Cannot restore ODF correctly from file exported by mtex #493

YoshieraHuang · 2020-06-27T12:40:51Z

YoshieraHuang
Jun 27, 2020

I use mtex to export a simple ODF to file and try to restore the ODF from the file. I have two version of mtex in my hand: 5.1.1 and
5.3.1. 5.1.1 can restore the ODF correctly but 5.3.1 cannot. Is this a bug or i use mtex incorrectly?

Here are the details:

I use this script to export ODF:

fname = '/home/jasper/Desktop/Cu.recrystal.odf.txt';
CS = loadCIF('Cu-Copper.cif');
SS = specimenSymmetry('1');
mod1 = orientation.cube(CS, SS);
odf = unimodalODF(mod1, 'halfwidth', 5*degree);
textureindex(odf)
h = [Miller(1,0,0,CS),Miller(1,1,0,CS),Miller(1,1,1,CS)];
% 
% % Pole figure
w1 = figure;
plotPDF(odf,h,'minmax')
mtexColorbar       % create a new colorbar\
set(gcf,'PaperPositionMode','auto');
ofname = [fname,'.PF.png']
print(w1,'-dpng',ofname);
 

% Inverse pole figure
w2 = figure;
plotIPDF(odf,[xvector,yvector,zvector],'minmax')
mtexColorbar       % create a new colorbar\
set(gcf,'PaperPositionMode','auto');
ofname = [fname,'.IPF.png'];
print(w2,'-dpng',ofname);

% define a equispaced grid in orientation space with resolution of 5 degree
S3G = equispacedSO3Grid(CS,'resolution',5*degree);
export(odf, fname, S3G, 'Bunge', 'generic');
%end

Here are the PFs and IPFs generated from the script above:

Here is my script to restore ODF:

%Import script
clear all;
fname = '/home/jasper/Desktop/Cu.recrystal.odf.txt';

CS = loadCIF('Cu-Copper.cif');
odf = loadODF_generic(fname, 'cs', CS, ...
    'ColumnNames', {'phi1', 'Phi', 'phi2', 'weights'});
textureindex(odf)
h = [Miller(1,0,0,CS),Miller(1,1,0,CS),Miller(1,1,1,CS)];

% % Pole figure
w1 = figure;
plotPDF(odf,h,'minmax')
mtexColorbar       % create a new colorbar\
set(gcf,'PaperPositionMode','auto');
ofname = [fname,'.PF.png']
print(w1,'-dpng',ofname);
 

% Inverse pole figure
w2 = figure;
plotIPDF(odf,[xvector,yvector,zvector],'minmax')
mtexColorbar       % create a new colorbar\
set(gcf,'PaperPositionMode','auto');
ofname = [fname,'.IPF.png'];
print(w2,'-dpng',ofname);

in 5.1.1, the generated PFs and IPFs is consistent:

However, in 5.3.1, the generated PFs and IPFs is weird:

the orientation densities(OD) of three equivalent directions <100>, <010> and <001> is not equivalent now.

Answered by ralfHielscher

Jun 28, 2020

It won't become better. Your initial ODF has a halfwidth of about 5 degree and you export it on a 5 degree grid. This like you want to interpolate a Gaussian from three points. You will need more points for higher accuracy. Ralf ******************************************************************** Ralf Hielscher Tel: +371-531-38556 Fakultät für Mathematik +371-531-22200 (Sekr.) Technische Universität Chemnitz Fax: +371-531-22109 Reichenhainer Str. 39 E-mail: [email protected] D-09126 Chemnitz http://www.tu-chemnitz.de/~rahi ********************************************************************

View full answer

ralfHielscher · 2020-06-27T21:05:06Z

ralfHielscher
Jun 27, 2020
Maintainer

Adding the option 'lsqnonneg' should resolve your issue, i.e.,

odf = loadODF_generic(fname, 'cs', CS, ...
    'ColumnNames', {'phi1', 'Phi', 'phi2', 'weights'},'lsqnonneg');

0 replies

YoshieraHuang · 2020-06-28T01:05:28Z

YoshieraHuang
Jun 28, 2020
Author

Adding the option 'lsqnonneg' should resolve your issue, i.e.,
odf = loadODF_generic(fname, 'cs', CS, ...
    'ColumnNames', {'phi1', 'Phi', 'phi2', 'weights'},'lsqnonneg');

Thanks for your reply! But the lsqnonneg option does not work. The results do not change.

0 replies

ralfHielscher · 2020-06-28T05:07:59Z

ralfHielscher
Jun 28, 2020
Maintainer

Could you please post the exact output of the loadODF command.

Thanks,

Ralf.

PS: I have tried it now several times but I could not reproduce your result when using the option lsqnonneg

0 replies

YoshieraHuang · 2020-06-28T05:50:06Z

YoshieraHuang
Jun 28, 2020
Author

Could you please post the exact output of the loadODF command.

Thanks,

Ralf.

