Spectrometer FEE specifications:
--------------------------------

Number of Channels: 27+31=58
-------------------

Event Rate: 100 KHz (maximum)
-----------

Dynamic Range: 12 bits (4096 ADC counts)
--------------
Motivation:
Top of Scale for one BPC: 25GeV = 4000 counts
                          6.25MeV/count
Bottom of Scale: 1count<10MeV

We want to be able to measure the Energy scale 
at 12GeV with  <0.5% error. This translates to 
<60 MeV error. This means that the sigma of the 
noise has to be 

3*sigma<60MeV => (sigma)N<20MeV

so that zero suppression does not affect 
the E scale setting. So, the requirement 

1 ADC count < 1 (sigma)N

sets the bottom of the scale.


ADC sampling requirement:
-------------------------

  10 MHz FADC (12 bit)

Baseline Subtraction:
---------------------

  After 96nsec the baseline should be restored to <1%.

  After 2 bunch crossings the baseline should be 
  restored to <0.1%.

  We want to have the ability to set a busy flag if any 
  channel sample crosses a predetermined threshold
  (say x% of the full scale).
  A schematic is shown below:


  PMT -> INT -> ADC -> BUFFER=======>triggered data
                         |       |
                         |       |(trigger)
                         |       |
                        SUMS -> FPGA -(busy)-> PC

  The FPGA should be able to set a (busy) bit every bcr
  which would let us know that the current bcr should
  not be used if the previous bcr has some sample 
  above threshold. So the FPGA has to be programmed in 
  such a way in order to look into the previous bcr 
  and depending on decisions on the previous bcr to set 
  (or not) a busy on the current bcr.

  We have a small chance of pulse pile-up because our
  expected MAXIMUM rate is 100 KHz = 1% of 10MHz:
  
  Occupancy < 1%.

Output PMT charge:
------------------

  For 700V voltage a 5.10^5 gain has been measured. 
  There are 200 photoelectrons/GeV typically produced.
  This means 16pC/GeV.

  Expected maximum Charge = 400 pC  (for 25GeV electron).


Maximum Voltage:
----------------

  Arbitrary: 2-5 Volts

Noise:
------

  Low frequency Noise << 10MHz  (transmission tests show a 
  50Hz noise). Low freq. noise amplitude should be <1% of 
  the full ADC scale. For 4V this is <40mV.

  High frequency Noise < 1 ADC count

Information needed for trigger decision:
----------------------------------------

  We need 4 energy sums: 2 for BPCs (X and Y strips) 
  and 2 for BPCn. For this probably only the central 
  strips are needed. These sums should be made for each
  bcr (ie 10MHz) and triggered events should be place in 
  buffers that can be readout at max frequency of 100kHz.
  These buffers are then written on a PC memory for further 
  analysis.

ZEUS busy:
----------

  We need to know for each bcr if ZEUS is alive (ZEUS busy).
  

6m tagger interface:
--------------------

  From the 6m tagger we need a "good tag" bit that flags an 
  event with an electron passing certain acceptance cuts.
  We also need the reconstructed energy and position for the 
  "good" events in order to perform calibrations.