[d2n-analysis-talk] LHRS Trigger Efficiency Study

[David Flay]

Hi Brad,

I've been working with the trigger efficiencies, and I have two methods
(see my latest talk for a description:
http://www.jlab.org/~flay/analysis/2010/LHRS_7_1_10.pdf):

First, we can consider the ratio:

for trigger 3:
eff_3 = bit3/(LT3+LT4)

for trigger 4:
eff_4 = bit4/(LT3+LT4)

However, shouldn't we also include trigger 8 here:

for trigger 3:
eff_3 = bit3/(LT3+LT4+LT8)

for trigger 4:
eff_4 = bit4/(LT3+LT4+LT8)

The results for this method are below:

========== Trigger Efficiencies (Method 1) ==========
Trigger 3: 97.1425%
Trigger 4: 0.0700378%
Trigger 8: 0%

(trigger 8 efficiency is zero since bit8 is prescaled at 100 for run
20676, and the bit8 variable yields 10 events.)

This method poses the question:

"Of all the triggers generated (setting the latch pattern) -- how many of
the good triggers pass through the prescale condition?"

We see with this method (method 1) that the efficiencies do not add to
100%.  I find this to be unsettling.  Then, I considered another way of
calculating the efficiencies:

for trigger 3:
eff_3 = bit3/(bit3+bit4+bit8)

for trigger 4:
eff_4 = bit4/(bit3+bit4+bit8)

The results for this (method 2) are below:

========== Trigger Efficiencies (Method 2) ==========
Trigger 3: 99.928%
Trigger 4: 0.0720461%
Trigger 8: 0%

We see here that the efficiencies <do> add to 100%, and subsequently, the
efficiency of T4 truly evaluates the inefficiency of the T3 trigger.

This method poses the question:

"For each trigger that passes the prescale condition, we set the bit
pattern.  Now, what percentage of those events are the good ones -- the
T3's?"

Clearly, these two methods examine two different ways of looking at the
inefficiency of the T3 trigger -- at the prescale (method 1) and after
(method 2).  Both points are of concern -- since we do not want to lose T3
triggers -- that was the point of having the prescale = 1 (method 1
examines the loss of T3's at the prescale => trigger supervisor).  Of
course, we should also be concerned with how many T3's are in the bit
pattern, compared to the whole -- especially when the prescale condition
on T8 is set to 100, and T3,T4 prescales are 1.  The fact that T3 is not a
perfect 100% in the latter case indicates an inefficiency.

I do have concern that method 1 yields efficiencies which do <not> add to
100%.  This is why I considered method 2.

However, would it be more accurate to consider some linear combination of
these two methods -- so that we effectively consider both issues in our
calculations?

We could also consider the L1A in the inefficiency as well.  However, I'm
not convinced that the speed at which (or equivalently, the time) a T3
gets to the L1A is cause for inefficiency. (see the talk I linked at the
top of the email).

What do you think?

Thanks,

Dave


-------------------------------------------------
David Flay
Physics Department
Temple University
Philadelphia, PA 19122

office: Barton Hall, BA319
phone: (215) 204-1331

e-mail: flay at jlab.org
            flay at temple.edu

website: http://www.jlab.org/~flay
              http://quarks.temple.edu
-------------------------------------------------