source: trunk/MagicSoft/TDAS-Extractor/Pedestal.tex@ 6623

\section{Pedestal Extraction \label{sec:pedestals}}

\subsection{Pedestal RMS}

The background $BG$ (Pedestal) 
can be completely described by the noise-autocorrelation matrix $\boldsymbol{B}$ 
(eq.~\ref{eq:autocorr}), 
where the square root of the diagonal elements give what is usually denoted as the ``Pedestal RMS''. 
\par

By definition, $\boldsymbol{B}$ and thus the ``pedestal RMS'' 
is independent of the signal extractor.

\subsection{Pedestal Fluctuations as Contribution to the Signal Fluctuations \label{sec:ffactor}}

A photo-multiplier signal yields, to a very good approximation, the 
following relation:

\begin{equation}
\frac{Var[Q]}{<Q>^2} = \frac{1}{<n_{phe}>} * F^2
\end{equation}

Here, $Q$ is the signal fluctuation due to the number of signal photo-electrons
(equiv. to the signal $S$), and $Var[Q]$ the fluctuations of the true signal $Q$ 
due to the Poisson fluctuations of the number of photo-electrons. Because of: 

\begin{eqnarray}
\widehat{Q} &=& Q + X \\
Var(\widehat{Q}) &=& Var(Q) + Var(X) \\
Var(Q) &=& Var(\widehat{Q}) - Var(X)
\end{eqnarray}

Only in the case that the intrinsic extractor resolution $R$ at fixed background $BG$ does not depend on the signal 
intensity\footnote{Theoretically, this is the case for the digital filter, eq.~\ref{eq:of_noise}.}, 
$Var[Q]$ can be obtained from: 

\begin{eqnarray}
Var(Q) &\approx& Var(\widehat{Q}) - Var(\widehat{Q})\,\vline_{\,Q=0}
\label{eq:rmssubtraction}
\end{eqnarray}

%\footnote{%
%A way to check whether the right RMS has been subtracted is to make the
%``Razmick''-plot
%
%\begin{equation}
%    \frac{Var[\widehat{Q}]}{<\widehat{Q}>^2} \quad \textit{vs.} \quad \frac{1}{<\widehat{Q}>}
%\end{equation}
%
%This should give a straight line passing through the origin. The slope of
%the line is equal to
%
%\begin{equation}
%    c * F^2
%\end{equation}
%
%where $c$ is the photon/ADC conversion factor  $<Q>/<m_{pe}>$.}

 One can then retrieve $R$ 
by applying the signal extractor to a {\textit{\bf fixed window}} of pedestal events, where the 
bias vanishes and measure $Var(\widehat{Q})\,\vline_{\,Q=0}$.

\subsection{Methods to Retrieve Bias and Mean-Squared Error}

In general, the extracted signal variance $R$ is different from the pedestal RMS. 
It can be obtained by applying the signal extractor to pedestal events yielding the bias and 
the resolution $R$.
\par
In the case of the digital filter, $R$ is expected to be independent of the 
signal amplitude $S$ and dependent only on the background $BG$ (eq.~\ref{eq:of_noise}).
%It can then be obtained from the calculation of the variance $Var[\widehat{Q}]$ 
%by applying the extractor to a fixed window of pure background events (``pedestal events'') 
%and get rid of the bias in that way. 
\par

In order to calculate the bias and Mean-squared error, we proceed in the following ways:
\begin{enumerate}
\item Determine $R$ by applying the signal extractor to a fixed window
    of pedestal events. The background fluctuations can be simulated with different 
    levels of night sky background and the continuous light source, but no signal size 
    dependency can be retrieved with this method. 
\item Determine $B$ and $MSE$ from MC events with added noise. 
%    Assuming that $MSE$ and $B$ are negligible for the events without noise, one can 
        With this method, one can get a dependency of both values from the size of the signal, 
        although the MC might contain systematic differences with respect to the real data.
\item Determine $MSE$ from the error retrieved from the fit results of $\widehat{S}$, which is possible for the 
    fit and the digital filter (eq.~\ref{eq:of_noise}). 
    In principle, all dependencies can be retrieved with this method, although some systematic errors are not taken into account 
   with this method: Deviations of the real pulse from the fitted one, errors in the noise auto-correlation matrix and numerical 
precision issues. All these systematic effects add an additional contribution to the true resolution proportional to the signal strength.
\end{enumerate}


