CODA Back End

CODA is the DAQ software system currently used by all experiments at Jefferson Lab and with some modifications could be used in Hall D. CODA is a flexible high performance collection of software components which allow dynamic construction of DAQ systems. There are four main components:

• ROC - The Read-Out Controller runs in a CPU housed in the same crate as the digitizing electronics. It is dynamically programmed to read-out events which are later sent to the EB.
• EB - The Event Builder collects event fragments from all ROCs outputs these events to the event transport system.
• ET - The Event Transport system moves events efficiently from one Unix process to another either on the same CPU or across the network. It frees CODA components from knowing the details of where the event stream has to go.
• ER - The Event Recorder gathers events from ET and writes them to permanent storage.

CODA comes with a lot of other tools for configuring and running a high energy/nuclear phyisics experiment (http://www.jlab.org/coda).

The data flow architecture is shown in Figure 3. This design is similar to that of Hall B at JLab except there are several EBs and ERs and a large farm. DAQ system of such a scale have been constructed at many High Energy labs around the world. Currently in CODA, the ROC reads out the electronics for each and every event. In Hall D, the event rate is too high for this scheme to work with a general purpose real-time operating system and CPU. There needs to be some other method that is either built into the electronics or into a single device in the crate to allow the ROC to read out events in blocks. The approximate data rate per crate is in the neighborhood of 20 MB/sec. This can be handled easily by a VME64x system and with sufficient computing power (250 SpecInt) by the ROC CPU equipped with gigabit ethernet. The total data throughput that the main DAQ switch needs is around 600MB/sec. This capacity is currently available in gigabit ethernet switches. The data filter program running on the farm nodes will connect to an EB event transfer system to read events and write them to an ER event transfer system. Control and monitoring software for these nodes and processes will need to be written and interface with the experiment control software. One important issue is how often calibration runs need to be done to keep the filtering efficiency high. The expected data rate out of the farm is about 20 $\,$kHz and < 100 MB/sec. Five of fewer Event recorder processes should be able to handle this. Note that the events will not be written to mass storage in any particular order. Dual ported fiber channel RAID disk will be used as in CLAS to allow the DAQ and the tape system to access the data in turn.

 

Figure 3: CODA Back-End

In summary, the CODA DAQ system can be adapted to work in Hall D. Parallel event building will be added next year. The run-control system is in the process of be rewritten. Hall D's needs will be incorporated into the design requirements. The collaboration should decide on a data format and ensure that it will work efficiently within CODA. The JLAB DAQ Group will ensure that CODA is ready to acquire data at 160 $\,$kHz when the Hall D collaboration is ready to make it so.