Date: Fri, 03 Sep 2004 15:45:47 -0400 (EDT)
From: Dave Mack 
To: Neven Simicevic 
Cc: Mark Pitt , tony@phys.latech.edu, grimm@jlab.org,
     finn@jlab.org, jmammei@vt.edu, nmorgan@vt.edu, carlini@jlab.org,
     smithg@jlab.org, opper@jlab.org, birchall@physics.umanitoba.ca,
     armd@jlab.org, yl0094a@jlab.org
Subject: Re: Qweak working group - next meeting


        Neven et al,

        Neven's email reminds me that to choose a final collimator design
we need to understand the impact on the absolute determination of 
and .

size/shape: Indeed some of the proposed collimation solutions have long
tails out to higher Q^2. Since we are ruled by the Mott cross section, all
our Q^2 distributions will be enhanced at low Q^2 with a tail toward
higher Q^2. But can we develop a simple criterion for how broad is too
broad? For the proposal, we WAG'ed that we didn't want the Q^2
distribution to be broader than +-50% of the mean. That already looked
very aggressive to me at the time. For a given average scattering angle
and two-body kinematics (ie, neglecting all radiation), this immediately
constrained the lower and upper edges of the ideal angle bite.  We may
have been too conservative, but I think gotchas like backgrounds and
mis-tracking will make it harder to measure the average Q^2 the broader
we allow this distribution to become.

absolute angle determinations: As I think Neven was getting at, we can't
achieve relatively clean angle definitions with collimation (and therefore
relatively clean Q^2 cutoffs) unless we drift to a collimator which is far
enough away to make the projected target size of roughly 5 cm
approximately "pointlike". In the space (and money) we have available, it
makes more sense to approximate this with the rear-most collimator. This
will reduce the sensitivity of the  determination to the absolute
cold length of the target cell and its longitudinal distance to the
collimator, Ztgt - Zcoll. For a hypothetical defining collimator at 1000
mm, the error derivative for  must be something like 0.2%/mm, and for
 it must be 0.4%/mm. (I'm assuming the collimator dimensions will be
known with sub-mil accuracy.) It is true that survey accuracy (in mm!) is
better over a shorter baseline, but I doubt if we win this way; we
increase the sensitivity of  to collimator skew/yaw, and we put
ourselves in the position where an unrecognized fat bubble on a target
window can generate 1% level errors in Qweak(proton).


        regards,

        Dave