source: trunk/MagicSoft/Mars/NEWS@ 8787

                                                               -*-*- END -*-*-

== <cvs> ==

;general
* fixed GroupBy options in plot*.C macros
* only display KVA points (not the Tuorla points) in plotoptical.C
* all programs now allow to add resources to the ones from the resource
file or overwrite them from the command line. This can be done like this
::ganymed --config=ganymed.rc --rc=Cut1.Param1:0.25 --rc=Cut1.Param3:5.5
:if something seems to be wrong you can debug what the program is doing
using the option --debug-env=3
* increased the diversity of debug levels. That also means that with -v3
you get now less output, i.e. only the output which is important for
an analysis. To get more output which helps for debugging try 4, 5 and 6.
* fixed a problem in the task synchronizing the reading of the reports
and the data stream. The effect was that in some circumstances the last
events of some trees could have been skipped. Since this is only a
very tiny part of the data no severe effect to your analysis results 
is expected.
* Dataset file now allow to include the sequence file directly within
the dataset file and also allow to exclude single runs from the dataset.
(for more details see Class reference of MDataSet)
* Sequence files now allow for excluding runs from the analysis
(for more details see Class reference of MSequence)
* added code to process the sum-trigger flag

;callisto
* in the pedestal processing now at least 50 processed events are
required for succeeding.

;ganymed/sponde
* Now data points are even shown correctly if they have error bars larger
than one sigma. Points with null or negative significances are omitted.

;sponde
*fixed resource files ('''Spectrum''' instead of '''MJSpectrum''')


 *** Version 2.0 (2007/09/03)

   - database: the position used for the pulse pos check was missing 
     in the callisto output and could thus not be filled into the DB.
     Also the old values seemed not exactly the PulsePos used for 
     teh check.

   - general: Resource file now allow an Include-Resource, i.e. you can
     read a resource file with default settings, include it in your
     resource file and overwrite the settings in your file. More than
     one include file is allowed. Inclusions can be iterative.
        Include: mydefaults.rc yourdefaults.rc
     The resources in the first file have higher priority than the
     second file.

   - general: Now the output files (calib*.root, etc) also contain the
     resource file (to check it open it in the TBrowser and choose Print()
     from the context menu)

   - general: Dataset files now allow to overwrite the default path
     to search for sequence and data files. Note, that the program
     option and individual resources will overwrite it. For example:
        SequencePath: /magic/montecarlo/sequences
        DataPath:     /magic/montecarlo/star

   - general: Fixed a problem with variables in MDataPhrase. It seems
     that only optimdisp (didn't work) was effected.

   - general: Some improvements to the optim and train processing. For
     example the ststud window now got a proper title.

   - general: The "Status Display: " was removed from the window title
     within an Eventloop.

   - general: added a new flexible histogram-class (MHn) which can be used
     for example in testing after optimizations or training to produce
     user defined histograms.

   - general: The Random Forest now allows attaching an evaluation
     function which is applied to its output. This allows to train
     on whatever you like (eg. train on log10(energy), but return
     energy)

   - general: The Random Forest training of energy (trainenergy.c) and
     disp (traindisp.C) has been improved by more graphical output to
     judge the result. 

   - general: The training of energy (trainenergy.C) has been improved
     a lot by training on a different quantity than energy. It gives
     a result which is highly independant of the source spectrum
     and shows lower bias than all other options.

   - general: Dataset files are allowed to contain collections of
     datasets. For more details see the class reference of MDataSet.

   - general: The default WobbleMode in a datset file now (if not
     overwritten) is "auto", i.e. wobble mode is set if the dataset
     doesn't contain off-sequences.

   - merpp: Merpp can read file format version 9 now. Please realize
     that for file format version 9 the extraction range is different
     and must be set to 0/49 instead of 15/64 in callisto_mux.rc until
     the first and last 15 slices have been removed from all MUX data.

   - mars: Similar to writing movies you can now select events which
     should be displayed in your display. Therefore add the following
     lines to your mars.rc:
         MFEvtNumber.FileName: ganymed00223552.root
         MFEvtNumber.Selector: ThetaSquared.fVal<0.04 && DataType.fVal>0.5
     The input file can be a ganymed-file after (ganymed*.root) or
     before (ganymed*-summary.root) file. For example to select all events
     from your on sample after cuts (excluding the ThetaSq cut) use
         MFEvtNumber.Selector: DataType.fVal>0.5
     To show all events from your on-sample after quality cuts
     use the summary file instead.

   - mars: new ways to call mars are implemented. Instead
     of calling
         mars filename.root
     you can now also use
         mars sequence.txt
     or
         mars sequence.txt inputpath
     or replace sequence.txt by the sequence number if the default
     sequence file should be used.

   - mars: The default size has been increased

   - mars: There are two new heckboxes which allow to switch off
     the calculated image parameters

   - callisto: was broken for MCs... fixed.

   - callisto: improved calculation of spline coefficients a lot. This
     leads to a further improvement of the event rate calibrating MUX
     data of about 15% (175evt/s instead of 150evt/s)

   - callisto: finally the raw data we read is now real 16bit (with
     10bit precision) for MUX data and not a stripped 8bit number
     anymore. Due to scaling of the numbers (without precision loss
     into an [0-256[ range, the final numbers for pedestal, pedestalrms
     and extacted signal don't change. However, at all places where
     raw data is directly accessed (eg. saturation and variation 
     limits in the signal- and pedestal-extraction) the numbers
     refer still to the data's 16bit range.

   - callisto: The maximum arrival time difference used in the bad
     pixel treatment is now in units of nanoseconds, which leads to 3.0ns
     for both, old FADC and MUX-FADC, data.

   - callisto: Be aware that callisto is now calibrating the arrival time
     in nanoseconds rather than time slices. Also all following tasks
     have been changed to assume ns units now. If you star data 
     calibrated with an older version it might be necessary to adapt 
     some timimng parameters (Muon Analysis and MFSoftwareTrigger)
     to obtain optimal results. The plot in the PulsePos tab 
     will now be in ns, too. This might also mean that the Muon
     analysis in star might not work as perfect as expected as long
     as old files read in.

   - callisto: The calibration constants of earlier updates got lost
     somehow. All constants have been updated.

   - star: The PSF is now determined from the profile of the ArcWidth
     instead of arcwidth/radius. The old way gave to much weight
     to the bins with low statistics. The reference lines have been
     updated.

   - star: For speed reasons events suitable for the muon analysis 
     are now also preselected by fConcCOG<0.1 which seems to be a very
     good preselector for muons.

   - star: is now calculating a new timing parameter the Slope along
     the major and minor axis of the shower. Therefore a numerical
     solution of a line-fit is used. To use MHillasExt.fSlopeLong
     and MHillasExt.fSlopeTrans don't forget to multiply it with
     the sign stored in MHillasSrc.fCosDeltaAlpha as usual.

   - star: The parameters fInnerSize, fInnerLeakage1 and fInnerLeakage2
     have been removed from MNewImagePar. They have never been good for
     anything.

   - star: Implemented a new recursive algorithm which is supposed to be
     faster. Also implemented new cleaning oiptions, which are: 
        + CleanLevel0:  
          The cleaning level above single core pixels can be kept
        + KeepIsolatedPixels:
          Define whether single core pixels should be kept
        + TimeLevel1:
          The coincidence window in nanoseconds to single used neighbors
        + TimeLevel2:
          The coincidence window in nanoseconds to two used neighbors
        + PostCleanType:
          MOde in which a time dependent post cleaning should be applied
           0: No post cleaning
           1: Require for each used pixels at least one used neighbor
              within a time-window of CleanLevel1
           2: Require for each used pixels at least two used neighbor
              within a time-window of CleanLevel2
           3: Require first two and second one used neighbor
              within a time-window of CleanLevel1 and fCleanLevel2 
              respectively
        + RecoverIsolatedPixels:
          - recoveres isolated core pixels if they have a used neighbor

   - mars/star: The new default cleaning is:
      + CleanLevel1: 6.0
      + CleanLevel2: 3.0
      + CleanTime1:  1.75
      + CleanTime2:  1.75
      + PostCleanType: 3

   - star/ganymed: The old spark-cuts have been replaced by new ones.
     These new spark cuts have been cross checked with the sequences
     84720 (before splitter), 101041 (after splitter) and 223539 (new
     FADCs). In all cases they seperate pretty well, but of course not
     perfect. Monte Carlos have suggested to tighten the cuts a little
     bit further at low sizes. This has been done.

   - ganymed: IMPORTANT - The spark cuts have been changed to match
     all available data more or less well. This might mean that for 
     YOUR data there are still sparks visible. 
     Whenever you do an analysis you have to make sure that NO sparks
     survive your cuts!

   - ganymed: The old hadronness and size cuts have gotten new indices
     (10, 11). The old indices 8 and 9 now mean an additional condition
     for ghostbusting based on the formula (dist-c[9])*c[8]-slope.
     If you want to switch off this cut just move the line far away enough,
     e.g. by setting c[9] to -500. Please update your ganymed.rc files
     accordingly!
     The unit of c[9] is deg. It is the dist at which your cut will
     cross the slope==0. The unit of the slope is ns/deg. So if
     you readin files calibrated with an older Mars-version you
     have to adapt the default value for c[9] to the FADCs:
        - Multiply by two for the MUX    FADCs    (run number > 200.000)
        - Multiply by 0.3 for the Siegen FADCs    (run number < 200.000)
        - Multiply by  42 for the meaning of life (run number = 200.000)

   - ganymed: The old Cut0 is now called CutQ and is calculated before
     the source position is caluclated. If you need a quality cut
     including source dependant data (be very carefull with this!)
     you have to use Cut0 instead.

   - ganymed: now properly supports three off regions in wobble mode.
     It will be the default from now on in ganymed_wobble.rc. To change
     the number of off-regions use
         NumOffSourcePositions: 1
     in your ganymed.rc. Also by default no off-theta cut will be done
     anymore so that you will get a background level higher than the
     signal level in your ThetaSq plot for ThetaSq>0.2.
     Due to this be carefull changing the scaling mode away from 
     "None". You will underestimate your signal if your scale interval
     is not properly set.
     (Rem: ThetaSqN is no longer needed)

   - ganymed: is now storing the events of all (on- and off-)
     source positions even in the case of more than one off-source
     position. This should allow sponde to work properly and the
     energy estimation is done correctly in all cases (without any
     "nearest position" trick)

   - ganymed: From 85340 (19.3.2006) on a real starguider calibration
     with a real pointing model for the starguider is done, calibrated
     with more than thousand tpoints taken from this date on.

   - ganymed: was broken for Wobble-MCs... fixed.

   - ganymed: The default cuts (MFMagicCuts) and the disp-parametrization
     has changed. Note that your old ganymed.rc will be incompatible
     with the new algorithms! Please check the ganymed.rc files and
     the class description of MFMagicCuts for more information.
     It is recommended at the moment to use identical Disp-Parametrizations
     for on/off- and wobble-mode. If cuts are optimized only
     the Area-Cut (parameters 2,3,4) and the theta-sq cut (parameter 1)
     should be optimized. All other values should be kept as they are.
     If you want to do an anylsis without timing a parameters, you can
     switch off the timing cuts by seeting parameter 7 to a very small
     value (eg. -99). To remove the slope parameter from the disp
     paremtrization set parameter 8 to 0.

   - ganymed: The distribution of observation time versus zenith
     angle is now shown for off- AND on-data.

   - ganymed: The contents of the source position plot are no longer
     averaged, thus a lot of (fake) events between the two wobble 
     positions have disappeared.

   - ganymed: There is a new tab "CutT" it contains the VsSize plots
     with the CutT, whihc is defined in ganymed.rc, applied. It
     is resonable to use your theta-cut for it, to see if there is
     something strange (sparks!) in your signal-region.

   - ganymed: To choose a dataset from a collection in a dataset file
     use the new command line option "--dataset=12345"

   - optim, sponde: should now properly support three off-regions.
     Just produce your ganymed summary files with three off-regions.
     optim and sponde will automatically get all off events from
     all off-regions from this file. Please do not forget to
     set the scaling in optimwobble.C properly:
        fit.SetScaleUser(1./3);
     it is not yet read in automatically (will follow soon).
     In the case of fixed Scale you can also use
        cuts.SetThetaCut(MFMagicCuts::kOn);
     no off-cut needed.
     Be carefull in case of three off-regions when switching on
     scale mode. Make sure that the scale interval is set correctly.

