High Throughput 2D SAXSWAXS Data Reduction

1
High Throughput 2D SAXS/WAXS Data Reduction
Peter Bösecke1, Theyencheri Narayanan1, Armando
Sole1, Rainer Wilcke1 1European Synchrotron
Radiation Facility (ESRF), F-38043 Grenoble
Cedex, France
2
Motivation
Strategy!

• Time resolved data
• Use of several detectors (SAXS/WAXS)
• Increasing data rates (20 Megabytes/s)
• Increasing amount of data (GB -gt TB)

• Data quality can only be estimated after basic
  corrections
• Detector corrections are only obvious to
  specialists
• Most of the SAXS data must be reduced before it
  can be used

• Online data correction
• Data is saved with all experimental information
• Data history is kept

3
Data Acquisition Rates

• 1 dim gasfilled detector
• 2 dim gasfilled detector
• 2 dim CCD detector

• increasing data rates
• danger of mistakes during data acquisition

4
SUN workstation 750MHz UltraSPARC III
data storage 1 TB
Windows PC (TACO device server)
start
Time Frame Generator (VISTA, VME)
data readout
data readout
trigger
SAXS CCD Detector (Frelon2000, Thomson
XRII, 2048x2048x14bit)
WAXS CCD Detector (Sensicam SVGA
1280x1024, Proxitronic MCP, fiber optic taper)
Sample (stretcher, stop flow device etc.)
5
SPEC (command line interface)
done
6
Corrections
identical exposure times
1. Dark image subtraction ADU
2. Spatial distortion correction ADU
3. Division by flatfield photons
4. Azimuthal averaging

• Subtraction and division are done pixel by pixel
• The spatial distortion correction consists of a
  horizontal and vertical displacement of each pixel

7
Timing

• Computing times (all in seconds)
• - correct correction time (dark image,
  distortion, ...)
• - azimuth azimuthal regrouping time
• - total total computing time including saving
  of output (raw data, corrected image and
  360 azimuthal regrouped image)
• 
  Frelon2000-bin
• correct azimuth total
• no correction 0.070 - 1.04
  
• dark image subtraction 0.158 - 1.37
• distortion correction 0.341 - 1.53
• flatfield division 0.378 - 1.59
• normalization 0.582 - 1.74
• background subtraction 0.647 - 1.82
• azimuthal regrouping 0.649 2.80 4.88

8
Summary
After calibration with a scattering standard each
detector frame is automatically converted to
absolute units. Depending on the performed
corrections the calculation takes less than 1
second per frame SAXS and WAXS data are corrected
in a similar way. The beamline user and his
local contact have additional information during
the experiment to optimize the set-up
For more details, see poster P. Bösecke et al.
High-throughput SAXS/WAXS Data Processing
9
Acknowledgement
ID2 beamline Stephanie Finet, Loys Goirand,
Jacques Gorini, Pierre Panine, Manfred
Rössle, Diego Pontoni, Volker Urban
ESRF support groups David Fernandez, Claudio
Ferrero, Jörg Klora
10
CCD based two dimensional detectors at
synchrotron radiation beamlines are steadily
increasing. At the High-brilliance Beamline
(ID02), ESRF, detectors with 1024 x 1024 pixels
and 16 bit data are in routine use. The hardware
can acquire up to 14 images per second.
Furthermore, these detectors often require
various corrections before the measured
intensities can be converted to absolute
differential scattering cross-sections. To
enhance the productivity and reliability of the
experiments, it is necessary to process the
experimental data online, simultaneously with the
data acquisition. A data analysis package has
been developed which can operate simultaneously
on several detectors. The program performs
corrections for detector artifacts such as dark
current, image distortion and spatial
inhomogeneities, as well as normalization by
incident photon flux, sample transmission and
angular acceptance. Each image is saved together
with the complete parameter set that describe the
experimental setup. Without azimuthal regrouping,
the correction time is of the order of 0.5
seconds per image (1024 x 1024 pixels, 16 bit) on
a 750MHz UltraSPARC III workstation. During
experiments where multiple fast acquisitions are
triggered by a time frame generator (e.g. 100
frames with alternating dead and live frames of
0.1s duration) the images are corrected after the
end of the last frame. In this case the delay is
acceptable because the whole experiment takes
only a few seconds. During less rapid experiments
single frames are repeatedly acquired and
corrected. The online application of this program
is demonstrated by a time-resolved SAXS/WAXS
experiment.