From:	MSUPA::WEERTS       "Man van Staal" 20-DEC-1995 13:37:00.06 
To:	@CONVENER,SPOKES 
CC:	ABOLINS,WEERTS 
Subj:	Large tile trigger problem.

                                    Michigan State Univ., Dec-20-1995

 Dear physics groups:

   over the last few days Dan Owen has found a error in triggering on jets 
with large tiles.  This problem exists for all data using large tiles ( 0.8 x 
1.6 in eta,phi). The cause of this error is the following:  when L1 triggers 
on a large tile a list of large tiles above threshold is transferred to L2. To 
enable the L2 jet algorithm to start from a seed trigger tower ( 0.2 x 0.2 in 
eta,phi), the Et weighted eta,phi position of trigger towers contributing to 
the large tile is calculated. The trigger tower in that position is used as 
the center for constructing a cone.  The energy in the cone is compared to the 
L2 threshold.   The problem is in the calculation of the eta,phi position from 
the large tile.  It has been found that this position sometimes ( frequency 
depends on Et and how close one is to the threshold of the large tile) is 
outside the physical eta,phi boundaries of the tile. This is obviously wrong 
and Dan has indeed found an error in the code that is used in L2 to calculate 
the position.  It is due to not using exactly the same Et in numerator and 
denominator, when mimicking the zero suppression in L1. However the problem 
only exists in L2; at L1 everything is done properly.

The result of this is that the trigger efficiencies you use for jets in your 
analyses may be wrong ( at least in the turn-on part). However it depends 
critically on how you determine those efficiencies.  So please check this and 
for more detailed information you should get your L2 representative involved. 
It is not at all clear how large the effect on your efficiencies will be, 
because it depends on how they are determined and on the cuts used in your 
final analyses.  As long as you safely stay away from the trigger turn on 
point, it most likely is not more then a few percent effect.  But it has to be 
checked.

                                         Harry 
<\PRE>
Gordon Watts -- 12/21/95