Opened 9 years ago

Closed 8 years ago

#153 closed defect (fixed)

2DSA does not support floating experiments

Reported by: demeler Owned by: gegorbet
Priority: normal Milestone: future
Component: ultrascan3 Version:
Keywords: review Cc:

Description

Floating experiments are experiments where the sedimentation direction is reversed. Those experiments are like others, except they have negative s-values. In the editing routine we already have a checkbox to designate an experiment as floating. While some of the existing routines are not easily suitable for adaptation (dcdt, vhw) without major changes in the algorithm, the 2DSA analysis should be easily adjustable. The only change that needs to be done is a change in the grid limits. The grid limits should cover -10000 < s < +10000, but the region -0.1 < s < +0.1 must be excluded - such low s-values do not cause any significant sedimentation at speeds possible in the XLA, so those signals would simply cross-correlate with the baselines determined in ti and ri noise calculations. Please make the changes in the desktop version and also in the LIMS instances. The LIMS2 instance by the way has faulty logic as well, and needs to be revisited (Dan, please take a look and ask me if anything is unclear).

Attachments (1)

float.png (171.6 KB) - added by demeler 8 years ago.

Download all attachments as: .zip

Change History (11)

comment:1 Changed 9 years ago by bdubbs

  • Milestone changed from 1.0 to future

comment:2 Changed 9 years ago by bdubbs

  • Owner changed from gegorbet, dzollars to gegorbet

comment:3 Changed 9 years ago by gegorbet

  • Status changed from new to assigned

comment:4 Changed 9 years ago by gegorbet

  • Keywords review added

The s value limits have been changes to: -10000 to 10000. This is true in both front- and back-end. Bruce had already added code in the generation of subgrid points to skip any s values in the -1e-14 to 1e14 range. This code formed the utility version of init_solutes() that is now in both FE and BE.

2DSA now needs testing with negative s values, as well as large s values.

Review-ready.

comment:5 Changed 9 years ago by gegorbet

Gary will work on modifications to us_astfem_sim so that it can produce floating experiments. Then Borries and others can test that 2DSA works for that case.

comment:6 Changed 8 years ago by gegorbet

  • Owner changed from gegorbet to demeler
  • Status changed from assigned to new

comment:7 Changed 8 years ago by demeler

  • Owner changed from demeler to gegorbet

I was able to successfully test floating data, which was kindly provided by Andre Musatov. The data wasn't great but was definitely able to prove the point that the analysis works (see attached). However, there are still some issues. For such data (to be found in the uslims3_cauma3 database under my name, called "float") the 3D plot does not automatically pop up, and most of the data are not saved. When attempting to model/display floating data in us_fematch, the program fails. Reports do not collect all images.

Changed 8 years ago by demeler

comment:8 Changed 8 years ago by gegorbet

  • Status changed from new to assigned

Overall, the runs by Borries and my own suggest that 2DSA handles floating data. The failure to pop up 3D data at Save and the limited suite of files saved are characteristics of 2DSA in general, not particular to floating data. Note that 2DSA is not exactly like Fematch in Save characteristics; and never will be. For one thing, there are a whole set of distribution plots in Fematch that are not part of 2DSA. For another, noise plots are handled differently. I think that users should be encouraged to use Fematch to produce reports, so that they are exactly equivalent whether the models come from backend or frontend. But, that is another issue and potentially another ticket.

The take-away from these tests on actual floating data is that there needs to be work done on Fematch to handle this case.

comment:9 Changed 8 years ago by gegorbet

Handling of floating experiments (negative s values) has been added to us_fematch. Output reports and plots now also handle floating models, as of version 1198.

Again review-ready

comment:10 Changed 8 years ago by demeler

  • Resolution set to fixed
  • Status changed from assigned to closed

I reviewed this ticket and am happy to report that everything seems to be functioning as it should! fematch nicely displays the same info as the GUI version of 2DSA, and the report manager picks up the report items correctly.

What does not work is the LIMS3 submission of this dataset as a floating experiment. I will set up a separate ticket for this. This ticket will be closed.
Displaying of floating experiment results from the backend still needs to be tested.

Note: See TracTickets for help on using tickets.