Trusted, Bureaucrats
4,228
edits
Changes
Formatting of exponentials.
* Shifted to using the faster numpy.linalg.matrix_power() function in lib.dispersion.ns_2site_star. This was originally using the lib.linear_algebra.matrix_power.square_matrix_power() function, however the numpy equivalent is faster.
* More speed ups of the [[NS CPMG 2-site star|NS 2-site star]] dispersion model. A number of calculations have been shifted to the target function initialisation code, avoiding unnecessary repetitive mathematical operations.
* Improvement of the error handling in the [[NS CPMG 2-site star|NS 2-site star]] model. The fA and p<sub>B</sub> parameters are no longer being checked. Instead a Mgx value of 0.0 is being checked for. This catches additional problems. And now instead of the R<sub>2eff</sub> value being set to zero, it is set to 1e991e<sup>99</sup>. This is because log of zero is -inf, and then multiplied by a negative constant gives positive inf.
* Docstring completion for lib.dispersion.ns_2site_star.r2eff_ns_2site_star(). Epydoc text was missing for some of the function arguments.
* Changed 'numerical integration' to 'numerical solutions' in the dispersion chapter of the manual.
* The Relax_disp.test_hansen_cpmg_data_ns_2site_expanded system test now passes. The test has been set up to match Relax_disp.test_hansen_cpmg_data_CR72. This allows the efficiency of each method to be compared by running the tests with the --time flag.
* Fix for the model nesting method of the relaxation dispersion auto-analysis for deselected spins.
* Added an upper constraint of 2e6 2e<sup>6</sup> rad/s for the k<sub>ex</sub> dispersion parameter. This is to prevent slow optimisation to values in the order of 1e20 1e<sup>20</sup> for models which fail.
* Updated the model lists for the relax scripts for optimising Flemming Hansen's CPMG data. The model lists now include the numeric models and the [[CR72 full]] model.
* The lib.software.grace.write_xy_header() can now set the legend fill pattern and font size.
* Changed the current data pipe in the relax saved states for Flemming Hansen's truncated CPMG data.
* Another change of the base relax files of the truncated CPMG data.
* Bug fix for the relax_disp.cpmgfit_input user function. The ν<sub>CPMG</sub> values need to be doubled and then divided by 1e3 1e<sup>3</sup> to obtain the 1/τ<sub>CPMG</sub> values in ms.
* Fix for the relax_disp.sherekhan_input user function for the recent changes.
* Updated all of the results for the truncated CPMG data from Flemming Hansen in the test suite. The results are now different as the errors are now much more precise as they come from all spin systems rather than just the truncated set of :4, :70, and :71.
* Bug fix for the [[MMQ 2-site]] dispersion model target function. The relaxation time was being taken as that of the first experiment for all experiments. This is a relic from the code being copied from a single experiment type model.
* Converted the [[MQ CR72]] dispersion model to handle MMQ data. This model can now handle proton-heteronuclear SQ, ZQ, DQ, and MQ CPMG-type data. Some debugging might still be required.
* Fix for the [[MQ CR72]] model for MQ-type data. The check to prevent acos of a number less than 1 has been changed to switch the sign rather than to set the back calculated R<sub>2eff</sub> to 1e991e<sup>99</sup>.
* Another bug fix for the [[MQ CR72]] dispersion model. The ν<sub>CPMG</sub> value rather than the relaxation time was being used to calculate the R<sub>2eff</sub> values as the division by 'n' was missing.
* The relax_disp.plot_disp_curves user function can now handle values of NaN. These are simply replaced by 0.0 to allow Grace to open the file.