The goal is to set up a cosmic test stand which can tag muons with
a timing precision of well below 100 ps. The ultimate goal would be
10 ps.
Once we have that we can start benchmarking other timing devices
agains this cosmic test stand.
This is what the sketch on page 15 of Arturs presentation illustrates :
https://indico.cern.ch/getFile.py/access?contribId=1&sessionId=0&resId=1&materialId=slides&confId=263667

A few more details :
To achieve order 100 ps from a crystal (which has been achieved
in current ECAL, so its known to be possible) we have to control all
aspects of the measurement very well :
1) Time pattern of the energy deposit from the muon. 
    This we can achieve by adjusting the trigger counter geometry such
    that the muons tresspass the crystal in a geometrically small region
    so that the passlength is uniform and equidistant to both ends of the crystal.
    -> Cedric is studying this in full complexity in the CMS environment.
        For the muon passing on a perpendicular trajectory through the crystal
        it is trivial.

2) Time pattern of the scitillation process. This is inherent to the crytal type.
    We can study by comapring eg. LYSO and BaF2 which are almost maximally
    different in this respect. But this pattern is invariant from event to event, modulo
    photo statistics. This we have to check how much we can ignore it.

3) Light propagation through the crystal. Again, for the very simple geometry
    of a muon perpendicular through the crystal this should be simple.
    -> Artur has studied this in detail for more complex situation like actual
        photon showers from a front hit high energy photon, including the
        Cerenkov component. This will become more challenging later in
        eg. a test beam.

4) The readout : photodetector, amplifier, DAQ, etc.
    Again for current ECAL, the systematic limit in the test beam was around
   a few 10 ps with an APD, pulse shaping, 25 ns sampling and subsequent
   timing reconstruction using 10 samples.
   This is what we briefly mentioned in the group meeting yesterday.
   Currently we have a phototube and a scope (with 5 GS rate ?).
   We have to understand this in more detail.

In practical terms it will be very important to have the crystal read out
by two equivalent branches (photo detector, any signal processing, DAQ, etc.)
on both sides. Then, with the muon crossing exactly in the center of the
crystal most of the above issues will cancel out and we can benchmark
both sides agains eachother and agains possible better equipment we want
to get (like fast SiPM) etc.
As an ultimat first test we want to move the crossing point of the muon along the
crystal axis and reconstruct the crossing point from the time difference on both
sides. The LYSO crystal is 10 cm long, this corresponds to 300 ps
Once we have achieved this we can start to benchmark agains another high
precision device (item 2 in the sketch on Arturs slides).
At this point the jitter in the trigger will become relevant, we may switch to
trigger on the crystal itself then. That can come later.