Originally Posted by cmjohnson
What you would propose would require an additional imaging optical pathway between the two DMDs and the image geometry requirements would be quite stringent. You'd have to have absolute flatness of field and your total tolerance for keystone, pincushion, linearity, or any geometric parameter would still have to be under 1/2 a pixel width over the total area of the DMDs. That would require multi element optics which will inevitably have some transmission losses, and they will add a certain amount of veiling glare which would probably reduce the available contrast ratio more than might be gained with stacked DMDs via this optical pathway.
And then you have another problem: What happens to your light output if DMD 1 is outputting 10 IRE (10 percent of max brightness) and DMD 2 is also set to 10 IRE? It's division. The net system output is 1 IRE.
Follow the math for just a few points.
DMD 1, 100 IRE, DMD 2, 100 IRE: Output is 100 IRE minus path losses.
DMD 1, 50 IRE, DMD 2, 50 IRE, output is 25 IRE minus path losses.
DMD 1, 20 IRE, DMD 2, 20 IRE, output is 4 IRE.
DMD 1, 10 IRE, DMD 2, 10 IRE, output is 1 IRE.
DMD 1, 1 IRE, DMD 2, 1 IRE, output is .01 IRE.
You have created a decidedly non-linear contrast modulator. In fact it appears to work on a square root law.
I don't think that would be at all useful unless you wanted to crush blacks like nobody has ever crushed them in all of history.
If you were to program DMD 2 to operate on a different contrast mapping principle, things would get quite interesting indeed,
but I'm in no way confident that stacking two DMDs by any practical means will simply provide the benefit of a deeper darker black level without other negative consequences.
Like I said, I don't know if it's possible/practical.
As you have figured out, you certainly couldn't just throw the chips into a machine with typical image processing software. The image processing would need to be designed from the ground up. Then there is the issue of alignment, but if they can manage it with 3 chip dlps, it may very well be possible with 2 in sequence. The bit about contrast loss due to optics is interesting, funny how theoreticals break down when figuring out practical application. While the static contrast would be multiplied by a factor of 1000-1500, ansi/fine detail contrast would probably take a hit.
I would never say it's viable/practical, but in theory it would be one hell of a high contrast performer.
I guess the main point I was trying to make here is Lcos does not represent all digitals. It is the closest in performance characteristics to crt but that is also it's primary failing. Crt's and Lcos are great for a lot of things but pronounced fine details in brighter scenes is definitely not one of them. A lack of contrast in fine details leads to a direct loss of perceived detail/resolution.
Sure when you get right up close to the screen with measuring equipment and a test pattern you can find 1080 lines on a crt, but things get a little murky when you are watching actual content at seating distance.