LoadSQW->SliceMD->SliceView loses information about high intensity peaks

[Alex Buts]

When tying to compare Mantid and Horace results, I am generating test horace dataset (see ​http://horace.isis.rl.ac.uk/Download_and_setup->run demo script, which I can not attach as ~700M) and view couple of slices.

They are generated by the command: w_template=cut_sqw(data_source,proj_100,[-0.4,0.2],0,0.05,3],-0.5,0.05,3],[30,40]); plot(w_template) lz 0 4 lz 0 300

and are attached below (HoraceDS_I004.png and HoraceDS_I300.png)

When I try to reproduce the same result in Mantid, I LoadSQW and do cuts by the following script:

LoadSQW(Filename='D:/Data/Fe/Aug10_v1/sqw/fe_demo.sqw',OutputWorkspace='Fe200',OutputFilename='D:/Data/Fe/Aug10_v1/sqw/Fe_demo.nxs')

w_template=SliceMD(InputWorkspace='Fe200',AlignedDimX='Q_
zeta,-0.4,0.2,1',AlignedDimY='Q_
xi,0,3,60',AlignedDimZ='Q_
eta,-0.5,3,70',AlignedDimT='E,30,40,1')

The results of visualizing w_template workspace with sliceViewer, normalized by number of events are presented on pictures: MantidFe200_I004.png and MantidFe200_I300.png

Mantid Images look "kind of" similar to Horace, but substantial part of information (particularly about the Bragg peaks -- which I was for) is absent

[Janik Zikovsky]

The fe_demo_slice_larger.png screenshot is the result of the following command:

w_template2=SliceMD(InputWorkspace='Fe200',AlignedDim0='Q_\\zeta,-0.4,0.2,1',
	AlignedDim1='Q_\\xi,0,6,60',AlignedDim2='Q_\\eta,-1,6,70',AlignedDim3='E,30,40,1') 

It looks like there is a scaling factor of about 2.2 in the coordinates read in from the SQW file. The peak that looks at (2,2) in the Horace screenshot is around (4.4, 4.4).

[Janik Zikovsky]

The coordinates saved in the file seem to be the raw Q coordinates, but looking at the Horace screenshots it looks like it is being plotted on a scale with units of "2.1893 Ang", whereas in MantidPlot we are using the same units as loaded (i.e. Ang) so this explains the discrepancy.

[Janik Zikovsky]

Resolved as invalid since it is a difference in plotting axes.

This probably goes back to the discussion Alex and Andrei have had concerning axis units etc. but it does not look to be a bug in either LoadSQW or BinMD/SliceMD.

[Alex Buts]

Horace keeps pixels (events) in Q but scales image (DND object) into whatever user requested.

As current design is not to keep image but deal with events (pixels) directly, the image obtained in slice viewed is indeed different.

Would agree, that resolution is correct at the current states of affairs, (Thanks, Janik) thought the state is a bit confusing.

When Andrei finishes his thicket on transformation matrix (#4813), changes and coordinate system scaling should be added to LoadSQW algorithm.

[Janik Zikovsky]

Created ticket #4846 for the issue you raise above.

[Alex Buts]

refs #4834 Minor changes to function factory allowing GaussMD work

with current function fitting framework

Changeset: 8a129d66cce5099f4431891baeebd30c95da5060

[Stuart Campbell]

This ticket has been transferred to github issue ​5681