   - sponde: sponde.rc and sponde_onoff.rc are now idetical

   - sponde: the resouce files have been rewritten with a lot of
     comments and a well working parametrization for an energy estimator

   - sponde: In the estimated energy versus monte carlo energy plot the
     contents where exchanged.... fixed.

   - sponde: the so called "accurate"-mode has been removed. It didn't
     give any improvement in accuracy, only decreased execution speed.

   - sponde: the so called "simple"-mode has been removed. It didn't
     give any improvement in simple.

   - sponde: the so called "refill"-mode has been removed. It was anyhow
     not implemented.

   - sponde: now checks whether the theta distribution of your on-data
     and the theta-distribution of your Monte Carlo sample (after
     weighting) fits. If it doesn't fit properly (eg. the Monte Carlo
     sample is incomplete) execution is stopped. Execution can be forced
     using the new option --force-theta. Use this option with care!

   - sponde: Proper collection areas can now be constructed also from
     Monte Carlo samples generated with different maximum impact
     parameters. Note that in previous version you neither got 
     a warning or failure, nor was there any obvious sign that the
     collection area was overestimated due to usage of files with 
     different maximum impact parameters.

   - sponde: If MC files with different lower energy limits are used
     the primary MC spectrum is artificially completed down to the
     lowest energy used at all. WARNING: that this gives correct 
     collection areas ONLY if none of the events in this region would
     survive your cuts at all.

   - sponde: the output file now contains more information about
     the spectrum (eg. the full 2D collection area histogram).
     Note, that this information can only be written to the file
     if it is stored automatically via command line argument.
     If you only store the status display from within the display
     the information is lost.

   - sponde: added a new tab "Disp". It shows the residual of the disp
     (Disp-Dist) versus several parameters and it is meant to judge the
     quality of the disp estimator. In the ideal case the residual is zero
     and doesn't depend on any variable. The tab is displayed after
     CutQ and Cut0.

   - sponde: added a new tab "Energy". It shows the residual of the
     energy (lg(estimated energy)-lg(monte carlo energy)) versus
     several parameters and it is meant to judge the quality of the
     energy estimator. In the ideal case the residual is zero
     and doesn't depend on any variable. In reality even a good estimator
     can show residuals versus Monte Carlo energy. The tab is displayed
     after all cuts.

   - sponde: added a new tab "EventDist" showing the unweighted real
     absolute number of events of your sample after cuts. This tells
     you how many events with this energy you had in your MC files.
     The same information you get from the error bars of the weighted
     histograms, but this is less intuitive.

   - sponde: The spectrum plots now show the crab- and 1553-spectrum
     for comparison. It is not meant to show these curves in
     publications, they are only for production.

   - sponde: The OriginalMC tree with the events produced by corsika 
     is now processed only once

   - sponde: Finally fixed most annoying bug, which has effected the
     last bin of the collection area. It was the problem that binnings
     of root histograms are from 1 to n (included) but there was a C-like
     loop in sponde from 0 to n (excluded). Now the loop also includes
     correctly under- and overflow-bins. The worse thing with this is,
     that for the events in the highest bin not only the bin-content
     was wrong, but also the weights applied to these events, which could
     in pricipal effect also other parts of the distribution (eg. due
     to energy estimation). Fortunately it effected only a few events
     in most cases.



 *** Version 1.2 (2007/05/14)

   - database: The database now has two new values Unsuitable50 and
     Unsuitable01. They express the number of pixels which are
     unsuitable for more than 50%, respectively more than 1%, of
     all calibrated event. It is a more accurate number than the
     previously used numbers, because they only take the first
     calibration (ignoring the interleaved calibrations) into account.

   - database: The database now has two other new values UnsuitableMax and
     DeadMax. They express the maximum number of pixels which were
     unsuitable, respectively dead, during the sequence. Because
     of high pedestal rms (cars passing) a few events with very high
     numbers of unsuitable pixels can happen. Not to suffer from this
     effect we don't take the highest 0.1% of the numbers into account.

   - general: fixed a bug which caused callisto and star to stop working
     properly because the callisto output was currupted

   - general: a script called scripts/makemovie has been added which
     simplified the process of producing several movies from a single
     sequence with different setup.

   - callisto: Some more code cleanup. Started to remove old obsolete
     code from cvs.

   - callisto: The extraction of the pedestal with the extractor was
     not random. It seems that extracting at one position is still biased
     (for example in the first or last slices we could still suffer from
     switching noise) Now it is completely random.

   - callisto: In MPedCalcPedRun (which is the pedestal extraction from
     pedestal files) individual pixels could be skipped due to high
     variations. This was already the case for the pedestal extraction
     from the lo-gains since the beginning and properly handled
     there, but was introduced in the extraction from the pedestal files
     a while ago. To calculate the average value it was still divided
     by the number of events procesed not by the number of summands
     really summed. This let to a pedestal which was a tiny amount too
     small (for a few piels in the order of 0.1%). This led to an slightly
     positive offset of the randomly extracted pedestal for a few pixels.
     I doubt that this has a big effect on the result, because the effect
     on the individual numbers is quite small. Due to this it is suggested
     that you rerun your calibration to make sure you are without any
     avoidable bias.

   - callisto: the ArrTimeRmsLimit is now is a check of the deviation from
     the median rms of the absolute arrival time and expressed in significance
     levels. This replaces the old meaning of an abolute upper limit. An
     absolute upper limit doesn't make much sense, because in our data 
     all arrival times are shifted at the same time (in fact it might be
     a shift of the artificial trigger w.r.t. to the light flash).

   - callisto: Added a new tab "BadPixTm" which shows the time evolution
     of the number of unsuitable pixels over the whole sequence. This
     is usefull mainly to judge if an intermediate calibration had
     problems.

   - callisto: Added a new tab "DeadPixTm" which shows the time evolution
     of the number of dead pixels over the whole sequence. Dead pixels 
     in this context are unmapped pixels, i.e. pixels which could not
     be interpolated, and thus are ignored in the further analysis.

   - callisto: It is now possible to use the position of the maximum
     of the spline as arrival time instead of the leading edge even
     when using the integral for the signal. Therefor set the extraction
     type to kIntegralRel ("IntegralRelative") and the relative height
     ("HeightTm") to -1:
       MJPedestalC1.ExtractSignal.ExtractionType: IntegralRelative
       MJPedestalC1.ExtractSignal.HeightTm: -1




 *** Version  1.1 (2007/04/27)

   - general: The new class MGeomCamDwarf contains the camera geometry
     for our prefered Dwarf camera and can be used to create general
     geometries of roundish or hexagonal cameras.

   - mars: The default cleaning is now abolute 8.5/4.0.

   - mars: The resource file mars.rc now contains examples for MUX data

   - merpp: Implemented file format version 8 (MUX FADC data). The 16-bit
     data containing only 10-bit information is still cooked down to
     8-bit, by cutting away the lowest 8-bit.

   - merpp: When reading raw data the data is now stored in a single
     array. The lo-gain array is obsolete. The interface stays the same.

   - callisto: To process the new MUX data use the resource file
        callisto_mux.rc

   - callisto: The -root and -raw options to read merpped raw data have
     been removed

   - callisto: If the -mc option is given the sequence-name can now be
     replaced by the MC sequence number

   - callisto: If the sequence file contains "MonteCarlo: Yes" the
     -mc option can now be omitted

   - callisto: Started further simplification by removing obsolete
     dependancies on MCalibIntensity*Cams.

   - callisto: removed unsused intensity calibration from MJCalibration
     for further simplification

   - callisto: for simplification removed references to
     MBadPixelsIntensityCam, MCalibrationIntensityQECam, 
     MCalibrationIntensityRelTimeCam, MCalibrationIntensityTestCam, 
     MCalibrationIntensityChargeCam, and MCalibrationIntensityBlindCam
     (this should at minimum change nothing at all, and at maximum
      improve thing if these containers where somewhere not handled
      properly)

   - callisto: Added the option to write an event movie. Please use
     this option with care (huge files) but exessively to understand
     our events. For further documentation please refer to the
     IACT wiki
     (currently at http://www.astro.uni.wuerzburg.de/mediawiki-1.9.0)

   - callisto: Improved binning for the pulse position check histogram

   - callisto: The code calculating the pulse position checknow takes
     the pedestal into account

   - callisto: The arrival time is now stored with full precision this
     helps to get rid of artefacts in histograms. For the signal this
     is not necessray because it is displayed in log-scale anyhow

   - callisto: Fixed a problem with internal exclusions. Pixels which
     have once been marked as unsuitable didn't get the necessary
     histograms filled anymore to be rechecked with each interleaved
     calibration.

   - callisto: The relative arrival time resolution (the rms of the
     relative arrival time offset) is now checked against its
     median and the median deviation, instead of the average 
     plus an absolute offset.

   - callisto: The relative arrival time offset is now calculated
     w.r.t. its median not w.r.t. to an artificial reference pixel.

   - callisto: The relative time calibration now marks pixels with
     outlaying offset as unsuitable

   - callisto: The interleaved relative time calibration has been switched
     on. The advantage is that pixels which are marked as outliers
     now can recover. An improvement over time is not expected, because
     the relative time calibration seems stable.

   - ganymed: If the dataset file containes "MonteCarlo: Yes" the "-mc"
     option can now be omitted.

   - ganymed: If either the dataset file containes "MonteCarlo: Yes" 
     or the option "-mc" is specified as default the monte carlo datacenter
     paths (/magic/montecarlo/...) are used if not overwritten in the
     dataset file.

   - ganymed: The output files now contain run- and event-number for all
     events, accessible as RunNumber.fVal and EvtNumber.fVal

   - ganymed: Added a plot showing the source position in the camera also
     for wobble mode

   - ganymed: In addition to the existing task which can be setup to
     be executed after Cut1 a final task called "EstimateEnergy"
     has been added, executed before writing the final output
     (ganymed00000000.root). The task setup in all three cases
     can be choosen individually. (For all three(!) cases you 
     can for example choose MRanForestCalc. For examples see all
     sponde*.rc and all ganymed*rc)

   - ganymed: If a container "MEstimatedEnergy" is found in the 
     parameter list (eg. because it was setup by MEstimateEnergy)
     it will be written to the output file.



 *** Version  1.0 (2007/02/16)

   - database: Website to plot values from the database has been added. 

   - database: The rate of spark events as found by star is now kept
     and the rate after cleaning is now with spark events subtracted

   - general: The changes to the interpretation of phrases lead to
     a problem with the access of MHMatrix objects (eg. "M[0]")... fixed.

   - general: Added example files how you can create your own calendar
      (eg. as christmas present) with Mars. Just run
           root macros/tutorials/calendar.C

   - general: Added a possibility to display grouped data in the plotdb.C
     macro (and similar macros). This allows to plot hourly, nightly,
     weekly, monthly and yearly averages. 
     For more details see the plotall function of the macro

   - general: fixed the error display of the camera display. It showed
     wrong results.

   - general: fixed a memory leak when reading many MStatusDisplays

   - general: added a macro (macros/tutorials/mirrordelay.C) plotting the delay
     between a spherical and a parabolic mirror

   - callisto: We have a new pulse position check which takes the properties
     of the extractor automatically into account. So no need to set 
     ExtractWinLeft/Right anymore. Again a check for pulses too much to the
     left is done. If possible the lo-gain extraction start and hi-gain
     extraction end is changed accordingly. If no proper extraction range
     can be set (hi-gain too much to the left) an error is raised.

   - callisto: New swapped pixels have been found and corrected

   - callisto: Redone the signal extraction this includes:
     + Unified the algorithms to extract the pedestal for all pedestal
       extractions done
     + Unification of hi- and lo-gain into one array
     + Unification of hi- and lo-gain extraction for Spline and
       Digital Filter (means now exactly the same algorithm is used,
       not a copy-and-pasted one)
     + Store pedestal subtracted data in a common place
     + Improved extraction algorithms. If the algorithms search a 
       maximum but don't find one they don't return the maximum anymore
     + Improved the extraction such that a information for signal
       and time which is consistent is returned
     + replaced the numerical spline with a (not slower) analytical
       solution
     + Lowered the saturation limit from 250 to 245 to make sure that
       we are far away in all pixels from problems with the clock-noise
     + The start of the lo-gain extraction is calculated from the hi-gain
       arrival time. In case of saturation an estimation is used.
     + the hi-gain window was extended to slice 16. This allows extraction
       of very late pulses in single pixels or of huge hadron showers.
     + Now uses the lo-gain extraction range for pedestal extraction,
       not the hi-gain range in the lo-gain

   - callisto: New hi-/lo-gain calibration constants have been produced
     for all periods from the statistic of the whole period. They are
     now used automatically depending on the periods. The old constants
     where deviation in avarega about 2%-3%. The constants are very
     stable from period to period (<<1%) except for exchanged channels.
     In a few cases the constants used so far were off by 1%-2% (maybe
     temperature effects?)

