The output voltages of each of the cathode preamp-shaper channel on each cathode FEB are sampled every 50 ns and stored in SCA's during the Level-1 latency. The readout of the stored samples is data-driven - they are digitized and read out only when a local LCT trigger associated with the sampled pulses occurs which is time-correlated with a Level-1 Accept. This requirement significantly suppresses data due to random background hits induced by neutrons and photons.
Given the Level-1 accept rate, the LCT rate, the SCA buffer size, and the digitization time of 26 microsec the deadtime of the of the system can be estimated and is found to be neglegible.
The digitized data is sent from FEB's on each chamber to the output buffer on the DAQ Motherboard (DMB) on that chamber. The buffer holds an event record for each and every Lev-1 accept. (The buffer size is designed to be 1 Mbytes) All event records in the buffer are sent by optical link to the DDU on the FED in the central DAQ system.
Based on the estimated (Lev-1)(LCT) coincidence rate, only 3% of the event records (3 kHz out of 100kHz of Lev-1 accept) from a chamber contain digitized data in them. Each of these records contains the Lev-1 ID and the motherboard ID (which indicates the chamber module from which the data comes from) plus typically (8 samples)x(2 bytes/sample)x(96 channels)=1536 bytes of digitized data. Such a record takes about 13 microsec to send via a 1 Gbits/sec link. The remaining 97% empty records takes much less time to transmit per record since they contain only Lev-1 ID and motherboard ID.
For a given event (or lev-1 accept), there will be chambers in which the track segment to be digitized crosses the boundary between two FEB's. There will also be chambers with more than one (Lev-1)(LCT) coincidences. For these "busy" chambers, more than one FEB will be digitized. However, the digitization time for such a chamber remains to be 26 microsec since digitization takes place simultaneusly on these FEB's. Thus the deadtime calculation is equally valid for these chambers. However, the data transmision time of a record from a "busy" chamber is longer. The maximum transmission time for a "busy" chamber record is 5x13 =65 microsec (from a chamber with 5 FEB's each generates digitized data).
The latched discriminator bits from each anode FEB will also be sent to the DMB. The amount of data is negligible compared with the cathode data - 96 bits from a chamber with 1 (LCT)(Lev-1) conicidence. The maximum anode data per chamber-record is 7x96 bits = 84 bytes.
Together with DAQ readout data, the LCT road and timing information that is sent to the Level-1 muon trigger will also be saved in the DAQ readout record.
In summary, there will be 432 optical links (one per chamber) going to the FED crate from the EMU system. The transmitted data is digital. For each Lev-1 accept, there will be a record sent to the FED from each optical link. Because of the low occupancy in the EMU chambers and because only the data due to a track segament in coincidence with Lev-1 accept is digitized and kept, 97% of the records sent on a given link are empty records. The size of the 3% non-empty records ranges from 1600 bytes to 8000 bytes, depending on how "busy" a chamber is.
The option of sending processed information (pulse height and pulse timing) instead of raw pulse samples (8 per pulse) could be considered if it is necessary to reduce the record size and transmission time.