Data Aquisition System (DAQ)
Given the characteristics of the detector, the data-aquisition system (DAQ in the following) has been designed to allow a continous data recording of the events occuring inside the drift chambers, avoiding as much as possible intristic dead time to meet the "continously sensitive" feature of the detector. Strong design issues are that the system must be able to operate in the absence of any external trigger signals and to aquire data from different independent sources containing the fragments of the same physics event.
DAQ LayoutThe data-aquisition system was based on the following assumptions:
- 96 electronics racks, hosting up to 55296 digitizing channels
- local buffering up to 4096 10-bit samples for each channel (1.6ms)
- network architecture merging data from multiple independent sources
- maximum sustained single hit rate of 1Hz/wire (2Mb/s/crate; ~200Mb/s for the whole detector)
- intermediate data storage able to decouple raw data acquisition from pseudo-on-line filtering
- at least four writing workstations to handle the data stream
The adpoted DAQ network is characterized by a two-level switching architecture.
The front-end electronics crates distributed on top of the cryostat have their local CPU connected through a 100Mb/s Fast Ethernet (FE) link to a fast 24 ports FE/GE switch. In this configuration each single chamber of the T600 is mapped to one single switch, allowing, if necessary, a natural topological segmentation of the detector. The final merging of all the data coming from the four chambers is performed through a central 1Gb/s Giga Ethernet switch, connecting all the local switches to the PC farm. A second FE/GE switch connects some utility machines to the central GE/GE switch, such as the slow control units, run control tasks, external trigger interface and tape storage unit. There is also an interface with the cryostat monitoring machines provided by Air Liquide.
The main feature of the adopted design is that the PCs can have flexible roles since they can simply write data on file or execute filtering procedures, or run the event display. Each workstation receives event data until its disk is full and then gives the control to another PC while running the data processing, or, following a round-robin scheme, can process one single event at a time. Such architecture is open for scalability since the number of PCs in the farm can be selected according to processing power requirements and, thanks to the switched network, the bandwidth can be easily distributed among multiple workstations.
The event builder has to merge fragments of data belonging to the same physical event appearing on any of the units and spread over a time window which can go up to one drift length (one or more miliseconds). The builder is based on a mxn node switched network which interconnects all DAQ components. The present layout provides single receiving and merging workstation handling a complete read-out of one T300 (24 read-out units) thus reaching a maximum expected input of ~50Mb/s and safely exploiting the link at one-half of the avaible throughput.