User:Boesecke: Difference between revisions

From canSAS
No edit summary
No edit summary
 
(27 intermediate revisions by the same user not shown)
Line 1: Line 1:
from Peter Boesecke (ESRF, Grenoble)
'''Go back to [[2012_Data_Discussion]]'''


The discussion in what format small angle scattering data should be saved is already going on for many years: XML, HDF5, NEXUS etc. I am convinced that this is not the most important question that must be answered. If somebody would decide: "It is BXHN" (just a name) I would not feel better because the most important part would still be missing: What data must be saved, what data are needed for analysis? As beamline scientist and local contact at X-ray scattering beamlines I have always tried to save as much useful metadata as possible together with the (2D) scattering data. I have always tried to make sure that the used geometrical parameters are not contradictory, i.e. well-defined. Many of these metadata are not needed very often, e.g. the proposal number, but some are practically always needed, e.g. the wavelength. The standard geometry is a small angle scattering experiment where the (1D, 2D) detector is perpendicular to the primary beam (required metadata: pixel sizes, wavelength, distance, beam center). I want to be able to analyze in a similar way standard geometry data and data from more complicated geometries, e.g. inclined detectors. I add parameters that default to the standard case when not given. An inclined detector is described by additional rotation angles that are zero when not defined. It can also be necessary to look to the data in a different way, e.g. to construct a physical detector mask, where distances in mm are much more adequate than in 1/nm.
Peter Boesecke (ESRF, Grenoble)


A description of my parametrization it is shown [[SX parametrization-ref-short 20120217.pdf | here]].  
The discussion in what format small angle scattering data should be saved is already going on for many years: XML, HDF5, NEXUS etc. I am convinced that this is not the most important question that must be answered. If somebody would decide: "It is BXHN" (just a name) I would not feel better because the most important part would still be missing: ''What data must be saved, what data are needed for analysis?'' As beamline scientist and local contact at X-ray scattering beamlines I have always tried to save as much useful metadata as possible together with the (2D) scattering data. The parameter set was evolving with time and I have tried to make sure that the parameters do not become contradictory and that they remain well-defined. Many metadata parameters are not needed very often, e.g. proposal number, but some others are, e.g. wavelength and beam center.


It must be possible to extract the described parameters from a standard SAS file (what are the corresponding NEXUS parameters?). How could/must I save them after refinement in the standard SAS file? I surely do not want to overwrite any raw data, e.g. the monochromator rotation. For me it would be sufficient to write all my own parameters. But I would prefer a common way.
As internal reference a standard geometry of a small angle scattering experiment has been chosen where the position sensitive detector (1D or 2D) is perpendicular to the primary beam (required metadata: ''pixel size'' in each direction, ''wavelength'', ''distance'', ''beam center''). The use of regions of interest and binning must be supported without complicated recalibration. It should be possible to analyze in a similar way standard geometry data and data from more complicated geometries, e.g. inclined detectors. Therefore, parameters are added that default to the standard case when not specified. An inclined detector is described by additional rotation angles that are zero when not defined. It becomes necessary to distinguish clearly between ''beam center'' and ''point of normal incidence''. Sometimes it can also become more comfortable to look to the data in a different way, e.g. to construct a physical detector mask. In this case distances in mm are much more adequate than in 1/nm.  


For statistical error propagation calculations I propose the optional use of a variance array that contains the variance of each pixel.
*Description: '''[[media:SX parametrization-ref-short 20120217.pdf | SX parametrization-ref-short]]'''.
 
It must be possible to extract the described SX-parameters from a SAS file independent of its format (task: what are the corresponding NEXUS parameters?). How could/must they be saved after refinement, e.g. in a standard SAS file? It is surely not desirable to overwrite any raw data, e.g. the monochromator rotation. It would be sufficient to save all ''SX-parameters'' or a SAS file standard parameter list that contains them, at least implicitly.
 
For statistical error propagation calculations I propose to store, optionally, the variance of each pixel (see: Peter Boesecke, "Reduction of scattering data" J. Appl. Cryst. (2007). 40, s423–s427)
 
In addition to describe a given experimental geometry it should be possible to save experimental conditions (''sample name'', ''temperature'', ''pressure'' etc.) and to have the possibility look-up experiments in a data base. But this is out of the scope of this contribution.
 
The above description is suitable for pinhole type experiments with linear and area detectors. Other detector types or different experimental configurations may need additional information. Whenever possible the descriptions should converge for limiting cases, e.g. the description of cylindrical detector geometry should converge to the description of linear detector geometry when the cylindrical radius approaches infinity. Linear and area detectors should be limiting cases of cylindrical and spherical detectors. The geometrical description of SANS data should converge with the description of SAXS data for gravity -> 0 (just a wish).

Latest revision as of 09:17, 19 July 2012

Go back to 2012_Data_Discussion

Peter Boesecke (ESRF, Grenoble)

The discussion in what format small angle scattering data should be saved is already going on for many years: XML, HDF5, NEXUS etc. I am convinced that this is not the most important question that must be answered. If somebody would decide: "It is BXHN" (just a name) I would not feel better because the most important part would still be missing: What data must be saved, what data are needed for analysis? As beamline scientist and local contact at X-ray scattering beamlines I have always tried to save as much useful metadata as possible together with the (2D) scattering data. The parameter set was evolving with time and I have tried to make sure that the parameters do not become contradictory and that they remain well-defined. Many metadata parameters are not needed very often, e.g. proposal number, but some others are, e.g. wavelength and beam center.

As internal reference a standard geometry of a small angle scattering experiment has been chosen where the position sensitive detector (1D or 2D) is perpendicular to the primary beam (required metadata: pixel size in each direction, wavelength, distance, beam center). The use of regions of interest and binning must be supported without complicated recalibration. It should be possible to analyze in a similar way standard geometry data and data from more complicated geometries, e.g. inclined detectors. Therefore, parameters are added that default to the standard case when not specified. An inclined detector is described by additional rotation angles that are zero when not defined. It becomes necessary to distinguish clearly between beam center and point of normal incidence. Sometimes it can also become more comfortable to look to the data in a different way, e.g. to construct a physical detector mask. In this case distances in mm are much more adequate than in 1/nm.

It must be possible to extract the described SX-parameters from a SAS file independent of its format (task: what are the corresponding NEXUS parameters?). How could/must they be saved after refinement, e.g. in a standard SAS file? It is surely not desirable to overwrite any raw data, e.g. the monochromator rotation. It would be sufficient to save all SX-parameters or a SAS file standard parameter list that contains them, at least implicitly.

For statistical error propagation calculations I propose to store, optionally, the variance of each pixel (see: Peter Boesecke, "Reduction of scattering data" J. Appl. Cryst. (2007). 40, s423–s427)

In addition to describe a given experimental geometry it should be possible to save experimental conditions (sample name, temperature, pressure etc.) and to have the possibility look-up experiments in a data base. But this is out of the scope of this contribution.

The above description is suitable for pinhole type experiments with linear and area detectors. Other detector types or different experimental configurations may need additional information. Whenever possible the descriptions should converge for limiting cases, e.g. the description of cylindrical detector geometry should converge to the description of linear detector geometry when the cylindrical radius approaches infinity. Linear and area detectors should be limiting cases of cylindrical and spherical detectors. The geometrical description of SANS data should converge with the description of SAXS data for gravity -> 0 (just a wish).