Beam Test Offline Code

1
Beam Test Offline Code

• Jianchun Wang
• Syracuse University
• 01/17/02

• Outline
• Status of the code
• Function modules
• Discussion topics

2
Status

• All main functions are done, but need more
  testing and polishing.
• The code was tested with Syracuse computers
  (Alpha with Linux and OSF1) during development.
• The code was submitted to CVS this week, we need
  to decide whether keep it as a new release or a
  new project (Rob and Lynn will tell us about
  CVS).
• The code was tested on fnsimu1 and fsgi03.
• GNUmakefile.standalone are used. We need to work
  out GNUmakefile.
• We need to solve some issues and decide on what
  else functions should be included.

3
General Description
Data Analysis Read in real data (include MC
track), convert ADC counts into charge, form
cluster and convert into coordinates, form rough
track and perform Kalman filter, write out ntuple
and track summary file for further study MC
Simulation Generate tracks, calculate
intersection points on each plane, simulate
charge collection and convert it into ADC counts,
write out in similar format as real data plus the
track information Auto Alignment Read in track
summary file (local coordinates and resolutions),
go through MINUIT with geometry conversion, for
different geometry calculate chi2 for each track,
minimize the sum of chi2 and write out new
geometry
4
Tasks

• Tasks are selected by tb_control and information
  inside input data file
• Names of output data file and histogram file can
  be defined or with default

5
Function Modules

• Data encode and decode
• Geometry manipulation
• ADC charge conversion
• Cluster formation and conversion into coordinates
• Track formation
• Track fitting with Kalman filter
• MC Simulation
• MINUIT alignment ( a.k.a. Auto alignment )
• CWN ntuple
• Parameter control through name list read

6
Data File Format

• A Header
• RAWD, SIMU, TRAK
• Length of the block
• Version
• Run ID
• Number of planes
• Time Tag
• other comments

• C MC Track
• Length of the block
• Number of tracks
• Local X for plane 1
• Local Y for plane 1
• Length of the block

• B Data Hits
• Length of data block
• Cell 1 data
• Length of data block

• D Track Summary
• Length of the block
• Plane ID
• Local X for plane 1
• Local Y for plane 1
• Error on X for plane 1
• Error on Y for plane 1
• Length of the block

Data format for one cell 1111 pppp 000c cccc rrrr
rrrr aaaa aaaa

• Raw data from DAQ (RAWD) ABBBBBB
• MC data generated (SIMU) ABCBCBC
• Selected track summary (TRAK) ADDDDDD

7
Sample of Data File
8
Endian on Different Platforms

• Within a given 16- or 32-bit word, bytes at lower
  addresses have higher significance (big_endian)
• Most processor use big_endian, some use
  little_endian, very very few use middle_endian
• The machine we used for DAQ, and most of machines
  that we are going to use for analysis are
  big_endian
• In order that the data can be used for both
  big_endian and little_endian machine, I add byte
  swap in data reading and MC data writing
• It was tested standalone, not with real code yet.
  And I need to work on GNUmakefile
• ( Most functions associated with coding is in
  testbeam_dio.c )

9
Geometry Manipulation

• The geometry manipulation has very little change.
• Initial geometry information is read in from
  geometry file (tb_geometry.dat).
• Parameters are derived for easier conversion
  between world coordinate (WC), and local
  coordinate (LC).
• In MINUIT fit, fit parameters are connected with
  selected geometry parameters. Each time the fit
  parameters changed, the derived parameters
  recalculated.

10
ADC Charge Conversion

• Testbeam 2000 use curve for FPIX0, and one set of
  4 thresholds for all FPIX1 cells.
• Testbeam 2002 use calibration for each cell, and
  directly use threshold value instead of curves.
• Each plane has its own sixy_adc_file name.
• The ADC information is also used in the MC
  simulation where charge in pixel cell is
  converted to ADC value.

11
Pixel Cluster

• Testbeam 2000 uses Pennys cluster formation
  initially wrote to work with simulation package
  we developed at Syracuse. The eta calculation
  using output from her code in 2-column case is
  wrong.
• Testbeam 2002 imported the one I use for
  standalone simulation.
• Cluster center of gravity is calculated and
  converted into local coordinate where the
  orientation of row and column is considered and
  can be easily controlled.
• World coordination is calculated here so track
  formation can use more precise position. In
  testbeam 2000, we use local coordinate shift with
  the center for track formation.

12
Track Formation

• Track formation is kept in same style.
• Use hits in two x seed planes to make a path, for
  other planes choose the closest hit (within a
  cut).
• Y information currently is not used.

13
Track Fitting

• Rob had updated Kalman filter code to work with
  current version MCFast.
• It can treat a pixel hit as two strip hits, and
  only fit on hits with good resolution.
• It performs fit and derives residual with or with
  out the hit, and also provide the chi2 of the
  fit.
• In MINUIT alignment, strip-equivalent is
  enforced.
• The interface subroutine is kept in same style,
  modification are done mainly associated with
  local coordinate.

14
MC Simulation

• One or more tracks are generated (with no
  multiple scattering at this moment).
• Intersection with each plane is calculated.
• Charge generation, noise simulation, ADC
  conversion were imported from standalone
  simulation with modification.
• Hits are encoded and written out.
• The electric field, charge mobility, diffusion,
  effects of magnetic field are calculated at
  initialization.

