E831  Data Acquisition              19 April 1995
                      General architecture summary

    E831 is a continuation of E687, with somewhat higher (x6) data rates.
    E831 uses the DART data acquisition hardware and software.

    Experiment data streams                         TCP/IP
    |      |      |      |                         ----------> Unix Hoist
    v      v      v      v                        |
    --------------------------> DDD <---> BIT3 <-> Workstation ----> tape/disk
          DART RS-485                VME                       SCSI

	Detector	Digitizer	Front End	Readtime(microsec)
        --------------------------------------------------------
	Target strips	...		DYC+		<10
	Vertex strips	Milano ADC	DYC+		<10
	OE Calorimeter	1881M  ADC	FSCC (2)	<30   ( .2*hits +12 )
	IE/Cerenkov	1881M  ADC	FSCC (2)	<30   ( .2*hits +12 )
	PWC		PCOS latch	DYC+ (4)	<10   ( .1*hits     )
	Straws		3377 FERA TDC	DYC+            <20 ? ( .1*hits     )
	Tag/Latch	misc		SCC/DYC+        <50   ( .6*word + 6 )

    Vertex, PCOS, FERA and SCC output is Differential ECLbus, on up to 50 foot
        cables, as used in E687 for input to LRS 1892 Fastbus memories.
    The DYC/FSCC front ends provide a few events of buffering (16 Kbytes),
	and feed a common RS485 DART cable, with 1 or 2 repeaters.
    The DART feeds a 128 MB 'DDD' DART standard FIFO memory for
        spill buffering. The DDD is located in a dedicated VME crate.
    The standalone Data Logging workstation makes large DMA data transfers
	from the DDD via a BIT3 DMA controller.
    The Data Logging workstation builds events, logs them to disk and tape,
        and distributes them to other workstations for online monitoring via
        the Unix Hoist software package.

   Peak input     - Front end interfaces read in parallel.
		     DYC+ = 20 MB/sec (PCOS/FERA/ECLbus limit)
		     FSCC = 20 MB/sec (Fastbus access)

   Average input  - 1000 events/second, 5 KByte/event = 5 Mbyte/sec.

   Average output - 1.5 Mb/s (10 times E687 1991/2)
		    Modest online trigger improvement or third level event
		    selection could reduce this.

   Output media   - 5 Gbyte 8mm tape is an appropriate medium. We would write
		    about 2000 tapes. We may spool output to disk first.

   Trigger/Log    - Level 3 software triggering will probably not be necessary.
	            Data logging consists of concatenating the event fragments,
		    building a simple index, and writing data to disk.

   Monitoring     - An Indy workstation will be used for Calibration, Control
		    and general monitoring of the hardware. We will check
		    physics results online with a 500+ VUPS "Expressline"
		    processing system.

   Calibration    - We may log interspill calibration data from DART to disk.



  Data Logging workstation considerations:

    Runs Run Control, Logging, Unix Hoist, limited monitoring.
    SGI is the only vendor presently supported by DART for data logging.
    Require VME access.
        Can use BIT-3 Adaptors for VME and GIO (Indy/Indigo,not Indigo-2)
        We grab data via BIT-3 in 1 Mbyte or larger chunks, as needed.
    Spare systems in the experiment are highly desirable.
    Logging to EXB-8505's 2 MB/sec.
    Should spool data to disk, then copy to tape.
        Probably put two or three 9+ GB disks on one SCSI, tapes on another.

    VME Challenges appear to be much too expensive, starting at over $50K
    Crimson - Minimal 64 Mb server too expensive at $40K retail.
        We will borrow a Crimson to replace a temporary 4D/35 for startup.
    Challenge S appears is affordable, at 12K, and has enough slots
        and SCSI busses to handle more ambition expansion. Its GIO bus
        is BIT-3 compatible .
    Indigo - Are short of compute power, and lack BIT-3 compatible GIO bus.
    INDY is ideally priced, has color display, is Bit3 compatible.
        We will have others in E831 for SCSI control.
        With disk spooling both GIO slots would be used, for BIT3/Tape SCSI.

  Calibration/Microstrip/Tagging workstation considerations:

    E687 successfully managed three CAMAC subsystems with independent
    workstations. We will replace three existing VAXes with INDY systems.
    The existing Jorway CAMAC branch drivers will get SCSI interfaces.