Michael Frank Deering: VR: The Virtual Portal
The Virtual Portal was (simultaneously with the CAVETM) the world’s first multi-screen projective head-tracked immersive stereo display environment. Initially based on three Sun GT accelerators, several versions of the Virtual Portal were built over the years using newer Sun 3D graphics accelerators and more advanced projector technologies.
I’ve always been interested in virtual reality since first hearing about Ivan Sutherland’s pioneering experiment in 1968. But for many years, the technology wasn’t up to the task. As detailed in the SPAR VR page, while at Schlumberger Palo Alto Research, as soon as the first handheld LCD TV’s became available, I immediately pushed for building a color head mounted display (HMD). Along with others, we built our own HMD, and also experimented with what became known as “fishtank VR” using a normal workstation color display screen in stereo. After arriving at Sun, I soon duplicated these efforts, but felt that the HMD’s were far too low on resolution and high on distortion. As an alternative to HMD’s, for many years I had considered the possibility of building multiple stereo walls, but now I had the technology to do so.
The idea was to build multiple floor to ceiling rear projection walls for stereo display. While some people in the past had considered curved display surfaces in an attempt to make the image seamless, I knew from the view projection equations built into real-time graphics rendering hardware (and most software renderers) that any non-flat display would be mathematically incompatible. So instead I went for stereo walls at 90 degree angles from each other, and trusted my projection math to make the resultant stereo image “seamless” to the viewer. Stewart Filmscreen built a custom three walled rear projection screen assemble, with only a 1/4 inch cable separating the screens. Electrohome eight inch three gun video projectors with special low persistence green phosphors were used to project onto the three walls. Each projector was driven by a separate Sun SPACRStation2 GT (GT) at 960×800 by 108 Hz stereo. All the Sun computers were networked together, with the center one designated as “master”, and the other two following.
Because the location and orientation of the head of only one user controls all the stereo projection equations, I always considered the Virtual Portal as a single user device. (Early on, I described it as “a headmount that wears you”.) Because Sun 3D hardware was so relatively inexpensive (compared to SGI), the total cost of a Virtual Portal was within the range of other high end single user devices. This allowed the total footprint of the implementation to be smaller; specifically the room to the user looked like a 6 foot by 6 foot square eight feet tall. With these vertical oriented projection walls close to the viewer, the lack of a ceiling or floor projector was much less noticeable. Conversely most CAVETM based displays had a user room ten feet by ten feet or larger, and a fourth projector onto the floor was important to correctly immerse the user. Because of the larger size and more expensive (e.g., usually SGI based graphics and computers), CAVEs were usually pitched as multiple simultaneous user devices. My general experience was that more than one user was mainly useful in demo mode, not for actual work. Another factor involved the particular application being run. To simulate the inside of an automobile, the closer screens of a 6 foot x 6 foot room worked better than a larger room, but when a full size mock up of the cab and fenders of earthmoving equipment was to be included inside the display space, a 10 foot x 10 foot room worked better. In any case, the number, position, and physical size of the projection screens that ran the Virtual Portal was fully controllable by a single configuration file read by Java 3D; the end user builder could construct whatever they wanted. (Presumably the same eventually because true for other similar software systems, such as VRJugglerTM). Some systems supported cylindrical or even spherical screens; these generally required more extensive post-warping of the rendered images, as none of the 3D hardware renderers could support these more complex viewing surfaces.
The first prototype of immersive projection displays were built internally at Sun before 1992, but the first external demonstration was on the show floor of SIGGRAPH 1992 in Chicago. As I was negotiating with SIGGRAPH management over show space, I learned that a group from the University of Illinois at Chicago was working on a similar concept called “The CAVETM” also to be displayed on the show floor at SIGGRAPH 1992. Both systems made their public debug there, with lines more than four hours long.
Sun eventually built two portable and three fixed installations of the Virtual Portal, and shared the technology with several customers who built their own implementations. The Sun implementations sequentially used the latest Sun graphics hardware as it became available: GT‘s, ZX‘s, Creator 3D‘s, and Elite 3D‘s. While the first two generations required three sperate networked computers, the later two were housed in a single Sun server. The software evolved as well: the first Virtual Portal was based on custom demo code, the next on HoloSketch, and the final on Java 3D. With the Java 3D implementation, all three (or more) display heads could be run off the same shared display list and window system.
The Virtual Portal is co-credited along with the CAVETM as the world’s first practical immersive projective display environments (technically the Virtual Portal beat the CAVETM to the patent office; there are several issued Virtual Portal patents, but few or none for the CAVETM). Because the Virtual Portal was built inside Sun as a demonstration of Sun’s 3D graphics capability, and was not considered a mass sales product, Sun did not do very much to publicize or push for customers to build their own versions. Because the CAVE and its follow-ons were built in university research environments, where there was considerable pressure to get even a few replications with industry, there tended to be a lot more publicity and additional implementations built. (Also several companies did commercial implementations, though generally as one-off custom products.) While immersive projection environments have proven superior to HMD’s in a verity of ways, the same general limitations of cost, resolution, tracker technologies, graphics generation, and supported software applications has precluded mainstream adoptions of each. This situation will likely continue to apply until these base technological limitations have been rectified.
This is the main technical publication on the Virtual Portal. Michael F. Deering “Making Virtual Reality more Real, Experience with the Virtual Portal”, in Proc. Graphics Interface ’93.
This is a flyer used for the 1993 version of the Virtual Portal using ZX graphics. (This image still shows the old GT towers by mistake.)
Below are links to some images of the Virtual Portal. (Eventually some image sequences shot inside the Virtual Portal will be added.)