\begin{figure}[htp]
\centering
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38993_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38995_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38996_RelMean.eps}
\caption{MExtractTimeAndChargeSpline with amplitude extraction: 
Difference in mean pedestal (per FADC slice) between extraction algorithm
applied on a fixed window of 1 FADC slice (``extractor random'') and a simple addition of 
2 FADC slices (``fundamental''). On the left, a run with closed camera has been taken, in the center
 an opened camera observing an extra-galactic star field and on the right, an open camera being 
illuminated by the continuous light of the calibration (level: 100). Every entry corresponds to one 
pixel.}
\label{fig:amp:relmean}
\end{figure}

\begin{figure}[htp]
\centering
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38993_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38995_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38996_RelMean.eps}
\caption{MExtractTimeAndChargeSpline with integral over 2 slices: 
Difference in mean pedestal (per FADC slice) between extraction algorithm
applied on a fixed window of 2 FADC slices (``extractor random'') and a simple addition of 
2 FADC slices (``fundamental''). On the left, a run with closed camera has been taken, in the center
 an opened camera observing an extra-galactic star field and on the right, an open camera being 
illuminated by the continuous light of the calibration (level: 100). Every entry corresponds to one 
pixel.}
\label{fig:int:relmean}
\end{figure}

\begin{figure}[htp]
\centering
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38993_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38995_RelMean.eps}
\vspace{\floatsep}
\includegraphics[width=0.3\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38996_RelMean.eps}
\caption{MExtractTimeAndChargeDigitalFilter: 
Difference in mean pedestal (per FADC slice) between extraction algorithm
applied on a fixed window of 6 FADC slices and time-randomized weights (``extractor random'') 
and a simple addition of 
6 FADC slices (``fundamental''). On the left, a run with closed camera has been taken, in the center
 an opened camera observing an extra-galactic star field and on the right, an open camera being 
illuminated by the continuous light of the calibration (level: 100). Every entry corresponds to one 
pixel.}
\label{fig:df:relmean}
\end{figure}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsubsection{ \label{sec:ped:fixedwindow} Application of the Signal Extractor to a Fixed Window
of Pedestal Events}

By applying the signal extractor to a fixed window of pedestal events, we 
determine the parameter $R$ for the case of no signal ($Q = 0$)\footnote{%
In the case of 
extractors using a fixed window (extractors nr. \#1 to \#22 
in section~\ref{sec:algorithms}), the results are the same by construction 
as calculating the RMS of the sum of a fixed number of FADC slice, traditionally 
named ``pedestal RMS'' in MARS.}.
\par
In MARS, this functionality is implemented with a function-call to: \\ 

{\textit{\bf MJPedestal::SetExtractionWithExtractorRndm()}} including \\
{\textit{\bf MExtractPedestal::SetRandomCalculation()}}\\

Besides fixing the global extraction window, additionally the following steps are undertaken 
in order to assure an un-biased resolution.

\begin{description}
\item[\textit{MExtractTimeAndChargeSpline}:\xspace] The spline 
maximum position -- which determines the exact extraction window -- is placed
at a random place within the digitizing binning resolution of one central FADC slice. 
\item[\textit{MExtractTimeAndChargeDigitalFilter}:\xspace] The second step timing 
offset $\tau$ (eq.~\ref{eq:offsettau}) is chosen randomly for each event. 
\end{description}

\par

The following figures~\ref{fig:amp:relmean} through~\ref{fig:df:relrms} show results 
obtained with the second method for three background intensities: 

\begin{enumerate}
\item Closed camera and no (Poissonian) fluctuation due to photons from the night sky background
\item The camera pointing to an extra-galactic region with stars in the field of view
\item The camera illuminated by a continuous light source of intensity 100. 
\end{enumerate}

Figures~\ref{fig:amp:relmean} through~\ref{fig:df:relmean} 
show the calculated biases obtained with this method for all pixels in the camera 
and for the different levels of (night-sky) background applied to 1000 pedestal events.
One can see that the bias vanishes to an accuracy of better than 2\% of a photo-electron
makefor the extractors which are used in this TDAS. 

