Converting Linear Polarized Light to Circular

Linearly Polarized Light (LPL)

The projector uses a "film", which is an active matrix of tiny transparent LCD's in front of a bright bulb to project a digital image.  The computer or s-video input controls the opacity and color of each LCD as it would a pixel on a monitor screen.  With the right lenses in front of the film, you get a bright projection of a digital image.

The VREX has the ability to project two independent images in (1024 x 768 true XGA image resolution) which are interleaved, line by line.  In this mode, we will project two versions of the same scene, where the viewpoint for each is offset slightly.  One image is intended for the right eye and one intended for the left eye to produce a stereoscopic effect.

For example, the left-eye image may be in all the odd scan lines and the right-eye image may be in all the even scan lines of the projected image.  A second film in front of the LCD matrix will linearly polarize each line, meaning it will cut out all light except that which is going in a certain direction or its opposite--along a certain line.  Furthermore, it will polarize all the odd lines in one direction and all the even lines in a direction turned 90 degrees from the odd lines.

In a pair of VREX polarizing glasses, each "lens" is a flat piece of polarizing film.  The same material is used for each "lens", but they are turned to 90 degrees of each other.  The cross formed by the axies of the two lenses is turned 45 degrees to the cross formed by the axies in the projected left and right eye images.  Image a plus sign overlaid onto the letter "X".  I do not really understand it, but this causes the left eye to see only the image intended for it and the right eye, likewise.

Advantages

Facing a single projected stereoscopic image, using LPL, the viewer can turn his head from left-right and up-down with no problem.  For any application this is a good thing, and for many applications the tilt problem is not an issue.  This allows a no-floor cave to have a seamless transition between adjoining screens.

This allows us to make our own stereoscopic glasses with full peripheral vision.  We simply make very large lenses from linear polarizing film that bends horizontally all the way around the subject's eyes, at whatever distance is appropriate.  At any point along this bend, the angle of the bend in the glasses with respect to the cave does not matter.  As long as the vertical orientation of the LPL film remains the same, we can bend it as much as we want horizontally.

When we are doing certain tests with no floor projection, we may prefer to use LPL for this reason.  The VREX projectors and the glasses the come with them provide LPL by default.
 

With linear polarization, the separation of the left and right eye images is dependent on the orientation of the glasses with respect to the projected image.  In the usual setup, the glasses have to be level for optimal separation.  The more you tilt your head (or the glasses), the more of the left eye image bleeds over to what your right eye see, and vice versa.  Up to about ten degrees of tilt, I didn't notice any problem, but any more than that incurs significant ghosting.  At 45 degrees, you see a perfect double image.  At 90 degrees, the left and right eye scenes are swapped, which is quite disturbing.

The tilt problem is a significant problem for us, because many people with balance disorders tend to hang their heads at odd angles.

The Floor Mismatch Problem

Figure one, below, shows the projection screens for the cave in their relative positions, with lines drawn on to indicate the orientation of the scan lines in the projected image.

As you can see, the scan lines in the middle and side panels are vertical, so the polarized glasses will give a good stereoscopic image across all three of them.  The left and right eye images are interleaved, line by line, on the screen and the are both polarized, but at 90 degrees from each other.  The polarized glasses also have their lenses turned 90 degrees from each other.  Both the glasses' lenses and this "lenticular" image are lined up so that each eye sees the scene it's supposed to.

This allows the viewer to rotate his head left and right as seen in figure two, with no problem.

However, the lines on the floor projection are necessarily perpendicular to the lines on the front projection screen, the one usually facing the subject.  We can correct for this by reversing the right and left eye scenes on the floor projection.  The perspective effects of the left and right eye scenes would also have to reversed, in software, or the viewer will see divergent perspective which is disturbing.

So now, the subject can also look up and down, as long as he keeps his head centered, as suggested in figure three.  Guess what happens if he looks down and left?  The image on either side screen will have reversed left/right eye imagery with respect to the floor projection, and with divergent perspective.  The seam between the floor and the side screens will look pretty bad.  The main problem is this:

We can't have the floor image aligned with both the side screens and the front screen.  :-(  We can only have one or the other.

Circularly Polarized Light (CPL)

Generally, circularly polarized light is polarized in some circular or spiral pattern, which can be right-handed direction or the left-handed.  (I really do not know how this works, but I know how to use it.)  If you project two overlapping images, one in left-handed CPL and the other in right-handed CPL, you can view them through a piece of circularly polarizing film and thereby extract one of those images.  So, passive-stereo glasses made with one "lens" being film for left-handed CPL and other for right-handed CPL should separate the right and left eye scenes for properly.

This approach eliminates the tilt problem entirely, and allows the viewer to turn his head in any direction, up to a point, without degredation.  Imagine an axis coming out of the center of the projected CPL double image, and another axis from the center of a given "lens" for one eye in the CPL stereoscopic glasses.  As long as the the view axis is no more than about 45 degrees from parallel to the projection axis, the image is fairly good.  Past that, degredation sets in, seen as ghosting as each (eye) image bleeds over to the other. For me, the illusion of stereopsis does not fall apart until about 65 degrees.

