The CTT Project

Electronic Mixer Box

Mixer Box Design Documents:

  • Mixer Box General Specification - [pdf]
  • Mixer Box Input Definition - [pdf]
  • [Excel WB (5.4MBytes)]
  • Mixer Box General Design Tool [ESE Document] - [Excel WB (3.5MBytes)]
  • Link To ESE MB Page
  • General Description

    The scintillating fibers of the Central Fiber Tracker, CFT, are arranged on the surfaces of eight barrels. These fibers are gathered into ribbons, each of which has two layers of 128 fibers. Each ribbon has a single connector to a clear-fiber light-guide. Click here for a PS picture of the ribbons as viewed from the South end of the Fiber Tracker. [PDF] This light guide transmits the light to its cassette end where it is split into two 128-fiber connectors that plug into the top of a cassette. Each Analog Front End, AFE, board therefore sees signals arranged by order of phi in each of the eight layers.

    The trigger requires the fiber channels be arranged into trigger sectors, which are wedges of all eight layers (nine with the central preshower) in phi. The rearrangement of fiber channels from ribbons to trigger sectors is done by the Mixer Box.

    The ribbons are arranged on the detector into a five-fold symmetry, and each of the five parts is called a Super Sector. The pattern of wave-guide bundles from each Super Sector is identical and the pattern of destination cassettes is identical. The routing of signals within each super sector is very complex. Click here for a picture of the ribbons at the end of the detector.

    The discriminated outputs from the AFE are multiplexed into groups of 140 bits per crossing. That's 20 bits per 53MHz-clock tick. These signals are input to one of 4 mixer boards (4 per super sector, 20 total) within the single mixer crate. After sorting, the signals are output, in groups of 189 bits per crossing, to the Digital Front End, DFE, boards.

    There are two constraints on the mixer box. First, it must sort the signals from the orientation by ribbons to the orientation expected by the DFE boards. Second, it must add less than one crossing of latency to the trigger time. This second constraint is to insure that the track lists are received by the Muon L1 processors early enough to not delay their trigger decision time.

    The mixing box sorts the signals in two steps. In the first step, the signals are exchanged over the back plane. In the second step, the signals are arranged by color group for output to the DFE boards. Click here for a PS picture of the Mixer Box data flow. In the above picture, which shows the data flow for an entire super sector, the data from the AFE is input along the left hand edge. Each of the four mixer boards has 17 input links, 15 for the CFT and 2 for the CPS. These links are shown on the diagram in the four groups into which they naturally fall. The B and H layer fibers are located in, and just fill, a pair of cassettes; the C and G in another; and the D and F in another. The E layer just fills one cassette and the A layer just fills 1/2 of a cassette.

    In the first step, the data is sorted across the back plane between neighbor boards. The A layer signals are only shared in the two pairs shown. The B-H, C-G, and D-F are shared between each pair of boards. After this step, the signals have been sorted onto the correct board for output. In the second step, the signals are sorted into the correct link for output. These links are coded by color after the definition devised for the DFE board design. In this color coding scheme the signals in each pair of trigger sectors in coded into six colors. Two of the six are only used in the home sector, two are shared in the internal home sectors and are thus shared within a DFE Mother board, and the last two are shared with the neighbor sectors/sector-pairs. The two that are shared accross DFE Mother boards are the RED and Green groups. Click here for documentation on the DFE and its input data formats.

    The mixing box serves as a router, each input signal is routed to one unique output signal. On each 53MHz-clock tick 1/7th of the data arrives. And on average after 4 clock ticks that 1/7th is output to the DFE. But, only on average. Some of the signals, which are input on clock tick m, are output before the m+4th clock tick and some after. The non-repetitiveness of the geometry and the moving of signals from clock bins means that each of the signals for 1/5th of the detector must be treated separately. Thus, each of the 7,680 bits must be coded through the mixing box individually and by hand.

    A spread sheet outlining the data I/O of the mixing box at a higher level can be found at this site. This document, prepared by Fred Borcherding, shows how the wave guides are plugged into a super sector of cassettes and the routing on the signals from the cassettes to the mixer board. It also shows which signals must be passed over the back planes. And it also shows the number and sizes of the output links.

    A spread sheet , prepared by John Anderson, giving a detailed mapping of the signals from cassette channels through the AFE, on through the mixing box and finally into the DFE is given at this site. WARNING - THIS SPREADSHEET IS 8.7MBytes IN SIZE. He also has notes on the mixer box and his spreadsheet at this site.

     -----------------------------------------------------------------------------------------------------------------------------------------------

    Last updated 2-Sep-99

    Fred Borcherding

    fredob@fnal.gov