\section{Results \label{sec:results}} Based on the previous investigations, we summarize the obtained results in table~\ref{tab:result}. The following criteria have been used to compare the extractors: \begin{itemize} \item Stability of the reconstructed charge in the calibration. Extractors with more than 5\% of the pixels excluded from the calibration with any colour or intensity are considered as too unstable. Also extractors yielding more than 0.1\% mis-reconstructed events are excluded. Moreover, the correct number of photo-electrons should be reconstructed at least for the standard calibration pulses (10\,LEDs\,UV). \item The extractor should yield stable results against slight modifications of the pulse shape. \item The extractor must also yield the correct charges for the low-gain pulses on average. \item The reconstructed charge must be linear to the input signal charge for all signals above the image cleaning level and below the low-gain saturation level. \item The resolution of the reconstructed charge should not depend significantly on the signal amplitude, especially should never become comparable to the intrinsic Poissonian signal fluctuations. \item The resolution of the reconstructed charge should not exceed twice the resolution of the best extractor. \item The extractor should not have a charge bias bigger than its charge resolution. \item The time resolution should not be worse than twice the one obtainable with the best extractor. \item The number of mis-reconstructed times should not exceed 1\% on average (including the FADC jumps). \item The needed CPU-time should not exceed the one required for reading the data into memory and writing it to disk. \end{itemize} Table~\ref{tab:result} shows which extractors fulfill the above criteria. One can see that there are a handful extractors which are excluded by only one criterion: \begin{itemize} \item The sliding window extracting 8 slices in high- and low-gain which is only excluded by the poor resolution at very low intensities. \item The integrating spline over 2 FADC slices which still needs to be optimized somewhat for speed. \end{itemize} The digital filter passes all tests, where the extraction over 4 FADC slices yields still superior resolutions compared to the one over 6~FADC slices. \newcommand{\no}{\textcolor{red}{\bf NO\xspace}} \newcommand{\nt}{\textcolor[rgb]{0.9,0.75,0.}{\scriptsize NOT\xspace}} \newcommand{\test}{\textcolor[rgb]{0.9,0.75,0.}{\scriptsize TESTED\xspace}} \newcommand{\ok}{\textcolor[rgb]{0.,.7,0.}{OK\xspace}} \newcommand{\best}{\textcolor[rgb]{0.,.4,0.}{\bf BEST\xspace}} \begin{table}[htp] \small{% \rotatebox{90}{% \centering \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|c|} \hline \hline \multicolumn{12}{|c|}{\large Tested Characteristics} \rule{0mm}{6mm} \rule[-2mm]{0mm}{6mm} \hspace{-3mm}\\ \hline \hline Nr. & Name &Stab. &Stab. & Stab. &Lin. & Res. & Res. & Bias & Res. & Stab. & Speed \\ & &$\widehat{S}$&pulse & low- &$\widehat{S}$&$\widehat{S}$&$\widehat{S}$& $\widehat{S}$&$\widehat{T}$& $\widehat{T}$ & \\ & &calib. &shape & gain & size & const. & & & & & \\ \hline \hline 1 & Fix Win. (4,4) & \no & \no & \no & \no & \no & \no & -- & n/a & n/a &\best\\ 2 & Fix Win. (4,6) & \no & \no & \no & \no & \no & \no & -- & n/a & n/a &\best\\ 3 & Fix Win. (6.6) & \ok & \ok & \no & \ok & \ok & \ok & -- & n/a & n/a & \ok \\ 4 & Fix Win. (8,8) & \no & \ok & \ok &\best& \no & \ok & -- & n/a & n/a & \ok \\ 5 & Fix W. (14,10) & \no & \ok & \ok &\best& \no & \ok & -- & n/a & n/a & \ok \\ \hline 6 & FW. Spl. (4,4) & \ok & \no & \no & \no & \ok & \no & -- & n/a & n/a & \ok \\ 7 & FW. Spl. (4,6) & \ok & \no & \no & \no & \ok & \no & -- & n/a & n/a & \ok \\ 8 & FW. Spl. (6.6) & \no & \ok & \no & \no & \ok & \ok & -- & n/a & n/a & \ok \\ 9 & FW. Spl. (8,8) & \no & \ok & \ok & \ok & \no & \ok & -- & n/a & n/a & \ok \\ 10 & FW. Spl (14,10) & \no & \ok & \ok & \ok & \no & \ok & -- & n/a & n/a & \ok \\ \hline 11 & FW. Pk S. (2,2) & \no & \no & \no & \no & \no & \no & -- & n/a & n/a & \ok \\ 12 & FW. Pk S. (4,4) & \ok & \no & \no & \no & \no & \no & -- & n/a & n/a & \ok \\ 13 & FW. Pk S. (4.6) & \ok & \no & \no & \no & \no & \no & -- & n/a & n/a & \ok \\ 14 & FW. Pk S. (6,6) & \ok & \ok & \no & \no & \ok & \ok & -- & n/a & n/a & \ok \\ 15 & FW. Pk S. (8,8) & \no & \ok & \ok & \ok & \no & \ok & -- & n/a & n/a & \ok \\ 16 & FW Pk S (14,10) & \no & \ok & \ok & \ok & \no & \ok & -- & n/a & n/a & \ok \\ \hline 17 & Slid. W. (2,2) & \ok & \no & \no & \no & \ok & \no & \ok & \no & \no & \ok \\ 18 & Slid. W. (4,4) & \ok & \ok & \ok & \no & \ok & \ok & \ok & \no & \no & \ok \\ 19 & Slid. W. (4.6) & \ok & \ok & \ok & \ok & \ok & \ok & \ok & \ok & \no & \ok \\ 20 & Slid. W. (6,6) & \ok & \ok & \ok & \ok & \no & \ok & \ok & \no & \ok &\best\\ 21 & Slid. W. (8,8) & \ok & \ok & \ok & \ok & \no & \ok & \ok & \ok & \ok &\best\\ 22 & Slid. W. (14,10) & \no & \ok & \ok & \ok & \no & \no & \ok & \no & \ok &\best\\ \hline 23 & Spline Ampl. & \no &\nt & \ok & \no & \no & \no & \no &\ok & \no & \ok \\ & & &\test& & & & & & & & \\ 24 & Splne Int. (1,1.5) & \ok &\nt & \ok & \no & \no & \no & \ok &\best& \ok & \ok \\ & & &\test& & & & & & & & \\ 25 & Spline Int. (2,3) &\best&\nt & \ok & \ok & \ok & \ok & \ok &\best& \ok & \no \\ & & &\test& & & & & & & & \\ 26 & Spline Int. (4,6) & \ok &\ok & \ok & \ok & \ok & \ok & \ok &\best& \ok & \no \\ 27 & Spline Int. (6,9) & \ok &\ok & \ok & \ok & \no & \ok &\best&\best& \ok & \no \\ \hline 28 & Dig. Filt. (6,6) & \ok &\ok & \ok & \ok & \ok & \ok & \ok &\ok & \ok & \ok \\ 29 & Dig. Filt. (4,4) & \ok &\ok & \ok & \ok & \ok &\best& \ok &\best& \ok & \ok \\ \hline \hline \end{tabular} } \caption{The tested characteristics for every extractor: Bias, Resolution (``Res.''), stability (``Stab.''), linearity (``Lin.'') for both charge signal $\widehat{S}$ and time $\widehat{T}$, and speed. \ok\ means that the extractor has passed the test, without being among the best. \no\ means that the extractor has severely failed the test and should not be taken because of that reason. \best\ means that the extractor(s) have come out as best of the particular test.} \label{tab:result} } \end{table} %%% Local Variables: %%% mode: latex %%% TeX-master: "MAGIC_signal_reco" %%% TeX-master: "MAGIC_signal_reco" %%% End: