IVAS Feedback/2.3/Meeting

From Northwestern University Center for Atom-Probe Tomography
Jump to navigation Jump to search

Thursday, Decmber 9th there will be a software meeting in the Bodeen Computer Lab (Tech building, Northwestern University), during which most of these points will be addressed. Please volunteer for feedback by writing your name after the title of the point to bring up.

NUCAPT applications presentations (10 AM-12 PM)

This may take less than two hours. Most of our presentations have been placed under either new IVAS 2.3 features or under upcoming IVAS 3.0 features.

Overview of new activities

Range files (Dieter)

  • RNG export doesn't retain ranges for non-periodic table names
I can't replicate. Dieter will try to. -- Rick 13:25, 20 November 2005 (CST)
  • Updating a range file in an analysis creates duplicate type with the new ranges while retaining the old one as well
See above screenshot. Observe the numerous duplicate ranges listed for the file. Two Th ranges are expanded. The lower one was the original range and the upper one is the new set of ranges. --Rick 13:25, 20 November 2005 (CST)

Current & anticipated software needs

Please see under roadmap for prioritized feature requests.

IVAS 2.3 release overview (12 PM-2 PM)

Note: Our experience and the present documentation may prompt the descriptions in this section to be concise. This discussion will be led by Imago, but certain subtopics are quite related to bugs/feature requests we've formulated, so more info on these appears below the subtopic (with someone from NUCAPT who can lead discussion of our questions/concerns). This may run over 2 hours.

Full complex ion support (N/A)

This does not really need to be discussed.

3D filter (Chantal)

Smoothing

What is the proxigram delocalization parameter doing, exactly? Is it smoothing merely the grid, or is it allowing single atoms to contribute to more than one voxel? Also, what function is used for the smoothing?

The grid delocalization method is similar to the method discussed at NWU in May 2005. The main difference is that a gaussian function is used in both the first pass (creating the grid) and, if applicable, the second pass (smoothing the grid). In creating the grid, a kernel is applied to distribute the ion to various voxels. This kernel is created using the integral of a 3D gaussian function with 2*pi sigma equal to the voxel length scale (labeled "Voxel Size" in IVAS). If smoothing needs to be applied ("Delocalization" greater than "Voxel Size"), another kernel is created and used to redistribute voxel values from the first pass grid to the smoothed, second pass, grid. This kernel used for the second pass represents the integral of a 3D Gaussian function with 2*pi sigma now equal to the value specified under "Delocalization".
The second pass smoothing can be pretty CPU intensive. Runtimes (assuming a fixed "Voxel Size") will scale as a function of ("Delocalization"/"Voxel Size")3. The implementation is multithreaded and will benefit from additional CPUs. Please note that using hyper threading on Intel CPUs can introduce a runtime penalty as you'll find the multiple threads created for the virtual CPUs bottlenecked by the existence of only one floating point unit/CPU.
There will be future conversation about how this differs from the method NU has used via Olof Hellman's ADAM/Apex & the pros and cons of both techniques.

x,y,z independent voxel size

With the grid delocalization, independent x, y and z length scales complicate the implementation of algorithms and the creation of kernels to accommodate the various algorithms. I would like to know if there is a compelling reason to utilize independent length scales in voxels since the introduction of grid smoothing. I was under the impressing that independent length scales were solely used to mimic smoothing...

If IVAS's smoothing was a bit better, we might not miss this feature. Dieter may also have particular examples for why this is compelling. Independent x,y,z voxel sizing did seem useful for visualization of alloys with low partitioning ratios & tips that had features which were strongly directional. -- Rick 13:25, 20 November 2005 (CST)

Interface analysis (Rick)

Be brief....

  • What does it mean to analyze a smoothed isosurface, as if it were the interface?
  • How is this feature being used by others?

Image Capture/Movie Making (Rick)

  • have a strange clipping plane
Clipping Bug in Image Dumps
Screenshot of IVAS (unclipped multirange).PNG
Shows what we see at the time of image dumps.
Screenshot from IVAS (small clip).PNG Screenshot from IVAS (extreme clip).PNG
Note the clipped top and bottom. Only the back corner of the bounding box is visible.
  • sometimes do not work (especially for shots with many spheres).
Usually it fails in a null pointer exception. -- Rick 15:35, 21 November 2005 (CST)
These are probably not video memory issues, and they are present in both the OpenGL and DirectX versions of Java3D. -- Rick 13:25, 20 November 2005 (CST)

Improved POS export (N/A)

Imago will lead this.

