Even though the author of this technique graciously included many long bits to explain this technique, it still goes flying over my head:


Let's start at just the first step in creating the nodal Surface network that results in the effect:

Click image for larger version. 

Name:	rayLength.jpg 
Views:	70 
Size:	221.2 KB 
ID:	134586

As you can see from this screengrab, with the Grid Size at 1mtr, the camera is at z=-3.

As per the video, I've connected the RAY LENGTH nodal output to the Transparency channel of the Surface. Now, I'm assuming that the value for a given spot on the sphere in question will be roughly ~2.5, since the sphere isn't a point, somewhere in that neighborhood, and measured from the Camera. As you can see, values >1 mtr on the Transparency channel result not in "more perfect transparency" (as that's 1 or 100%), but just a white unshaded Surface. (This also implies this technique may have problems if the camera is closer or equal to 1 meter, but we shall see.)

So my question is: why are values greater than 100% on the Transparency channel resulting in some visibility? I realize this is not physics and just a bunch of equations, but..... huh?

(One of the charms of LW is that it doesn't necessarily clamp values to numbers that would strictly make sense: this is A Good Thing, as innovative users have found all sorts of applications for nonsensical values, such as "negative Brightness" for lights.)

Also, am I correct in assuming that Ray Length is measured from the Camera?