Note from Marvin (24 Nov.) about matching of Si sectors to CFT sectors:

I will describe the latest ideas for the SVT processor.

The trigger sector is 4.5 degrees.  This must somehow match into the 15   
degreee sectors of the SVT.  The proposal made at NIU was to compress the   
trigger information into 8 sectors of 45 degrees each so that they could   
be sent to L2 and to the AND-OR network for the L1 trigger.  AT NIU this   
system used the backplane of the front end crates and some daughter   
boards.  It makes much more sense to do this compression in a separte   
crate.  I had planned on doing a 5 to 1 compression and then another 2 to   
1.  This still makes sense but we will now try to put it on one board.    
Note that it is not certain that it will fit on one board.

If this is successful, then we have all the information for a 45 degree   
wedge on one board.  This information normally goes to the L1 AND-OR   
generator and to an MBT if there is a L2 accept.   But it also contains   
all the information needed for the SVT.  Thus, I suggest that we   
implement a 4 way split of the fiber that goes to the MBT and send one   
fiber to each of the 3 SVT sector processors (three 15 degree sectors per   
45 degree one).  Barsotti's group has measured the signal loss in a 4 way   
split and it is around -8.5 dbm.  D0 has an informal spec of -9 dbm and   
the Finnesar low limit is -13 dbm.  Bit error rates increase sharply as   
one goes towards the -13 dbm limit.  -9 dbm looks ok in tests done at the   
CD but we may need to increase the optical power from 0 dbm to 1 or 2 dbm   
to get additional noise margin.

Each SVT sector trigger will get a lot more information than it needs.   
However, the window comparator on each SVX readout fiber is designed to   
only store roads that intersect its ladders.  The rest of the information   
is discarded.  While it is true that 3 times as much information is   
thrown away, it should have no affect on the original design.  Everything   
should work.

Cracks are exclusively a problem for the silicon.  By the time the data   
reaches the compression card, the road is in one or another of the fiber   
sectors even if it projects into a different silicon sector, i.e., a   
silicon sector that is not below a given 45 degree wedge.  This is not   
the case for the silicon.  A track can have a road that spans two   
sections or even is in a completely different section.  The best way to   
handle this is to split the SVX readout signal into 3 and send the same   
data to sectors on both sides of the crack.  Then the there is no   
difference between an edge track and one in the middle - the same hardware
 can process everything.  Alternatively, one could think of sending data   
across the trigger backplane but this will require a special section on   
the trigger boards.  Also, it is likely that sometimes the crack will   
span crates which will then require a special board.  It will likely be   
simpler and cheaper to build more receiver boards.

We may want to ignore sections where the overlap is small such as the   
areas where the silicon detectors overlap.  Note that because there are   
positive and negative tracks, data must be sent to both ways;  one cannot   
just go in one direction.

The main thing that this does to the fiber tracker is to require a board   
that processes 10 sectors at a time.  There is no special part for SVT.    
Tracks are selected for transmission to L2 and sent out over the fiber   
optic line.  The SVT simply gets this information and uses it for track   
roads.  It is possible that the SVT will require more information than L2   
would.  In that case, the larger data set will need to be transmitted.