Projector configuration

Construction on my 180 degree curved projection screen is complete, at least for now – there’s stuff to finish later, but only once I’m happy that everything is working as it should.

Now I’m moving on to placing and configuring the projectors. This will be a relatively lengthy process, so I thought I’d document a few of the pitfalls first, before I really get started.

The projectors for this project are mounted to the rear of the screen using a gantry – image below – which is just a pair of cross-beams mounted at the back of the screen and above the line of the top of the screen. The correct projector placement point is where an imaginary line perpendicular to the centre of each half of the screen crosses the gantry, or thereabouts.

I did my first test of projector placement yesterday, and that showed me that I was in fact mounting it too far back from the screen. My calculations were incorrect. So today I’ve been working out the correct distance via a degree of trial and error. Once I had it, I cut down the gantry beams to the right length and re-mounted the gantry rails. Finally, I mounted one projector and fixed it down so I could begin tests.


Each projector is mounted on a ceiling-mounting bracket which lets me hang it upside-down. The part of the bracket that would usually fix to the ceiling then goes up through the gantry rails and across them, and is screwed down into the rails. This is pretty secure, although the whole structure is able to flex a little and when I bounce around the floor, it does. I can’t really make stuff much more rigid without using much harder wood, or a steel frame. In any event, some movement is inevitable. When I’m seated and playing games, it’s unlikely to be much of an issue.

The projector mount allows for a degree of motion in three axes – tilt in both directions and rotation around the axis of the mount shaft. So fine adjustments can be done here.

I fired up the projector and, having adjusted the focus, I was able to turn on the test pattern. This is a simple grid which shows you the extent of the image and lets you spot any warping.


As you can see from the photo, the grid is very warped, and that is to be expected, because the screen is curved and not flat, while the projector lens is designed to correct light to a flat plane. The reason why it’s not warped at the top, BTW, is that the projector lens is designed to throw light asymmetrically, so that the bottom of the screen is at the level (or just above the level) of the bottom of the projector. Otherwise, you’d have to have the projector mounted directly behind the centre-point of the screen, which is usually not convenient.

This means that the warping is most prominent at the bottom of the image, and indeed as you see the grid spills off the bottom of the screen in the centre of the image, and appears highly curved. This is a result of the curve of the projector screen, though, not the image itself.

One thing you should realise straight away, if you intend to do this yourself, is that you’re not going to get a pixel-perfect image at the end of it. The geometry of the thing means that there will always be a portion of the projector image lost beyond the bottom of the screen, and possible the sides too. You will use screen warping software to correct for this, but after this a pixel in the source image will no longer be a pixel on the projected image – it may be multiple pixels as it’s stretched, or indeed missing as it’s squeezed out of existence. You wouldn’t want to do word-processing on this screen, but it’s fine for games where areas of detail are usually in the front center, where the image is pretty close to unchanged.

Also, to blend the images of multiple projectors you need to have some overlap between them – this is necessary to get things to line up visually without it being obvious that multiple image sources are in use – so the image has to be big enough to spill over the half or third of the screen that each projector covers, and this means you’ll also lose some image off the far sides of the screen.

When I did my test, I found out a couple of interesting things. First, that my screen is not precisely circular; in fact, it’s actually slightly elliptical, and this is a result of some design decisions I made when building the screen frame. It’s only a few degrees out, not enough to make me want to start again, but it does mean that the image is most vertically compressed at the centreline of the screen, so I have to use the zoom to bring the image out so that it fills the screen depth at the centre, and this in turn means I’m getting some further overspill towards the bottom right and left of the screen.

Second, that my ratio calculation is now out by a distance, because of the screen being out of true. So my width x height calculation is wrong, and at the height I need to fill the screen at the centreline, the image is overlapping to each side of the screen-half by about 15%.

The upshot is that, when I’m done, I’ll probably have about 2/3 as many projected pixels on the visible screen half as the source image, and so there will certainly be some visible softness and aliasing. There are a variety of graphics card tricks I can use to fix this.

I’ve avoided using the keystone corrections built in to the projector. These are digital corrections rather than optical, so they work in a similar way to the warping software, by squeezing and stretching pixels from the source to the output; since I’m going to be doing that in software, I don’t want to make it any worse. And I’ve been careful to mount the projector as near to straight and true as I could, so there’s very little keystoning on the actual projected image.

That’s it for now. I need to hang the second projector, line it up alongside the first, and then get with the warping. I’ll post some details of that as I’m doing it. I’d hope to have a first video test done maybe over the weekend… watch out!