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%1

\begin{figure}[htp]
\centering
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38993_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38995_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Amplitude_Amplitude_Range_01_09_01_10_Run_38996_RMSDiff.eps}
\caption{MExtractTimeAndChargeSpline with amplitude:  
Difference in RMS (per FADC slice) between extraction algorithm
applied on a fixed window and the corresponding pedestal RMS. 
Closed camera (left), open camera observing extra-galactic star field (right) and 
camera being illuminated by the continuous light (bottom). 
Every entry corresponds to one pixel.}
\label{fig:amp:relrms}
\end{figure}


\begin{figure}[htp]
\centering
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38993_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38995_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeSpline_Rise-and-Fall-Time_0.5_1.5_Range_01_10_02_12_Run_38996_RMSDiff.eps}
\caption{MExtractTimeAndChargeSpline with integral over 2 slices:  
Difference in RMS (per FADC slice) between extraction algorithm
applied on a fixed window and the corresponding pedestal RMS. 
Closed camera (left), open camera observing extra-galactic star field (right) and 
camera being illuminated by the continuous light (bottom). 
Every entry corresponds to one 
pixel.}
\label{fig:int:relrms}
\end{figure}


\begin{figure}[htp]
\centering
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38993_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38995_RMSDiff.eps}
\vspace{\floatsep}
\includegraphics[width=0.47\linewidth]{MExtractTimeAndChargeDigitalFilter_Weights_cosmics_weights.dat_Range_01_14_02_14_Run_38996_RMSDiff.eps}
\caption{MExtractTimeAndChargeDigitalFilter:  
Difference in RMS (per FADC slice) between extraction algorithm
applied on a fixed window and the corresponding pedestal RMS. 
Closed camera (left), open camera observing extra-galactic star field (right) and 
camera being illuminated by the continuous light (bottom). 
Every entry corresponds to one pixel.}
\label{fig:df:relrms}
\end{figure}



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Figures~\ref{fig:amp:relrms} through~\ref{fig:df:relrms} show the 
differences in $R$ between the RMS of simply summing up the FADC slices over the extraction window 
(in MARS called: ``Fundamental Pedestal RMS'') and 
the one obtained by applying the extractor to the same extraction window 
(in MARS called: ``Pedestal RMS with Extractor Rndm''). One entry of each histogram corresponds to one 
pixel of the camera.
The distributions have a negative mean in the case of the digital filter showing the 
``filter'' capacity of that algorithm. It ``filters out'' between 0.12 photo-electrons night sky 
background for the extra-galactic star-field until 0.2 photo-electrons for the continuous light.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\subsubsection{ \label{sec:ped:slidingwindow} Application of the Signal Extractor to a Sliding Window
of Pedestal Events}

By applying the signal extractor to a global extraction window of pedestal events, allowing 
it to ``slide'' and maximize the encountered signal, we 
determine the bias $B$ and the mean-squared error $MSE$ for the case of no signal ($S=0$). 
\par
In MARS, this functionality is implemented with a function-call to: \\ 

{\textit{\bf MJPedestal::SetExtractionWithExtractor()}} \\

\par
Table~\ref{tab:bias} shows the bias, the resolution and the mean-square error for all extractors using
a sliding window. In this sample, every extractor had the freedom to move 5 slices, 
i.e. the global window size was fixed to five plus the extractor window size. This first line 
shows the resolution of the smallest existing robust fixed window algorithm in order to give the reference 
value of 2.5 and 3 photo-electrons RMS for an extra-galactic and a galactic star-field, respectively. 
\par
One can see that the bias $B$ typically decreases 
with increasing window size, while the error $R$ increases with 
increasing window size, except for the digital filter. There is also a small difference between the obtained error 
on a fixed window extraction and the one obtained from a sliding window extraction in the case of the spline and digital 
filter algorithms.
The mean-squared error has an optimum somewhere in between: In the case of the 
sliding window and the spline at the lowest window size, in the case of the digital filter
at 4 slices. The global winners is extractor~\#29 
(digital filter with integration of 4 slices). All sliding window extractors -- except \#21 -- 
have a smaller mean-square error than the resolution of the fixed window reference extractor. This means 
that the global error of the sliding window extractors is smaller than the one of the fixed window extractors
even if the first have a bias.
\par
The important information for the image cleaning is the number of photo-electrons above which the probability for obtaining 
a noise fluctuation is smaller than 0.3\% (3$\sigma$). We approximated that number with the formula: 

