<P> <BR><BR><FONT SIZE=2><B>Paul A Hsieh</B></FONT><BR><FONT SIZE=2>04/10/2000 01:34 AM</FONT><BR><BR> <FONT SIZE=2>To:</FONT> <FONT SIZE=2>vtkusers@public.kitware.com</FONT><BR> <FONT SIZE=2>cc:</FONT> <BR> <FONT SIZE=2>bcc:</FONT> <BR> <FONT SIZE=2>Subject:</FONT> <FONT SIZE=2>exploring vde2000 dataset 2--fantastic!!!</FONT><BR> <BR>This is a resend of a message I posted last night. Apparently</P><P>it didn't get through. Maybe this is because I attached some</P><P>graphics files, which made the message too long. Below is</P><P>just the message. The graphic files plus the avi's can be</P><P>downloaded via anonymous ftp as described below. -- Paul</P><P><BR></P><P><FONT FACE="Courier New">Over the weekend, I explored the vde2000 dataset 2 using vtk.</FONT><BR><FONT FACE="Courier New">What I had initially thought was an exercise in handling </FONT><BR><FONT FACE="Courier New">a large data set (1.7 gigabytes) turned out to be a fantastic</FONT><BR><FONT FACE="Courier New">data set with beautiful and rich complexities. If you have</FONT><BR><FONT FACE="Courier New">the facilities to download and handle this data set, I would</FONT><BR><FONT FACE="Courier New">highly recommend it. Even if you work with only one of the 127 time</FONT><BR><FONT FACE="Courier New">slices (about 14 Mb each), there is still plenty to explore. </FONT><BR><FONT FACE="Courier New">I think this is a great data set for the vde2000 "Toolkit showdown". </FONT><BR><FONT FACE="Courier New">If other vtk users are visualizing this data set, please post your </FONT><BR><FONT FACE="Courier New">results. I hope we can help Will put up a good showing at vde2000.</FONT><BR><BR><FONT FACE="Courier New">I'll summarize below the initial results of my data exploration.</FONT><BR><FONT FACE="Courier New">I am attaching a few gif files. However, to see the temporal</FONT><BR><FONT FACE="Courier New">evolution of this data set, it is necessary to look</FONT><BR><FONT FACE="Courier New">at some animations. I created a few avi files from sequences</FONT><BR><FONT FACE="Courier New">of images but these are too large for email attachment. If you</FONT><BR><FONT FACE="Courier New">are interested in viewing them, you can download them via</FONT><BR><FONT FACE="Courier New">anonymous ftp from: mprcamnl.wr.usgs.gov. Go to the "pub"</FONT><BR><FONT FACE="Courier New">directory and get the file cdusk.zip.</FONT><BR><BR><FONT FACE="Courier New">Here is the description of the "cdusk" data set provided by</FONT><BR><FONT FACE="Courier New">the vde2000 web page:</FONT><BR><BR><FONT FACE="Courier New"> ----------------------------------</FONT><BR><FONT FACE="Courier New"> Accretion Disk Around a Black Hole</FONT><BR><FONT FACE="Courier New"> ----------------------------------</FONT><BR><FONT FACE="Courier New"> The "cdusk" data is from a simulation of tidal effects on an </FONT><BR><FONT FACE="Courier New">accretion disk around a black hole.</FONT><BR><BR><FONT FACE="Courier New"> Each data file contains 2 variables, density and velocity,</FONT><BR><FONT FACE="Courier New"> on a grid of 150 x 150 x 75 points. The velocity has large</FONT><BR><FONT FACE="Courier New"> values outside of the disk, but since the density is low</FONT><BR><FONT FACE="Courier New"> at those points the high velocity is not interesting. The</FONT><BR><FONT FACE="Courier New"> mass flux flowing into/out of the disk is given by the</FONT><BR><FONT FACE="Courier New"> product of these two variables. The spiral wave pattern</FONT><BR><FONT FACE="Courier New"> induced by the tidal pull of the companion star should</FONT><BR><FONT FACE="Courier New"> be evident in an Z-slice of the velocity.