next up previous
Next: Analysis Software Up: Design of a New Previous: Electronics

Data Acquisition

The design of the data acquisition system for the µLan detector is based on the one used in the 2 experiment, described in reference [22]. The software underlying the system is UNIDAQ, originally developed for the SSC Laboratory. It provides a distributed buffer management service as well as a set of tools for collecting, transferring, and storing data.

The design is also motivated by the very high data rate that this experiment will produce. We anticipate that there will be approximately 750000 events per second. We will use waveform digitizers (WFD) to record a snapshot of about 100 ns for each event, binned in intervals of 2 ns. There will be about 60 bytes of data per event. This implies a data collection rate of 50 megabytes per second. A DLT7000 tape drive can sustain a rate of 5 megabytes per second (assuming no compression); consequently, we would require at least ten tape drives writing in parallel. In the course of recording 2  x  1012 events, we would then generate 120 terabytes of data, requiring approximately 3500 tapes. It would not be entirely impossible to collect this much data, but we intend to avoid doing so.

We will group the 50 required WFDs into five VME crates [23]. Each crate will also contain a fast single-board computer running the vxWorks operating system. Approximately 10 times per second, the embedded computer will cycle through all the WFDs in the crate, draining the data from their FIFO buffers. At that point, it will process the data to determine the parameters that describe each pulse. In particular, it will compute the time, area, height, and width of each pulse. The samples for any events which seem to be anomalous in any way will be retained. Also, we will retain the entire set of pulse samples about 1% of the time in order to allow for systematic studies. For all other pulses, only the fit parameters will be kept.

The remaining data will be forwarded to an event-builder computer. The total data rate at this stage that is implied by the previous discussion is less than 10 MBytes per second. It will be feasible to transfer this amount of data over an ordinary 100 MHz Fast Ethernet network to a industry-standard computer running the Linux operating system. This machine would collect the data from the various crates and write it to tape. In fact, it should be possible for it to fill a set of histograms for the pulses which could be fit in a straightforward manner. These histograms would be flushed to tape every minute instead of the entries from which they were made.

We anticipate that a need may arise to read out electronics components other than the WFDs. The data rate for these components should be very small when compared to the data rate in the main detector. It should also be noted that the UNIDAQ system already provides the tools to read out many common CAMAC modules as well as the custom electronics developed for the 2 experiment.

Because of the FIFO memory design in the WFD electronics, the data acquisition process will introduce no dead time in the experiment. Every muon that is delivered can be recorded.


next up previous
Next: Analysis Software Up: Design of a New Previous: Electronics
Gerco Onderwater
1999-05-25