The layout of the Data Acquisition system is shown in Fig. 6, its subsystems are compiled in table. The basic design of the system, with the exception of the FADC's which only take intermittent samples, is to take a deadtime free sample of every channel of the system for a contiguous period of 10-100 ms duration. All modules are started simultaneously and stopped simultaneously whenever a single module signals that it is full.
|
A DMA transfer of this data, which we call an ``event-block'', is then initiated to the PPC cards while the next event is collected. All modules have sufficient buffer memory such that data collection and readout of the previous event is done simultaneously. The data rate is around 30 MB/s total.
Simultaneous with the DMA transfer of the data to the PPCs over the VME bus, the PPC's, slightly preprocess the data and transfer it to the other crate or desktop PC's over the PCI and PVIC busses. The PVIC cards can handle a sustained transfer rate of 100MB/s.
Once in the Desktop PC's a major data compression is done. The large TPC data set is scanned for muon tracks. Just the end points and some quality parameters are stored for each muon track and compressed information about any associated MWPC and scintillator hits. In this way the initial 30 MB/s data flow is reduced to a rate that can be recorded on a DLT data tape, 1-2 MB/s. Based on our testrun data, a single 1GHz processor can reduce the data based on our current algorithms. However, we wish to concurrently take sub samples of uncompressed raw data based on a variety of important triggers. For this purpose we envision at least a second processor that can simultaneously scan the same data for various important event types before the raw data is disposed of.
The WFD's are used for the analog readout of the scintillator hodoscope and will have their own VME crate as described in Ref [18]. The system will be be the responsibility of the Boston group, which has constructed similar modules for the g-2 experiment and is now designing a new generation of 500 MHz WFD's.
A three weeks run should result in about 60-100 DLT's (Type III 10/20GB), an upgrade to DLT IV tapes is considered.The primary data stream to tape consists of the end points of clearly tracked, unambiguous muon tracks, and any associated MWPC and hodoscope hits within the time region of interest. Complicated events such as crossed tracks, or uncertain tracking results will be tagged and either read out raw for offline processing, or the entire ambigous region will be disposed of.
