It seems to me that ECRB format don't quite convenient for our data
structure. I don't know is it final decision about ECRB format but let me
express some points.
First, for single hits we will spend 1/2 bank space for 0 information.
Second and more important, for reconstraction results will be
more convenient to write hits only and mark each hits as part of shower using
status word. Can be situation when we have 1 inner hit and 2 outer hits from
one shower. How should we store info for such case ? Will be better to write
3 hits one after other and mark them as parts of the same shower. In result
we'll have several groups of hits = showers and then several single hits.
Third: in following reconstruction ( using info from other detectors ) we
can decide that some single hit is also part of some shower. Or some hit
which connected already with shower should be connected with another shower.
For such cases we'll change status word for such hit without reordering all
info in ECRB bank.
In additional will be useful to keep separate strip info in the end of
ECRB or in separate bank.
In present ecfit_hits_showers hits->showers procedure give us probability
for each couple of hits. Using this method we can connect hits with big
distance between them and put little probability for such couple. For one
particle per sector we'll have good result combaining all energy. For many
cluster events we must check probability and make decision using this value.
In this method we have 1 separate hit maximum, all other hits are connected
in couples with different propability. In new ec_fill_bos I'm using distance
between hits centers as parameter for shower identification. This parameter
can be TCL control parameter or internaly fitted. Playing with different
GEANT files we'll chouse the best method, but reasonable output bank structure
can make our algorithm more convenient.
Serguei