With the StEmcSimulatorMaker recently cleaned up and streamlined (thanks again, Adam), it wasn't a difficult task to simulate the effect of cross talk between the fibers in the BSMD readout.  Nominally this would involve each strip losing some percentage of it's energy into the surrounding strips due to cross capacitance in the transmission lines from the strips to the readout.  In particular (eta) strips in the center of the barrel have the longest transmission lines, hence the largest cross capacitance and the largest cross talk.  Phi strips have short enough transmission cables that the effects of cross talk should be negligible.  In accord with Oleg's advice, this effect is taken to decrease linearly from its maximum at |eta| = 0 to zero at |eta| = 1 (in the plots shown below "Cross Talk" refers to the maximum value at |eta| = 0).

Because I'm interested in the effects on photon showers I proceeded slightly differently.  In particular consider a shower with one very high strip and two very small (if not zero) surrounding strips.  Due to cross talk the high strip leaks 2n% of it's energy into the surrounding strips, and since the initial energy in these strips was so small this 2n% increase dominates.  Moreover, the n% in each neighboring strip could then "cascade" into the next neighbor.  The thought was that for narrow enough showers this might be a factor in the final shower shape.

Upon studying the energy deposited by various McTracks I found that this scenario isn't as implossible as one might think.  Many tracks, especially lower energy tracks, deposit energy in only one or two SMD strips and leakage due to cross talk leads to a nontrivial increase in strip multiplicity.

Immediately following are the gory details of the implementation, feel free to skip ahead to the pictures.

For each final state particle in StMcEvent the BSMDE strip with the highest energy deposition due to said particle is found.  2n% of the energy of this strip is leaked to the immediately neighboring strips, n% in each.  The energy in the neighboring strips is recalculated, and n% of this energy is leaked into the next neighbor.  Everything is repeated for the BSMDP strips, and the detector response continues through the rest of the StEmcSimulator (conversion to ADC, conversion back to energy) per usual.

For testing purposes StGammaCandidates were found using a prompt photon sample with partonic pT between 9 and 11 GeV.  Only those candidates coincident with prompt photons in the PYTHIA were included in the histograms.

Now for the results.  First, let's check out the multiplicity of the eta strips.  Even with small cross talk (in bench tests the maximum cross talk was measured to be around 3-4%), there is a nontrivial change in the multiplicity distribution,

This can also be seen in the average shower shape in both the BSMDE and BSMDP.  As expected energy is lost in the peak strip and redistributed to the neighboring strips, with the dominant effect being seen in the overestimate of 9%.  If the current estimate of 3-4% is low, and the actual value approaches 5-6% then cross talk could have a significant impact on the shoulders of the photon showers.

The new code has been written with modularity in mind.  Cross talk can be readily implemented across BEMC towers and preshowers, provided a criteria for included neighbors is decided.  Moreover the exact algorithm used above can be tweaked as desired with minimal effort.