% Created 2021-01-11 Mon 17:29
% Intended LaTeX compiler: pdflatex
\documentclass[presentation]{beamer}
\usepackage[utf8x]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{graphicx}
\usepackage{grffile}
\usepackage{longtable}
\usepackage{wrapfig}
\usepackage{rotating}
\usepackage[normalem]{ulem}
\usepackage{amsmath}
\usepackage{textcomp}
\usepackage{amssymb}
\usepackage{capt-of}
\usepackage{hyperref}
\usepackage{pdfpages}
\usepackage{xcolor}
\usepackage{graphicx}
\usepackage{xspace}
\usepackage{libertine}
\usepackage{fontawesome}
\usetheme{Boadilla}
\setbeamertemplate{navigation symbols}{}
\hypersetup{colorlinks}
\setbeamerfont{alerted text}{series=\bfseries}
\usetheme{default}
\author{Brett Viren}
\date{\today}
\title{Advanced 2.5D Trick}
\hypersetup{
 pdfauthor={Brett Viren},
 pdftitle={Advanced 2.5D Trick},
 pdfkeywords={},
 pdfsubject={},
 pdfcreator={Emacs 27.0.91 (Org mode 9.3.6)}, 
 pdflang={English}}
\begin{document}

\maketitle

\begin{frame}[label={sec:org8b26e7c}]{2.5D trick basics}
Calculate \(\mathcal{R}^{2D}_i, i \in (1, ..., N)\) on 2D slice across strip/wire at coordinate \(z_i\) along the strip/wire direction.  

\begin{description}
\item[{wires}] each slice represents a different 2D infinite-wire model with different pitches per plane and different inter-plane phase.
\item[{strips}] different sampling of the ``2D single hole with micro wires'' geometry.
\end{description}

The ``trick'' is to combine \(N \{\mathcal{R}^{2D}_i\}\) into one
\(\mathcal{R}^{2D}_{sim}\) and then average that over impact positions
to form \(\mathcal{R}^{2D}_{sp}\).

What is the *optimum combination``?
\end{frame}

\begin{frame}[label={sec:org4973b1d}]{Signal processing of simulation}
Simulation is fine-grained 2D convolution:

\[  \mathcal{M}_{sim} = \mathcal{N}_{sim} + \mathcal{S}_{sim} \circledast \mathcal{R}_{sim}  \]

Signal processing is course-grained 2D deconvolution:

\[ \mathcal{S}'_{sim} = F_{sp} \circledast \mathcal{M}_{sim} \circledast \mathcal{R}_{sp}^{-1} \]

Substituting.

\[ \mathcal{S}'_{sim} = F_{sp} \circledast \mathcal{N}_{sim} \circledast \mathcal{R}_{sp}^{-1} + F_{sp} \circledast \mathcal{S}_{sim} \circledast \mathcal{R}_{sim} \circledast \mathcal{R}_{sp}^{-1} \]

Ignoring noise.

\[ \mathcal{S}'_{sim} \approx F_{sp} \circledast \mathcal{S}_{sim} \circledast \mathcal{R}_{sim} \circledast \mathcal{R}_{sp}^{-1} \]
\end{frame}

\begin{frame}[label={sec:orgaf2b26a}]{What is \(\mathcal{R}_{sim} \circledast \mathcal{R}_{sp}^{-1}\)}
Reminder of domain of the \(\mathcal{R}\)'s:

\[\mathcal{R}_{sim} \to \mathcal{R}_{sim}(i, w, t), \mathcal{R}_{sp} \to \mathcal{R}_{sp}(w, t)\]

\begin{description}
\item[{\(i\)}] counts impact positions around wire \(w\)
\item[{\(w\)}] counts over in-range wires, eg \(\in {1 \pm 10}\)
\item[{\(t\)}] counts drift ticks
\end{description}

When converted to frequency space:

\[\mathcal{R}_{sim} \circledast \mathcal{R}_{sp}^{-1} \to \frac{R_{sim}}{R_{sp}}\]

\(\mathcal{R}_{sim} \circledast \mathcal{R}_{sp}^{-1}\) is a kind of normalization of \(\mathcal{R}\) in sample bins of impact positions.
\end{frame}

\begin{frame}[label={sec:org4ddb5dd}]{SP+sim approximately invariant under change of \(\mathcal{R}\)}
Ignoring noise, to first order, speaking loosely, etc:

\[ \mathcal{S}'_{sim} \sim F_{sp} \circledast \mathcal{S}_{sim} \]

\begin{center}
\alert{Output of SP on sim is approximately independent of \(\mathcal{R}\)}
\end{center}

This is like the ``too perfect SP on sim'' problem with LArSoft 1D
model.  LS has 2 ignored dimensions, WCT has only 1.

\vfill

So what?  I'm not sure.  It means that det/sim comparisons can not
decide \alert{optimum 2.5D combination}. And, that gives an idea how to use
data to pick an optimum.
\end{frame}

\begin{frame}[label={sec:orgdb6f542}]{Simple metric on signal sensitive to \(\mathcal{R}\)}
Invent some metric \(M_x = f(\mathcal{S}'_x)\) of SP output which should be near 1.0.

\begin{itemize}
\item Eg: ratio of total induction charge to total collection charge for a small ``blip'' of ionization such as from \(^{39}\)Ar decay.
\end{itemize}

We may form:

\begin{description}
\item[{\(\mathcal{R}^{2D}_i\)}] 2D response calculated on slice \(i\) of \(N\) of some 3D geometry.
\item[{\(M^i_{det}\)}] the metric from SP detector data event using response \(\mathcal{R}^{2D}_i\)
\end{description}

We then try to form a constraint on relative weights of each \(\mathcal{R}^{2D}_i\):

\[ |N_{det} - \sum^{N_{slice},N_{det}}_{i,det} w_i M^i_{det}|; \sum^{N_{slice}}_i w_i = 1.0 \]
\end{frame}

\begin{frame}[label={sec:org740b646}]{What does it mean to find \(\{w_i\}\)?}
\begin{itemize}
\item We calculate \(\mathcal{R}^{2D}_i\) in a pure 2D space (unrealistic).

\begin{itemize}
\item This includes use of unrealistic 2D drift paths.
\end{itemize}

\item Then \(\{w_i\}\) are calculated using \alert{true data 3D drift paths} to determine some kind of importance of each slice.

\item This is somewhat a fiction as the \(\{w_i\mathcal{R}^{2D}_i\}\) model is not perfectly realistic, but it is close.
\end{itemize}
\end{frame}
\end{document}