At the moment my main focus is the off line calibration of the barrel EMC using neutral pions from 2006.  My (very) rough plan of how to do this is as follows:

1. Understand the pion mass distribution in all its glory.  This includes the pion mass peak, the eta mass peak, and all background contributions.  I believe this will be best done using Monte Carlo, so I am currently exploring these simulations.  See plots below.

2. Once I understand my backgrounds, I can calculate where the pion peak should be.  That is to say, the Etas and the background will tend to shift the pion peak from its true (pdg) value.  I suspect this shift will be towards higher mass.  Therefore, we should not expect our pion peak, in real data, to align perfectly with the pdg value of the pion mass, but, if we understand our backgrounds well enough, we will know where a the peak would lie for a well-calibrated machine.
3. Run my pion finder over all of the long2 and transverse data.

4. For each of the 4800 towers, plot the invariant mass of any pion whose decay photon hits that tower.  For most of the unmasked towers there should be enough pions in each tower to resolve a pion mass peak.  We can then fit these mass spectra with a function constructed from simulations and calculate how well the mass spectrum (and especially the pion peak) conforms to the expected mass spectrum.

5. Calculate a 'correction factor' that would bring each tower to the expected form.
6. Apply this correction factor to the gains for each of the towers.
7. repeat steps 3 - 6 until the desired precision is reached.

So, how does this sound as a plan?

The first step is to look into the simulations, which I have been doing.  Unfortunately, the first pass has not been encouraging.  For example...

 
Sorry about the poor axis labeling.  The X axis is invariant mass (in GeV) and the Y axis is counts.  This is the mass spectrum for reconstructed pions that correspond to thrown pions.  By 'correspond' I mean that a pion is reconstructed within a radius of .05 (in eta-phi space) of a thrown pion.  All of my reconstructed pions have to pass my  usual set of analysis cuts which are:

Fires the L2-gamma trigger*
Energy  of each photon > 0.1 GeV
P_t > 5.2 GeV
Asymmetry < .8
No charged track association.
Z vertex is between -60 cm and 60 cm.
At least one SMD strip in each plane is good.

Note that the width of the pion peak is far too large (note that this plot does NOT include any thrown Etas.)  Compare this to a similar plot using real data.



Sorry about the poor labeling.  However it's easy to see how much narrower this mass peak is, and this plot includes Etas and background.  Taking for example, the valley in between the pion peak and the eta peak (at, say, M = 3.5 GeV) we see that the peak-to-valley ratio is greater than 10:1.  For the simulation plot this ratio is more like 3:1.  I'm not sure why this is, but I am looking into it.


UPDATE (August 8th)

While combing my monte carlo code for bugs that would cause the wide mass peak seen two plots above, I noticed that the towers for the two daughter photons were often far apart, even for events with thrown pions.  So I looked at what would happen if I limited my pion sample to those in which the two daughter photons were no more than one tower apart (R < .05 in Eta-Phi space.)  I hadn't done this previously because I mistakenly thought my trigger was only looking at events in which this was the case anyway.  The results (seen below) are encouraging.



As you can see the mass peak is much narrower and looks much more like the mass peak from data.  Also, and even more encouraging, is that while the total number of found 'pions' is vastly reduced, the pions in the expected peak region seems to be not reduced at all.