# E391a DAQ meeting, Dec.13, 2001

Number of channels and data sizes.

 * PMT signal out   : 1000ch
 * HV supply        : 1000ch, 4 CAEN crates + 2 LeCroy crates
 * Amp+Discri. out  : ADC 1000ch 
                      ADCsum 1000/8x2=250ch 
                      TDC 1000ch 
                      -> 2250ch in total
 * Data size    : 4Byte/ch -> 10kB/event(without pedestal supression)
                  100Hz -> 1MB/sec = 80GB/day
                  -> 7 TB of data size for 300 shifts of beam time.


ADC

 * FASTBUS 1885N(96ch) x 12 modules.
 * 2 crates (7+5 modules) due to the maximum-current limitation
    of the power-supply module.
    (40A for -5.2/+5V,  10A for -2V, 6A for +-15V )
    (35A for -5.2/+5V,    A for -2V,  A for +-15V )

  -> 1 crate with High Power supply will be simple and easy to handle.
     Is there a High Power supply module ? 


TDC

 * TKO 64ch x 16 module.
 * 2 TKO crates (10+6 modules) due to the power-supply module.
    (50A for +5.2/-5.4V,   A for +-8V,   A for +-15V )

  -> 1 crate system will be simple.
     TKO crate can stands so high power supply ? May be not.
  -> 1 CPU can handle multiple crates via multiple SMPs on the VME slot.
     No problem with 2 crates. 


Readout System

 * SPARC54 CPU with Solaris8

 * NGF(SIS4100) as the FASTBUS-VME interface.
    o Channel-by-channel data reading is OK with SPARC54.
    o Block-transfer not achieved yet.
    o Pedestal suppression not tested yet.

 * (FPI-DPM (Belle system) still available.)
    (o Channel-by-channel data reading is OK with SPARC54.)
    (o Block-transfer not achieved yet.)

 * SCH-SMP as the TKO-VME interface.
    o Channel-by-channel data reading is OK with SPARC54.
    o Sparse-scanning(no data suppression) by SCH is OK.
    o Sparse-scanning(data suppression) by SCH not tested yet.
    o Block-transfer not tested yet.

 * One-CPU system or multiple-CPU system? (see system.txt)
    MIDAS  ... Can handle multiple frontend ?
    Unidaq ... multiple frontend not supported. Need event collector.

    How to make sure the event number consistency from the multiple frontend?




Computer and Network (see network.txt)

 * Sun workstation as a disk server for SPARC54.
 * Local address. (172.21.xxx.xxx)
   Can use kl-6(Linux) as a gateway.
 * Need PCs for storage server, run cotroller, online monitor, HV controller,..


Storage (see system.txt)

 * Three ideas.
  1) Send data to KEKCC storage system via network. 
  2) Prepare a local DLT drive. 
  3) Two local DLTs. Redundant, easy to maintain than one system. 
 * One DLT drive costs ~1 millon yen.
 * One DLT tape cartridge costs ~ 10,000 yen.
   7 TB of data = 175 tapes ~ 1.75 millon yen.



Trigger

 * Simple multiplicity logic? (NIM, 32-input logic available)
 * FPGA? (Any expert?)
 * Need several trigger condition:
     PhysicsRun, cosmic, beamtest, random, ...



Amp+Discri. module test.

 * Need to study for 
   o Linearity for charge output.
   o Cross talk.
   o Low threshold test ... works as we expect ?
   -> Should be done by Moriya-san (Yamagata Univ.)

 * Need to fix the circuit design. (Need to discuss with Elec. staffs.)
   o DC or AC coupling ?
   o Automatic correction mechanism for offset voltage?


HV control
 
 * Midas can manage to control them.
 * Unidaq ... ?




















Cables see (cables.txt)

 * PMT -> (feedthrough for vacuum) -> AmpDiscri
    - 50¦¸ coaxial cable (RG174U, 40yen/m)
    - ~several meters x 1000ch. 

 * Amp+Discri. -> Counting-hut
   o Shielded with the copper housing from the detector-hall to counting-hut.

   o Charge and charge_sum outputs:
     - 50¦¸ coaxial cable.
     -  90 meters x 1250ch = 100,000m
         Low noise, low attenuation required ? 
           -> Several choices, but expensive, 400yen/m=40 million yen in total.
         Noise and attenuation not so serious ?
           -> RG58C/U (70yen/m, 7 millon yen).
     -  LEMO type connectors for each edge. (LEMO or SUHNER)

     -  ADC1885N has a special connector ... 48ch paddle card x 2.
         For now
           Paddle card -> coaxial(~3m) x 48 -> 17pair flat-connector x 3.
         Better to change it to 
           Paddle card -> coaxial(1~2m) x 48 -> LEMO type connectors.
     -  Need a patch panel before the ADC input.
      
   o Discriminated timing signal:
      - Twisted 17_pair cable. (430yen/m)
      - 30 meters x 64 = 2,000m (860,000yen)
      - 40 pin flat-connector each edge. (64 x 2 = 120 connectors)
      -> Shielded twist-pair might be better.


Grounding, Noise test.

 * Create a good GND line at the detector hall.
 * All the GND line of the signal cables, HV, electronics, power supplies,
   should be connected to this GND line.    
 * Leave the possibility of other grounding way.
 * Need to check the noise level of the signal line with the real 
   setup of cabling.



Layout of the electronics (see layout.txt)