Comments on: Nanomachines: How the Videos Lie to Scientists

By: Molecular machine animations in the New York Times

Molecular machine animations in the New York Times — Tue, 16 Nov 2010 23:00:08 +0000

[...] Nanomachines: How the Videos Lie to Scientists [...]

By: Accelerating Future » Atomistic Small Bearing: Dynamics Performed in NanoEngineer-1, Visualized in Blender

Fri, 30 Apr 2010 19:36:41 +0000

[...] of the atomistic bearing described by Eric Drexler in Nanosystems. Remember to read Drexler’s article or watch this video to understand, “the rotation-induced speed of the shaft surface is [...]

By: Accelerating Future » Molecular Rotor in Motion

Accelerating Future » Molecular Rotor in Motion — Sat, 21 Nov 2009 19:39:06 +0000

[...] Vladimir Nesov points out in the comments, all videos like this should be taken with a grain of salt — they’re sped up to many times what the actual speed would be. This rotor, if it were [...]

By: Productive Nanosystems: The Ribosome Videos

Productive Nanosystems: The Ribosome Videos — Fri, 17 Jul 2009 16:00:57 +0000

[...] Nanomachines: How the Videos Lie to Scientists [...]

By: Will Ware

Will Ware — Tue, 17 Feb 2009 18:47:40 +0000

No joy with "img" tag, but the image is here.

By: Will Ware

Will Ware — Tue, 17 Feb 2009 18:46:30 +0000

Python code for ellipsoids is here. It turns out the math does look like PCA after all, though simplified a bit. Regrettably Google Groups reformatted my code in spots, hopefully the line foldings are obvious. Assuming I'm allowed to use an HTML "img" tag, an example appears below, with major and minor ellipse axes shown in red and green for a 2D case.

By: Will Ware

Will Ware — Thu, 12 Feb 2009 15:57:06 +0000

Maybe then a ray-tracer is not the best tool to use. It's overkill to worry about getting the shadows in the right places, and whether the reflections and refractions through/around a carbon atom are correct, so a raytracer like POV-Ray really attacks the wrong problems. Way back when I wrote some much simpler code for rendering molecules, and it could easily be modified to handle ellipsoids. It's basically a Z buffer, so transparency would be a bit messy, each pixel requiring (I think) a tree structure of the surfaces, terminating a branch either when it hits something opaque or when any remaining transparency gets too dim to matter. Above, I suggested PCA as a good way to determine positions and orientations for ellipsoids. Having since taken a little time to look at what PCA does, I think that too is probably overkill. Averaging a few products of deviations should suffice to construct the ellipsoid for each atom. I'll give this some thought, and see if I can find time to make some tweaks in my old code.

By: Vladimir Golovin

Vladimir Golovin — Wed, 11 Feb 2009 16:38:53 +0000

Will — yes, hit-testing rays against ellipsoids is very cheap, but unfortunately this will work only in raytracers that explicitly support them (as far as I know, modern renderers like Mental Ray and VRay don’t expose their support of procedural ellipsoids — they must be tessellated into polygons before rendering).

POV-Ray does support procedural spheres, but it cannot render Ambient Occlusion and contours, which are essential to create the Qutemol look that Eric prefers for molecular visualizations.

Also, procedural ellipsoids aren’t supported by GPUs — they also accept polygons only, so if we use a GPU-accelerated renderer ellipsoids won’t help. Yes, there are GPU-based raytracers like the Intel’s one, and they must support procedural ellipsoids, but I’m not sure about their availability (both commercial or open-source).

Regarding transparency: yes, of course POV-Ray does support it, but my hunch is that we won’t gain much compared to the proper motion blur / averaging stroboscopic frames. Also, I expect that transparency will make proper shadowing (let alone Ambient Occlusion) impossible or very difficult.

By the way, just found a command-line tool that can average specified images together (scroll to the bottom):
http://www.imagemagick.org/Usage/layers/