\begin{equation}
N_{\mathrm{phe}}^{\mathrm{thres.}} \approx B + 3\cdot R
\end{equation}

Table~\ref{tab:bias} shows that most of the sliding window algorithms yield a smaller signal threshold than the fixed window ones, 
although the first have a bias. The lowest threshold of only 4.2~photo-electrons for the extra-galactic star-field and 5.0~photo-electrons 
for the galactic star-field is obtained by the digital filter fitting 4 FADC slices (extractor~\%29). 
This is almost a factor 2 lower than the fixed window results. Also the spline integrating 1 FADC slice (extractor~\%24) yields almost 
comparable results.

\begin{landscape}
%\rotatebox{90}{%
\begin{table}[htp]
\vspace{3cm}
\scriptsize{%
\centering
\begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|c|c|c|c|c|}
\hline
\hline
\multicolumn{16}{|c|}{Statistical Parameters for $S=0$} \\
\hline
\hline
 & & \multicolumn{4}{|c|}{Closed camera} & \multicolumn{5}{|c|}{Extra-galactic NSB}  & \multicolumn{5}{|c|}{Galactic NSB} \\
\hline
\hline
Nr. & Name         &  $R$  & $R$ & $B$ & $\sqrt{MSE}$ &  $R$ &$R$  & $B$ & $\sqrt{MSE}$ & $B+3R$ & $R$ &  $R$& $B$ & $\sqrt{MSE}$ & $B+3R$  \\
    &              &  (FW) & (SW)& (SW)& (SW) & (FW) &(SW) & (SW)& (SW) & (99.7\% prob.) & (FW)&(SW) & (SW)&(SW) & (99.7\% prob.) \\
\hline                                                     
\hline                                                     
4   & Fixed Win. 8  & 1.2  & --  & 0.0 & 1.2  & 2.5  & --  & 0.0 &  2.5 & 7.5 & 3.0 &  -- & 0.0 & 3.0 & 9.0 \\   
\hline                                                     
--  & Slid. Win. 1  & 0.4  & 0.4 & 0.4 & 0.6  & 1.2  & 1.2 & 1.3 &  1.8 & 4.9 & 1.4 & 1.4 & 1.5 & 2.0 & 5.7 \\
17  & Slid. Win. 2  & 0.5  & 0.5 & 0.4 & 0.6  & 1.4  & 1.4 & 1.2 &  1.8 & 5.4 & 1.6 & 1.6 & 1.5 & 2.2 & 6.1 \\
18  & Slid. Win. 4  & 0.8  & 0.8 & 0.5 & 0.9  & 1.9  & 1.9 & 1.2 &  2.2 & 6.9 & 2.2 & 2.3 & 1.6 & 2.8 & 7.5 \\
20  & Slid. Win. 6  & 1.0  & 1.0 & 0.4 & 1.1  & 2.2  & 2.2 & 1.1 &  2.5 & 7.7 & 2.6 & 2.7 & 1.4 & 3.0 & 9.5 \\
21  & Slid. Win. 8  & 1.2  & 1.3 & 0.4 & 1.4  & 2.5  & 2.5 & 1.0 &  2.7 & 8.5 & 3.0 & 3.2 & 1.4 & 3.5 & 10.0 \\
\hline                                                                              
23  & Spline Amp.  & 0.4  & \textcolor{red}{\bf 0.4} & 0.4 & 0.6  & 1.1  & 1.2 & 1.3 & 1.8 & 4.9 & 1.3 & 1.4 & 1.6 & 2.1 & 5.8 \\
24  & \textcolor{red}{\bf Spline Int. 1} & 0.4  & \textcolor{red}{\bf 0.4} & 0.3 & \textcolor{red}{\bf 0.5} & 1.0 & 1.2 & 1.0 & 1.6 & 4.6 & 1.3 & \textcolor{red}{\bf 1.3} & 1.3 & 1.8 & 5.2 \\
25  & Spline Int. 2 & 0.5  & 0.5 & 0.3 & 0.6  & 1.3  & 1.4 & 0.9 &  1.7 & 5.1 & 1.7 & 1.6 & 1.2 & 2.0 & 6.0 \\
26  & Spline Int. 4 & 0.7  & 0.7 & \textcolor{red}{\bf 0.2 } & 0.7  & 1.5  & 1.7 & \textcolor{red}{\bf 0.8} &  1.9 & 5.3 & 2.0 & 2.0 & 1.0 & 2.2 & 7.0 \\
27  & Spline Int. 6 & 1.0  & 1.0 & 0.3 & 1.0  & 2.0  & 2.0 & \textcolor{red}{\bf 0.8} &  2.2 & 6.8 & 2.6 & 2.5 & \textcolor{red}{\bf 0.9} & 2.7 & 8.4 \\
\hline                                                                              
28  & Dig. Filt. 6 & 0.4   & 0.5 & 0.4 & 0.6  & 1.1  & 1.3 & 1.3 &  1.8 & 5.2 & 1.3 & 1.5 & 1.5 & 2.1 & 6.0 \\
29  & \textcolor{red}{\bf Dig. Filt. 4} & 0.3   & \textcolor{red}{\bf 0.4} & 0.3 & \textcolor{red}{\bf 0.5}  & 0.9  & \textcolor{red}{\bf 1.1} & 0.9 & \textcolor{red}{\bf 1.4} & \textcolor{red}{\bf 4.2} & 1.0 & \textcolor{red}{\bf 1.3} & 1.1 & \textcolor{red}{\bf 1.7} & \textcolor{red}{\bf 5.0 }\\
\hline
\hline
\end{tabular}
\vspace{1cm}
\caption{The statistical parameters bias, resolution and mean error for the sliding window 
algorithm. The first line displays the resolution of the smallest existing robust fixed--window extractor 
for reference. All units in equiv. 
photo-electrons, uncertainty: 0.1 phes. All extractors were allowed to move 5 FADC slices plus 
their window size. The ``winners'' for each row are marked in red. Global winners (within the given 
uncertainty) are the extractors Nr. \#24 (MExtractTimeAndChargeSpline with an integration window of 
1 FADC slice) and Nr.\#29 
(MExtractTimeAndChargeDigitalFilter with an integration window size of 4 slices)}
\label{tab:bias}
}
\end{table}
%}
\end{landscape}