   - callisto: To gain obsolete space we do not write the 
     MPedestalFundamental anymore. Also calibration and bad pixel
     treatment is not done for it anymore because it has never been used.

   - callisto: Fixed a bug in the seuqence handling which caused the
     light condition to be ignored

   - star: Added two new tabs "Sparkless" and "Sparks" these tabs show
     basically the same as "Cleaned" but the distribution devided into
     the events fullfilling a spark cut or being rejected by it.
     This cut can be changed from star.rc (see example inside) by
     the "SparkCut" directive

   - star: Star displays now the average individual pixel-rate (IPR) versus
     time and the average discriminator threshold (DT) of all pixels.

   - optim: fixed a problem with the optim-macros which was due to a bug
     in the new phrase parsing

   - ganymed: To get a nice FalseSource plot (instead of the all-information
     debug plot displayed as a standard plot) use the context menu
     somewhere between the two upper plots, click on "DrawNicePlot"
     and enjoy the result. (It is the same context menu which contains
     the other options)
     PLEASE ALWAYS USE THIS PLOT FOR ANY KIND OF PRESENTATION!

   - ganymed/optim/train: The cut program, trainings- and optimization
     macros could run with less files found than expected from the
     sequence-files. No the execution is stopped in this case.

   - ganymed: now stores the dataset in the output file with all information
     contained (as sequence number, sequence path, data path) so that the
     used dataset file can be recreated complete from the output

   - sponde: now also allows the uasge of random forest

   - sponde: added a plot showing the cut efficiency

   - sponde: Using root versions < 5.13/04 still resulted in wrong binomial
     errors (eg. effecting the collection area) because the errors of the
     two histograms were swapped in the calculation, which makes a difference
     in the case the histograms contains weighted events. Aditionally the
     totalarea was incorrectly taken into account, which results in too small
     errors, which makes a difference at low statistics were the errors
     or the collection area dominate.



 *** Version  0.10.3 (2006/10/23)

   - database: Implemented ObservationMode flag in websites.

   - database: For sequences there is now also the stoptime available on the
     websites. 

   - database: Removed ManuallyChanged flag from sequence-info websites, 
     as it is not needed anymore with the new sequence building algorithm.

   - general: MRolke.[cc,h]: Modification of TRolke from root_v5.12.00b.
     Class for calculation of confidence intervals (upper limits). 
     Use model 3 or 4 for upper limit calculation for MAGIC observations.
     Please read the class instruction of TRolke for more information.

   - general: The programs now return 0xfe if the requested resource file
     doesn't exist.

   - general: Errors comming from root itself are now not output to
     stderr anymore but to our default (colored) log-stream.

   - general: Changed the interface to TFormula such that now everything
     is parsed by TFormula, which should allow to use phrases even
     without paranthesis correctly parsed.

   - callisto: replaced the AdHoc number of effective integration slices
     in the digital filter  by an average of the correct number, averaged
     over all possible set of weights. This also gives correct numbers
     for the lo-gains which were totally wrong before.

   - callisto: drastically reduced memory consumption by not storing
     all intermediate calibration results in memory anymore (we might
     write them into a file in the future)

   - ganymed: To get a nice Theta-Sq plot (instead of the all-information
     debug plot displayed as a standard plot) use the context menu
     somewhere between the two upper plots, click on "DrawNicePlot"
     and enjoy the result. (It is the same context menu which contains
     the "DrawAll" option)
     PLEASE ALWAYS USE THIS PLOT FOR ANY KIND OF PRESENTATION!



 *** Version  0.10.2 (2006/10/15)

   - general: changed the building of the sequence files such, that
     the source-/project name in case of wobble-sources is not the
     source/project of the first data run anymore but the new generalized
     wobble source/project

   - general: Added some functions to MAstro to calculate sun and
     moon properties

   - general: Added two new macros (starvisday.C and starvisyear.C)
     which allow to plot the visibility of several stars in one day
     or of one star along one year



 *** Version  0.10.1 (2006/10/11)

   - general: HTML logging output was distorted... fixed.

   - general: Changed the writing of the sequencefiles such that only
     as many pedestal files as giving at least 1000 pedestal events which
     are nearby the first calibration run are used as pedestal runs

   - sponde: In the calculation of the collection area(s) and the 
     distribution for MOnte Carlo and estimated energy the error
     calculation was wrong because root didn't take the errors
     properly into account... fixed.



 *** Version  0.10 (2006/10/10)

   - general: all executables now check consistency of the root version
     installed with the root version used for compilation. The return
     code in this case is 255.

   - general: The algorithm to build seqeunces has been redesigned
     completly. It is now more flexible in terms of a resource file
     resources/sequences.rc which can be used to partly overwrite
     the default algorithm and defined the important part of the
     algorithm. It also takes correctly into account several ways
     of taking wobble-data. Currently the new algorithm doesn't take
     the number of pedestal events anymore into account (so a sequence
     could have much less than 1000 pedestal events and data runs are
     no-more used instead)

   - general: a lot of small changes to make the code compatible with
     root 5.12/00 and gcc 4.1

   - general: Fixed some warnings thrown if more warnings are switched
     on in the compiler

   - general: Added a function to MMath which returns the Median of
     the distribution abs(y[i]-median) but at 68:32 division instead 
     of 50:50, which is somehow the median counterpart of the rms.

   - general: MAstroCatalog allows now to read calatog files (as bsc5)
     even if they are compressed with gzip (recommended compression
     is with "-1")

   - showplot:
     + batch mode creation of image files now works with root 5.12/00
     + added support for writing bmp (though it never produces
       bmp which can be read by programs like xv or gimp)
     + added support for xml
     + added new option to start the display with a different size, by
        --display-width=nnn or --display-height=nnn
     + added options to define the canvas size when the display is started
        --canvas-width=nnn or --canvas-height=nnn
       the option can also be used for batch-mode conversion
     + added an option which automatically fits the display size to the
       desktop-size: --auto-size

   - merpp: didn't recognize files with the extension .raw.gz - fixed.

   - merpp: allow the reports to have 1000ms due to a bug in the camera
     program.

   - callisto: New swapped pixels have been found and corrected. For
     details see MRawRunHeader::FixAssignment()

   - callisto: Callisto now raises an error if one of the input files could
     not be opened.

   - callisto: In some rare cases it could happen that in the signal
     extraction of the digital filter the extractor returned nonsense
     results due a wrong initializion of a variable (this happens in
     cases in which the extracted signal is exactly zero or the
     digital filter couldn't slide because of a too small extraction
     window, which could happen in the lo-gain extraction)

   - callisto: The lo-gains have been marked as valid in some cases 
     when they got not extracted at all (deltsumlo=deltatimelo=0) because
     the allowed extraction range (fLoGainLast-fLoGainFirst) was
     smaller than the number of slices (fLoGainWindowSize) needed
     by the lo-gain extractor

   - callisto: Updated absolute calibration constants (muon calibration)

   - callisto: fixed a bug which caused the hi-gain saturation not
     to be handled properly in the spline (the position of the
     saturation was calculated incorrectly)

   - callisto: fixed a bug which gave weird results if hi- _and_
     lo-gain could not be extracted properly. In some cases hi-
     and lo-gain have not been extracted at all and the random result
     got calibrated. In other cases the hi-gain was saturating and the
     lo-gain has not been extracted. For this a completly arbitraty
     scale factor was applied to the result from extracting the
     charge of the saturating hi-gain.

   - callisto: improves handling of extracted signal if lo- and/or
     hi-gain could not be extracted properly. If no appropriate
     calibration of either the signal or the arrival time is possible
     the pixel is now markes as unsuitable instead of estimated.

   - callisto: Fixed a problem in the spline extraction which gave
     weird results if the signal was extracted from the first (and
     highest) slice (as a workaround extraction from the first
     slice has been forbidden)

   - callisto: Fixed a bug regarding the use of the QE which doesn't
     effect the result because the default is used anyhow.
     (The bug was there since 14/07/2005)

   - callisto: Changed the default for fgLoGainStartShift to -2.4

   - callisto: Fixed a bug which caused all hi-/lo-gain calibration
     factors to be always 10 for all pixels as soon as the first
     calibration update has happend.

   - callisto: Fixed the hi-/lo-gain offset for the Digital Filter. It
     was off by more than one slice (which gives wrong values for
     pixels which have saturated the hi-gain) This was a problem
     mainly for timing studies, not so much for the automatic standard
     analysis. Also the values for the Spline have been changed slightly:
        + Digital Filter from 1.70 to 0.95
        + Spline         from 1.39 to 1.30
     The numbers were sorted out by an extensive test in the
     test-database.

   - callisto: Removed the tab showing the average calibrated arrival
     time. It never gave useful information.

   - callisto: Now tab "HiLoCal" which shows the ratio between hi- and
     lo-gain pulse for data pulses if hi- and lo-gain have saturated
     and both were extracted because the hi-gain was above the 
     LoGainSwitch.

   - callisto: Now tab "HiLoOff" which shows the offset between hi- and
     lo-gain pulse position (after application of LoGainOffset, which
     means in the ideal case it is 0) for data pulses if hi- and lo-gain
     have saturated and both were extracted because the hi-gain was above
     the LoGainSwitch.

   - callisto: After extensive tests a new threshold level for determination
     of the pulse position has been found at 50phe. This is due to the
     fact that for lower pulses the arrival time changes its average
     behaviour dramatically. It tends to the average of the extraction
     range and starts sticking to non-floating point numbers (5, 6, 7, ...)

   - callisto: Added two more plots to the calibration part. The average
     extracted signal and the average extracted arrival time as
     extracted by the signal extractor (This can help debugging problems
     because it doesn't involve a fit)

   - star: changed the fit for the effective on time such that initial
     values are calculated automatically now instead of using build
     in values. This makes the fit more indepedant of the underlaying
     rates. With a test of ~350 sequences in the test database it could
     be shown that the new fit gives the same result +/-1sek.
     The highest deviation was +5s the lowest -10s. You can now control
     the bahaviour with two resource values:
        MHEffectiveOnTime.FistBin:   3
        MHEffectiveOnTime.NumEvents: 12000
     Use FirstBin=1 and NumEvents=120 to fit the effective on time
     from the resulting gamma candidates

   - star: The Center histogram in MHHillas is now by default displayed
     with the colz draw-option

   - ganymed: now produces valid error codes if failed

   - ganymed: MFMagicCuts now supports a hadronness cut. Switch it on
     with the option "hadronness" instead of "area" or "both" if
     you want to use both cuts. The hadronness cut includes a fixed
     cut in hadronness (index 8) and a fixed cut in size (index 9).

   - ganymed: MFMagicCuts: fixed abug introduced 2005/7/14 which caused
     the M3Long of the anti source to be not properly evaluated 
     (actually is was evaluated as a unique random number) in case
     of optimization with anti-theta cut switched on. This might have
     lead to improper, but not necessarily wrong, optimization results.

   - ganymed: ganymed_onoff.rc, ganymed_wobble.rc - updated with comments
     how to use random forest

   - ganymed: use the same spark cuts as in ganymed_wobble.rc as in
     ganymed_onoff.rc. Seems to be unintentional that they were different.

   - sponde: Since the introduction of the possibility to define a
     normalization energy instead of using the integral the default
     was not using the integral anymore but a normalization energy of
     1GeV. It is now the integral again.

   - sponde: Fixed a problem in the weighting which did not allow
     a MC slope of -1.



 *** Version  0.9.6 (2006/05/24)

   - general: changed the binning of the azimuthal profile of the camera
     such that it is aligned with the 60deg boundaries.
     
   - merpp: better handling of problems with the TH, TD part of the 
     CC-REPORT for files older than 200507190 and 200412210 
     respectively

   - merpp: improved handling of RECEIVERS-COM-ERROR in CC-REPORT

   - merpp: implementd latest changes in starguider reports (sinc 9th May)

   - callisto: if an arrival time is extracted outside the fadc range is
     was set to the boundary before. Now a random number is assigned
     instead to avoid unfortunate coincidences of arrival times just
     by this arbitrary value.

