Digitizer Data

The µLan set-up involves about 100 waveform digitizers in 10 VME crates (each digitizer has 4 inputs per module). The digitizer data for tile events comprise (i) a 32-bit header word with a 16-bit start time and a 14-bit fill number and (ii) several 3-word data blocks each with channel number, time stamp, and 8 bytes of FADC data. The header word defines the end-of-fill or start-of-measuring time. The 3-word data blocks each contain 16 ns of 2 ns-binned pulse-shape data. A typical event contains three 3-word data blocks (i.e. 48ns of pulse-shape) yielding ten 32-bit words or 40 bytes of data. The data is stored temporarily via a 1 MB fast FIFO memory in the waveform digitizer. The digitizer logic allows for storing of data in both ``singles mode'' (i.e. if either the inner tile or outer tile is above threshold) or in ``coincidence mode'' (i.e. when both corresponding inner and outer tiles are above threshold).

At 4 ×105 tile events/sec, with 80-bytes of inner tile data and 80-bytes of outer tile data, the total data rate is 64 MB/s and per-crate rate is 6.4 MB/s. For 1 ×1012 Michel electrons the total amount of data is 160 TB or about 1500 110-GB SDLT tapes. If the t0 counters are also recorded in the digitisers the data rate and total data will double to 128 MB/s and 320 TB. Also a VME crate with two t0 counters must handle a rate of 64 MB/s (this option of recording t0 counters in waveform digitizers during regular data taking is unlikely).

Compressed Data

Using front-end processors the raw waveform data will be compressed to an energy, time, tile number and maybe other useful quantities (chi-squared, zero-level, etc). Assuming an 80-byte raw event yields an 8-byte compressed event the total rate to the back-end processor is 6.4 MB/s and the total storage for 1 ×1012 Michels is 16 TB or about 150 110-GB SDLT tapes. All these numbers will increase by 50% if similarily compressed data from the t0 counters is also recorded.

Histogrammed Data

Using the back-end processor the data will be histogrammed. Histogramming by tile number will be important in understanding the disortions from precession. Histogramming by tile energies will be important for understanding efficiency variations due to gain and offset.