I will now take you, reader, on a journey involving both technology and aesthetics intermingled with projection practices that are a century old and how the ways that the digital cinema revolution impacts this, combined with concerns about stereoscopy. Stick with me and it may all make sense by the end. Along the way I’ll inform you about some engineering choices, and how this has impacted not only digital cinema projection, but stereoscopic projection.
Projecting motion pictures is a century-old art. A film projector uses a band of transparent material coated with picture that has perforations for an indexing function to properly place the film in the projector’s optical path. The film is held at rest in what is called the “gate” while a frame is projected, one at a time, on the screen. A projector light source shines through that transparent material with its image, and the tiny image is projected by a lens onto a screen. The image itself, less than an inch wide, makes it onto a screen that could be 60 feet wide. This is amazing considering that, if done correctly, the results are beautiful despite the enormous magnification.
As the film rides through the gate, which is the mechanical device to hold the film in place, the mechanism that drives the film through the projector must accomplish a lot. Each frame has to be held at rest in the gate, and each successive frame has to be indexed in the same relative position and the plane of the film has to be held perpendicular to the lens axis. The gate consists of two major components: a pressure plate and an aperture plate. They hold the frame in place.
The aperture plate is closer to the lens than the pressure plate. It forms a rectangle around the frame, and aperture plates are often interchangeable to accommodate how much of the frame is used. For example, when projecting motion pictures a larger aperture plate can be used for Panavision or CinemaScope movies, because more of the frame is used. The film and the aperture plate, which is in front of the frame, cannot both be brought into sharp focus on the screen. If you bring the aperture plate into sharp focus, the film is out of focus, and vice versa. There’s no choice; you have to focus on the film, because who cares about a sharp aperture plate?
However, if you focus on the film, the aperture plate–or that which is surrounding the film–is going to be blurry. That is why movie projection is set up so that the image bleeds onto the black rectangle that surrounds the screen. If you don’t do it this way, you’ll have blurry edges for the vertical and horizontal portions of the projected edges, and for aesthetic reasons that would is deemed to be unpleasing. Therefore, some of the image is wasted. It is sacrificed in order to make a nice, crisp, sharp surround.
Here is another thing to think about with regard to motion picture projection. Projectors in most theaters are usually higher than the screen, angled downward. What this means is that, rather than a rectangle frame being projected onto the screen, a trapezoidal frame is projected onto the screen. It’s typically not a big deal, even if the angle is rather steep, because the material on the edges of the screen can get lost or cropped as described above. Usually the distortion isn’t significant because of the nature of the subject matter being projected. If the plane of the screen were perpendicular to the lens axis there wouldn’t be any trapezoidal distortion. But you can’t do that, because in order to have the best illumination, even from a matte screen, you want to tip the screen in the direction of the audience. The theatrical film industry has lived with losing a little bit of picture due to the difference in the distance between the aperture plate and the frame, and also with losing some image because of the trapezoidal distortion effect.
When setting up digital projectors, you don’t have the same concerns. There is no aperture plate in a DMD digital projector. So there is no reason that you should have to lose a single pixel, because there is no conflict between focusing on the aperture plate (which is nonexistent) and the imaging engine.
What Texas Instruments did in the design of their digital projectors was to specify what I call a perspective control lens–because that’s what they were called when used in cameras. Lenses like this have been used in road warrior and other projectors. I remember them being used in slide projectors years ago. By mechanically moving the lens so that the lens axis remains perpendicular to the plane of the screen but shifts either vertically or horizontally, you can shift the image upward, downward, or to the side without inducing trapezoidal distortion. This means you need to have a lens that “covers” more than it otherwise would. A normal lens designed just to cover the size of the frame or the image engine will have vignetting; that is to say, the corners or edges of the projected image will get dark. Designing a lens that has more coverage makes for a more difficult and expensive lens design, but it has benefits because trapezoidal distortion can be eliminated.
When digital projectors are installed, a couple of interesting things happen. In the first place, when the digital projector is added to a facility that has an existing film projector and the film projector remains in place, the digital projector is now placed off-axis. The film projector has the preferred on-axis location, so the digital projector is going to be off to one side. Coupled with the fact that most projectors are going to look downward at the screen, you can see that you will get even more trapezoidal distortion from digital projectors just because of the facts of life in projection booths.
