set Up USE Test cmd Test Suite FETCH MAC PHI2 disk DAISY Programs Event builder COM RUN 21 I I I I I I I I 3 VME to ND link The transfer of data and handshake signals between the VMEbus front end system and the ND 5800 computer is realised using a VME to VME link A description of hardware and software is presented 3 1 Operating the link The software running in the link between the front end and the MD 5800 computer requires no user interaction Hence this task should be invisible to the user It is none the less necessary to check that the system is running Thestatus of the task is easily checked by inspecting the LEDs BR3 and AS onthe CES VMDIS 8003 VMEbus monitoring unit The LEDs will flash with an interval of approximately 3 seconds when the DOMINO controller is running If no ac ti
29. nd crate 2 3 Software sss 2 3 1 Event Builder 2 3 2 Memory allocation 2 3 3 Data format 2 3 4 Pattern bit allocation 2 3 5 Compilation and loading serene 21 3 VME to ND link 3 1 Operating the link 3 1 1 Restarting the DOMINO controller 3 2 Hardware 32 1 The DOMINO controller 3 22 Internal VMEbv crate 3 2 3 VME VME link 3 2 4 Architecture 3 3 Software 3 3 1 Memory mapping 3 3 2 Compilation and loading esee 28 4 Sorting and data Storage mersseraversavarsersavernvrarsnensrnsensevesrenserernsrarrarensnsvasensersnesevenee 30 4 1 Controlling the data acquisition 4 1 1 The commands 30 4 2 Program architecture 33 4 2 1 The link segment 34 4 2 2 The local control program 4 2 3 Magnetic tape transfer task 36 4 2 4 Exabyte transfer task eese nenne tente tentent tnnt 36 4 2 5 Dummy handshake task roerssereerersarseververersenserssrsserserastessererseneeseveensen 36 42 6 Data rate monitor task 37 4 2 7 On line sorting 37 4 2 8 Survey of the ND 120 5000 files cccscsscssssereceesosessssversaveasanesnansessone 39 5 Handshaking seseesserersnvsessnsssasvnsensesserenserastassorssnevenennaensvsesenessenneverassetasnrennerseesennese 5 1 The VMEbus message box m 5
30. ops 4 1 1 The commands DAISY LOAD UNLOAD This command must be issued prior to any other DAISY command The com mand is particularly useful when a new sorting application routine is to be loaded DAISY START Starts up the data acquisition The acquisition may run with NONE MAGTAPE or EXABYTE storage medium DAISY PAUSE Temporarily stops the data acquisition No filemark is written The acquisition may be restarted with DAISY START DAISY STOP Stops the data acquisition A filemark will be written to the output medium The user will be asked if the medium should be unloaded DAISY STATUS Prints out a page of status information An example is shown below Acquisition last started 1991 02 26 09 53 54 SHIVA acquisition is RUNNING 3k VME acquisition is RUNNING Storage Medium EXABYTE 3k EXABYTE Position 8324 3 Records accumulated 238 Number of bad records 0 k Fraction of records sorted 100 Average datarate 81 events s The EXABYTE Position is the absolute record number counted from beginning of tape on the Exabyte cartridge Filemarks are not counted The fraction of events sorted depends on theevent rate During low rate particle gamma runs all data will be sorted In singles runs it will typically drop to 50 DAISY DUMP Dumps the sorted spectra histograms in memory to disk file on the user area KJF SWAP Two dump modes are available A singles dump will write
31. ors The cable should not be touched unless absolutely necessary An extra Branch Highway cable 2 m is available 2 2 5 Power supply and crate The VME crate is delivered by FORCE Computers 11 It is equipped with full 21 slots P1 and P2 backplanes supporting 32 bit data and address Each signal line is terminated with two 3302 470Q resistor networks The VME system is powered by a FORCE SYS68K PWR 20 primary switched power supply 12 It supplies the following output voltages 5V 90A 12V 20A 12V 5A A DC DC voltage converter fitted on the rear provides for the voltages 5 2V and 2 0V needed for ECL circuitry See 8 14 2 3 Software The software of the front end system is developed using the VALET Plus system 13 Valet Plus is a modular VMEbus based microcomputer system developed at CERN for use in applications like testing of electronics equipment control and data acquisition The VALET Plus architecture isolates instrumentation specific hardware and software ina VMEbus crate which is called the VALET The VALET is driven by a MC68020 processor which runs application programs under a resident monitor MoniCa using ROM based PILS Portable Interactive Language System and standard libraries It can access instrumentation buses like CAMAC FASTBUS VMEbus and GPIB Support for standard peripherals as well as the interface to control the VALET are provided by a personal computer in our case a Macintosh This personal
