Angular detectors data generated by save_nxspe are incorrect for rings

[Alex Buts]

Two pictures, attached to the ticket illustrate this for LET one is obtained using precaluclated phx file and another one -- from data, located in nxspe.

[Alex Buts]

correct theta

[Alex Buts]

wrong theta

[Alex Buts]

refs #8095 presumably fixed issue

but the unit tests still need some work on

Changeset: b5a40a1637a5c8c928a744d71cdf8664e549fa0b

[Alex Buts]

refs #8095 Unit tests fixed

and hopefully correct

Changeset: 8dab6c20308412974afb28ea1638a718f7e4b64e

[Alex Buts]

refs #8095 Make it run multithreaded

Changeset: 0d2b6c383cd4ff0c21b720b437d9f6482a1fd1d2

[Alex Buts]

refs #8095 fixing Unix warnings

and correct parallel interrupt region defined

Changeset: 4b01211d4a1e28d1b10760834576bba482277047

[Alex Buts]

refs #8095 Another Unix warning

Changeset: d3f0dc8cd08e6486d4a09a9d938b1db83a9aae5d

[Alex Buts]

refs #8095 Comments and Doxygen

Changeset: 6efb23d3c0223f72813605b2b6b3482bbc699878

[Alex Buts]

It is not really easy to test this ticket unless you know mslice and can easy reduce some ISIS inelastic data into energy transfer.

If you can, then you can reduce some inelastic data using rings map for this cycle, generate nxspe file and load it into mslice using and not using correspondent par file. (take par from nxspe). The result should be (almost) identical. This would be an easy and comprehensive testing and I may do it with an instrument scientist. Actually, current LET system test uses rings map so one may only need correspondent .par which can be found at ​https://svn.isis.rl.ac.uk/InstrumentFiles/trunk/let (I think any rings.par or phx will do)

The error was not in SaveNXSE itself but in the child algorithm, FindDetectorsPar. The problem was the reliance of this algorithm on calculating centre of mass for rings and using this value to identify the ring itself and its derived parameters.

This would not work for rings which do not cover symmetric (vrt. the folding transformation around the plain, containing the beam line) areas of scattering angle (for incomplete rings of equal theta angels).

This is why I've completely rewritten the algorithm. The code become slower, but I have compensated it by multithreaded execution.

Tester not familiar with reduction and mslice can generate par or phx file (SavePAR or SavePHX) for an instrument (this will run FindDetectorsPar and save results into correspondent file) and compare R Phi and Theta values for Mantid detectors (Show Detectors) with R Phi and Theta in par or phx file. This is a sanity check and values should coincide. Then tester should group detectors into blocks and do the same for the instrument with combined detectors. Angular parameters for combined detectors should be very similar to the Mantid one.

Main part of FindDetectorsPar algorithm deals with calculations of angular sizes of the simple or combined detectors as seen from the sample. I do not know how these values are currently used so can not check them thoroughly (there are also number of approximations made when calculating these values) but these values should look reasonable.

[Alex Buts]

refs #8095 fixing parallel issue on MAC

Changeset: cca2486cd3b0da4f62ce686a896be6b286b20e1b

[Alex Buts]

refs #8095 save check for undefined detector.

Changeset: b5bca483a43da3e68801464a4126867fe7adc9b0

[Alex Buts]

refs #8095 use reset instead of = to free the shared pointer

in the expression written to avoid MAC hung-up problem

Changeset: 870f1fb8e93ae208f8bb0dcae33f6acdca910016

[Andrei Savici]

How do you generate the rings? If you use GenerateGroupingPowder to generate the xml grouping file, it automatically generates the par file to be used with SaveNXSPE

[Alex Buts]

There are number of map files used in ISIS to construct rings. They are generated by matlab programs which I have never seen. The references for LET and other instruments are above but be careful with them as (e.g. MAPS rings (parker rins)) will not group virgin workspace correctly. It needs some additional manipulations. Let rings seems will work on fresh workspace.

but if you have ring-grouped workspace and do not provide par file when saving nxspe, the algorithm should generate and save to nxspe detector coordinates similar to one in the par file you have. This will be good and simple check.

There were some issues when using GenerateGroupingPowder for ISIS and I am not sure if they are resolved (something to do with 4 to 1 mapping or other oddities)

[Alex Buts]

refs #8095 NXSPE version have changed to 1.2

to indicate to clients that detector's information is processed differently.

Changeset: 94c5e346795ab5d3561a1cd0e1ae42a939d3f0fe

[Owen Arnold]

Merge remote-tracking branch 'origin/bugfix/8095_wrongAngles'

Full changeset: ​11a8af5f1ef27b8982a0a7a8d234d2ce135228a2

[Stuart Campbell]

This ticket has been transferred to github issue ​8940