[Halld-physics] PID & Vertex Reconstruction

[Matthew Shepherd]

Hi Dave,

I think Paul and I concluded yesterday (in a few private emails) that we were using different words to talk about an operationally identical procedure.  Re: "PID agnostic tracking chamber", of course you're right, especially for low momentum stuff, hence the "(to some extent)" in my original message.  My past experience is that synching to the machine RF and bunch finding is easily done in the inner layers of the tracker without any PID.  That was at CESR with a 14 ns bunch crossing period instead of 2 ns. 

-Matt

On Sep 20, 2011, at 6:55 AM, David Lawrence wrote:

> 
> Hi All,
> 
>     Just to add my 2 cents...
> 
>     First off, I'm not sure Matt's statement about the tracking chamber 
> being PID agnostic is entirely correct. At lower momenta it becomes very 
> important which is why we fit multiple hypotheses for each track. The 
> statement is definitely true though for higher momenta.
> 
>     I think I understand what Matt is saying about the "native time 
> frame" of the detector. However, I'm not sure I'm seeing what is 
> ultimately gained from doing the matching there. Once you determine 
> there is a match, you have the same problem of how you assign a time to 
> a specific point on the track so that you can calculate the 
> time-of-flight. In either Matt's or Paul's method, one still has a 
> mismatch that is handled by either saying:
> 
> 1.) The reconstructed TOF time is assigned to the POCA of the track to 
> the reconstructed TOF coordinates.
> 
> or
> 
> 2.) Another track fitting pass is done to include the reconstructed TOF 
> coordinates as a constraint (possibly a loose one).
> 
> It seems to be 6 of one, half dozen of the other for this particular 
> aspect.
> 
>     In my opinion, there is an added benefit to matching at the vertex 
> in that the PID chosen for one track may depend on that for another. In 
> the same vein as the "native time frame" argument, I would say that the 
> vertex time projects all reconstructed particles onto a single plane 
> allowing one to compare not only different mass hypotheses for a given 
> track, but different tracks.
> 
> Regards,
> -Dave
> 
> On 9/19/11 2:17 PM, Matthew Shepherd wrote:
>> Hi Paul,
>> 
>> Right -- the point is that the BCAL and TOF hits might not really be along the trajectory but scattered nearby.  You would rather project one track and match to many hits then try to take many hits and project each back to one track vertex.  I'm not sure latter is even defined -- I wouldn't know how to take a BCAL hit that is displaced from a track and projected it back to the vertex.  I don't know what path to follow.  But given a track I can make a best guess extrapolation to the BCAL and compute a time.  I assume that in many cases it will be easy to define a vertex and tracks, but there will be ambiguity in putting BCAL hits or TOF hits on those tracks.
>> 
>> Note that information derived from tagger, SC, and tracking chamber (for the most part) is PID agnostic.  Seems like ti would good to define an event and vertex based on this information first and then do PID by tracking out to the PID devices.
>> 
>> I don't have any proof this is correct, just my instinct.  Since you are writing the code, I think you should go ahead do whatever you feel is best and we'll find out how good it works.
>> 
>> -Matt
>> 
>> 
>> On Sep 19, 2011, at 2:02 PM, Paul Mattione wrote:
>> 
>>> I'm not sure that there is an advantage in using that method.  For the comparison, there is no difference between first projecting the RF time to the vertex and then to each of the FCAL/BCAL/TOF hits, and projecting the RF and each of the FCAL/BCAL/TOF hits to the vertex.  No matter where you compare the track times along the trajectory, you'll have to take the material scattering into account (this is already done when you swim the tracks with the DReferenceTrajectory routines, by the way).  Actually I think comparing the times at the vertex is more convenient, because during the analysis you want to know the time of each particle there anyway to compare the track times to each other.
>>> 
>>> The reconstruction software does save all of the BCAL, FCAL, and TOF hits that position-match each track, so the user can use them to do custom PID later.  But I don't think you want to use the RF timing information to select the BCAL/FCAL/TOF hits during reconstruction, because it may bias your PID.  If you have two hits that position-match your track but have significantly different times, maybe the best thing to do is to use the reconstructed time in the drift chambers to do PID first, and then pick the position-matched hit with the closest time.
>>> 
>>> - Paul
>> _______________________________________________
>> Halld-physics mailing list
>> Halld-physics at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/halld-physics
> _______________________________________________
> Halld-physics mailing list
> Halld-physics at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-physics