TITLE: The Transistors are Coming
NAME: Richard Sutherland
COUNTRY: US
EMAIL: rich@brickbots.com
WEBPAGE: http://supermegamulti.com
TOPIC: Great Inventions
COPYRIGHT: I SUBMIT TO THE STANDARD RAYTRACING COMPETITION COPYRIGHT.
JPGFILE: rpaz0404.jpg
RENDERER USED: 
    Mental Ray

TOOLS USED: 
    Maya 5, Mental Ray, Photoshop

RENDER TIME: 
    11.5 hours

HARDWARE USED: 
    P4 2.4ghz 2gb ram

IMAGE DESCRIPTION: 

The first Transistor, surrounded by its offspring through the years.  On the
left hand side of the image is a model of the original point contact transistor
made by the folks at Bell Labs.  It is surprisingly small prototype at less
than 10cm high.  Point contact transistors were quickly replaced by field
effect transistors, which were more practical to produce and required no
adjustment, so most of the 'offspring' pictured are of the field effect type. 
However, the point contact was the first of the new generation of solid state
switching devices and has earned well deserved recognition as the first
transistor.  

Several historic transistors are pictured, along with a few IC's to show the
rapid advance since the transistor was invented.  The original Pentium CPU in
the mid ground/right contained over 3,000,000 transistors, not to mention the
thousands of other solid state components to support them, and takes up
significantly less volume the the first transistor.  


DESCRIPTION OF HOW THIS IMAGE WAS CREATED: 


MODELING/TEXTURING:

I started by replicating the transistor prototype, which took the lions share of
the creation time.  I figured that such a historic object would be well
documented but it was difficult to find good measurements and pictures of the
device from various angles.  Ultimately I worked out about how large the lexan
holder (the transparent C shaped bit) was and tried to deduce the scale of
everything else from pictures I could find.  Each drop of solder and wire was
modeled individually, some with polygons, others with Nurbs surfaces.  

The holder was especially challenging. The modeling was straightforward, but the
materials took a lot of time.  I think they cut the holder from a thick piece
of lexan.  The front and back are polished and pretty transparent, the sides,
edges, and drilled holes that were cut are scratched and scatter light making
it much less transparent.  I developed several different materials and applied
them to the various faces to get the look right.  The front and back clear
portions have a hand painted transparency and texture map to duplicate some of
the abuse the holder took during its creation.  The edges looked chipped and
the section where they attached the front and back electrodes were quite
scratched.  All the materials for the object are refractive (1.6 IOR) and had
to be raytraced.  The scratched portions have some translucence with a very
small depth and focus, and a special back light to help the scattered light
effect.

I modeled all of the rest of the transistors one at a time either from examples
I found or bought or from pictures of the historic types.  All the texture
maps, except the Pentium label, were created by hand and were as true to the
originals as I could manage.  For some I created complex shading networks,
others were pretty simple.  I threw in a generic IC with my mark on it and
IRTC0404 (far right), which ended up a bit too blurry to see.  I had an old
Pentium chip lying around and have always thought it was aesthetically
pleasing, so I included that as the last element.  For the texture map I just
placed it on the ol' scanner and touched up the results for use.

The set is a nurbs sphere flattened at the bottom.  I used the global
illumination and final gather features of Mental Ray, so the set had to be
enclosed.  It has a material that is flat grey for the lower half of the sphere
(the portion flattened out that everything sits on) which fades gradually to a
slight fractal texture with a bluish tinge towards the top of the sphere.

LIGHTING:
I tried a lot of lighting setups for this one.  Ultimately I ended up with a
pretty photographic type setup due to the physical lighting simulation of
Mental Ray.
-The key light has raytraced soft shadow and is slightly yellow to offset the
blue-grey of the set, all the other lights are pure white. 
-There is a back light that only affects the scratched lexan.  
-A light to create the bright spot on the back of the dome which also cast
photons (10,000) for the global illumination.
-A light pointed up at the top of the dome to cast photons (10,000) for the GI
to reflect light back onto the scene for ambient illumination.

RENDERING:
I rendered this at 1280x720 (US HDTV aspect ratio) with these settings as far as
I can remember:

Max Reflected Rays: 5
Max Refracted Rays: 5
Max total rays: 8
-This was probably a bit overkill...

Global Illumination on: 99 accuracy, 0 (auto-calc) for radius
Final Gather on: 200 rays, min and max radius 0 (auto-calc).
-Also again a bit overkill, most of my test renders used 10 final gather photons
and worked reasonably.  I kept upping the QTY to get finer and finer detail in
some of the FG/GI shadows.  Especially around the pins of the Pentium model. 
100 seemed to work fine, but for a final render I just doubled it to 200. 
NOTE: If you use both GI and FG in mental ray and have a larger number of FG
Rays, turn on the option for Photon Precaching in the FG section. It takes
longer for the GI step, but the FG is much faster.

Samples: min 2, max 7, gauss filter size 3
-The min samples are key for the DOF.  If you leave it at the default for
production (0 min, 2 max) the DOF blurred parts will look very spotty.  You can
still see a bit of this in the further objects.  Setting the min to a higher
value forces more samples for the blurred areas (well ALL areas).  3 would have
been better, but going from 1 to 2 took my render time from 3 hours to 11
hours.  I shudder to think what 3 would do, but I bet it would look real nice.

POST PROCESSING:
I then adjusted brightness and contrast a bit in Photoshop and saved it out
again.  I created a 1280x720 noise image in photoshop (white background,
filter->add noise, 100% uniform) to be used for film grain and saved that out
as well.  I then took both back into Maya to composited them together on a self
illuminating plane with the layered shader.  I could have done this in seconds
in Photoshop but wanted to stick as closely to the rules as I could. 
Ultimately, the effect is just about identical with a lot more work.  Hmmmm...
was this really the intention of the rules?  Would anyone have objected if I
just did it in Photoshop?  Comments?  I find that some fake 'film grain' or
noise adds some depth to an image.  It is very, very subtle (so much so that
the JPEG encoding might loose it), but it helps with the photo-real illusion.
Basically, renderers are just a little too perfect :-)
Finally I added the top black bar with the text in Photoshop.  

