Date: Sat, 19 Nov 2011 23:05:29 -0500
From: Leo Piilonen <piilonen@vt.edu>
Subject: KLM trigger calculations

Dear Gary and Gerard and Iwasaki-san:

Attached please find a figure that shows the locations of the existing FEE crates on the yoke periphery.  

16 crates for the endcap (8 at each end) and 16 crates for the barrel (8 at each end).  

For the barrel, this means that the Indiana FEE boards (for RPC layers 2-14) and Hawaii FEE boards (for scintillator layers 0-1) will sit in the same crate.   
I use this for the calculations below on trigger inputs.

I assume that each FEE transmits only correlated theta-phi (z-phi) hits within its detector module to the KLM trigger, as Gerard had proposed. 
(The trigger cannot use 1D hits for its track reconstruction.)  

This coincidence requirement cuts down on rate from neutron hits and from SiPM intrinsic noise.   I expect << MHz rate on the trigger fibers.

Endcap:
  2 endcaps x 14 layers/endcap x 4 modules/layer = 112 detector modules
  150 SiPMs/module / (1 SiPM/preamp x 15 preamps/carrier x 10 carriers/FEE) = 1 FEE/module --> 112 FEE boards
  15 FEE/crate --> 8 crates [8 crates are unused]

  Option a) One trigger fiber per FEE board --> 112 trigger fibers from endcap
  Option b) One trigger fiber per crate with 15 FEE/crate --> 8 trigger fibers from endcap

Barrel:
  2 ends x 15 layers/end x 8 modules/layer = 240 detector modules
  Assume 1 FEE/module, whether RPC module or scintillator module --> 240 FEE boards  (This implies ~30% wastage of FEE real estate for scintillator modules, using the EKLM FEE design; the 3/10 unused sections don't have to be populated on these boards.)
  15 FEE/crate --> 16 crates

  Option a) One trigger fiber per FEE board --> 240 trigger fibers from endcap
  Option b) One trigger fiber per FEE crate --> 16 trigger fibers from endcap

Combined:
  Option a) 352 trigger fibers to KLM trigger
  Option b) 26 trigger fibers to KLM trigger

The UT3 board has 24 GTH RocketIO inputs and, optionally, 40 more GTX RocketIO inputs on a daughter card.

The associated merger board, designed for CDC trigger info and rates, concentrates 8 optical inputs (~3 GHz) to 2 optical outputs (~10 GHz).  This is overdesigned for the KLM trigger (meant to handle high-volume CDC data), and has insufficient concentration.

Option a1) 352 trigger fibers / 4 level-1 mergers = 88 L1 merger boards.  88 L1 merger boards / 4 level-2 mergers = 22 L2 merger boards
                    --> 110 merger boards and one UT3 without a RocketIO daughter card.  This is not realistic ($10K/merger board x 110 = $1M)
Option a2) 352 trigger fibers --> 6 independent UT3 boards.  Not an easy match to the KLM geometry (8 octants in barrel, 4 quadrants in endcap, 2 ends, ...)
Option a3) Discard ~1/4 of trigger fibers (parts of backward endcap and/or outer layers) to get
                    256 trigger fibers --> 4 independent UT3 boards (one for each KLM section: forward/backward endcap and forward/backward barrel).
Option b1) 26 trigger fibers --> one UT3 board with a RocketIO daughter card.  Algorithm can find tracks that cross from one part of KLM to another.
Option b2) Discard 2 of the trigger fibers (parts of backward endcap and/or outer layers) to get
                    24 trigger fibers --> one UT3 board without a RockerIO daughter card.  Algorithm can find tracks that cross from one part of KLM to another.

Regards,
  Leo