You put your bulb at the center of the spherical reflector. This gives you twice as much point-source(-ish) light headed towards the fresnel and the LCD. Solid plan, seems to be the method of choice...

What if you put the bulb at the focal point instead of the center? The reflector could cover the entire back of your enclosure, as big as the whole LCD. The result would be collimated light from the reflector and point-source light from the filament.

My very approximate 2D napkin math says that the collimated light will be from the rear 160 degree arc of the bulb, while the point light will be from the front 45 degree arc. Extend that to 3D space and I think it means that the collimated light will make up the overwhelming majority(*) of the light reaching the display. Incidentally, the point light will be 25% brighter at the center of the LCD than at the edges, while the collimated light will be only 10% brighter at the center.

If you had some sort of magical two-way-mirror-ish fresnel that could get all the light at the LCD going the same way, this would result in a massive increase in brightness. Assuming you can't do that, I want to know what would happen if you just left out the collimating fresnel. The overwhelimg majority(*) of the light hitting the LCD is collimated by the reflector. That would all reach the projection lens properly. The un-collimated light that came straight from the bulb would be approximately collimated when it reached the front fresnel, which means only a small bit of it would hit the projection lens, and only a small bit of that (the center) going the right direction. I expect you would get a very dull haze covering your whole front wall, or a hot spot, im not sure which.

I am very interested in others' opinions of this idea, or variations thereof. Especially anyone who can comment on mistakes I may have made with my optical assumptions, or improve my math (or lack thereof).

(*) I can't remember enough trig to do the math for how much sphere surface, and thus light output, the mirror would get, or the fresnel gets now, but I know the mirror would get a hell of a lot more.

In the image below the green is a reflector (doesn't have to be spherical, any curved mirror has a focal point, right?) and the projection lens, blue is the LCD and front fresnel (combined for simplicity), red is "bad" light, and yellow is "good" light.

People have experimented with this before, except you need a parabolic reflector to create collimated light, not a spherical. As you mentioned, the majority of the useable light will be reflecting, while a majority of light from the front of the bulb will be trash. The problem is that the bulb itself blocks light from reflecting back off the center of the reflector, leaving a dim spot in the middle of your collimated light cylinder. Every orientation of the bulb has been attempted, and even with some parallel light exiting the front of the arc, the problem always remains.

My horrible artistic skills, inability to draw a half-parabola, and poor representation of reflection and refraction aside, how about this? Not only does it get the bulb out of the way, it puts EVEN MORE of the light to use, maybe as much as 75% of the output of the bulb. It also makes the hot spot (what little is left, this should have less than 1/4 as much hot spotting as a normal design) center on the bottom of the image instead of the middle, which could look a lot better in some situations.

That has a better chance of working. None of the red lines will make it into the projection.. don't know if you were including those in the 75%. I don't really see a benefit from this over a standard setup though unless the angle os useable light is greatly increased. You use a decent amount of the light w/ a spherical reflector and a 220mm fresnel. Theoretically you could gain brightness since there will be one less fresnel blocking light, but overall I think you would see better results with a spherical reflector, fresnel, and pre condenser lens.

With enough fiddling you could tweak this design to use upwards of 80% of the light from the bulb. You would need a parabola with a very low focal point, and the mirror would stretch almost all the way to the LCD on the top side of the projector. A 15" LCD 10" from the filament receives less than 10% of the light. Double that with a reflector, 20%. 20% vs 80% is a staggering difference. Forget making the projector brighter, existing designs are plenty bright as it is. You could get the same output brightness out of bulbs that are cooler, cheaper, easier to maintain, longer life, or any combination thereof.

Hi sparr
Have you got a parabola reflector in mind ?..
It's always been an issue finding a quality parabolic reflector of the right size...

I do not. Finding one this extreme would be quite difficult. It is virtually impossible with glass, it would have to be a metal bowl of some sort. If someone put together some solid plans for this, I could see having a mold and 30-50 aluminum reflectors the appropriate size made. Chrome the inside and polish to your heart's content. I have been trying to find a 3D optics package to experiment with, nothing short of Mathematica seems available

what? 80%??
The best I could calculate with your design was 11.5% with bulb below the lcd (hot) and 8 % with the bulb beside the LCD. This is similar to a standard setup. howver there may be advantages to hotspotting as you suggested plus. increase these numbers by another 1% because you lose the first fres.

Kirk

Even crappier drawing this time, since all I have is MS Paint... You need a very 'deep' parabola, with a focal point close to its vertex, and stretching all the way to the LCD (the closer the better). The parabolic mirror can 'cover' an area of the light output approaching 50% (in physics experiment land, zero-thickness bulb), and a spheric below it doubles that. There is a very slight loss where the half-circle cross section of the parabola intersects the shape of the LCD. I really need to put together a 3D model of this to do the math precisely. In 2D world in the illustration, with probably-not-objectively-placed light rays, 8/10 of the rays hit the LCD collimated, 1/10 hit it at an undesirable angle, and 1/10 are wasted against the back of the box.

going the wrong way dude.
your box just got a lot longer, your hotspot got 10 times worse, and the reflector is 20 times harder to do. Your fisrt crack was the closest. IMHO

Your looking for the most light but also at the same time you want an even "quality" light. With optimization I think you have the potential of 15% good quality light instead of the standard 10% centerloaded light.
A 50% increase in brightness and a more even lighting patteren than the stock version....This sounds good to me! I'll make a mock-up this week and try it.


Kirk