PS: I have tried it now several times but I could not reproduce your result when using the option lsqnonneg

I used wrong pictures for PFs and IPFs of restored ODF in the first comment of this issue. I edit it. The difference of ODs of three equivelant directions is not so large but is still not negligible.

In my last comment, I mistake the lsqnonneg option to lsqnonneq option. 😓. The output changes with lsqnonneg option but still not consitent with the input ODF:

Here is the output of the loadODF command without lsqnonneg option:

Interpolating the ODF. This might take some time...
   lsqr tolerance cut back: 0.013
   lsqr tolerance cut back: 0.017
   lsqr tolerance cut back: 0.022
 
odf = ODF (show methods, plot)
  crystal symmetry : Copper (m-3m)
  specimen symmetry: 1
 
  Uniform portion:
    weight: 0
 
  Radially symmetric portion:
    kernel: de la Vallee Poussin, halfwidth 5°
    center: 2432 orientations, resolution: 5°
    weight: 1

Here is the output with lsqnonneg option:

Interpolating the ODF. This might take some time...
 
odf = ODF (show methods, plot)
  crystal symmetry : Copper (m-3m)
  specimen symmetry: 1
 
  Uniform portion:
    weight: 0
 
  Radially symmetric portion:
    kernel: de la Vallee Poussin, halfwidth 5°
    center: 9 orientations, resolution: 5°
    weight: 1

The textureindex of input ODF, restored ODF without lsqnonneg and restored ODF with lsqnonneg are ~202, ~331 and ~166, respectively.

My version of MATLAB is R2020a.

0 replies

ralfHielscher · 2020-06-28T06:06:41Z

ralfHielscher
Jun 28, 2020
Maintainer

It won't become better. Your initial ODF has a halfwidth of about 5 degree and you export it on a 5 degree grid. This like you want to interpolate a Gaussian from three points. You will need more points for higher accuracy. Ralf ******************************************************************** Ralf Hielscher Tel: +371-531-38556 Fakultät für Mathematik +371-531-22200 (Sekr.) Technische Universität Chemnitz Fax: +371-531-22109 Reichenhainer Str. 39 E-mail: [email protected] D-09126 Chemnitz http://www.tu-chemnitz.de/~rahi ********************************************************************

…

On Sun, Jun 28, 2020 at 7:50 AM Yoshiera ***@***.***> wrote: Could you please post the exact output of the loadODF command. Thanks, Ralf. PS: I have tried it now several times but I could not reproduce your result when using the option lsqnonneg I used wrong pictures for PFs and IPFs of restored ODF in the first comment of this issue. I edit it. The difference of ODs of three equivelant directions is not so large but is still not negligible. In my last comment, I mistake the lsqnonneg option to lsqnonneq option. 😓. The output changes with lsqnonneg option but still not consitent with the input ODF: [image: Cu recrystal odf txt lsg PF] <https://user-images.githubusercontent.com/38752027/85939083-34ec3a00-b945-11ea-9b88-c2c73a2686ac.png> [image: Cu recrystal odf txt lsg IPF] <https://user-images.githubusercontent.com/38752027/85939079-31f14980-b945-11ea-9316-4b5aae413593.png> Here is the output of the loadODF command without lsqnonneg option: Interpolating the ODF. This might take some time... lsqr tolerance cut back: 0.013 lsqr tolerance cut back: 0.017 lsqr tolerance cut back: 0.022 odf = ODF (show methods, plot) crystal symmetry : Copper (m-3m) specimen symmetry: 1 Uniform portion: weight: 0 Radially symmetric portion: kernel: de la Vallee Poussin, halfwidth 5° center: 2432 orientations, resolution: 5° weight: 1 Here is the output with lsqnonneg option: Interpolating the ODF. This might take some time... odf = ODF (show methods, plot) crystal symmetry : Copper (m-3m) specimen symmetry: 1 Uniform portion: weight: 0 Radially symmetric portion: kernel: de la Vallee Poussin, halfwidth 5° center: 9 orientations, resolution: 5° weight: 1 The textureindex of input ODF, restored ODF without lsqnonneg and restored ODF with lsqnonneg are ~202, ~331 and ~166, respectively. My version of MATLAB is R2020a. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/mtex-toolbox/mtex/issues/493#issuecomment-650701185>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACBYJ2JWCR6GTJ2MDOHZYWLRY3KZRANCNFSM4OKAOS2Q> .