</FONT><BR><BR><FONT FACE="Courier New"> Each data file corresponds to a separate time slice, with</FONT><BR><FONT FACE="Courier New"> all files together making a smooth time sequence showing</FONT><BR><FONT FACE="Courier New"> the evolution of the disk. The 127 HDF data files are </FONT><BR><FONT FACE="Courier New">named in sequence from cduskcD1334.hdf through </FONT><BR><FONT FACE="Courier New">cduskcD1460.hdf, and each one has a file size of</FONT><BR><FONT FACE="Courier New"> 13,502,145 bytes, for a total dataset size of</FONT><BR><FONT FACE="Courier New"> 13,502,145x127 = 1,714,772,415 bytes.</FONT><BR><BR><FONT FACE="Courier New">My first attempt to explore this data set was to create a few</FONT><BR><FONT FACE="Courier New">isosurfaces of the density distribution. These isosurfaces had</FONT><BR><FONT FACE="Courier New">a somewhat wrinkled and wavey look, but they didn't seem that interesting.</FONT><BR><FONT FACE="Courier New">I thought to myself: this is not such a difficult data set to visualize.</FONT><BR><FONT FACE="Courier New">BIG MISTAKE! </FONT><BR><BR><FONT FACE="Courier New">It turns out that a much better starting point is to look at 2D slices </FONT><BR><FONT FACE="Courier New">of this data set (see attachment figure_1.gif). In fact, such an </FONT><BR><FONT FACE="Courier New">illustration was included with the cdusk data set, but I didn't bother to </FONT><BR><FONT FACE="Courier New">study it, thinking it was too "primitive." Another lesson learned.</FONT><BR><BR><FONT FACE="Courier New">If you look at figure_1.gif, you can see on Plane "A" how the </FONT><BR><FONT FACE="Courier New">yellow and red colors are interlaced with each other in a </FONT><BR><FONT FACE="Courier New">"strand-like" or "wispy" manner. This feature can be easily overlooked</FONT><BR><FONT FACE="Courier New">because the lookup table (color bar) is applied over the entire density</FONT><BR><FONT FACE="Courier New">range from -6 to 3.3 (don't know why density can be negative!). There is</FONT><BR><FONT FACE="Courier New">almost not enough color to see the details of the density distribution.</FONT><BR><BR><FONT FACE="Courier New">In figure_2.gif, which is a head-on view of Plane "A", the lookup table </FONT><BR><FONT FACE="Courier New">is applied to a narrower density range, from 1.2 to 2.2. In other words, </FONT><BR><FONT FACE="Courier New">locations where density is less than 1.2 is clamped to blue; locations </FONT><BR><FONT FACE="Courier New">where density is greater than 2.2 is clamped to red). One can now clearly </FONT><BR><FONT FACE="Courier New">see the spiral pattern. The picture has a "fractal" look, and is reminiscent</FONT><BR><FONT FACE="Courier New">of what happens when paint of two different colors are stirred together. </FONT><BR><FONT FACE="Courier New">Prior to complete mixing, strands of one color intermingle with strands </FONT><BR><FONT FACE="Courier New">of another color. If you view the time evolution of density on this plane </FONT><BR><FONT FACE="Courier New">(see the file figure_2.avi from the cdusk.zip download), you can see that </FONT><BR><FONT FACE="Courier New">the red (higher density) is trying to spiral outward while the yellow </FONT><BR><FONT FACE="Courier New">(lower density) is trying to spiral inward. Note, however, that the spiral </FONT><BR><FONT FACE="Courier New">patterns are three-dimensional and actually go out of the plane.</FONT><BR><BR><FONT FACE="Courier New">I found it very difficult to cleanly capture these 3D spiral patterns </FONT><BR><FONT FACE="Courier New">using isosurfaces. Figure_3.gif is such an attempt that doesn't do an </FONT><BR><FONT FACE="Courier New">adequate job. Here the lookup table spans the full density range from </FONT><BR><FONT FACE="Courier New">blue = -6 to red = 3.3. The orange isosurface represents density = 2.1. </FONT><BR><FONT FACE="Courier New">The cyan isosurface represents density = -3.5. (The top of the cyan </FONT><BR><FONT FACE="Courier New">isosurface is "capped" by a piece that is clipped from the top of the grid.) </FONT><BR><FONT FACE="Courier New">One can see the 2 main features of this dataset:</FONT><BR><FONT FACE="Courier