Re: [bug #22008] Interpolation with relax.plot_disp_curves, creates un-even number of CPMG blocks.Re: [bug #22008] Interpolation with relax.plot_disp_curves, creates un-even number of CPMG blocks.Hi, Ok, I'll leave it to you to revert r22937. For the commit message, remember to include the name of the revision reverted, the command used (svn merge -r22937:r22936), as well as the old svn commit message indented by 2 spaces and surrounded '.....'. This is required for me to keep track of things - this is obviously a weak point when compared to git! Cheers, Edward On 4 May 2014 17:38, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:
Hi Edward. B14 is no longer in MODEL_LIST_NUMERIC_CPMG. And I printed out the powers from the target function, and the relax_disp.plot_curves interpolate with odd numbers. So, ... bum bum. But I am sure, that an even ncyc number becomes handy! Best Troels 2014-05-04 16:42 GMT+02:00 Edward d'Auvergne <edward@xxxxxxxxxxxxx>:Hi, The commit r22937 was rather your bug fix (http://article.gmane.org/gmane.science.nmr.relax.scm/20678). If you revert this, do you still see the problem? Maybe this fix is why you can no longer reproduce the bug ;) The relax_disp.cpmg_frq to relax_disp.cpmg_setup user function change is r22943 (http://article.gmane.org/gmane.science.nmr.relax.scm/20684). I think such a change should be performed now rather than later. This is because scripts are allowed to break between minor version numbers, i.e. between 3.1.x to 3.2.x relax versions. But scripts and the relax API should not break for a given minor version number, i.e. all 3.2.x releases should be compatible. Then I can release relax 3.2.0 with the large specific API changes, code clean ups, the B14 model, and all of the recent bugfixes. Regards, Edward On 4 May 2014 16:32, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:Yeah, that is probably best. It will break all current scripts at people places. Can you "store" the change for later user? Best Troels 2014-05-04 15:21 GMT+02:00 Edward d'Auvergne <edward@xxxxxxxxxxxxx>:Hmmm. I wonder what happened there. I've now renamed the user function (http://article.gmane.org/gmane.science.nmr.relax.scm/20684). This will nevertheless be useful if we wish to advance the numerical models in relax. Do you think you should revert r22937 then? http://svn.gna.org/viewcvs/relax/trunk/specific_analyses/relax_disp/data.py?r1=22937&r2=22936&pathrev=22937 Regards, Edward On 4 May 2014 12:48, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:Hm. Now I cant reproduce the bug? An odd number of NCYC, gives fine interpolated graphs. Hm... 2014-05-04 12:24 GMT+02:00 Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx>:Sound good with a flag, default to True. That means less code interruption. :-) Best Troels 2014-05-04 12:12 GMT+02:00 Edward d'Auvergne <edward@xxxxxxxxxxxxx>:I might try implementing this user function change, and store the cdp.ncyc_even flag. Then you could use it for the interpolation. Regards, Edward On 4 May 2014 11:59, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:The defaults for a relax_disp.cpmg_setup could be: relax_disp.cpmg_setup(spectrum_id=None, cpmg_frq=None, ncyc_even=True) This can then be expanded in the future for special CPMG dispersion experiment types (CW decoupling vs. pi pulses, etc.) where the numeric model would require changes. Regards, Edward On 4 May 2014 11:53, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:Sorry, that was a bad typo, it should be Flemming Hansen's dispersion pulse sequence! On 4 May 2014 11:52, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:Hi, It depends on the pulse sequence. Here is one I found written by Flemming Hanser where you can use odd numbers: http://nmrwiki.org/psdb/kaylab/vnmrsys/psglib/CaHD_cpmg_GLY_dfh_600_v1.c Look for the comment: "ncyc can be either even or odd :)" Such sequences are probably in the minority though. Anyway, maybe we need a new user function. It would be good to have a series of user functions for specifying the experimental information. We already have that with: spectrometer.frequency relax_disp.exp_type relax_disp.relax_time relax_disp.cpmg_frq So keeping along these lines, maybe we need to have a relax_disp.cpmg_even_ncyc user function? Or we rename relax_disp.cpmg_frq to relax_disp.cpmg_setup and have that user function reserved for all CPMG pulse sequence info? What do you think? Renaming user functions does require a relax minor version number change though, so introducing it before relax 3.2.0 is a good idea. Regards, Edward On 4 May 2014 11:24, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:Hi Edward. The number of CPMG blocks has to be an even number. I remember this clearly, since I once did an CPMG experiment, with some ncycs equal 4, 10, 15, 20, 25, ... And the intensities for all eksperiments with odd NCYC number was horrible. Kaare told me, that ncycs always has to be even. That was something that Mikael Akke also have insisted on. But that day, I could not easily find it in the literature, so I left it and accepted just another fact of NMR. And when I did the interpolated graphs with an odd-number of NCYC, that looked weirdo. Sig-saw all over the place. I actually think it could be a input check in relax, warning the user if the number of CPMG blocks are not equal? And, I would be very happy to find it in the literature. :-) Have you ever come around this? Best Troels 2014-05-04 11:06 GMT+02:00 Edward d Auvergne <NO-REPLY.INVALID-ADDRESS@xxxxxxx>:Follow-up Comment #2, bug #22008 (project relax): If this is a restriction of only the B14 analytic CPMG model (http://wiki.nmr-relax.com/B14), it would be best if only the B14 is affected. There is no need to restrict the numeric models based on the artificial limitation of an unrelated model. Maybe the best solution would be to check if the model is B14, and if so skip odd interpolation points? _______________________________________________________ Reply to this item at: <http://gna.org/bugs/?22008> _______________________________________________ Message sent via/by Gna! http://gna.org/