\clearpage

Figures~\ref{fig:sw:distped} through~\ref{fig:df4:distped} show the 
extracted pedestal distributions for some selected extractors (\#18, \#23, \#25, \#28 and \#29)
 for one exemplary channel (pixel 100) and two background situations: Closed camera with only electronic
noise and open camera pointing to an extra-galactic source.
One can see the (asymmetric) Poisson behaviour of the 
night sky background photons for the distributions with open camera.

\begin{figure}[htp]
\centering
\includegraphics[height=0.43\textheight]{PedestalSpectrum-18-Run38993.eps}
\vspace{\floatsep}
\includegraphics[height=0.43\textheight]{PedestalSpectrum-18-Run38995.eps}
\caption{MExtractTimeAndChargeSlidingWindow with extraction window of 4 FADC slices: 
Distribution of extracted "pedestals"  from pedestal run with 
closed camera (top) and open camera observing an extra-galactic star field (bottom) for one channel 
(pixel 100). The result obtained from a simple addition of 4 FADC 
slice contents (``fundamental'') is displayed as red histogram, the one obtained from the application of
the algorithm on 
a fixed window of 4 FADC slices as blue histogram (``extractor random'') and the one obtained from the
full algorithm allowed to slide within a global window of 12 slices. The obtained histogram means and 
RMSs have been converted to equiv. photo-electrons.}
\label{fig:sw:distped}
\end{figure}