15
MINUIT Alignment

• This is to determine the real geometry
  configuration by minimize the sum of chi2 from
  many track fit.
• CPU is a big issue here since we need to perform
  Kalman filter many many times.
• It deals with summary of pre-selected tracks.
• It provides interactive mode although it can also
  run in batch mode.
• Due to CPU limitation, only small portion of all
  geometry parameters can be FREE at a time. This
  is controlled in mn_skip.dat file, with very
  simple format.
• The best geometry is written in mn_geometry.dat
  with same format as tb_geometry.dat.

16
Histogram ID

• 11100 Geometry parameters
• 11200 Geometry parameters after
  alignment
• 11300 Global parameters and Detector
  parameters
• 11400 Simulation parameter maps
  
• 11500 Charge simulation parameters
• 21000 Simulated track and hits
• 31000 Track summary
• 32000 MINUIT fit
• 41000 Hit information
• 42000 Cluster information
• 43000 Difference of reconstructed hit
  position and generated position
• 51000 Form track
• 61000 Kalman filter

17
Column Wised Ntuple

• CWN has many advantages over RWN especially when
  the number of element is uncertain.
• With CWN, when we add more pixels or shift to
  study other plane, we need only small
  modification on index.

Benefit mentioned in Robs message a) does not
waste disk space on information which is absent
for this event  ie you can declare very large
arrays for hits but it only uses disk space for
the ones which are present.b) faster interactive
response if you are only referencing a small
subset of the ntuple variables.  This comes
partly from a) but also from a "smarter" layout
of how data is stored on disk.c) much easier
maintenance when you change the number of
detectors
18
Variables in CWN

• Track Block
• ievt, ntrks, n_klm_ok, itrk, klm_ok, x_int,
  x_slp, y_int, y_slp,
• chisq, cl, ndof, fixyslp, yclass, nplanes
• Plane Block
• (ngeom elements except for ngeom)
• ngeom, itype, iort_row, nclus, iclus, npix, nrow,
  ncol, rowb, colb, rowq, colq,
• eta, sumq, xg, yg, xl, yl, xsig, ysig, xw, yw,
  zw, rxi, srxi, spxi, ryi, sryi, spyi,
• rxo, srxo, spxo, ryo, sryo, spyo, xf, yf, rhof,
  xslp, yslp, infitx, infity
• Hit Block
• (nhit elements except for nhit)
• nhit, ipln, icol, irow, iadc, icls, qval

19
Name List Variables
DETECTOR sixy_carrier(1) 2
sixy_ort_r(1) 1 sixy_ort_c(1) 1
sixy_ort_z(1) 1 sixy_u_app(1) 140
sixy_u_dep(1) 105 bfield_x(1)
0.0 bfield_y(1) 0.0 bfield_z(1)
0.0 sixy_thick(1) 280
sixy_siz_r(1) 50 sixy_siz_c(1) 400
sixy_noise_val(1) 0.0 sixy_gain_err(1)
0.0 sixy_sig_cpl(1) 0.0
sixy_thr_val(1) 0.0 sixy_thr_dsp(1)
0.0 sixy_thr_err(1) 0.0 END

• CONTROL
• max_evts 10000
• tb_control 'event'
• run_mc 999
• ncycle 1000
• binary_weights 0
• override_seeds 0
• xref1o 1
• xref2o 2
• yref1o 5
• yref2o 6
• dx_cuto 0.06
• y_slp_seed 0.000055
• max_miss_planes 1
• drop_plane(9) 1
• END

FILES geom_file 'tb_geometry.dat'
data_in_file ' ' data_out_file ' '
hbk_out_file ' ' sixy_adc_file(1)
'adc_plane1.dat' END
20
To Be Discussed

• Data format and byte swap
• CVS and terminology (Rob and Lynn)
• Do we need to include SSD?
• Column wised ntuple
• ???

21
Analysis Flow

• call read_namelist
• status open_old_file(data_in_file)
• status read_data_header(run_tmp,
  num_tmp, type_tmp)
• call histo_init(hbk_out_file)
• status read_geometry(0)
• call fill_derived_geom
  call sixy_geom_hfill(11100)
• call sixy_detector_parameter call
  sixy_dpar_hfill
• if ( want_summary ) status
  open_new_file(data_out_file)
• if ( want_summary ) status
  write_data_header(runId, nPlanes, runType)
• 10 status read_event(1)
• if ( status .eq. 0 ) then
• call sixy_adc2q
• call form_clusters
  call hit_hfill
• call form_tracks
• call fit_tracks
  call fit_hfill
• call fill_ntuple
• if ( want_summary ) call
  mn_write_track
• if ( not_enough ) goto 10
• endif

22
MC Generation Flow

• call read_namelist
• call histo_init(hbk_out_file)
• status read_geometry(0)
• call fill_derived_geom
  call sixy_geom_hfill(11100)
• call sixy_detector_parameter call
  sixy_dpar_hfill
• status open_new_file(data_out_file)
• status write_data_header(run_num,
  n_pix_planes, type_tmp)
• call ranecu_init
• call set_beam_properties
• call sixy_drift_map
  call sixy_map_hfill
• call sixy_charge_parameter
• do evt_num 1, max_evts
• call mc_track
• call mc_signal
• call mc_write
  call mc_hfill
• enddo
• output histogram

23
Alignment Flow

• call read_namelist
• status open_old_file(data_in_file)
• status read_data_header(run_tmp, num_tmp,
  type_tmp)
• call histo_init(hbk_out_file)
• status read_geometry(0)
• call fill_derived_geom
• call sixy_geom_hfill(11100)
• call sixy_detector_parameter
• call sixy_dpar_hfill
• call mn_prepare_fit
• call mn_read_track
• call mn_fit_geo
• call sixy_geom_hfill(11200)
• output histogram