Fix bugs in Tragectory calculations in MD normalization for direct geometry

[Alex Buts]

Build LET event workspace using script attached and data files available in system tests Data directory (it will fail with current reduction but usable event workspace LET1_spe will be produced

Then run script below to produce normalization workspace and do normalization.

The picture provided clearly shows that the trajectories calculations is wrong

SetUB(Workspace='LET1_spe', a=2.73, b=2.73, c=2.73)
SetGoniometer(Workspace='LET1_spe', Goniometers='Universal', Axis0='psi,0,1,0,1')
AddSampleLog(Workspace='LET1_spe', LogName='Psi', LogText='20', LogType='Number')
ConvertToMD(InputWorkspace='LET1_spe', QDimensions='Q3D', Q3DFrames='HKL', QConversionScales='HKL', OutputWorkspace='mdWs1', MinValues='-0.1,-0.1,-0.1,-5', MaxValues='1,0.05,0.05,10', SplitInto='20', MaxRecursionDepth=1)
MDNormDirectSC(InputWorkspace='mdWs1', AlignedDim0='[H,0,0],-0.1,1,20', AlignedDim1='[0,K,0],-0.1,0.05,20', AlignedDim2='[0,0,L],-0.1,0.05,20', AlignedDim3='DeltaE,-5,10,20', OutputWorkspace='mdWs1a', OutputNormalizationWorkspace='mdWs1Norm')

DivideMD(LHSWorkspace='mdWs1a', RHSWorkspace='mdWs1Norm', OutputWorkspace='mdWs1b')

[Alex Buts]

LET shortReduction

[Alex Buts]

map file for LET reduction

[Alex Buts]

Errors in trajectories

[Andrei Savici]

Seems to be a small confusion in the process. Will put a big note in the algorithm help

[Andrei Savici]

More warnings in the help file. Refs #10829

Changeset: 550c3ef90f8c1971fd26daa04a400f971959acc1

[Andrei Savici]

Usage example for MDNormDirectSC. Refs #10829

Also enabled the parallel processing. Refs #10830

Changeset: e9e0dd6659761abae296e5ed10a8f99928625887

[Andrei Savici]

Usage example. Refs #10829

The MDNorm concept page will be expanded as part of ticket #9105

Changeset: f981f5ee946b5bea67cacb7cd42cd9bcad1c85d9

[Andrei Savici]

Normalization is calculated correctly for event workspaces only if the vent workspace is cropped and binned to the same interval. The normalization is using the limits from Rebin to calculate minimum and maximum final energy. However, if the event workspace is not cropped to the same interval, one might have MD events where the normalization is calculated to be 0. Maybe ConvertToMD for events should automatically ignore the events that fall outside the binning range.

Not too much to test. Check if the explanation makes sense, maybe play with the usage example. A note about same limits for cropping and binning should be in the help file.

[Andrei Savici]

cont_nonzero not found on older numpy version. Refs #10829

Changeset: 890003d4fe4af7081cc85d109580d411c3383036

[Alex Buts]

Merge remote-tracking branch 'origin/bugfix/10829_MDNormDirectSC_help'

Full changeset: ​08f4c050566677dcdc2d25c12c38b49eb1b600f2

[Alex Buts]

I have briefly looked through the changes and they are all positive. I am also reasonably satisfied with the explanation on why event workspace normalization behaves weird (We can fixing this later modifying e.g. ConvertToMD)

I still need to run more tests to be sure that new normalization is correct (or some IS will certainly do that)

But as experimental algorithm this all certainly belongs to the release 3.3

[Stuart Campbell]

This ticket has been transferred to github issue ​11671