\begin{figure}[htp]
\centering
\includegraphics[height=0.43\textheight]{PedestalSpectrum-23-Run38993.eps}
\vspace{\floatsep}
\includegraphics[height=0.43\textheight]{PedestalSpectrum-23-Run38995.eps}
\caption{MExtractTimeAndChargeSpline with amplitude extraction: 
Spectrum of extracted "pedestals"  from pedestal run with 
closed camera lids (top) and open lids observing an extra-galactic star field (bottom) for one channel 
(pixel 100). The result obtained from a simple addition of 2 FADC 
slice contents (``fundamental'') is displayed as red histogram, the one obtained from the application 
of the algorithm on a fixed window of 1 FADC slice as blue histogram  (``extractor random'')
and the one obtained from the
full algorithm allowed to slide within a global window of 12 slices. The obtained histogram means and 
RMSs have been converted to equiv. photo-electrons.}
\label{fig:amp:distped}
\end{figure}

\begin{figure}[htp]
\centering
\includegraphics[height=0.43\textheight]{PedestalSpectrum-25-Run38993.eps}
\vspace{\floatsep}
\includegraphics[height=0.43\textheight]{PedestalSpectrum-25-Run38995.eps}
\caption{MExtractTimeAndChargeSpline with integral extraction over 2 FADC slices: 
Distribution of extracted "pedestals"  from pedestal run with 
closed camera lids (top) and open lids observing an extra-galactic star field (bottom) for one channel 
(pixel 100). The result obtained from a simple addition of 2 FADC 
slice contents (``fundamental'') is displayed as red histogram, the one obtained from the application 
of time-randomized weights on a fixed window of 2 FADC slices as blue histogram and the one obtained from the
full algorithm allowed to slide within a global window of 12 slices. The obtained histogram means and 
RMSs have been converted to equiv. photo-electrons.}
\label{fig:int:distped}
\end{figure}

\begin{figure}[htp]
\centering
\includegraphics[height=0.43\textheight]{PedestalSpectrum-28-Run38993.eps}
\vspace{\floatsep}
\includegraphics[height=0.43\textheight]{PedestalSpectrum-28-Run38995.eps}
\caption{MExtractTimeAndChargeDigitalFilter: Spectrum of extracted "pedestals"  from pedestal run with 
closed camera lids (top) and open lids observing an extra-galactic star field (bottom) for one channel 
(pixel 100). The result obtained from a simple addition of 6 FADC 
slice contents (``fundamental'') is displayed as red histogram, the one obtained from the application 
of time-randomized weights on a fixed window of 6 slices as blue histogram and the one obtained from the
full algorithm allowed to slide within a global window of 12 slices. The obtained histogram means and 
RMSs have been converted to equiv. photo-electrons.}
\label{fig:df6:distped}
\end{figure}

\begin{figure}[htp]
\centering
\includegraphics[height=0.43\textheight]{PedestalSpectrum-29-Run38993.eps}
\vspace{\floatsep}
\includegraphics[height=0.43\textheight]{PedestalSpectrum-29-Run38995.eps}
\caption{MExtractTimeAndChargeDigitalFilter: Spectrum of extracted "pedestals"  from pedestal run with 
closed camera lids (top) and open lids observing an extra-galactic star field (bottom) for one channel 
(pixel 100). The result obtained from a simple addition of 4 FADC 
slice contents (``fundamental'') is displayed as red histogram, the one obtained from the application 
of time-randomized weights on a fixed window of 4 slices as blue histogram and the one obtained from the
full algorithm allowed to slide within a global window of 10 slices. The obtained histogram means and 
RMSs have been converted to equiv. photo-electrons.}
\label{fig:df4:distped}
\end{figure}

\clearpage

\subsection{ \label{sec:ped:singlephe} Single Photo-Electron Extraction with the Digital Filter}