   - callisto: added a new option to callisto.rc which allows
     to set the maximum arrival time difference allowed for the 
     badpixel algorithm to determin whether a pixels belongs to a shower:
       MJCalibrateSignal.MBadPixelsTreat.MaxArrivalTimeDiff: 0.9

   - callisto: changed the old value (0.5) for MaxArrivalTimeDiff 
     to 0.9 as suggested by Stefan due to a recent study on the fraction
     of pixels surviving the image cleaning

   - callisto: switched off error in case of "pulse is too much to the left"

   - callisto: The extraction window is not adapted anymore. This means that
     the pulse is extracted from the full range. The draw-back is that the
     background (pedestal) will become slightly higher. Switching on/off
     the PulsePosCheck in the callisto.rc only effects filling the
     histograms.
     WARNING: If the pulse is too much to the left data is calibrated!
     Check the PulsePosCheck-PulsePosition AND the Inhomogeneity parameter
     in the database to make sure that your data is valid!

   - star: Show also the number of stars correlated correctly by the
     starguider

   - ganymed: implemented a new class (MHThetaSqN) which allows to use more
     than one off-source region in wobble-mode. To use it add the following to
     your ganymed_wobble.rc:
      + MJCut.NameHist: MHThetaSqN            (switch the new feature on)
      + MHThetaSqN.NumOffSourcePos: 3         (define number of off-regions)
      + MHThetaSqN.DoOffCut: Yes,No           (switch on/off the off-cut)
      + Cut1.ThetaCut: None                 
      + MHThetaSqN.SignificanceCutLevel: 2.0  (increase off-cut by 2.0/1.7)

   - ganymed: in the case the latest report is older than a default (currently
       one minute) and the current report will be skip the starguider
       correction and calibration will be reset. The maximum age can be
       setup from ganymed.rc by (minutes):
         MPointingDevCalc.MaxAge: 1.5

   - ganymed: A first implementation showing all size-bins for the theta-sq
     resp. alpha-plots in a single tab automatically



 *** Version  0.9.5.1 (2006/05/05)

   - general: Added a new option "zdiff" to the Makefile which runs
     "cvs diff" as "make diff" would do, but with compression

   - general: accelerated MTFillMatrix (used for eample in all training
     and optimization macros) by skipping some obsolete calles in the
     eventloop the first and second loop could be accelerated by ~20%

   - general: Added a missing feature in the MFilterList class which
     prevented MFEnergySlope from working correctly in trainenergy.C

   - general: Accelerated the random forest training and usage a bit

   - merpp: Adapted to new raw data file format version 6

   - merpp: fixed a problem with reading the starguider reports 
     after 6.3.2006

   - callisto: At some times six pixels had to be exchanged manually in the
     data. If the pixels to be swaped are both not found, an error was raised
     and calibration was stopped. Now it is assumed they are intentially not
     mapped and the Error has been changed to a Warning.

   - star: Added new image cleaning based on the arrival time of the shower.
     The new cleaning can be accessed using the parameter "Time" instead of
     the the old one (like "Absolute) in the star.rc-file.
     Recommended parameters:
      + MImgCleanStd.CleanLevel1:  8.2
      + MImgCleanStd.CleanLevel2:  5
      + MImgCleanStd.CleanRings:   2
      + MImgCleanStd.KeepSinglePixels: No

   - star: Simplified the calculation of the effective on time. By letting
     the fit start in the third bin the result should become more stable.
     The fit quality is now ignored for the theta-plot which is not used
     in the analysis later. Still better pre-cut which, for example,
     remove events triggered by car-light (rate changing on a short
     time-scale) are missing.

   - macros: optimonoff.C, optimwobble.C: implemented a new strategy
     for optimization (kWeakSource) which is suggested by the Berlin
     people for weak sources.

   - macros: plotdb.C: 
       + if no dataset and no time-interval is given only sequences with
         an effective on-time above five minutes are plotted anymore
       + added a new tab "RelTime" displaying the relative on-time
         (effective on-time divided by the observation time)

   - macros: optim*.C: by skipping some obsolete calles in the eventloop
     the optimization became a lot faster

   - ganymed: In the second loop the MHNewImagePar histograms disapeared.
     To display them just delete the pads displayed on top.

   - ganymed: fixed a bug which could cause strange behaviour in wobble
     mode if no starguider information was available due to an incorrect
     initialization of the starguider calibration

   - ganymed: The scale region in the False Source Plot (MHDisp) can now
     be changed from within the plot. It can also be setup from the
     resource file using:
       + MHDisp.fScaleMin: 0.325
       + MHDisp.fScaleMax: 0.475

   - ganymed, showplot: Fixed a problem with the fit to the off-data
     (HistOff) which could cause a time lag or an inifinite loop.
     The bug did not effect the results at all.

   - sponde: Added a new tab showing the effective collection area after
     trigger (including the software trigger)

   - sponde: MMcSpectrumWeight has a new option which allows to define
     the scaling factor not by the integral of the events but at
     a defined energy:
        e.g. MMcSpectrumWeight.NormEnergy: 200

   - MTFillMatrix, sponde: by skipping some obsolete calles in the
     eventloops the loop could be accelerated by ~20%



 *** Version  0.9.5 (2006/03/13)

   - macros: fixed a typo optimwoble.C (SetThetaCut instead SetAlphaCut)
     and added the standard Disp-parametrization

   - macros: added a new macro datacenter/macros/plotstat.C allowing
     to plot statistics about the data processing

   - macros: added a new macro to plot the muon correction coefficients
     versus period (macros/plot/mucal.C)

   - showplot: path inflation for ganymed files was broken

   - callisto: Fixed an important bug. When the number of saturating pixels
     were calculated the bad pixel information already available from the
     calibration was ignored. This yields wrong values in case pixels
     are broken and saturate all the time or randomly.

   - callisto: Updated Muon calibration constants. Spline data now also uses
     different values for different periods

   - star: The time-evolution shown in the effective on-time plot now displays
     the axis range (if there are no holes it should be similar to the
     effective on-time)

   - star: Added two new image parameters:
       + MNewImagePar.fConcCOG: The ratio of the equivalent signals of
         the three pixels next to the center of gravity and the size
       + MNewImagePar.fConcCore: The ratio of the signals of all
         pixels inside or touching the ellipse and the size

   - ganymed: now stops in requested source isn't found in catalog file

   - ganymed: Many improvements to the phi-plot in wobble mode:
      + with a template the ideal background is displayed
      + a possible camera inhomogeneity is taken into account
      + the anti theta cut is better supported
      + the phi cut is automatically calculated from the theta cut
     See the examples in ganymed_wobble.rc for more details.

   - ganymed: fixed plot for E^2*dN/dE. The spectrum was multiplied with two
     times the bin-width instead of the absolute energy

   - ganymed: New cuts for on/off and wobble mode

   - ganymed: The smearing of the False Source plot has been increased 
     to 0.06

   - ganymed: The starguider calibration can now be changed from the 
     resource file.
      See the examples in ganymed*.rc for more details.

   - ganymed: added a new option for on-/off-observations
       MJCut.RandomSourcePosition
     If switched on the distribution of observation time of the source
     position in the camera is calculated from the on-data and randomly
     applied with this distrbution to the off-data.

   - ganymed: by skipping some obsolete calles in the eventloop the
     first and second loop could be accelerated by ~20%

   - ganymed: The default in ganymed_onoff.rc is now to determin the source
     position in the off-data randomly according to its distribution in the
     on-data.



 *** Version  0.9.4.3 (2006/02/13)

   - general: Fixed a bug in most of the programs which prevented the
     compile date in the startup message to be displayed

   - general: Simplified program calls. The following abbreviations 
     are now possible:
        showlog callisto 55599
        showplot signal 55599
        star 55599
        callisto 55999
        ganymed 2
     In this cases the corrsponding root-, log-, sequence- or dataset-
     filename is automatically compiled using the default datacenter
     paths ("/magic/*")

   - general: Fixed a long outstanding bug which made axis tick marks
     disappear if MDataFormula was used (eg. abs(x))

   - general: Updated MMath with new functions to calculate the results of
     a exponential, logarithmic and powerlaw fits analytically.

   - general: Updated some macros with comments:
      + macros/optim/optimdisp.C
      + macros/optim/optimenergy.C
      + macros/optim/optimwobble.C:

   - general: MTFillMatrix (the class to fill one or two MHMatrix from
     files) now allows adding a pre-cut like in the optimization. E.g. this
     is useful to perform g/h-separation cuts before training the random
     forest.

   - general: Between the rund 53300 and 68754 the pixels 553-558 were not 
     correctly assigned in the raw data assignement table. A fix to correct 
     the assignment tabel have been implemented.

   - general: MAstroCatalog now allows rotating of the sky-grid and stars
     by 90, 180 and 270 deg.

   - general: the macro plotdb.C now allows to highlight sequences belonging
     to a datset. For more details see the descritions in the macro.

   - StatusDisplay: Fixed the algorithm to calculate the size ratio
     of the display, it gave wrong results espacially for big sizes.
     Now the embedded canvas should always have a ratio width/height=3/2.
     The "Size" options have been updated with values fitting the 3/2 
     size ratio.

   - RanForest:
      + Updated the random forest classes to support also the
        regression method implemented by Thomas H.
      + added new tutorial macro how to train the random forest
        for energy estimation (macros/optim/rfenergyest.C)
      + new classes to train the random forest (still in development)
        mjtrain/MJTrainEnergy, mjtrain/MJTrainDisp, mjtrain/MJTrainSeparation
      + new tutorial macros for random forest training in macros/train
        trainenergy.C, traindisp.C, trainseparation.C

   - mars: now allows to setup a cut in the resource file mars.rc to choose
     the events to be displayed, eg:
       Cut.Condition: {0} && {1}
       Cut.0: MHillas.fSize>100
       Cut.1: MHillas.fSize<900

   - merpp: Made aware of the new .rep format 200510250. This includes
     interpretation of the Receiver Board Temperatures (MCameraRecTemp),
     the discriminator thresholds (MCameraTD) and the disciminator
     delays (MCameraTD). All of them come with the CC report and are therefor
     written to the CC tree.

   - callisto: fixed a bug in the calculation of the arrival times in
     the Digital Filter. They were all shifted by the same amount
     fHiGainFirst. This caused an offset in the order of the position of
     the left border of the extraction window in the times calculated
     from high-gain and from low-gain. The bug effected only the absolute
     arrival times. The bug was there since 30.1.2005.

   - callisto: Changed the ExtractionWindow for the pulse position check
     such that the camera inhomogeneity becomes as less as possible and
     comparable to the pulso position check switched off. More details
     on the study will be given at our result web-page.

   - ganymed: ganymed.rc, ganymed_onoff.rc:
     changed to new cuts trained with 1ES1218 and Crab-data
     for maximum sigma*log10(excess)

   - ganymed: In addition to the Hadronness calculator (CalcHadronness)
     a new option was implemented to estimate Disp (CalcDisp)

   - ganymed: Implemented two new options which allow
     to overwrite the default path to search for the sequence files
       --ins=/magic/sequences
     and where to search for the image data
       --ind=/magic/data/star

   - ganymed: MSrcPosCorrect now allows overwriting of the misfocussing
     dx/dy from the resource file by 
       MSrcPosCorrect.Dx
       MSrcPosCorrect.Dy

   - ganymed: fixed reading the source catalog. Only as many chars
     as the requested source name had were compared (If a source
     "1ES1218" and "1ES1218+304" were in the catalog and "1ES1218+304"
     were requested it could happen that the coordinates of
     "1ES1218" were read) The source coordinates are now printed
     in addition to its name to the logging output.

   - ganymed: changed the background fit in the case of a ThetaSq function
     from a polynomial to a exponential, this should give a big improvement
     for on-only analysis'

   - ganymed/sponde: Calculation of the error of single size-/energy-
     bins now uses Li/Ma (5) instead of LiMa (17)

   - sponde: Added a plot E^2*dN/dE

   - sponde: The energy estimator plot should now show values like
     they are commonly used.

