Re: questions regarding data fitting in the class "TBLRecon"

From: Timothy Smith (tim_smith@MIT.EDU)
Date: Fri Aug 31 2001 - 13:43:43 EDT


Hello Aaron,

        I think I can answer these two.

On Thu, 30 Aug 2001, Aaron Joseph Maschinot wrote:

>
> Hello:
>
> I have some questions regarding the class TBLRecon, for anyone who happens
> to know.
>
> 1) There are multiple member functions of this class that perform fits of
> the data (e.g. WCfit1, WCfit2, and WCminuitDit). Which one is "the best"?
>

        WCfit1 is the method I generally use, it is fast, and gives
                   a reasonable fit most of the time.

        WCfit2 is an idea for a fitter which has not been followed up
                   on. Maybe someday.

        WCminuitFit is a robust fitter, but it is also MUCH slower then
                     WCfit1. I can imagine a situation where if WCfit1
                     failed, one could try WCminuitFit.

> 2) To identify a scattered electron, you need six pieces of info: three
> vertex coordinates (x0, y0, z0) as well as the three spherical coordinates
> of that electron's initial momentum vector (p, theta, phi). When you
> reconstruct the electron's path, it is only possible to determine p,
> theta, and phi. information regarding x0 and y0 is gone due to the fact
> that the target has a nonzero rho value. however, it is still possible to
> detemine a corrected zc value. My question is: is this zc value as
> determined by the TBLRecon class the z coordinate of the point of closest
> approach to the z-axis of the reconstructed track? Or is there some other
> definition being used in the class. I tried to look at the actual code,
> but it's hard to decipher due to a lack of comments and the fact that i'm
> not sure what exact calculations were done, anyway.

        It is true that the vertex might be described by (x0, y0, z0),
        but in general a tracks vertex can at best be described by
        (d0, z0), where d0 is the distance of closes approach to the
        z-axis (rho). So in general there are five parameters to
        fit (p, theta, phi, z0, d0). However, in the fitter I have
        assumed that d0 -> 0 because the vertex is not constrained
        by the target, rather it is constrained by the intersection
        of the target and the beam line.

                The best fit we could hope for if we fit d0 is about

                sigma(d0) ~= 3*sigma(z0)

        for the same reason that sigma(phi) ~= sigma(theta). That
        means we could only fit d0 to 1 cm, but we know the beam to
        a micron.

                So I have assumed that we know d0 best from the beam
        position. This also means that we only have four parameters
        to fit (p, theta, phi, z0) which is much faster then four
        parameters.

>
> Thank you,
>
> Aaron
>

                                Tim Smith

P.S. Ask more questions!
____________________________________________________________________
 Timothy Paul Smith Research Scientist
 MIT Bates Lab tim_smith@mit.edu
 21 Manning Rd. tel: (617) 253-9207
 Middleton, MA 01949 fax: (617) 253-9599



This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:28 EST