Figure~\ref{fig:df:sphespectrum} shows spectra 
obtained with the digital filter applied on three different global search windows.
One can clearly distinguish a pedestal peak (fitted to Gaussian with index 0)
and further, positive contributions.
\par
Because the background is determined by the single photo-electrons from the night-sky background,
the following possibilities can occur:

\begin{enumerate}
\item There is no ``signal'' (photo-electron) in the extraction window and the extractor 
finds only electronic noise. 
Usually, the returned signal charge is then negative.
\item There is one photo-electron in the extraction window and the extractor finds it.
\item There are more than one photo-electron in the extraction window, but separated by more than 
two FADC slices whereupon the extractor finds the one with the highest charge (upward fluctuation) of both.
\item The extractor finds an overlap of two or more photo-electrons.
\end{enumerate}

Although the probability to find a certain number of photo-electrons in a fixed window follows a 
Poisson distribution, the one for employing the sliding window is {\textit{not}} Poissonian. The extractor 
will usually find one photo-electron even if more are present in the global search window, i.e. the 
probability for two or more photo-electrons to occur in the global search window is much higher than 
the probability for these photo-electrons to overlap in time such as to be recognized as a double 
or triple photo-electron pulse by the extractor. This is especially true for small extraction windows
and for the digital filter.

\par

Given a global extraction window of size $\mathrm{\it WS}$ and an average rate of photo-electrons from the night-sky 
background $R$, we will now calculate the probability for the extractor to find zero photo-electrons in the 
$\mathrm{\it WS}$. The probability to find any number of $k$ photo-electrons can be written as:

\begin{equation}
P(k) = \frac{e^{-R\cdot \mathrm{\it WS}} (R \cdot \mathrm{\it WS})^k}{k!}
\end{equation}

and thus:

\begin{equation}
P(0) = e^{-R\cdot \mathrm{\it WS}} 
\end{equation}

The probability to find one or more photo-electrons is then:

\begin{equation}
P(>0) = 1 - e^{-R\cdot \mathrm{\it WS}} 
\end{equation}

In figures~\ref{fig:df:sphespectrum},
one can clearly distinguish the pedestal peak (fitted to Gaussian with index 0), 
corresponding to the case of  $P(0)$ and further 
contributions of $P(1)$ and $P(2)$ (fitted to Gaussians with index 1 and 2). 
One can also see that the contribution of $P(0)$ dimishes 
with increasing global search window size.

\begin{figure}
\centering
\includegraphics[height=0.3\textheight]{SinglePheSpectrum-28-Run38995-WS2.5.eps}
\vspace{\floatsep}
\includegraphics[height=0.3\textheight]{SinglePheSpectrum-28-Run38995-WS4.5.eps}
\vspace{\floatsep}
\includegraphics[height=0.3\textheight]{SinglePheSpectrum-28-Run38995-WS8.5.eps}
\caption{MExtractTimeAndChargeDigitalFilter: Spectrum obtained from the extraction 
of a pedestal run using a sliding window of 6 FADC slices allowed to move within a window of 
7 (top), 9 (center) and 13 slices.
A pedestal run with galactic star background has been taken and one exemplary pixel (Nr. 100).
One can clearly see the pedestal contribution and a further part corresponding to one or more 
photo-electrons.}
\label{fig:df:sphespectrum}
\end{figure}

In the following, we will make a short consistency test: Assuming that the spectral peaks are 
attributed correctly, one would expect the following relation:

\begin{equation}
P(0) / P(>0) = \frac{e^{-R\cdot WS}}{1-e^{-R\cdot WS}}
\end{equation}

We tested this relation assuming that the fitted area underneath the pedestal peak $\mathrm{\it Area}_0$ is 
proportional to $P(0)$ and the sum of the fitted areas underneath the single photo-electron peak 
$\mathrm{\it Area}_1$ and the double photo-electron peak $\mathrm{\it Area}_2$ proportional to $P(>0)$. We assumed 
that the probability for a triple photo-electron to occur is negligible. Thus, one expects:

\begin{equation}
\mathrm{\it Area}_0 / (\mathrm{\it Area}_1 + \mathrm{\it Area}_2 ) = \frac{e^{-R\cdot WS}}{1-e^{-R\cdot WS}}
\end{equation}