   - sponde: Now MMcSpectrumWeight also excepts formulas with two X
     (a powerlaw with cutoff didn't work before)



 *** Version  0.9.4.2 (2005/10/01)

   - general: added new tutorial macro macros/tutorials/threshold.C:
     how to calculate the threshold

   - general: added a new game. Start it from the interpreter with
        MagicJam j; (other games are: MineSweeper, MagicSnake, MagicDomino
                     and MagicReversi)

   - general: neighbor pixels in MGeoms are now sorted clockwise

   - merpp: made aware of new format 20050829-0

   - callisto: New scale factors from muon analysis, stored in new 
     resource file resources/calibration.rc

   - callisto: MBadPixelsCalc now raises an error instead of simply stopping
     the eventloop if something went wrong. This is necessary for the
     automatic processing

   - callisto: Implemented a new treatment for the time information 
     developed by S.Ruegamer, this should give much better response than
     the old very primitive algorithm. This is especially important
     for bad pixels due to bright stars as Ceta-Tauri (btw: the old
     treatment was buggy! As neighbor pixels to interpolate the 
     arrival times always the pixels 0 to 5 were taken)

   - callisto: In addition to the new treatment of arrival times the
     decision of the treatment whether a pixel contains shower
     information or not is used to improve the signal interpolation
     further. Previously interpolated pixels showed a non-zero signal
     do the peak searching signal extractors. Having a signal pixel
     beside raised the signal above the cleaning level. Thus the
     pixel survived cleaning to often but with a too low mean signal.
     By replacing empty pixels with a gaus (mean and rms from the
     corresponding Area entries in MPedPhotCam) this effect can
     be supressed.

   - callisto: The status "unmapped" (for pixel which cannot be interpolated)
     got lost when the events were stored by the calibration... fixed.
     (This was no problem for normal image cleaning because the pixel 
      content was artificially set to 0 not surviving standard cleaning
      algorithm)

   - callisto: Fixed a bug which caused the random and peak-search pedestal
     extracted with an extractor to be exchanged for the first (roughly)
     500 events (*Extractor and *ExtractorRndm). 
     (They were simply exchanged in callisto.cc) This bug might have been
     introduced in Mars 0.9.4 when the order of calculation of the two
     types of pedestal in callisto got exchanged. The bug only effects
     the first seconds of data of each sequence.

   - star: Taking the arrival time of Muons into account the calculation
     of the parameters could be made independant of the noise 
     (pedestal rms). The dependance was due to the bias of the peak search
     extractor. Now only a ZA dependamcy is left which is not yet fully
     understood.

   - ganymed: The sequences were still not sorted correctly in any case, 
     because they were sorted by the full qualified path name. Now they
     are sorted by their sequence number.

   - ganymed: new values for the spark-event-cut

   - sponde: required missing MEnergyEst from ganymed.root... fixed.



 *** Version 0.9.4.1 (2005/08/18)

   - callisto: fixed some problems with the calibration in case of
     inteleaved events. Therefor the final and some intermediate
     fits are skipped which would take place on partly filled #
     histograms and gave wrong results or failed completely.

   - callisto: If the intermediate finalization of the histograms
     calculating the mean charge of the calibration signal fails
     it is counted now and printed in PostProcess of MCalibCalcFromPast.

   - ganymed: Fixed some bugs which caused problems in On-only mode.
     Still the false source plot doesn't give reasonable results.



 *** Version 0.9.4 (2005/08/05)

   - general: Fixed the ZA binning. It did not correctly fit the
     MC binning

   - general: Added a class to correct for the missfocussing in
     April (MSrcPosCorrect)

   - general: Random Forest energy estimator should now work properly

   - general: MHCamera now displays the profiles in deg instead of mm

   - general: MH::SetPalette offers a lot of new palettes

   - general: MHillas - the case of CorrXY==0 is now handled properly

   - general: implemented the possibility to change the line and
     marker style of a sky-grid drawn by MAstroCatalog

   - general: Runs in sequences and datasets are now automatically 
     ordered (it is assumed that the order of their filenames correctly
     represents the order of observations) to ensure correct reading
     of the subsystem data

   - general: Runs cannot be added to a sequence twice anymore. Also
     Sequences cannot be added twice anymore to a dataset.

   - general: fixed a bug in MAstro::Deg2Hms reported by Abelardo. 
              It gave wrong results by 1./2pi

   - showplot: got a new option to start a root interpreter, too

   - mars: now displays a rough estimate of Disp and the third moment

   - mars: now displays the event time MTime if available

   - mars: show muon parameters graphically

   - mars: now the file to open can be given as commandline
     argument

   - merpp: Added support for new arehucas version >= 200507190
     implementing values for three temperature sensors in the
     counting house (MCameraAUX) and the name of the HV settings file
     (MCameraHV).

   - merrp: Also fixed some problems with wrong versions numbers stated
     by arehucas and implemented some fixed for old versions (<200407070)

   - callisto: MCalibrationHiLoCam can now be printed from its context
     menu, eg in the TBrowser

   - callisto: fixed logain offset (fgOffsetLoGain) from 1.7 to
       - 1.39 (MExtractTimeAndChargeSpline) 
       - 1.40 (MExtractTimeAndChargeDigitalFilter)
     This is important mainly for timing studies.

   - callisto: Changed limits in MHCalibrationChargeCalc from
       - -100.125 to -98   (fgChargeHiGainFirst)
       - 1899.875 to 1902. (fgChargeHiGainLast)
       - -100.25  to -99   (fgChargeLoGainFirst)
       - 899.75   to 901.  (fgChargeLoGainLast)
     Introduced new limits:
       - fgNumLoGainBlackoutLimit: 0.05
       - fgLoGainBlackoutLimit: 3.5
       - fgLoGainPickupLimit: 3.5

   - callisto: The handling of automatic weights for the digital filter 
     has been improved. There are now new calibration weights special
     for Monte Carlo calibrations.

   - callisto: Updated Muon scale with latest results (1% difference
     for most data, 5% for Mar/Apr'05)

   - callisto: use a new criterium to exclude bad pixels: If the high-gain
     was saturated and the blackout-events in the low-gain exceed the 
     fNumLoGainBlackoutLimit, the pixel is declared unsuitable.
     This excludes those pixels which have a saturating high-gain 
     channel, but the low-gain switch does not switch often enough 
     to make the distribution reliable.

   - callisto: fix a bug counting the number of saturated events. Up to now,
     the number of saturated slices was counted (which is one for a not too
     high number), but for some (pathological) pixels, many more slices
     saturated and produced wrong limits.

   - callisto: New options in in callisto.rc for MCalibrationRelTimeCalc:
     + MCalibrationRelTimeCam.CheckFitResults: Yes
     + MCalibrationRelTimeCam.CheckDeviatingBehavior: Yes
     + MCalibrationRelTimeCam.CheckHistOverflow: Yes
     + MCalibrationRelTimeCam.CheckOscillations: Yes

   - callisto: introduce max. number of events for intercalibration, 
     the rest gets skipped. Default: 4096
     The calibration causes too many un-reliable pixels if more 
     than about 5000 events are treated (@500 Hz) because of the
     mode hopping of the VCSels. However, in the past, some 
     calibration runs have been taken (erroneously) with more 
     than 5000 events, especially the intensity scans where  
     a good precision is needed.

   - callisto: Set UseExtractorRes to kTRUE, will raise the number of 
     photo-electrons for the spline extractor by 1-2% (correctly).

   - callisto: The intercalibration is now done using a full f-factor
     method instead of a simple update. This makes things slower but
     more reliable in terms of gain-fluctuations and sudden VCSEL gain
     changes.

   - callisto: MPedestalY3 now uses the same extraction range as the data
     as calculated by MPedestalY2

   - callisto: A check of the fraction of allowed unsuitable (10%) and
     unreliable (30%) pixels has been addded when the conversion is 
     calculated from interleaved events

   - star: fixed a bug which caused MEffectiveOnTime containers not to
     be written. In the worst case they were skipped almost all, in the
     best case they were all ok. It depends on how often the new value
     was assigned just with an event which was skipped later by another
     task

   - star: new histograms MHPointing and MHRate

   - star: New tab (Currents) displaying the time evolution of the mean 
           DC currents

   - star: New tab (MeanRms) displaying the time volution of the mean
           Pedestal Rms

   - star: Output now contains MReportStarguider

   - star: Improved muon analysis. Due to the absense of a constant
           for the gaus fit for arc width and the fact that the arcwidth
           plot was a histogram instead of a profile the resulting psf
           became pedestal rms dependant

   - ganymed: added new plot (MHPhi) for wobble mode as described in
     http://www.astro.uni-wuerzburg.de/results/ringmethod/

   - ganymed: corrects now (correctly?) for the misspointing in 
              May/April 05

   - ganymed: Two new default templates for ganymed.rc are avilable:
              ganymed_onoff.rc and ganymed_wobble.rc

   - ganymed: The first version of MFMagicCuts has been released

   - ganymed: the Conc1 plot was incorrectly scaled in MHVsSize

   - ganymed, sponde: There are two new options to determin the number
     of bins for the signal region (NumBinsSignal) and the number of
     total bins (NumBinsTotal) in the MHThetaSq histogram

   - ganymed: optimized palettes for MHDisp

   - ganymed: Fixed an important bug when using ThetaSq as fit function:
     The result for the sigma of the gauss(sqrt(x)) fit was a factor
     sqrt(2) to large

   - ganymed: The false source plot (MHDisp) is now based on Disp
     and a background model determined in the first loop is
     subtracted

   - ganymed: MFMagicCuts are now available in the cvs. The correct
     solution for the disp ambiguity problem is now calculated using
     the third moment. The third moment cut therefor is skipped.

   - ganymed: Output now contains "Disp [MParameterD]" which sign is
     already determined by the third moment along the major axis and
     so its actual sign is aligned along the major axis of the shower.

   - ganymed: If starguider data is available it is taken into account
     when calculating the source position. This is especially important
     for wobble data at high zenith angles and culmination data.

   - ganymed: If available the starguider data is taken into account as
     good as it is possible without a starguider calibration. To switch
     it off add
        MPointingDevCalc.MaxAbsDev: -1
     to your ganymed.rc

   - sponde: the zenith angle distribution is now weighted instead of
     choosen by default. This gurantees higher statistics in the highest
     and lowest energy bins. To switch this behaviour off use...

   - sponde: now automatically displays the energy threshold



 *** Version 0.9.3 (2005/06/03)

   - general: MMcEvt now derived from MMcEvtBasic which should 
     have no influence on compatibility with older camera files

   - general: Implemented a new option to MReadReports which allows
     to force reading of the first event of some trees (eg. Drive-tree)
     before any other tree.

   - general: Made a fix to MPointingPosCalc which assumes that the
     run-type is real-data if Process is called before ReInit, such
     that also the very first Drive-events in a run are treated 
     correctly.

   - shoplot now displays the filename in the title bar of the window

   - merpp: new options: --auto-time-start --auto-time-stop which
     allow to cut updating with slow-control at the start- or 
     (previousle --auto-time only allowed: and) stop-time of
     the run stored in the runheader.

   - callisto: changed default of fgFallTimeHiGain in 
     MExtractTimeAndChargeSpline from 1.5 to 0.5. 0.5 was
     already set as default in callisto_Dec04Jan05.rc

   - callisto: the digital filter (MExtractTimeAndChargeDigitalFilter)
     now changes the filename for new weights automatically for
     cosmics, MC and for different calibration colors.

   - callisto: a problem with the order of some code cause the
     name of the weights file in MPedestalY2 to be ignored.
     Wrongly the calibration_UV weights were used instead
     of the cosmics weights set in callisto.rc.

   - callisto: should allow ct1 pulser data when the color is
     explicitly set also for pretty new runs

   - callisto: a new option (MCalibrateData.ScaleFactor) allows
     to introduce an additional scale factor (should be around 1)
     to adapt for slight deviations in different extractors from
     the MC calibration. For MC calibrations the scale factor
     is fixed to 1

   - callisto: MFCosmics should handle very high signals a bit
     more accurate now

   - callisto: when calculating the mean pedestal rms for different
     pixel sizes or sectors also kUnsuitableEvt is now taken into
     account - not only kUnsuitableRun

   - callisto: '-raw' option is now the default
   
   - callisto: A dead pixel check has been implemented in MJPedestalC2.
     This has the advantage that mean values in the calibration (eg
     arrival times) are calculated more accuratly.

   - callisto: In the resource file callisto_Dec04Jan05.rc
     MJPedestalY2.ExtractWinRight has been reduced from 4.0 to 2.0

   - callisto: new Hi-/Lo-Gain intercalibration constants
     hilocalib_sp1.root and hilocalib_sp1_mc.root

   - callisto: changed default for MJPedestalY2.MaxEvents
     from 2000 to 5000 like in callisto_Dec04Jan05.txt

   - callisto: in MCalibrationChargeCalc the limit fgPheErrLowerLimit
     has changed from 9.0 to 6.0. This is necessary to exclude two very
     ringing pixels in September which - when kept - destroy the image
     cleaning around them (Markus G.)

   - callisto: (!) The Muon analysis of all data taken so far has
     suggested that different extractors give slightly different
     absolute results after calibration, due to there different
     handling of the calibration- and cosmics pulse shape. Although
     the over all calibration seems to be a bit to pesimistic (or 
     the MC seems to be a little bit - 3-5% - too much attenuated).
     There for correction factors have been introduced which correct
     the calibration constants by
         0.956937799 for Digital Filter (callisto.rc)
     and
         0.904159132 for Spline (callisto_Dec04Jan05.rc)
     calibration. The main advantage is, that both data types can
     be easily analysed together and compared to the same MC sample.
     The factors are choosen such, that the reference is now a 
     Monte Carlo sample calibrated with Digital Filter (callisto.rc)

   - star: fixed a bug which caused the first bin of the histograms
     vs time to disapear in PostProcess. This gave wrong results
     on the screen. The output and all following programs not
     directly accessing the histograms haven't been affected

   - star, ganymed: Made sure that always the first Drive-tree event
     is read and Processed before the first data event was read.

   - ganymed: implemented the possibility to replace the standard 
     false source plot with a different one. For example: If
     you want to use a Disp based FS plot use the resource:
       NameHistFS: MHDisp
     For details about the requirements of these plots look
     at the corresponding classes

   - MJOptimize: For On/Off and Wobble optimization MJOptimize
     has been replaced by MJOptimizeCuts. To use MHThetaSq instead
     of MHAlpha call MJOptimizeCuts("MHThetaSq"). Make sure in this
     case MAlphaFitter is setup correctly. For On-Only optimization
     MJOptimize has been Replaced by MJOptimizeCuts, too. Call
     RunOn() instead of Run(). For energy estimation MJOptimize
     has been replaced by MJOptimizeEnergy

   - sponde: The input MC spectrum can now be weighted to fake a
     different spectrum. This is done via MMcSpectrumWeight. For
     more details see the class description and sponde.rc

   - sponde: The paremeter comparsion plots are not scaled by
     their entries anymore. Instead the MC plot is scaled by using
     the result spectrum of the analysis. If the input MC spectrum
     and the result spectrum has different slopes the absolut 
     normalization is normally wrong.



 *** Version 0.9.2 (2005/05/20)

   - general: Changed the default paths for calibrated data and image
     files. (The implemented access to these files doesn't yet exist)

   - general: support for MRunIter has been removed from the job classes
     (use the setter functions of MSeqeunce instead)

   - general: ProcessFile has been renamed to Process in all job classes,
     because ProcessFile is missleading

   - general: The storage of the palette in MHCamera is now possible

   - general: new bits describing unsuitable and unreliable pixels have
     been introduced in MBadPixelsPix: kPreviouslyExcluded, 
     kDeadPedestalRms, kFluctuatingArrivalTimes

   - general: fixed reading of CC-REPORTS (weather data) for Arehucas file
     versions newer than V200404070 

   - general: support for cc file versions newer than V200504130. Added
     support for active loads (MCameraActiveLoad) and central pixel
     (MCameraCentralPixel)

   - general: added new histograms for image parameters versus size
     (MHVsSize)

   - general: The eventloop now automatically calls PrintStatistics

   - Status Display: Improved support for printing:
     + A default can now be set in .rootrc (for more details see 
       MStatusDisplay::PrintPS)
     + A print dialog is opened so that printer and command line
       can be canged

   - showlog: improved such that it handles different verbosity levels
     more accurate now. This mean you can suppress output when
     using showlog.

   - showplot: improved
     + The new graphics formats have been implemented 
       (pdf, svg, png, jpg and xpm)
     + Implemented an interface for printing. Because you can change
       the printing command from the command line you can use all
       the nice postscript tool. To print a booklet for example call: 
       showplot -b --print --print-cmd="cat %f" filename.root | lpr
       showplot -b --print --print-cmd="psbook %f | psnup -2 | lpr" filename.root
       showplot -b --null --print --print-cmd="psbook %f" filename.root > book.ps

   - callisto: fixed a bug in the recognition of the calibration bit
     pattern. Concerns only intensity calibration. (The strength of the 
     intensity has been updated in some cases)

   - callisto: add a filter against PIN Diode events when running over
     calibration runs since some of the recent calibration runs have
     been taken with Pin Diode.

   - callisto: added support for the runs 39942, 39944, 44834, 39941,
     39943, 44833 in the calibration (MCalibColorSet)

   - callisto: MJCalibration.MHCalibrationChargeCam.ProbLimit has
     been set to 1e-18 in callisto_Dec04_Jan05.rc

   - callisto: in callisto_Dec04Jan05.rc: set value for 
     MJPedestalY2.MaxEvents from 2000 to 5000 to get better 
     statistics for the calculation of the mean pulse time 
     of the cosmisc
       
   - callisto: the charge limit in the calibration has been raised 
     from 2.5 to 4.5

   - callisto: new setup read from callisto.rc
     + MCalibCalcFromPast.fNumEventsDump
     + MHCalibrationChargeCam.NumHiGainSaturationLimit
     + MHCalibrationChargeCam.NumLoGainSaturationLimit

   - callisto: MExtractTimeAndChargeSpline extractor:
     + The integration ranges were shifted by 0.2 FADC slices to the right. 
     + unreliable handling of floating point values caused different
       results on different platform or with different compiler options:
       The new algorithm is numerically exact and more accurate because
       it doesn't calculate the integration position by multiple times
       adding the step-size (which results in numerical uncertanties
       exspecially if multiplied with large numbers)
     + A lot of fixes have been introduced which effects integrating the
       spline at the edges of the valid range. In this case any memory
       was randomly accessed. This behaviour can be tested replacing
       all MArrayF by TArrayF which perform a range check (but of course
       they are a lot slower)
     ! No result obtained with the Spline before can be trusted! Due to
       random memory access it might by completely random!

   - callisto: set new defaults in MExtractTimeAndChargeDigitalFilter:
     + extraction range in the high gain to 4.
     + weights file to cosmics_weights46.dat

   - callisto:
     + introduce the check for fluctuating arrival times.
     + set charge limit to from 2.5 to 4.5
     + introduce asymmetric exclusion due to "deviating number of phes":
       Default: -9.0 +4.5
     + make deviating number of phes unsuitable (see David's findings).

   - callisto: fixed the display of unsuitable pixels. The old change
     to a real profile in MHCamera broke it.

   - callisto: The cosmics filter is used in MJCalibration now 
     independant of the pulser color (formaly it was used only for 
     CT1-pulser data). The threshold of number of max empty pixels 
     has been raised from 2% to 5% in MJCalibration. The maximum
     fraction of rejected event is 50% otherwise the calibration has
     failed. The filter now ignores unsuitable pixels.
     Setup the filter from callisto.rc using "ContCosmics"
     (see MFCosmics::ReadEnv)

   - callisto: changed default for fgOffsetFromLogain back
     from -2.8 to -1.8 in digital filter. 

   - callisto: for the common extractors 
      + MExtractFixedWindow
      * MExtractTimeAndChargeSlidingWindow
      * MExtractTimeAndChargeSpline
      * MExtractTimeAndChargeDigitalFilter
     the hi- and lo-gain reolution is stored in fResolutionPerPheHiGain
     and fResolutionPerPheLoGain

   - callisto: Updated the hi-/lo-gain intercalibration constants
     hilocalib_sp1.root, hilocalib_df46_mc.root, hilocalib_df4.root, 
     hilocalib_df6.root

   - callisto: The default lo-gain extraction window for the
     MExtractTimeAndChargeSlidingWindow extractor has been increased 
     from 6 to 8 to take the stretch of the pusle into account

   - callisto (!!!): fixed an important bug in the setting of the 
     incoming pedestal pointer to the signal extractor extracting
     pedestals. This bug caused the pedestals not to be subtracted 
     correctly in the calibration afterwards. The effect was about 10% 
     too high conversion factors for the digital filter and the spline
     and more than 40% for the sliding window. 
     The bug existed since end of February! (Markus G.)

   - star: added muon support to star. A new tab "MHMuonPar" is
     displayed. The lower right plot is an estimate of the point spread
     function. A reference value is displayed. To get the real PSF
     you must compare with MC. The setup for the muon analysis
     is done in star.rc. More informations can be found in
     MHSingleMuon and MHCalibParCalc

   - star: The weather data is now displayed in star

   - star: MEventRateCalc handles the calculation of the event rate more
     accurate now in case of the start of a new run inside a sequence

   - ganymed: a bug in MSrcPosCalc has been fixed (The changes were taken
     from a bugfix of Wolfgang to Loc0LocToCam; Wolfgang: the results are
     changed only minimally

   - ganymed: (MJCut and MJOptimize) now displayes the number of 
     excess events versus size. The energy estimation is done in
     MJSpectrum (sponde)

   - ganymed: got support for using other variables than Alpha, eg. Theta.
     Therefor you need a class deriving from MHAlpha which supports
     this variable (one is already existing: MHTheta, which is now the
     default in ganymed). It is setup through ganymed.rc

   - added a full featured spectrum program (sponde) which reads
     ganymed output and monte carlos and compiles a spectrum



 *** Version 0.9.1 (2005/04/14)

   - Fix bug on the arrival time reconstruction when using Digital
     Filter (in some marginal cases (10%) times were reconstructed shifted
     by 0.2 slices)

   - all executables now return 2 where in previous versions -1 has
     been returned (gave problems with shell-scripts)

   - the Hi-/Lo-Gain intercalibration for monte carlo files is now
     also done with a intercalibration input file

   - a bug in the time fine adjustment of the digital filter has been fixed

   - a lot of small fixes to calling TPad::Divide with a margin==0

   - MStatusDisplay now shows the memory usage

   - added a memory leak statistic after the eventloop

   - MHadronness replaced by MParameterD

   - MEnergyEst replaced by MParameterD

   - wobble mode can now be switched on from the data-set

   - switched on bad pixel detection for a pedestal level
     which is more than 5 times the variance

   - Fixed default paths to be correct for the new data center

   - Fixed handling of drive data. The pointing position in now
     copied to MPointingPos inside star for real and MC events

   - made sure, that the sequences in a dataset are sorted by
     their number to ensure the correct synchronization

   - added all resource files necessary to handle spline extraction
     of signal for the late 2004 and the early 2005 data

   - update the reference lines for the december data

   - changed default for fgHiGainFirst from 2 to 0 in spline extractor
     This is important for the Dec. and Jan. data where the calibration
     pulse position changes much.

   - changed default for fgOffsetFromLogain from -1.8 to -2.8 in digital
     filter. This corrects only an defective extraction of the MC at 
     low-gain signals lower than 90 phes.

   - fixed some slowdown in calibrating data with interleaved events from
     Jan and Dec. As a workaround made the pretty palette the default. 
     The inverse Deep Blue Sea has a bug which slows down long calibration
     runs a lot.

   - implemented new values from new starguider report as defined
     in TDAS00-07 V7.1 (eg SkyBrighness)

   - Fixed the formerly wrong unit of azimuth and zenith angle deviation
     in starguider report (If you merpp the same reports with this version
     you'll get the correct units (arcmin) but a different result
     than with old versions (wrong units))

   - Now the status display can also write PNG, JPEG and XPM (other file
     formats are still unsupported by root)

   - added a lot of scripts and macros in directory datacenter used
     for automatic processing



 *** Version 0.9 (2005/03/23)
   
   - Calibration is now done autmatically into photo-electrons like 
     decided by the Collaboration Board
 
   - Reference lines in the display can be read from resource file

   - Interlaced calibration events can be handled by callisto 

   - Hi-Gain vs. Lo-Gain calibration constants can be calculated
     and applied in callisto.

   - class MJCalibTest can be called by callisto

   - callisto can now run with three data types: rawdata, rootdata, MC

   - added two classes mreport/MReportStarguider.cc and 
     mreport/MReportStarguider.h for implementing starguider-data
     
   - the output of merpp-update now contains starguider information:
     MReportStarguider and MTimeStarguider
   
   - Fixed bug in MHHadronness.cc which made that the displayed Q-factor 
     values were lower than they should. This was more visible for high
     energy (Size) events.