This effect will avoid the Screen Mismatch Problem, which for a floor-projected cave is a show-stopper.  It may cause some fuzziness or ghosting at the corners, as the subject will usually be looking at one wall joining the corner at a different angle than the other, as shown below.

        (insert figure here)

Because the view angles are different, there will be a different amount of ghosting or fuzzing on the two screens, creating a seam of sorts.  However, there is some overlap in the optimal viewing angles of both screens, which allows the viewer to look directly into a corner with no problem.

(insert figure)

We are helped more by the fact that the standing walls of our cave meet at a 99.5 degree angle.

Put very simply, all you have to do is put a clear piece of the quarter-wave retarder film (looks like clear plastic) in front of the projector lens and two more in front of the the lenses on one of those plastic sunglasses from VREX. Make sure that all three bits of plastic are correctly oriented, and looking at a stereographic image.  There is a little more ghosting, but very little light loss.  I really do not understand how this works, and even the working knowledge I have is difficult to explain.  Feel free to peck at my prose.

The Procedure

To get the correct orientation for the retarder film, you will have to experiment by looking at the effects.  Just do this:
 

1. Take a piece of the retarder film and tape it to the projector, so it covers the main lens.  For reference, find an identifiable edge on the origonal sheet, and make sure that the same edge on the piece over the lens is up--parallel to the ceiling.  See figure.
2. Play the VREX demo tape on a VCR, which is hooked up to a projector and look at the image, through one of the regular polarizing glasses by VREX.
3. Does the image look right?  Can you get fusion with the stereographic image you see?  If so, you got lucky the first time.
4. Otherwise, you will have to untape the film over the projector lens, rotate it 45 degrees, in any direction, tape it back down and look again.  You should not have to do this more than three times, before the image looks right.  This is your reference point.
5. Now cut another piece from the sheet of retarder film, as shown in the figure above.  Make sure that part of it has a bit of that same identifiable edge as the piece over he projector lens.  Make sure it is in the same orientation as the film over the projector lens.
6. Turn the piece of film over the projector lens by 90 degrees, and look at the screen.  You will get a stereographic image, and if it looks right you are done.  However,  there is a 50-50 chance that the left and right eye images might be reversed.  This creates an ugly/wierd stereographic image where the left and right eye images have their perspective convergence reversed!
7. If the left and right eye images are reversed, simply rotate the piece of film over the lens by 180 degrees, and it should be fine.


            (insert figure)

Note that the images that come on the VREX demo tape are meant to be viewed far away.  If it's projecting to fill one of the nearly eight foot tall screens in the cave,  you have to stand about ten feet away to easily get fusion.  You also have to hang your head sideways, because the disparity in the images is vertical with the projector on its side!  The disparity needs be horizontal for people who's eyes are side-by-side.  ;-)  This problem goes away if your cave software uses the head tracker to always keep the disparity in line with your eyes.

More details

To better understand what's going on, here is a list of the different effects you can get with the retarder film.
 

1. If a 1/4 retarder film is in the vertical LPL beam, oriented at 0 degrees, the end result is just vertical LPL, again.  No useful effect.
2. If the film is oriented at 45 deg to LPL, the result is left hand circular polarized light (CPL).
3. Film oriented at 90 degrees -> HORIZONTAL LPL
4. Film oriented at 135 deg -> Right hand CPL
5. Film oriented 180 deg -> Vertical LPL


At angles between the 45's, the light becomes elliptical polarized light, which is not pretty.  The odd and even lines are linearly polarized at 90 degrees to each other.  So one piece of 1/4 wave retarder film will, properly oriented, will turn one eye image (left or right) into right handed CPL and the other into left handed CPL.

Edmund Industrial Optics makes 1/4 wave (lambda) retarder film that "can be used to convert linearly polarized light oriented at 45 deg from the direction of the fast/slow axis into circularly polarized light".  Whatever that means.  They sell sheets 14" X 24" for $230.00.  It makes the images darker and blurrier, but hardly enough to notice.  Another detail:  The retarder film is optimized to transmit 560 nm light (green) the best (92% transmission).  For wavelengths above and below this, there is some degradation (though not much) in the amount of transmission.
 

First, we can adjust the virtual environment itself, so that there is usually something extremely bland in the floor/sides interface.  A field of one color on both, near where they join, would look seamless, because each eye sees exactly the same thing.  This might work for something like Tunnel World, a derivative of Box World.  The user is simply  moved back and forth down the tunnel and not having any pattern at the corners might not make much difference to the experiment.  However, in a world where the user is more free to move and look about, this approach won't work, unless you really want to surround the user with some kind of virtual moving box.

Second, we could track the user's head movement and let the side screens "fade to flat" whenever the subject's head tilt gets too extreme to see any depth.  Most of the time the subject will be looking straight ahead, and we don't have stereopsis in our peripheral vision, anyway.  We could even switch the left-right eye views (with perspective correction) if the user is looking down and to the side.  That would create conflict between the front and side screens, but it probably wouldn't be as visible.

Making the bottom of the cave concave or bowed will not solve the problem.  It would only spread the ugly seam between bottom and sides over a larger area.  This sort of "blurring" might make the contrast less jarring or maybe make it worse.  It would also introduce some other problems, like how to focus the top projector.