New">(1) an hour-glass shaped region (cyan/blue) of "low" density</FONT><BR><FONT FACE="Courier New">(2) an disk shaped region (orange) of "high density"</FONT><BR><FONT FACE="Courier New">Although one can see the forest, one cannot see the trees! </FONT><BR><FONT FACE="Courier New">The avi file figure_3.avi shows the time evolution of these 2 isosurfaces. </FONT><BR><BR><FONT FACE="Courier New">The nested structure of the inward and outward spiral is what makes</FONT><BR><FONT FACE="Courier New">this dataset so complicated and interesting. This can be illustrated by</FONT><BR><FONT FACE="Courier New">looking at the velocity distribution. The dataset contains only the</FONT><BR><FONT FACE="Courier New">radial component of velocity, and not the complete 3-component vector.</FONT><BR><FONT FACE="Courier New">Thus, one can only see regions where the radial velocity is positive </FONT><BR><FONT FACE="Courier New">(which I assume to mean outward spiral) versus regions where the </FONT><BR><FONT FACE="Courier New">radial velocity is negative (which I assume to mean inward spiral). </FONT><BR><FONT FACE="Courier New">Figure_4.gif is an attempt to show this on 2 cuts. Here, red represents </FONT><BR><FONT FACE="Courier New">regions where the radial velocity spirals outward (positive), and blue </FONT><BR><FONT FACE="Courier New">represents regions where the radial velocity spirals inward (negative).</FONT><BR><BR><FONT FACE="Courier New">In figure_5.gif, I try to show the magnitude of the radial velocity</FONT><BR><FONT FACE="Courier New">separately for positive direction (left figure) and negative directions</FONT><BR><FONT FACE="Courier New">(right figure). Finally, in figure_6.gif, I try to show the 3-D regions </FONT><BR><FONT FACE="Courier New">in the lower half of the grid where the radial velocity is positive (left </FONT><BR><FONT FACE="Courier New">figure) or negative (right figure). The idea is that if you put both </FONT><BR><FONT FACE="Courier New">regions back together, they'll fit neatly to form the entire lower half </FONT><BR><FONT FACE="Courier New">of the grid. It is really interesting how the outward (positive) and inward </FONT><BR><FONT FACE="Courier New">(negative) spirals are arranged with each other. (By the way, if the full </FONT><BR><FONT FACE="Courier New">vector field is available, this would make a neat test of streamline </FONT><BR><FONT FACE="Courier New">computation. Poor extrapolation might overshoot from the inward spiral</FONT><BR><FONT FACE="Courier New">to the outward spiral, reversing the direction of the streamline!) </FONT><BR><FONT FACE="Courier New">I have also made a side-by-side animation of the temporal development </FONT><BR><FONT FACE="Courier New">of the two velocity regions. This is figure_6.avi.</FONT><BR><BR><FONT FACE="Courier New">Well, that's about it. If anyone has neat ideas of how to visualize this</FONT><BR><FONT FACE="Courier New">data set, please share them. (I briefly tried volume visualization but</FONT><BR><FONT FACE="Courier New">didn't get very far.) Finally, thanks again to the vtk authors</FONT><BR><FONT FACE="Courier New">and all the dedicated folks at Kitware. Having used vtk for several</FONT><BR><FONT FACE="Courier New">years, I am no less impressed by its capabilities today than the</FONT><BR><FONT FACE="Courier New">first time I saw that spinning cone on my computer screen.</FONT><BR><BR><FONT FACE="Courier New">Paul Hsieh</FONT><BR><BR></P>
--------------------------------------------------------------------
This is the private VTK discussion list. Please keep messages on-topic.
Check the FAQ at: <http://public.kitware.com/cgi-bin/vtkfaq>
To UNSUBSCRIBE, send message body containing "unsubscribe vtkusers" to
<majordomo@public.kitware.com>. For help, send message body containing
"info vtkusers" to the same address.
--------------------------------------------------------------------