This week I spent a majority of my time preparing results to present at DNP06.
My goal was to spend more time analyzing the drift time to distance conversion
and fitting to the extreme vertical regions in the magnet in both the vertex
reconstruction and momentum reconstruction.  These produced some surprising 
results which will require some careful study.  I also examined the preshower
and total shower energy deposition to try and help identify different types of
particles.  Finally, I have put together a rough version of my talk for DNP.

To reexamine the drift time to distance conversion, I wanted to separate the
tracking away from the fit method on the chance that the tracking was
inherently missing certain regions or had some type of "fitting efficiency".
To do this I worked with the raw drift time spectrum and cut away hits that 
were outside the shower window (to help select on tracking events).  I fit to
the background at some range of times below t=0 and looked at the distribution
above this.  The time spectrum above background was then forced to a flat
distance distribution and then fit the time to distance to a function.

page 1

This is the time spectrum for hits within the shower region.  The black line
represents the background fit.

page 2

This is the fit from time to distance.  The triangles represent the points
found such that the time spectrum represents a flat distance distribution. A
narrow range of angles were chosen (-0.05 < x' < 0.05) and maximum drift
distance of 5mm was used.  In red is a fit of a straight line (representing a
constant drift velocity) that allowed for a universal t0 offset for the entire
plane.  This was used to help the parameterization of the function:

d = c0*Tanh((v*(t-t0)+c2*(t-t0)**2)/c0)

c0 represents the asymptotic value of distance (the maximum drift distance)
v is the constant drift velocity for low t
t0 is the universal time offset for the plane
c2 roughly represents the "sharpness" of the bend from constant v to asymptotic

Values for these fits for each plane can be found in fitparams.txt in this
directory.  Note that most planes have a few ns offset perhaps suggesting that
our t0 offsets may need to be rethought.

page 3

This is the drift distance spectrum.  While most planes are relatively flat,
there are slopes at low distances.  These may be due to the fact that there is
some inefficiency at low drift distances or that the t0 offsets may be 
incorrect.  Improvent of the offsets may improve this plot.  The spike at
higher distance on most plots is probably due to some background getting into
the fit since we are forced to asymptotic distance.

page 4

This is the tracking efficiency measured against drift distance.  The first
chamber seems to have a drop at lower drift distance by almost 10%.  This does
not seem to occur as dramatically in the other two chambers (perhaps due to the
lower rates).  An allowed range of -10 to 210 ns time after offset applied was
allowed to consider hits for efficiency.  The hit was only required to be
present in the cell the track passed through and not necessarily used in the 
fit.

page 5

These are the fits of residuals after this drift time to distace conversion
was used.  Our spacial resolution is somewhere around 300~350um.  A two 
gaussians of two different widths were fit.  The resolution is quoted as sigma
of the more narrow gaussain.  The wider gaussian's amplitude decreases 
dramatically when pairs of wire are ignored or crosstalk cuts are introduced,
suggesting that this wider distribution occurs when a wrong wire in a pair is
chosen.  The wider distribution's amplitude is smaller by about a factor of 10.

page 6

Corrections were made to the wire angles using survey results for rotations
about the z_detector axis.  Variations between chambers should show up as
slopes for the residuals as a function of y, shown in this plot.  There is no
significant contribution from this now.

page 7

Thise are residuals as a function of wire after all corrections were applied
and newer drift time to distance conversion was used.  There appears to be some
jitter from wire to wire, slopes, and other effects that could be due to poor
t0 offsets, poor representation of z position in the database, and nominal
shifts in wire position.  It is my feeling that perhaps there should be
corrections to the position of wires by looking at the centroid of this on a
wire by wire basis and forcing the to be 0 for all wires.

I spent some time looking at fitting the vertex and momentum in the more
extreme vertical positions in the drift chambers.  There have been some
deviations in both of these fits that need to be corrected for in able to take
all statistics.  Currently the region we can fit to only represents about 70%
of the data.

It is important that we first get our vertex reconstruction correct.  If it is
not, we do not have confidence that our scattering angle is measured correctly
which is necessary to fit momentum properly.

page 8

This the vertex reconstruction as a function of the vertical position in the
magnetic midplane in BigBite (where we assume all of our interaction takes
place).  As you can see there are devations when |x_bend|>0.5.  In fact, there
appears to be two separate distributions

page 9 and 10

These are the plots broken up about the left half (y>-4cm) and right half
(y<-4cm) of the drift chambers.  There appears to be antisymmetric effects
occuring as a function of horizontal position in the drift chambers.  This
appears to be a very dramatic shift as a function of position and appears to
take place over a distance of less than 1cm in y.

page 11 an 12

Corrections were made to the position to help this distribution go flat.
For low xbend a quadratic correction was used.  For high xbend a quartic 
correction was used.  These fits were done by hand.  Scripts need to be written
to automate and optimize the process.

page 13

Corrections were made for the momentum reconstruction using the vertex 
reconstruction corrections.  Both corrections were taken to be quadratic and
there appears to be no dependancy on y.  For low x there appears to be a 
fuzzy region where we have 'superelastics'.  This only appears for the y>-4cm
and is possibly due to a poor fit.  Once again scripts need to be written to
automate and optimize this process.

page 14

This is the current momentum resolution using the new drift time to distance,
chamber rotation corrections, and the FULL area of the drift chambers.  It is
currently about 0.9%.  Superelastics still appear and and believed to be due
to poor fits in the vertex reconstruction, which needs to be more fully
examined.  When coplanarity cuts are applied these events reduce in size
suggesting that there is something wrong with the measured scattering angle.

On Bogdan's suggestion I looked more closely at the shower/preshower energy
deposition signature for events in an attempt to help distinguish between
different particles.  I optimized the linear combination of shower and
preshower to give the best measure for momentum on elastic events.

page 15

This is the total shower energy vs p.

page 16

Total shower energy vs p for ps > 500 (electrons).  Note the lack of outliers.

page 17

Total shower energy vs p for ps < 500 (electrons and pions).  There is a second
distribution around 0.6GeV for total shower energy.  This is the "pion"
distribution and heavily overlaps with the electron distribution.

page 18

p-total shower energy for elastics.  This is roughly a gaussian distribution.
Devations are likely due to poor calibration of individual blocks. Resolution
is roughly 9% sigma.

page 19

p-total shower energy for all events.  Note that the distribution is now
skewed due to the addition of the pion distribution.  Note that it does not
separate out well.

page 20
Total shower energy vs preshower energy.  This still appears to give the best
separation between the pion distribution and electron distribution.

I will continute efforts on looking at fixing the vertex and momentum fits, as
I think they represent the most crucial thing to be done in BigBite.  I think
the drift time to distance conversion does not require significant amount of
work left and that t0 offsets need to be looked at, which I think Brandon and
Nilanga will spend some time on.

I am also thinking about spending some time on producing an assymetry for the
final Q2 point in tandem with Rob.  I think that independent analysis on this
will be helpful in judging believablity for our preliminary results for DNP.
While this is restricted by the amount of time I have, I think that, given the
replays are finished with sufficient time before the conference, I should be
able to come up with a realistic number working under Gregg's guidance.

Seamus Riordan
riordan@jlab.org
October 3, 2006