We estimated the effective window size $\mathrm{\it WS}$ as the sum of the range in which the digital filter 
amplitude weights are greater than 0.5 (1.5 FADC slices) and the global search window minus the 
size of the window size of the weights (which is 6 FADC slices). Figure~\ref{fig:df:ratiofit}
shows the result for two different levels of night-sky background. The fitted rates deliver 
0.08 and 0.1 phes/ns, respectively. These rates are about 50\% lower than those obtained 
from the November 2004 test campaign. However, we should take into account that the method is at 
the limit of distinguishing single photo-electrons. It may occur often that a single photo-electron 
signal is too low in order to get recognized as such. We tried various pixels and found that 
some of them do not permit to apply this method at all. The ones which succeed, however, yield about 
the same fitted rates. To conclude, one may say that there is consistency within the double-peak 
structure of the pedestal spectrum found by the digital filter which can be explained by the fact that 
single photo-electrons are separated from the pure electronics noise. 
\par

\begin{figure}[htp]
\centering
\includegraphics[height=0.4\textheight]{SinglePheRatio-28-Run38995.eps}
\vspace{\floatsep}
\includegraphics[height=0.4\textheight]{SinglePheRatio-28-Run39258.eps}
\caption{MExtractTimeAndChargeDigitalFilter: Fit to the ratio of the area beneath the pedestal peak and 
the single and double photo-electron(s) peak(s) with the extraction algorithm
applied on a sliding window of different sizes.
In the top plot, a pedestal run with extra-galactic star background has been taken and in the bottom, 
a galatic star background. An exemplary pixel (Nr. 100) has been used.
Above, a rate of 0.08 phe/ns and below, a rate of 0.1 phe/ns has been obtained.}
\label{fig:df:ratiofit}
\end{figure}

Figure~\ref{fig:df:convfit} shows the obtained ``conversion factors'' and ``F-Factor'' computed as: 

\begin{eqnarray}
c_{phe} &=& \frac{1}{\mu_1 - \mu_0} \\
F_{phe} &=& \sqrt{1 + \frac{\sigma_1^2 - \sigma_0^2}{(\mu_1 - \mu_0)^2} }
\end{eqnarray}

where $\mu_0$ denotes the mean position of the pedestal peak and $\mu_1$ the mean position of the (assumed) 
single photo-electron peak. The obtained conversion factors are systematically lower than the ones 
obtained from the standard calibration and decrease with increasing window size. This is consistent 
with the assumption that the digital filter finds the most upward fluctuating pulse out of several. Therefore,
$\mu_1$ is biased against higher values. The F-Factor is also systematically low (however with huge error bars), 
which is also consistent 
with the assumption that the spacing between $\mu_1$ and $\mu_0$ is artificially high. 
Unfortunately, the error bars are too high for a ``calibration'' of the F-Factor. 
\par
In conclusion, the digital filter is at the edge of being able to see single photo-electrons, 
however a single photo-electron calibration cannot yet be done with the current FADC system because the 
resolution is too poor. These limitations might be overcome if a higher sampling speed is used and the artificial
pulse shaping removed. We expect to improve this method considerably with the new 2\,GSamples/s~FADC readout of MAGIC.

\begin{figure}[htp]
\centering
\includegraphics[height=0.4\textheight]{ConvFactor-28-Run38995.eps}
\vspace{\floatsep}
\includegraphics[height=0.4\textheight]{FFactor-28-Run38995.eps}
\caption{MExtractTimeAndChargeDigitalFilter: Obtained conversion factors (top) and F-Factors (bottom)
from the position and width of 
the fitted Gaussian mean of the single photo-electron peak and the pedestal peak depending on 
the applied global extraction window sizes.
A pedestal run with extra-galactic star background has been taken and 
an exemplary pixel (Nr. 100) used. The conversion factor obtained from the 
standard calibration is shown as a reference line. The obtained conversion factors are systematically
lower than the reference one.}
\label{fig:df:convfit}
\end{figure}



%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "MAGIC_signal_reco"
%%% TeX-master: "MAGIC_signal_reco"
%%% TeX-master: "MAGIC_signal_reco"
%%% End: