New MINOR release 01.04.00

Changes since 01.03.07:
* improvement: apply calib should not discard visible scan data

   like that quality of applied calibration function becomes transparent to operators

* fix: debug messages for data loss

* feat: timing drift feedback

New PATCH release 01.03.07

Changes since 01.03.06:
* fix: array sizes in BunchTiming
* feat: x2timer sync with bunchesStartTimeIndex
  in order to conveniently select odd bunches

New PATCH release 01.03.06

Changes since 01.03.05:
* fix: R^2 calculation

New PATCH release 01.03.05

Changes since 01.03.04:
* feat in EodScan: user given range cut
  for time axis fit
  This is needed because currently server cannot decide whether bunch
  peaks are from actual bunches or from wake fields

New PATCH release 01.03.04

Changes since 01.03.03:
fixup

New PATCH release 01.03.03

Changes since 01.03.02:
x2timer settings: avoid adding up of rounding errors

New PATCH release 01.03.02

Changes since 01.03.01:
* add less important bunch metrics in time domain
* improve: EodScan should disard invalid data
* improve BunchDetection output
  set 'bunchPeakDetected' true only if we actually detected peak of non-zero width

New PATCH release 01.03.01

Changes since 01.03.00:
* fix: avoid crash due to empty or constant time axis segment

New MINOR release 01.03.00

Changes since 01.02.01:
EOD scan
* improve: let EodScan explicity check that image present
* feat: auto fine scan
  allows automatic redution of step size if EO signal was detected
* fix: exclude forbidden x2timer delay range

Bunch detection
* improve: if EO signal not detected, do not output bunch metrics
* improvement: evaluate R^2 of fit
  sometimes fit does not match at all, e.g. when we try to fit bunch
  shape at edge and need to make assumption about unchanged bunch width.
  Bad fit model can be detected by R^2 < 0. Make use of this check and do
  not output fit line then, in order not to confuse.
* fix: skewSimple metric was covered by numerical noise
* feat: allow setting expected peak sign

Bunch timing
* fix: flash destination filtering
* fix: allow correlation with 'odd' timing samples
  startTimeIndex user input should allow capturing odd-time bunches,
  if x2timer triggers configured for that

New PATCH release 01.02.01

Changes since 01.02.00:
* calculate tFirstLineUs automatically
     - before tFirstLineUs was setpoint
     - now, we assume BUNCHES.start + offset is set to point on first electron bunch
     - usually, offset=0, but can set by user

* EodScan: optimize setLaserTime
     first update x2timer setpoints first,
     then laser_lock setpoints,
     then wait at least once, on next macro pulse, until settings accepted

     this scheme is more predictable and faster than previous one.

* fix: noise level in BunchDetection

New MINOR release 01.02.00
* new feature: EodScan

New PATCH release 01.01.04
fixup

New PATCH release 01.01.03
*  do not mask bad pixels in raw data
   was decided in order to let operator see reason/correctness
* more consistent handling of bad pixel values
  should improve statistics outputs which currently suffer from NaN values

New PATCH release 01.01.02
* feat: allow disabling of bunch signal processing via config
* minor improvement for bad values
  avoid NaN; instead use 0 in running statistics.
  Contrary to NaN, it does not destroy the series but it's a compromise as well.
* enable/disable for advanced EO features
     currently this is
      - EOD timing scan and readback of timing settings
      - communication with laser lock and motor sps for EO system supervision

     This should be disabled if server should only read out Kalypso images.
* fix : moving average reset button
* feat: moving average for selected line
* feat: fit also bunch profiles when bunch is at line edge
   assuming bunch width is same as previous fit.
* feat: turn off FFT if fBandStopWidth=0
* fix: use clamp to correct frequency range to allowed
* redefine debugLevel -> logLevel

New PATCH release 01.01.01

Changes since 01.01.00:
* support FLASH config

New MINOR release 01.01.00

Changes since 01.00.00:
- feature: BunchTiming module for correlation of signal lines with expected bunch times
- refactoring/optimization: create new module BunchDetection from ImgProcessingMod
  this should help when FFT filtering causes high CPU load in ImgProcessingMod
- feature: valid pixel value range
  for auto-detection of bad pixels and saturation effects

New MAJOR release 01.00.00

Changes since 00.09.00:
*     add pixLower, pixUpper inputs

*     make bunch peak detection more robust

*     discard peaks which intersect time axis ends

*     improve data validity handling
      - exclude division-by-zeros values
      - if not enough values remain, anything is invalid
      - otherwise, signal classification metrics (e.g. bunchDetected) are set
        valid, even if bunch was not detected
      - only if bunch signal valid and bunch detected, final outputs
        like bunch position, are marked valid

*     output signal classification metrics even if bunch was not detected
*     add signal sanity check
      also check that laser power is significant on bunch signal lines before
      evaluation of bunch profile metrics.
      Before, we only check laser power on (unmodulated) laser lines.

*     Allow subset selection all|odd|even for laser lines

*     data initialization also in case that input parameters are invalid,
      in order to avoid random utputs

New MAJOR release 00.09.00

first release, for testing only