Since Texas Instruments has specified perspective control lenses in their DMD projectors (these lenses are made mostly by Konica-Minolta and some by Schneider), you would expect that you wouldn’t have any trapezoidal distortion in projection. The preferred method for setting up a digital projector with such a lens is to maintain the lens axis perpendicular to the plane of the screen. If that condition is fulfilled, then the horizontal and vertical shifting of the lens will result in an image that fills the screen, with no trapezoidal distortion. As long as the effective plane of the image engine is parallel to the plane of the screen, which can be achieved if the projector is not tilted, there will be no trapezoidal distortion. If you set everything up correctly there is no reason why you need to lose a single pixel, because you don’t need to bleed the projected image onto the black surround to counter aperture plate focus or trapezoidal distortion effects.
Unfortunately, this is not the way digital projectors are set up. I don’t have the figures, and I haven’t visited every theater but I can tell you that in almost every projection booth I have been in, and in every theater I know about (which is admittedly a small sample when you consider that there are five thousand digital projectors out there), I have not seen it done correctly. The installers are angling the projector downward like a film projector, and they are bleeding the image over onto the screen surround. That’s because the guys who set up the projectors know how to set up film projectors and don’t seem to know how to set up digital projectors. This means that we are losing pixels that we don’t need to lose. This means that we are getting trapezoidal distortion that we don’t need to get.
What does any of this have to do with the stereoscopic cinema? Here’s what: For the floating window that is so effective in increasing the parallax budget or the depth of projected stereoscopic images, trapezoidal distortion will spoil the shape and the intention of the floating window. Next, and more important, is that if you crop the image by bleeding it over onto the screen surround (following convention film practice) you will be cropping off the floating windows and totally destroying the effect. The practices in the field of setting up digital projectors have a material impact on the stereoscopic cinema and it is not a good impact on the planar cinema, for that matter. It’s just foolish not to take advantage of an intrinsic advantage of the digital projector design by losing pixels and by adding distortion where you don’t need to. It is destructive to louse up the projection of a stereoscopic image by ruining the floating window effect that is so incredibly effective.
My feeling is that, Texas Instruments, you have a lot of smart guys there, but you made a practical mistake; but I think I would have made the same mistake (I am a smart guy too). By increasing the complexity of the projection lenses, which already have the complexity of zoom capability (with a small range for tweaking the magnification), and by having to use a relay design because the DMD engine is buried so deep in the projector you got one wickedly complex pile of glass. And you’ve added more complexity to the design by adding a perspective control or lens shifting design, which is what I am concerned with here. This last requirement must have increased the cost of goods of the projection lens. And as I have stated with such vehemence, the feature is generally not being used.
Is there a cure? The best cure is to educate the people who set up the extant projectors. A another way might be to use a software distortion correction, so that the projectionist could use the same old practice of angling the projector downward at the screen. Then using the kind of correction available in Photoshop (for example) with “Transform,” he or she would dial in a rectangle that would actually meet the edges of the screen surround. That way one could “pre-distort” the image to eliminate any trapezoidal distortion on projection.
There are a couple of arguments against this idea. One is the computational power that would be required on a frame-by-frame basis. I am not enough of an expert to be able to tell you about this. My gut feeling is that it is not a big deal, and would cost a lot less than what has been required to add extra optical capability to the projection lenses. For all I know, the existing projectors with their little built-in computers can handle the job, or maybe the next gen can.
The next argument against my idea is that such corrective pre-distortion is going to move us away from square pixels and change the effective resolution on the screen on a point-by-point basis. People who are purists might object to this, because we are changing the image magnification and pixel density across the plane of the screen in order to correct the trapezoidal distortion. But what’s the big deal? I don’t think anybody could ever see the difference. You’d have a system that would result in an image that had no distortion; I think you would be able to lower the cost of the lens; and you would have something that would work a lot better given the existing practices of projectionists. Plus, the benefits for stereoscopic projection and the floating window would be enormous.