Simple cluster counting often results in more hits than the real number of particles due to a wide spread of a cluster, and the trigger decision must be made reasonably fast.
An extensive GEANT-based Monte Carlo simulation study was performed in order to avoid such troubles, and a new algorithm which counts only the number of isolated clusters (ICN) was developed. The algorithm is shown in Fig. . This logic is implemented for each TC (denoted "0"). With this logic, only TC at the upper most and right most corner of a connected cluster produces a final hit signal. Other TCs do not give any final signal because they are vetoed by adjacent TC hits. The logic implemented on a Xilinx FPGA chip provides a clear seperation between Bhabha and other hadronic physics events, and works exceptionally well.
This simple clustering logic is applied to over 132 TC input signals and the number of isolated clusters is then tallied. In addition, we delay the 132 input and 16 output signals to record in a set of FIFO pattern register on the board (CCM). The recorded cluster and ICN bits from the pattern register are read out through the VME-bus to allow a continuous monitoring of the operation of CCM. FPGA counts the number of clusters asynchronously and the timing latency of CCM turned out to be about 50 ns. Five CCM modules are needed to cover all the TCs and each CCM output is merged at the ECL trigger master (ETM) that provides the final number of ICNs. It consists of 3-bits and 1 carry-bit and is used in the physics trigger.
On the other hand, a simulation study indicates that the physics trigger design might not be adequate due to a relatively large contribution frm cosmic rays to the trigger. Thus, we have designed a very simple cosmic veto logic, in which ICNs are divided into four quadrants and arranged to make sure that particles are produced from the interaction point. With this condition, the contribution from cosmic rays could be lowered to about 10 Hz as shown in Fig. (b). ICN3 has been used to trigger physics events in real data taking runs.