0 replies

YoshieraHuang · 2020-06-28T06:44:55Z

YoshieraHuang
Jun 28, 2020
Author

Thank you for your generous replies! I change the 5 degree grid to 1 degree grid, and here is the output:

The ODs are smaller than input ones, and the difference shoule be thought negligable? Could you please explain why the difference happens? Thank you

0 replies

ralfHielscher · 2020-06-28T07:56:55Z

ralfHielscher
Jun 28, 2020
Maintainer

May I ask what you want to achieve? If you simply want to store an ODF variable on disk and reload it later you can simply take the Matlab commands save and load. This works without any error whatsoever. The export command is thought to export ODFs to an external program. And yes, this reconstruction is not perfect. But in your case it is quite ok. The maximum value of an ODF, which you seem to compare, is the worst of all ODF characteristics. As it is extremely sensitive to all kinds of parameters choice. I would never ever use this value to characterize an ODF. Ralf. ******************************************************************** Ralf Hielscher Tel: +371-531-38556 Fakultät für Mathematik +371-531-22200 (Sekr.) Technische Universität Chemnitz Fax: +371-531-22109 Reichenhainer Str. 39 E-mail: [email protected] D-09126 Chemnitz http://www.tu-chemnitz.de/~rahi ********************************************************************

…

On Sun, Jun 28, 2020 at 8:45 AM Yoshiera ***@***.***> wrote: Thank you for your generous replies! I change the 5 degree grid to 1 degree grid, and here is the output: [image: Cu recrystal odf 1 txt lsq PF] <https://user-images.githubusercontent.com/38752027/85940157-53563380-b94d-11ea-8149-fabe3671c41b.png> [image: Cu recrystal odf 1 txt lsq IPF] <https://user-images.githubusercontent.com/38752027/85940158-53eeca00-b94d-11ea-8ef5-ac5b410bd99a.png> The ODs are smaller than input ones, and the difference shoule be thought negligable? Could you please explain why the difference happens? Thank you — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/mtex-toolbox/mtex/issues/493#issuecomment-650706206>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACBYJ2K476LWIREFBOKWRTLRY3RHFANCNFSM4OKAOS2Q> .

0 replies

YoshieraHuang · 2020-06-28T08:14:41Z

YoshieraHuang
Jun 28, 2020
Author

May I ask what you want to achieve? If you simply want to store an ODF variable on disk and reload it later you can simply take the Matlab commands save and load. This works without any error whatsoever. The export command is thought to export ODFs to an external program. And yes, this reconstruction is not perfect. But in your case it is quite ok. The maximum value of an ODF, which you seem to compare, is the worst of all ODF characteristics. As it is extremely sensitive to all kinds of parameters choice. I would never ever use this value to characterize an ODF. Ralf. ******************************************************************** Ralf Hielscher Tel: +371-531-38556 Fakultät für Mathematik +371-531-22200 (Sekr.) Technische Universität Chemnitz Fax: +371-531-22109 Reichenhainer Str. 39 E-mail: [email protected] D-09126 Chemnitz http://www.tu-chemnitz.de/~rahi ********************************************************************
…
On Sun, Jun 28, 2020 at 8:45 AM Yoshiera @.***> wrote: Thank you for your generous replies! I change the 5 degree grid to 1 degree grid, and here is the output: [image: Cu recrystal odf 1 txt lsq PF] https://user-images.githubusercontent.com/38752027/85940157-53563380-b94d-11ea-8149-fabe3671c41b.png [image: Cu recrystal odf 1 txt lsq IPF] https://user-images.githubusercontent.com/38752027/85940158-53eeca00-b94d-11ea-8ef5-ac5b410bd99a.png The ODs are smaller than input ones, and the difference shoule be thought negligable? Could you please explain why the difference happens? Thank you — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#493 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACBYJ2K476LWIREFBOKWRTLRY3RHFANCNFSM4OKAOS2Q .

I use MTEX to generate ODF for the input of my own program.

I compared the intensities of three equivalent crystal direction <100>, <010> and <001> in PF, which shoule be identical for cubic in my opinion. The arrows in this pictures show what i concerned.

The max intensities in IPFs are also related to this.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cannot restore ODF correctly from file exported by mtex #493

{{title}}

{{editor}}'s edit

{{editor}}'s edit

Replies: 8 comments

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

Cannot restore ODF correctly from file exported by mtex #493

YoshieraHuang Jun 27, 2020

Replies: 8 comments

ralfHielscher Jun 27, 2020 Maintainer

YoshieraHuang Jun 28, 2020 Author

ralfHielscher Jun 28, 2020 Maintainer

YoshieraHuang Jun 28, 2020 Author

ralfHielscher Jun 28, 2020 Maintainer

YoshieraHuang Jun 28, 2020 Author

ralfHielscher Jun 28, 2020 Maintainer

YoshieraHuang Jun 28, 2020 Author

YoshieraHuang
Jun 27, 2020

ralfHielscher
Jun 27, 2020
Maintainer

YoshieraHuang
Jun 28, 2020
Author

ralfHielscher
Jun 28, 2020
Maintainer

YoshieraHuang
Jun 28, 2020
Author

ralfHielscher
Jun 28, 2020
Maintainer

YoshieraHuang
Jun 28, 2020
Author

ralfHielscher
Jun 28, 2020
Maintainer

YoshieraHuang
Jun 28, 2020
Author