Data Acquisition system for E896 at the AGS

Paper: 448
Session: B (poster)
Presenter: Schambach, Joachim, University of Texas at Austin
Keywords: data acquisition systems, event building

Data Acquisition system for E896 at the AGS

Jo Schambach
Physics Department, University of Texas, Austin, TX 78712
E896 collaboration


E896 is a search for short-lived composite strange matter
(in particular the H0 di-baryon) and an investigation of
hyperon production in 11.6 A GeV/c Au-Au Collisions. The
experimental setup contains a distributed drift chamber
(DDC), a scintillator hodoscope neutron detector (MUFFINS),
three scintillator Time-of-Flight walls (TOF) and a silicon
drift detector array (SDDA).

The experiment is logically divided into two almost
independent detector systems, each with its own data
acquisition (DAQ) system. The first system will readout
and archive data from the DDC, all beam counters, the
MUFFINS detector, and the TOF walls. The second system is
responsible for the readout and archiving of data from
the SDDA, and is coupled to the first system mainly through
a common trigger system and event tagging scheme.

The main data volume will come from the DDC. The design is
driven by the requirement to read out approximately 8000
drift chamber electronics channels and to use existing
Fastbus electronics. This is accomplished by using 84
multihit Fastbus TDC modules (LeCroy 1879), distributed
over 6 Fastbus crates. Two more Fastbus crates contain
Fastbus TDC (LeCroy 1875A) and ADC (LeCroy 1885F) modules
to readout the electronics of the TOF and MUFFINS. The data
of each Fastbus crate will be read out via a LeCroy 1821
Segment Manager/Interface (SM/I), which includes a fast
sequencer and the capability to make the data available to
several flavors of "Personality Card" on its auxiliary
connector. The personality card transmits the data via 32
differential ECL data line pairs over 100 feet to the
counting house. The data sinks for the ECL cables in the
counting house are ECL/VME dual access memory boards
(LeCroy 1190). Each Fastbus crate is connected to two
1190 memory boards, each housed in a separate VME crate.
This allows the data for consecutive events to be latched
in a "ping-pong" fashion in alternate VME crates, thus
reducing the data rate requirements in each individual VME
crate. Data in each VME crate is buffered in VME memory,
and then archived to an Exabyte Mammoth running at a rate
in excess of 3 MB/sec. Communication between the 2 VME
crates' CPU's is accomplished via a reflective memory
interface (Systran's SCRAMnet). This interface also
connects the two DAQs of E896 by communicating event
information between the two systems. This architecture is
capable of recording about 1000 events per AGS spill
(about 1 second of beam with a 2 second pause).

The design of this DAQ is based on the
understanding that the use of Silicon Drift detectors in
E896 provides the opportunity to test many components that
will finally be used in the vertex detector (SVT) of the
STAR experiment at RHIC. In order to align the activities
for the E896 SDDA with the requirements for the SVT, the
proposed DAQ is based on the DAQ system to be used in the
STAR SVT system test, only multiplied to provide for the
appropriate amount of SDD channels.

For 27 SDD's the system consists of 3 readout (RDO) systems
currently being developed for STAR, 3 VME receivers (STAR
DAQ prototypes) [connected to the readout system via
optical fibers], and a trigger/clock distribution board
(TCD) to interface to the trigger system.

Each RDO system is custom built to service up to 9 SDD's,
for a total of up to 27 detectors. Each RDO is connected
via an optical fiber to an optical receiver card, the
"ROSIE" card developed for STAR as a prototype of the
receiver cards for DAQ. Each ROSIE is hosted by a VME
card (Cyclone CVME964) with an Intel 80960 processor on it,
which is supposed to simulate the role of one of the
processors of the STAR receiver cards. In E896 these
processors perform the zero suppression on the data.
From simulations we expect only about 1% of these data to
be relevant for off-line analysis. It would therefore
greatly reduce the demands on the DAQ, if only data above
a set threshold is archived. We are currently testing the
speed of algorithms to do this. The requirement on these
algorithms should be that they perform the zero-suppression
in at least the same amount of time it would take to
readout all of the data to a storage device. During CHEP97
we will present results from these investigations.

In addition to the SDDA data, a CAMAC system connected to
the VME system duplicates beam information into the SDDA
data stream, so that SDDA data can be analyzed relatively
independent of the data from the DDC DAQ described in
another talk.