Rainer supplied me with a test configuration for the L3 trigger changes
needed to select Y(3S) -> pi0 pi0 Y(1S) [-> invisible]. I have generated
100,000 events in 24.2.1a and I find that out of all events (including
non-fiducial events) the trigger has a total efficiency of 16%. This 
is to be compared to the ~0% total trigger efficiency prior to these
changes.

I apply a loose event reconstruction to these events:

* photons: E_lab>0.030 GeV and LAT<0.8
* pi0s: mass = [0.090, 0.165] GeV/c2 and p*<0.8 GeV/c
* dipions: mass = [0.2,1.0] GeV/c^2 and recoil mass = [9.35,9.8] GeV/c^2

This has an efficiency of 80% on the triggered events.

The pi0s require further refinement to remove junk multiple candidates,
but this is already a great improvement over the status quo.

The total signal efficiency at this point is 13%.

Separate L3 Line Efficiency
=====================================================================

The L3 trigger is broken into two lines which are OR'd to
create the L3OutPi0Pi0Invisible line. These are:

* L3OutPi0Pi03Invisible3Gamma
* L3OutPi0Pi03Invisible4Gamma

It is interesting to see the efficiency on each of these separately.

                    3 Gamma                   4 Gamma
=====================================================================
Trigger Eff         11%                       5%
(relative to all
generated events)
---------------------------------------------------------------------
Reco. Eff           84%                       90%
(relative to
trigger)