32. s Absolute VMEbus addressing is not utilized as calls to the VMEbus library introduce considerable overhead However one factor must be observed As the VMV link operates through a 1 MByte window the buffers must be contained within a 1 MByte boundary in memory The Event_Builder features a command C Frint STATUS to examine the start and end addresses of the data buffers FIC8230 on board DRAM 20002000 201FFFFF BUFFER 1 address space 2010A660 2012A65C BUFFER 2 address space 2012A660 2014A65C Figure 2 5 The current data buffer addresses Figure 2 5 shows the current start and end addresses for the two buffers Any changein the Event_Buildercode may cause the buffers to be moved in workspace MC 68020 MC 68881 MoniCa Begin gt SYSCONF wik SYSTEM CONFIGURATION ok Total RAM space 2040 kbytes First RAM address 200020 Last RAM address 201FFFFF Percentage of RAN for PILS workspace 55 Additional Histogram space 0 kbytes Number of extra RPC messages allocated 1 CAMAC Branch driver without DMA for branch 1 Figure 2 6 The VALET Plus system configuration 17 If the buffers should fall outside such a I MByte boundary this can be adjusted by changing the percentage of RAM used for PILS workspace This is examined using the MoniCa command SYSCONF Figure 2 6 displays the current configu ration Message Box Unlike the data buffers the Message Box is locked in memory ata given location in VMEbu
33. s address space The start address of the Message Box is set to 201F7FD0 The first location of the Message Box contains the data buffer start address The layout of the Message Box is described in section 5 2 3 3 Data format The data format is fully under program control It may thus easily be changed to fit future application A sketch of the data format is presented in figure 2 7 Note that only the lower 16 bits are actually used in the processing despite the fact that the Event Builder operates with 32 bits wordlength DATA FORMAT zm m ocu azm lt m Figure 2 7 The data format 18 The first word of the event gives the event length The example shown in figure 2 6 has a length of 11 words The bits 12 15 are set to 1 as a tag for the event header Note that the identification tag itself does not provide a unique identifi cation pattern words may though unlikely have these bits set however since the event length is known problems are avoided Following the event header is a TPU identification word Each TPU has a associated code TPU 1 has the code 800A TPU 2 is 800B and TPU 3 is 800C The next word isa pattern word each bit identifying a given detector Thescheme is that each bit set will give rise to two datawords That is read out of two digitalization devices will take place 2 3 4 Pattern bit allocation As mentioned in the previous section each bit of the pattern words are associated with
34. s due to their lower cost per channel For the particle counters and Ge counters high resolution NIM ADCs have been used Each detector gives rise to one energy parameter and one time parameter Thus a total of about 80 parameters are present The counting rate was estimated to reach 100 kByte s for the highest beam intensities In order to meet these demands a new data acquisition system had to be designed The system was named DAISY an acronym for Data Acqulsition SYstem Emphasis was put on modularity both in the software and in the hardware The VMEbus system was chosenas the building stone for the front end system The host computer a ND5800 from Norsk Data was already present An overview of the system is presented in figure 1 1 VME my Domine mp 2 Be MC68020 rom Je MC68020 Figure 1 1 Overview of the data acquisition system 2 VME front end system The front end part of the acquisition systems is responsible for the read out of digitalization devices pattern word generation and the creation of event buffers First an introduction to the operation of the front end data acquisition system is given The following chapters describe the hardware components and the soft ware of this system 2 1 Getting started The front end system is operated from the Apple Macintosh First of all ensure that the VALET crate is powered on and that theconnection between the FIC 8230 CPU board and the Macintosh is present On
35. ters for the SILENA 4418 V and SILENA 4418 T CAMAC ADCs and TDCs 1 2 These devices have a parameter memory which holds information on Upper Level Discriminator ULD Lower Level Discriminator LLD Offset Common Threshold The latter is applicable for ADCs only The parameters loaded will survive a CAMAC reset Z cycle but will be zeroed after a power off The program is named CAMPARI an acronym for CAMac PARameter Input To launch the program issue the command USE Campari cmd The following menu will appear Reading set up data from file SETUP DRT i Oslo Cyclotron i I Experiment setup Set CAMAC ADC parameters Set CAMAC TOC parameters SAVE setup to file Initialise CAMAC Crate EXIT from setup Choose 000D CAMPARI gt a Selecting A pops up the sub menu for ADC set up Campari i CRHRC ADC Parameter Input i A Set COMMON threshold B Set Upper Level Discriminators C Set Lover Level Discriminators D ERE OFFSET values Choose CAMPARI The use should be self explanatory There is a choice between individual i e channel by channel set up or common set up To see the current values select individual mode Typing CR in reply leaves the current value unchanged typing a K moves you one level up in the hierarchy Afterthe valuesaresuccessfully entered they may be saved toa file SETUP DAT found in the folder DAISY The contents of this file will automatically be load
