I have an idea that I think might improve uniformity and I thought I'd run it by you guys. As far as I can tell, the main problem with uniformity is that the T15 arc is too big to fit in the condenser's focal point. Towards the edges of the condenser, light generated outside of the focal point (at the perimeter of the arc) is bent farther from normal to the collector (less "columnated") than than the light generated at the arc's epicenter -- this light is mostly wasted. At the center of the condenser, though, most of the light is already fairly close to columnated (even with no condenser at all), so less light is wasted. The amount of wasted light, therefore, increases linearly as the distance from the center of the condenser increases. Assuming this is true, then perhaps the effect could be mitigated by somehow intentionally bowing the condenser fresnel, with the bulge on the groove-side. It's kinda hard to explain, but if you think of the arc as a spherical surface rather than a point, then bowing the fresnel inward would allow it to better conform to that surface. Theoretically, light reaching the edge of the condenser would become more columnated, while light at the center would become less columnated. The net effect would be a more uniform screen, with the same overall brightness. Anyway, I don't know if I explained that well, or if I'm just talking nonsense. Either way, I'm interested in your opinions.

PS, I'd like to emphasize that I'm talking about bowing the condenser with the groove-side out. The fresnel has a natural tendency to bow on the smooth-side (in the absence of appropriate counter measures). I'm talking about somehow bowing the lens in the "opposite" direction. In other words, if my theory is true, then this is not something that would've been discovered already by accident.

Well as you've stated the fresnel works by taking a point light source hitting the center of the fresnel and using the fresnel grooves to spread that light evenly across the entire fresnel surface. But as a result you lose light the farther you get from the center.

So it's not really about curving the surface to get light to hit better on the outside edges because light is not being used that hits the outside edges. The only light being used is the light that hits the center of the fresnel. And by curving the fresnel you are actually wasting light because then it is not going to be going parallel through the Lcd but at an angle thus not being collimated properly.

But it would be easy for you to try and post results.

The light doesn't hit the center then get spread outward (this is somewhat closer to how an LCD backlight works). Rather each concentric ring takes light coming from the direction of the point source, and refracts it so that it becomes normal to the fresnel surface. By design, the outer fresnel rings refract light at much sharper angles than those in the center, and thus the edges are less accommodating of light generated outside the focal point. But yes, I should be able to do the experiment. The problem is introducing a spherical bow into the fresnel. I could make a cylindrical bow quite easily, but it needs to be spherical to fit my theory. I'd need to heat it up and wrap it around something very large and round. hmmm....

To go with your theory it would need to be very large and round, you almost would not see the bowing...

Another thing to consider is that even if it does work, there's the matter of keeping the fresnels bowed. I can't even think of how this might be done. I'm going to try to set something up this weekend. I think, initially, I'll just do a simple cylindrical bowing, and measure the difference in light uniformity along the center horizontal (which should be geometrically identical to that of a spherical bowing). If there's an improvement, I'll know I'm on to something; if not, then I'll know my theory is flawed.

I tried this, and the image didn't appear significantly different. I realize now that I wasn't considering the inverse square law and fresnel line density; both of which significantly degrade fresnel performance towards the perimeter. The inverse square law is the most significant of the two, and the the best way to fix it is by using a longer focal length collector/reflector. Obviously, this implies an even longer box (which isn't practical), so the next best thing is probably a precondenser.

Still, the experience wasn't a complete waste. I found that putting a piece of electrical tape in the center of my 18"fl Beseler reduces vignetting by making the center dimmer. Obviously this isn't ideal, as a bright uneven image is preferable to a dimmer, even one.