   - Made MExtractTimeAndChargeDigitalFilter the default for callisto

   - added a new loop to callisto to support the correct extraction of 
     the pedestal and pedestal rms through the signal extractor

   - added class to perform fourier tranformation of a hexagonal structure
     (MHexagonalFT)

   - added class to perform fourier tranformation of a hexagonal structure
     in a tasklist (MHexagonalFTCalc) example: hft.C

   - added class to enhance the MCerPhotEvt from the Magic camera geometry
     MGeomCamMagic to a enhanced geometry MGeomCamMagicXT having only
     small pixels, example: extendcam.C

   - added possibility to write data to memory (TTree) using MWriteRootFile

   - added possibility to read a TTree stored only in memory by MReadTree

   - new classes to deal with the interleaved calibration events:
     MTriggerPattern, MTriggerPatternDecode and MFTriggerPattern

   - implemented filters to callisto to skip calibration events

   - changed callisto to calculate three kinds of pedestals and pedestal rms:
     + fundamental pedestal (MPedPhotFundamental): extracted directly from
       the slices by statistics                              
     + pedestal from extractor (MPedPhotFromExtractor): extracted with the
       signal extractor without randomization
     + random pedestal from extractor (MPedPhotFromExtractorRndm): extracted
       with the signal extractor with randomization (extraction window is 
       randomly fixed)
     The three kind of pedestal is not yet implemented (for speed reasons) as
     long as nobody really needs it:
     + pedestal from extractor with randomizatiuon (MPedPhotFromExtractorRndm): 
       extracted with the signal extractor with randomization

    - callisto now finds the pulse position in the data automatically

    - implemented data reduction in calibrated data by reducing the precision
      of the stored floating point values (pedestal, rms and signal) to
      roughly 0.5%

    - fixed showplot to make file format conversion (eg. converting a
      status display stored in a root file into a postscript file) in
      batch mode possible:
      showplot -b --save-as-ps[=outfile.ps] infile.root

    - Allow class names as part of filter rules, eg
      (MFSupercuts) && (MHillas.fSize>100)

    - fixed a lot of stuff to allow all programs to process MonteCarlo
      camera files.

    - enhanced possibility to debug reading of resource file

    - MDataChain should handle priorities of operator more 
      accurate now

    - completely revised the container in which the signal is stored
      MCerPhotEvt/Pix, MArrivalTime --> MSignalCam/Pix
       * the new implementation is faster and
       * needs less storage space

    - FIRST version which can process Monte Carlos through the 
      whole chain of programs (callisto, star, etc)

    - implemented new image parameters for the border length of
      the shower image (MNewImagePar2)

    - implemented wobble-mode in ganymed

    - new class (MJOptimize) taking care of all optimization processes



 *** Version 0.8.6 (2004/10/05)

   - New combined signal and time extractor 
     MExtractTimeAndChargeDigitalFilter

   - The Report for the trigger section in the CC run summary has been
     updated, and now the IPRs (Individual Pixel Rates) are available.

   - new macros for database interaction: filldotrun.C, filldotrbk.C

   - added new program which can convert colored output which was redirected
     into a file back into colored text (for example to display it with
     more): showlog

   - added new program to display an MStatusArray of plots (stored with
     MStatusDisplay) from a file: showplot

   - Added support for Sequences (MSequence). Sequences are blocks of run
     files containing the same calibration and belonging to the same 
     source

   - Added a new task (MTaskEnv) which can be used in setup files to switch
     between different tasks (for example to choose a signal extractor)

   - Implemented automatic file splitting in MWriteRootFile

   - After image cleaning an island index is assigned to all used pixels.
     The index corresponds to the order of the islands in size.

   - added a cleaning which takes signal height _and_ arrival time into
     account: probability cleaning (for more details see MImgCleanStd)

   - added a cleaning which takes takes the absolute number of photons
     times the area ratio as cleaning level 
     (for more details see MImgCleanStd)

   - Changed calculation of image parameters such, that it can be done
     for all used pixels or the pixels with a defined island index only

   - implemented new image parameters displaying the number of islands,
     saturated hi-gain and lo-gain pixels (MImagePar, MHImagePar)

   - event display in executable changed to support also calibrated files
     (done with MJCalibrateSignal)

   - added program doing the calibration and the signal extraction
     and signal calibration: callisto Most part of the setup is done
     by a resource file: callisto.rc

   - implemented support for pixels with negative IDs (Raw file format
     version 4)

   - implemented an new option in merpp (--auto-time) to get the time
     range for cc and/or caco data from the run header of the raw data

   - New star-program (star and MJStar). Taking calibrated data from
     callisto/MJCalibrateSignal as input. The setup is done through
     a resource file: star.rc

   - Added possibility to setup the eventloop in the mars-executable
     Camera Display by a resource file: mars.rc

   - The the mars-executable Camera Display can now also read
     calibrated data files written by callisto/MJCalibrateSignal

   - Added a new class which can calculate the source position in
     the camera from a pointing model known to be correct and
     the pointing model used for observations (MSrcPosFromModel)

   - added a new version of the class calculating the effective on-time
     MHEffectiveOnTime. It unifies the old classes MHEffOnTime, 
     MHEffOnTimeTime, MHEffOnTimeTheta, MHTimeDiff, MHTimeDiffTime and
     MHTimeDiffTheta.

   - added the possibility to count how often a signal is above (or 
     below) a certain threshold in MHCamEvent and MHCamEventRot (this
     can replace MHTriggerLvl0)

   - added ore convinient ways to set the axis-titles of your histograms
     in your macro to some MH-classes like: MH3, MHVsTime, MHSectorVsTime,
     MHPixVsTime, MHVsTime, MHCamEvent and MHCamEventRot



 *** Version 0.8.5 (2004/08/06)
 
   - new Signal Extractors: MExtractFixedWindowSpline 
     and MExtractAmplitudeSpline

   - implementes interleave in merpp to be able to merpp parts of a file

   - implemented force-functionality to merpp

   - star display to display star catalogs (MAstroCatalog)

   - star display to 'mirror' stars to the camera (MAstroCamera)

   - implemented a new cleaning option which allows cleaning scaled with
     the real ratio of the rms depending on the pixel size, not only the
     square root of the pixelsize (MImgCleanStd, MCameraData)

   - removed obsolete MBlind* classes

   - Signal extractor taking the AB flag into account implemented (this makes
     an odd number of slices for signal extraction possible
     (MExtractSignalABcorr)

   - implemented a class extracting the pedestal from the raw-data (events)
     itself using empty lo-gain (MpedCalcFromLoGain)

   - Added mean values to MPedPhotCam

   - Added Makefile for Darwin (this OS is NOT officially supported!)



 *** Version 0.8.4 (2004/04/19)

   - new fast arrival time extractor using cubic splines: 
     MExtractTimeFastSpline

   - implementes multi-argument support in MDataChain
     ("pow(a, b)")

   - implemented double reference support in MDataChain
     ("MCameraLV.fPowerSupplyA.fVoltagePos5V")

   - added support in MF for expressiond like 
     "MHillas.fWidth<2*<MHillas.fLength"

   - MDataChain is now able to support variables like [0], [1], ...
     which can be used in fit functions as parameters. The interface
     is implemented through the new virtual function 
     MParContainer::SetVariables

   - added new class MArrivalTimeCam/MArrivalTimePix:
     Storage containers to hold the ArrivalTime of the camera/pixel

   - added new classes for the calibration of relative arrival times:
     MCalibrationRelTimePix and MCalibrationRelTimeCam
  
   - added new classes to extract the signal of the blind pixel and the 
     PIN Diode separately: MExtractBlindPixel and MExtractPINDiode

   - added new classes for the bad-pixels treatment (MBadPixels*)
     which are more powerfull than the old ones (MBlindPixel*)
     and will replace them.

   - added filter against cosmics: MFCosmics

   - added new class MArrivalTimeCalc2:
     Calculates the arrival time as the signal weighted mean of the time 
     slices in the maximum window

   - added a new macro call pointspreadfucntion.C that fits with a 2D 
     gaussian the DC spot for an star. It gives you the RMS of
     the PSF and the center of the star, very useful for misspointing 
     studies.

   - added support for multi-telescope camera files. Added classes 
     MStereoCalc and MStereoPar, classes to calculate and store shower
     parameters calculated from the data of two telescopes, for the MC
     studies regarding the MAGIC clone.

   - added MMcCalibrationCalc, MMcCalibrationUpdate, classes to obtain
     MC data calibrated in photons instead of ADC counts.

   - added three macros to show the calibration and analysis chain of 
     MC data. The macro starmc.C does both calibration and analysis, producing
     an output containing Hillas parameters for each event. The macro 
     mccalibrate.C does just the calibration and writes out a file containing
     pixel information in photons instead of FADC data. The macro starmc2.C
     reads in such a file and produces an output with Hillas parameters.

   - added MPedPhotCalc to compute pedestals in units of photons, see
     macro pedphotcalc.C

   - added all class to perform the calibration, see macro calibration.C

   - added class MFFT to perform Fast Fourier Transforms

   - added class MSimulatedAnnealing to perform simulated annealing 
     minimizations

   - added new macro bootcampstandardanalysis.C which holds the skeleton
     of the analysis chain discussed in the december bootcamp. 

   - added support for DAQ raw data file format V2

   - implemented an arbitrary time stamp for the events from raw-data
     files with file version<3. It is calculated from the run number, 
     the date of start and the FADC clock tick from the first module

   - implemented the correct decoding of the event time-stamp
     coming from the DAQ for raw-data files with file-version>2

   - added a new classes storing the pedestals in units of photons: 
     MPedPhotPix, MPedPhotCam

   - added support for DC current files from the camera control

   - added support for start- and stop-time when merpping report files

   - added MReportCC storing the weather data from the CC-REPORT

   - added class to fill trainings and test matrices (MTFillMatrix)

   - added a filter performing a software trigger (MFSoftwareTrigger)

   - Added classes to perform parts of the standard analysis (Jobs):
     + MJPedestals
     + MJCalibration
     + MJExtractSignal
     Note: All of them are preliminary!

   - Added Iterator to iterate over files giving only the run-number

   - Added class (MHSectorVsTime) and example (sectorvstime.C) to plot
     the evolution of a sector vs time



 *** Version 0.8.3 (2003/12/05)

   - adapted Random Forest classes to allow the use with MC
     multi-telescope files which can be produced with the coming
     camera version 0.7.

   - added classes to read and interprete central control report files
     (Directory: mreport, Base-Class: MReport, Reader: MReportFileRead, ...)

   - enhanced merpp (the executable) to support .raw and .rep files.

   - added classes for light calibration (MCalibration* and 
     MHCalibration*) by Markus Gaug. Further work has to be and will
     be done soon.

   - Added a preliminary 'synchronizer' (MReadReports) which will 
     replace MReadMarsFile as soon as it is finished.

   - Added a tutorial macro demonstrating how report files are merpped
     and how they can be read from the root file in the correct order
     of time (readrep.C)

   - Added a generalized histogram (similar to MH3) called MVsTime
     which allows to plot variables (or rules - MDataChain - of
     variables) vs time. For an example have a look at tempvstime.C.

   - Changed the name of the event trees:
       + Events    --> Events
       + PedEvents --> Pedestals
       + CalEvents --> Calibration

  - changed name of MTime in the event trees from MRawEvtTime to MTime

  - added displaying the sector indices to the mars event display

  - Magic Online Analysis (MOnA) implemented in a first version
    (see mona.cc, MOnlineDump and MOnlineDisplay)

  - added classes to calculate event rate
    (preliminary: MEventRateCalc, MEventRate)

  - added new executable 'readdaq' to read *.raw files from the DAQ directly

  - added classes dealing with the telescope pointing position (MPointing*)

  - implemented the 'final' Mars time stamp based on some experiences with
    the new 'Synchonizer'



 *** Version 0.8.2 (2003/11/10)

   - accelerated standard analysis by a factor of 6-8 ...
     + doing precalculations in MGeomCam for the pixel area ratio
     + creating a look up table for the single pixels in MCerPhotEvt

   - added task to calculates weights in case you want to 'change'
     the energy spectrum of a monte carlo file 
     (MMcWeightEnergySpecCalc - Great work from Marcos Lopez Moya)

   - added a new executable star which implements the
     star.C macro as precompiled executable. Currently a flexible
     setup is not possible, but a 'input card' setup will follow.

   - Starting to implement Multi-Telescope-Support based on the 
     upcomming camera version 0.7. The base for a general support
     is implemented and the star-functionality is proven to work.

   - Replaced Event-Display in and Camera-Display in the Mars GUI
     executable by a more convinient display base on MStatusDisplay.