Mass linearization (Kevin)

  • When the experiment covers wide voltage range, the calibration isn't working well. The mass vs. pulse (depth) graph shows you that the mass changes as the experiment progress.

License dongles (N/A)

I don't anticipate questions & this isn't of general interest, so perhaps we should exclude this topic.

IVAS 3.0 Roadmap (2 PM-4 PM)

Imago's initial ordering was:

  • Proxigram
  • Spectral Analysis
    • Background Subtraction
    • Peak Deconvolution
    • Auto-ranging/peak identification
  • ROI generalization
  • Radial Distribution Function
  • Error calculations for various analyses
  • Increased 3D hardware support
  • GUI improvements (improve consistency and efficiency)
  • Development API improvements
  • Reconstruction improvements

I've reordered these by our priority, below

Proxigram (Dieter)

Will show .applescript with these parameters & will run it through. Imago's comments: Please develop your requirements for this feature (Minimum resolution, maximum distance, etc). Do you want to calculate and sum against all interfaces/precipitates or is it important to be able to isolate and calculate against one or more specific interfaces/precipitates...

We'd be most comfortable if these parameters were user-adjustable. It should definitely be able to do it as a sum against all interfaces.
Segmentation of the isosurface into separate features & being able to pick one or more of these could be useful. This has a much lower priority, though. We'd also need to discuss what should happen when precipitates are excluded from a proxigram: atoms in/near an excluded precipitate could be either included (in which case they'd seek out the NEXT-closest isosurface) or excluded from the proxigram.

Error calculations for various analyses (Prakash)

  • Counting statistics (show binomial equation, etc.)
  • Assumption of infinite reservoir, independence
  • Estimated is better than nothing

Increased 3D hardware support (Rick)

  • Fixed screenshot/movie bug (not apparently a hardware issue)!!!
  • ActiveX Java3d (This isn't of general interest, and should possibly be dropped)

ROI improvements (Rick)

Segmented Selection

  • Export POS
  • Analyses

Intelligent Selection

Additional ways to make selection:

  • Criteria-based
  • Non-orthogonal cutting
  • Additional shapes

Generalization

  • use the contents of the 1D cylinder, precipitate analysis ellipsoids, and isosurface.

In particular:

  • Cut out a new POS file (or files in the case of segmented export) from any of these (Thanks for the ROI-export of the cylinder!)
  • Run a ladder diagram along the 1D cylinder
By ladder diagram, do you mean a cumulative histogram or a something linked to evaporation order?
Rather than linking it to evaporation order (increasing z in the POS, as the ladder.py currently does), I want to link it to the z of the 1D cylinder. The current workaround is to export the cylinder & use ladder.py on the new POS. But this work-around can be tedious, particularly as you may want to adjust the orientation of the cylinder.
  • Density-analysis, planar spacing,etc.

Rescaling of data (Marsha)

  • x,y,z independent resizing

Radial Distribution Function (N/A)

Development API improvements (Rick)

  • Documentation!!!!

Reconstruction improvements (Rick & Kevin)

  • Allow user to edit default evaporation field (and other parameters) for a reconstruction (e.g. we always use 30 V/nm for Al at 30 K & the default of 19 isn't good)
  • Allow to have the expert panel show by default for the reconstruction step

GUI improvements

Allow user to choose parameters to display the bounding box and scale bars (Rick)

(eg. color and line width of bounding box and position of scale bars so that they are not "backwards" when data set is rotated).

  • Change spacing major, minor ticks
  • Units
  • Font

Aspect vs Aspect Ratio (Keith)

Output window resize->visualization window oddness (Marsha)

Spectral Analysis (Dieter, Keith)

Background Subtraction

Peak Deconvolution

Auto-ranging/peak identification

  • Peak fit

LCC improvements (Dieter? Rick?)

This discussion might be moved to Friday.

  • Make loadlock turbo pump settings only adjustable by an admin.
  • Move display of golden/multi/partial/bad into DaVIS.

No Discussion Needed

System JVM & system Java3D library

We do plan on allowing an option for the Direct3D renderer with the IVAS Windows installer in the next IVAS release.

Please save the choice of OpenGL vs. DirectX, so that whichever was used in the last version is the default for upgrades.

Movies

  • claim to export, but no file is created.
Tom at Imago acknowleged this bug and will send a work-around. -- Rick 13:25, 20 November 2005 (CST)
Patch sent & deployed -- Rick 16:39, 1 December 2005 (CST)

Ray tracing

VRML->POV-Ray converters seem to exist.