36. the 1 D spectra 1 S52 only The multiparameter dump will write all 1 and 2 D spectra except 51 532 list of the spectra is given in Table 4 1 DAISY CLEAR Clears the spectra in memory The spectra on disk files will not be affected DAISY SNAPSHOT Prints out some few events from the current databuffer The data is shown in hexadecimal representation C f chapter 2 3 3 for interpretation In order to make the spectra accessible for the SHIVA system they must first be dumped to disk files using the DAISY DUMP command This will take 40 60 seconds message notifies the user whenthedumping is finished The data may then be imported to SHIVA using the READ SPECTRUM command for 1 di mensional spectra and 2DIM READ for the 2 dimensional ones First ofall define the default directory and user by giving the command 31 SET DEFAULT DIRECTORY PACK ELEVEN KJF SWAP As an example consider the 2 dimensional spectrum THICKSP First we must define a 2 dimensional spectrum in SHIVA SET 2D SPECTRUM 2048 8 We have now defined the spectrum SP2DIM Next issue the command 2D READ THICKSP SP2DIM to move the spectrum into SHIVA workspace X Dim Y Dim Comment ESP E Counter DESP AE counter EDESP Particle telescopes THICKSP i Particle identification GESP Ge counters TGESP Ge time spectra NASP Nal counters TNASP m _ Nal time spectra PSP KE Particle spectrum P1SP Particles in coincidence GE1SP Ges
37. the VMEbus side cable 9051 should be plugged into the serial port marked TERMINAL This cable should then be connected to the modem port of the Macintosh Now it is time to launch the Bridge application The program is found in the folder DAISY To start it double click the Bridge4 0 icon After a few seconds the VALET Plus login dialogue window will appear The settings shown are the appropriate ones Generally there should be no need to change tem Click on the OK button or give RETURN to continue E Welcome to the VALET Plus Bridge V4 0 Connection to VALET via Communication on Mac Modem Port O7 bits O Mac Printer Part 8 bits Transmission baud rate RTS CTS protocol 619200 O Enabled O 9600 9 Disabled 4800 O Trace Selection DJ Issue Login Request o NI amp s Bridge4 0 What then happens will depend on the state of the VALET There are three scenarious Cold start i e a VMEbus reset has been performed warm start and finally warm start with a running application First consider a cold start The dialogue will then look like HC 68020 HC 68881 MoniCa Version 1 1 from 89 4 7 Cold start Begin gt PILS Welcome to the DAISY front end system I I I The following programs are available i USE Event cmd Data Acquisi tio I l l l USE Campar end CAMAC RDC TDC Set up USE Test cmd Test Suite Command clear editstyle for proper operation with the Mac Leaving immediate
38. tially on two twist and flat cables Each signal line is terminated in both ends by an active terminator the plug on module VBT 8214 Up to 15 crates may be connected each crate is identified by a hexadecimal number 1 F In our set up the internal VME crate in ND computer is number 1 and the front end acquisition crate is number A 3 2 4 Architecture Figure 3 1 shows the architecture of the VME to Nord link All transfers between the two systems are initiated from the Nord crate holding both the VBR 8212 and the VBE 8213 units The FIC 8230 processor module of the front end crate acts as a VME slave during transfers Figure 3 1 Block diagram of the VME to Nord link 3 3 Software The software ofthe DOMINO Controller runs asa stand alone task The software is developed in the programming language PLANC 20 a member of the ALGOL PASCAL family of block structured languages PLANC is mainly inteded for writing system software such as operative systems and compilers PLANC allows for in line assembler code This is important in the current application as access tu aosolute physical memory addresses is performed It is vital to make the buffer transfers as fast as possible The VME front end Buffer 1 sem shore loop Nord semaphore loop Figure 3 2 Float diagram of the Move_Buffers task system is locked during data buffer transfer The actual transfer time has been measured to 65 ms for a buffer of 32
39. vity is found the controller must be restarted 3 1 1 Restarting the DOMINO controller The DOMINO controller is restarted from the ND 5800 computer It is necessary to be logged in as user SYSTEM in order to get access to the DOMINO MONITOR Ask the system administrator if the password is unknown The procedure will be as follows nd domino domi monitor DM open path server 50 Connected to MC68020 based DOMINO DM soft DM place kjf daisy ces Placing PACK TWO KJF DAISY CES DSEG Placing PACK TWO KJF DAISY CES PSEG DM run Type T M CR CR t m lt CR gt CR Data Buffer Eater Transfer of data buffers from VME system to ND 5800 computer e DM target status Controller status running Cache zon Memory pratection on DM exit Unfortunately it is not possible to run this procedure as a mode file This is due to a synchronization problem between the DOMINO controller and SINTRAN 3 2 Hardware 3 2 1 The DOMINO controller An interface between ND 5000 computers and VMEbus has been developed by Norsk Data 16 A prototype has been installed on the laboratory s computer It is implemented asa MFbus card located ina master slot in the MFbuscrate inside the ND 5800 computer The interface moduleis designed according tc the DOMINO concept 17 18 which is a standardization of I O controllers for the 5000 series The controller inter faces the MFbus the main bus of the ND computers and a serial
Download Pdf Manuals
Related Search