   - fixed the algorithm for the blind pixel treatment (MBlindPixelCalc)



 *** Version 0.8.1 (2003/10/20)
   
   - added signal subtraction for pure on data by means of fitting
     the background in the off region or by performing a combined
     signal/background fit. Provides necessary histograms for
     obtaining energy spectra and a light curve.

   - added classes to perform and study the selection of the 
     2nd Level Trigger on MC data (example in triglvl2.C macro)

   - added pedestal calculation

   - implemented "democratic cleaning" (tail cut value proportional 
     to pixel area) according to Wolfgang's suggestion.

   - added Sigmabar treatment (calculation and basic histogramming)

   - added basic routines to apply padding

   - added first implementation of MARS-based source visibility plotter
     (needs to be linked against slalib, preliminary version)

   - enhanced functionality of data chains: random numbers, conversion
     functions and access to MHMatrix objects
   
   - fixed some bugs in the CT1 Preproc file reader, which gave wrong
     numbers under certain circumstances (runs with 0 events included)

   - fixed a bug in the general filter MF (didn't work on some systems
     at all)
   
   - fixed a bug in chaining files, which caused the analysis to crash or
     ignore files

   - implemented more geometry classes describing the parameters used
     to simulate the PMT camera

   - added new Monte Carlo classes containing the setup of the different
     programs in the MC chain.

   - added a (preliminary) general task to evaluate chi-square of a given
     parameter in an eventloop.

   - first implementation of a container describing the observatory
     location (Long, Lat, etc)
   
   - added Random Forest - method for g/h-separation

   - made compatible with the latest PRO version of root (3.04/02 and 3.05/07)
     (this means, that it is compiling, but not yet fully tested)

   - added a new status display which can show the present status
     of histograms while an eventloop is running (MStatusDisplay,
     an example can be found at status.C)

   - reorganized directories:
     classes describing the image              --> mimage
     classes containing Monte Carlo Histograms --> mhistmc
     classes for Random Forest method          --> mranforest

   - added (event-)weights to fill histograms

   - colors in logging output to screen

   - changes to the Image Parameter declaration:
       + moved fConc, fConc1 from MHillasSrc to MNewImagePar
       + moved fNumUsedPix, fNumCorePix from MHillas to MNewImagePar
       + MHillasExt now derives from MParContainer instead of MHillas
     --> Files written with the old definition are more or less unreadable

   - reading of reflector files

   - display data from reflector files in the camera

   - implemented currents (and display them in the camera)

   - logarithmic scale in camera

   - added class to iterate through directory contents (MDirIter)



 *** Version 0.8 (2002/11/18)

   - added all subdirectories as Include-Paths for ACLIC (s. rootlogon.C)

   - corrected 48(!) wrong entries in the Next Neighbor table of the
     Magic Camera (all in the inner part of the outer ring, mostly 
     missing a single NN)

   - Fixed a bug in the table describing the Next Neighbours of the
     CT1 camera (a pixel 127 doesn't exist and the pixel 126 didn't
     have any neighbour)

   - changed trigrate.C: now it calculates the pure NSB triggers from
     a camera file (generated with no showers), so there is no need
     to introduce them by hand in the BgR array (for more details
     please ask Abelardo)

   - added MHHadronness (and MHadronness) as a general facility to evaluate
     the quality of a gamma-hadron seperation method

   - changed MWriteAsciiFile to accept rules (MDataChain)

   - Calls to MWriteAsciiFile::AddContainer and MWriteAsciiFile::AddRule
     must be replaced by calles to MWriteAsciiFile::AddColumn and
     MWriteAsciiFile::AddColumns. Be carefull, the arguments have slightly
     changed.

   - Implemented generalized event-matrices (one row per event)
     (MHMatrix)

   - implemented the one dimensional composite probabilities
     (MHCompProb, MCompProbCalc, comprob.C)

   - implemented the Multidimensional Distances (Next Neighbours, Kernel)
     (MMultiDimDistCalc, multidimdist.C, multidimdist2.C, MHMatrix)

   - Added a camera display displaying the pixel numbers to the
     event display gui.

   - Added Monte Carlo Informations to event display gui

   - Changed the camera display to display the pixel numbering
   
   - Added three buttons to the camera display to change the palette

   - Added the number of used and core pixels after image cleaning
     to MHillas

   - Changed the algorithm used in MHillas according to TDAS 02-03

   - Changed the Alpha range according to TDAS 02-03 to [-90°,90°]

   - Changed the algorithm used in MHillasSrc according to TDAS 02-03

   - Added fCosDeltaAlpha to MHillasSrc

   - The numbers of photons used to calculate fConc and fConc1 in
     MHillasExt are now scaled with the pixel size, so that one get
     a four times smaller value for the bigger pixels in the outer ring.

   - added new task to smooth the camera contents (MCameraSmooth)

   - added possibility to use interpolated pixel values for blind pixels
     instead of removing it completely from the analysis

   - Implemented the possibility to save the whole eventloop with
     its setup to a file. This information can be read and from
     it you can (re)create a Macro. See MEvtLoop::Read, MEvtLoop::Write
     and MEvtLoop::MakeMacro. If you find something not working, please
     report - this stuff is still in a beta phase.

   - MBlindPixelCalc set all Ceta-Tauri pixels to unused not only for the
     file having crab starfield, but also for all following files - fixed

   - MTaskList::PrintStatistics can now be instructud to print also the
     title, too: use PrintStatistics(0, kTRUE)

   - Changed the image cleaning so that pixels with to many 'used'
     neighbors are left used (to get rid of 'holes' in events)

   - Introduced more output to the camera displays

   - Added an array of histograms (eg one hist per run), MHArray

   - Added a _preliminary_ version of the reader for CT1 PreProc files
     (MCT1ReadPreProc)

   - Fixed the error calculation of MHMCollectionArea and MHMcIntRate
     (thanks to Raquel)

   - added the first energy estimator (MEnergyEstParam, estimate.C,
     estfit.C) using a parametrization

   - Added some new Monte Carlo paremeter containers used to store
     setup information from the MC chain (MMcConfigRunHeader, 
     MMcCorsikaRunHeader, MGeomPMT, MGeomMirror)



 *** Version 0.7 (2002/04/30)
 
   - added a bugfix to MCerPhotCalc. In older camera versions (<=40)
     the pedestal mean value was saved incorrectly. For files from
     this versions we substract 0.5 from the pedestal mean.
     WARNING: This may effect your results, so don't wonder...

   - First implementation of an algorithm using only triggered events
     for the collection area calculation

   - Ascii Output can now also be used for parameter containers which
     doesn't overload MParCointainer::AsciiWrite

   - The Ascii Output is now also capable of writing single data members
     of one container

   - You are now able to change the order of the values written to the
     ascii file

   - You can now specify a conversion factor for each data member written
     to an ascii file. This may be usefull to change the units of the 
     data member (eg. degrees instead of millimeters in case of the 
     hillas parameters)

   - Replaced old MHillas by a new structure which allows you to extend
     the parameters stored in MHillas very easily:
     see MHillas and MHillasExt

   - Added classes to handle source dependancy of image parameters:
     see MHillasSrc

   - Added container (MBinning) to have a standard input for the binning
     in different histograms (eg. the Energy bins should be the same in
     all histograms)

   - Changed Hillas histograms from mm to deg

   - Added the flexible binning to the hillas histograms

   - Added a filter for the alpha parameter (MFAlpha)

   - It is now possible to write single data members of a class object to
     an output stream instead of the whole container only

   - Added a generalized filter for a data member:
     MFDataMember filter("MHillas.fWidth", '<', 0.5);

   - Added a generalized Filter parser for more difficult filter rules
     MF filter("MHillas.fWidth<0.5 && MHillas.fLength>0.5");

   - Added new Monte Carlo histograms:
     MHMcEnergyImpact and MHMcEfficiency

   - Added many new histograms needed for the flux calculation.

   - Added a generalized histograms which can fill and show up to three
     data members in either a one-, two- or three-dimensional histogram, eg:
     MH3 mh3("MHillas.fWidth", "MHillas.fLength");

   - Added:
       * calculation of <Theta> as a function of Theta (MHThetabarTheta)
       * calculation of <Theta> as a function of time (MHThetabarTime)
     they are needed in the flux calculation in 
     order to select for a given bin in Theta (or time) the appropriate 
     effective collection area, which is assumed to be given as a function 
     of Theta and the true energy

   - Added calculation of the migration matrix in the energy 
     (MHMcEnergyMigration):
     the migration matrix describes the migration from the true (E_true)
     to the estimated energy (E_est); it is needed to determine the
     distribution of E_true from the measured distribution of E_est 
     by unfolding

   - changed the color palette in the camera display to DeapSea.



 *** Version 0.6 (2002/01/15)

   - Introduce the option of disable pixels (MBlindPixels and
     MBlindPixelCalc) for the analysis (eg MHillasCalc) to be 
     able to deal with Star Field NSB.

   - Fixed a bug in the pedestals substraction

   - Introduced verbosity levels

   - Introduced a task (MReadMarsFile) which reads the RunHeader tree, too.

   - Fixed several bugs in MFTriggerLvl1 which caused to filter not to
     work correctly (thanks to Abelardo)

   - Introduced correct treatment of the diffuse Night Sky Background 
     and NSB (MMcPedestalNSB)

   - Corrected treatment of the electronic noise (MMcPedestalCopy)

   - Introduced MMcRunHeaders which contains monte carlo run informations

   - Added a 'Print'-Task (MPrint) which calls the Print function
     of a parameter container for each event

   - Removed MHtml (was replaced by THtml of root 3.02) in dohtml.C

   - changes to be more efficient with root 3.02/06

   - root 3.02/06 fixes:
     + the memory leak when reading MRawEvtData
     + loosing more time than necessary to update the progress bar

   - MWriteAsciiFile is now capable of writing more than one container
     in a line. For example: To be able to write Monte Carlo and 
     Hillas values (energy and alpha) in one line.



 *** Version 0.5 (2001/11/06)

   - Added new GUI functionality (Analysis)

   - Added a camera display to be able to display events in the camera
     together with the calculated hillas ellipse

   - Added a 'Clone'-Task (MClone) which can clone a parameter container
     object at any state of the analysis, so that the output are several
     states

   - Fixed a bug in the collection area error calculation 
     (Thanks to Ciro and Abelardo)

   - Fixed a bug which causes merpp to crash in some environments

   - Implemented auto enabling scheme for Branches to speed up reading
     a lot (see MReadTree and MTask)

   - Fixed a bug in the Histogram classes which caused histograms to
     be written to a file by MWriteRootFile without request

   - changed the object I/O from the old stylish root I/O to the new
     'auto evolution scheme' This was done only for classes which are not
     yet used in Monte Carlo files, so that these files are still readable.

   - Added support for a progress bar in MEvtLoop and MReadTree 
     (see the Analysis Window in Mars for an example)

   - Fixed a bug in the package creation which disturbed the conversion
     between a mars version unpacked from a tar file and the repository.

   - Changed the code to writer merpped (converted into root format)
     raw files, to get faster (around 10%)



 *** Version 0.4 (2001/09/20)

    - This release is made to compile on a root version >= 3.00

    - Fixed a bug which causes MReadTree to read the first file only

    - Added a task to compute the trigger rate (MMcTriggerRateCalc)
 
    - Added a task to write a container to an Ascii file (MWriteAsciiFile)

    - Added a task to write several container to a root file (MWriteRootFile)
    
    - Added calculation of the Enegry Threshold (MMcThresholdCalc)
    
    - Added calculation of the collection area (MMcCollectionAreaCalc)
    
    - fixed some bugs in the Hillas calculation
    
    - added filters to be able to control the task execution dependent on
      a parameter (for example: the number of level 1 triggers in a MC-file)



 *** Version 0.3 (2001/04/23)
        
    - Many demo macros added    
  
    - MCT1ReadAscii is now able to handle more than one file (s. ::AddFile)
    
    - Hillas Calculation added
    
    - Standard image cleaning available
    
    - MReadTree is now able to handle more than one file
      (Remark: of the same structure)
    
    - Now we are able to display events in a first version of the
      Camera event display. To do this we must run the macro 
      "macros/readCT1.C" to read data from CT1. 



 *** Version 0.2 (2001/01/24)
 
    - First Implementation of a Gui to read the raw data. Start the
      gui with the command mars. 



 *** Version 0.1 (2000/12/20)
 
    - First official release 0.1
    
    - Changes:
    
       + changed calling style of MRawPixelIterator from do...while to 
         while-loop