Hi Shantanu, I was wondering which version of relax you are using? The bug looks quite familiar and if you are not using relax 2.1.2, I would recommend updating to the newer version and testing if the bug is still present (http://www.nmr-relax.com/download.html). I have checked the diff_min.py sample script and noticed that it was designed back in 2003 in the very early days of relax development. Today the script is useless! I have updated it in the relax subversion repository so you can use this version instead (http://svn.gna.org/viewcvs/relax/trunk/sample_scripts/model_free/diff_min.py?view=markup). I am guessing that you are reverting to the 2 decade old analysis methodology because you do not have data at two field strengths yet. Note that you will have to do a lot of work writing and debugging scripts to implement the old analysis types (even when using Art Palmer's Modelfree4), and I will not have much time to help you. If you do implement them, would you could consider contributing them to relax to be added to the sample script directory? You should be able to use Python's object oriented design to create one script implementing the entirety of the algorithm of Mandel et al., 1995 (the whole iterative process of the figure in http://www.nmr-relax.com/manual/methodology_Mandel_et_al_1995.html). I would recommend using the old data published by the Palmer group using this methodology to test that your super-script produces the expected results - such a test is essential to be sure that the full protocol is producing the correct results. If you are designing an entirely new model-free protocol to handle single-field strength data (or even if you are implementing the full Mandel et al., 1995 protocol), then it might be worth considering becoming a relax developer and implementing the protocol in the auto_analysis directory of relax. The front end script can be very similar to the dauvergne_protocol.py script, but the back end quite different. In such a case, you should create a set of test models of synthetic data and implement a relax system test before starting with the algorithm. Once you algorithm is complete and fully debugged, then the system test will pass. This should all be developed in a special relax branch. Cheers, Edward P. S. Do you intend on collecting data at a second field strength? On 6 January 2013 19:54, Shantanu S. Bhattacharyya Mr <ssbhatta@xxxxxxxxxxxxxx> wrote:
Shantanu