Received: from watserv1.uwaterloo.ca (watserv1.waterloo.edu) by karazm.math.UH.EDU with SMTP id AA17225 (5.65c/IDA-1.4.4 for ); Tue, 22 Oct 1991 00:35:01 -0500 Received: by watserv1.uwaterloo.ca id ; Tue, 22 Oct 91 01:30:55 -0400 Date: Tue, 22 Oct 91 01:30:55 -0400 From: Dave Stampe-Psy+Eng Message-Id: <9110220530.AA09571@watserv1.uwaterloo.ca> To: glove-list@karazm.math.uh.edu < I think the mailer barfed this back at me: if you got it already, please ignore...> I am posting this as a word of encouragement to garage VR folk, and as a benchmark to judge prospective equipment against. I apologize in advance if I seem to be disparaging about equipment or performance of software: I am stating numbers as I have them (mostly from the source below, and some from assorted research, technical documents too long misplaced to give references to, and personal experience). Feel free to correct me, but be sure you are looking at the same aspects I am (i.e average case vs. worst/best case). The following is from "Reality Built for Two: A Virtual Reality Tool" in Computer Graphics journal by C. Blanchard et al from VPL. It is a 2 page article, and the 2 most relevant paragraphs are quoted here in full, with some of the other equipment mentioned summarized later on. *>The Eyephone: *> *>The Eyephone consists of 2 color LCD monitors viewed using a LEEP optical *>system. Proprietary diffusion techniques are used to merge the distinct *>pixels od the LCD into a continuous image. and to reduce conflicts between *>depth of field and stereo cues. A high resolution dot pattern is *>superimposed over the image to improve percieved resolution. *> *>Each monitor is driven by the image rendered by one of the Irises. *>(Silicon Graphics 4D-80 GT, 2 per user). The monitors and optics are *>mounted in a soft, counterweighted headdress which has physical and *>psychological advantages over the more rigid and intimidating helmet *>mounted displays. I've had the opportunity to work with the LEEP optical systems. They give a WIDE (90 degrees is possible, 100 degrees is *quoted*) field of view, but distort the image. The distortion is fixed by using an undistorting lens and camera mounted on the IRIS monitor. Now, maybe VPL has gotten around this somehow, but they don't say... The pixels in a LEEP display look about as big as the nail on your pinky at arm's length. (1/2 to 1/3 of a visual degree) The resolution (actually USED area) of the LCD in the Eyephone is 280 by 240 pixels at most. The "proprietary diffusion" seems to be a rather simple diffuser, needed to smear the RGB bands in the LCD panel out, and to prevent contrast band effects having to do with the limited view angle of the LCD panels. The stereo conflict stuff just means that the diffused picture ALWAYS looks out-of-focus. Those dots will probably cause the problem all over again. As for the headband, when I worked with these displays I discarded it because it couldn't hold the displays steady enough, and replaced it with a frame and a pilot's helmet. Worked great, but was a little hard to get on and off. If I'm working with a $18,000 ($24,000 color) set of displays, I don't want them falling off my head! *>System Performance: *> *>Using Silicon Graphics 4D-80 GT's as rendering engines we are currently able *>to render worlds with approximately 1400 simultaniously viewable polygons, *>including a high percentage of many-sided and concave polygons, at *>interactive rates of 10 Hz or greater. The total number of polygons in *>a virtual world may exceed this limit when you break the world into *>visibly discreet chambers limiting the number of polygons you can see at *>one time. Hmm, that poly count is hard to interpret. If we translate it into 2000 quadrilaterals, we probably are not underestimating things. The "simultaneously viewable" phrase either means the *possible* viewable set of objects from any position (i.e the "chambers" mentioned later) or the number of polys not clipped by viewpoint and sent to the Irises to be rendered. I suspect it is the former, so the number of polys actually sent to the Iris for rendering might be 1000 or so. The idea of "chambers" is a primitive way of pruning the rendering input. I suspect that the 1000-poly count is *way* higher than what the renderer in a well-designed 3D video game would have to handle, given the same world to draw. (No solid data on this, but Jez San claims to handle a 20,000 poly world on a 386 PC at a goodly frame rate...). Also, by reducing the viewport size from 80 by 100 degrees down to a garage-VR size viewport (suitable for simple eyephone optics) of 40 by 50 degrees, the area (and number of polys) is reduced by a factor of 4. So, 300 polys MIGHT be a good upper-limit for the renderer. That still means we need good "front-end" software to tell the renderer what to draw. I believe this number is quite achievable on a home system. A SUMMARY OF OTHER EQUIPMENT MENTIONED: The DataGlove, of course. Its problems and frailty has been summarized by others. The world model is updated by a Mac II (sorry, no info on what model) which talks to the Irises by Ethernet. The glove position (palm and tip of index finger) and the Eyephone unit angle and position are returned by a Polhemus Isotrak magnetic tracking system. As I recall, the sampling rate on an Isotrak is 80/(# inputs) so that's about 26 samples/sec, assuming 1 isotrak per user. The Pohemus sensors suffer from some of the same noise and glitch problems as the Powerglove's sensors. They don't have to be pointed at the receiver array, but they suffer from distortion from any nearby metal object (screws, desks, you name it). CONCLUSION In relation to these numbers, achievable results from the garage VR project don't look terribly bad. I think the point is that any system has its drawbacks and tradeoffs, and current high-end VR systems have their share. The difference is that they can throw money at a problem and buy off-the-shelf equipment to save time, whereas the "garage" VR folk have to make do. In the end, I think we will compare well. DISCLAIMER: Consider this a first draft, shown to collegues for criticism and evaluation. There is NOTHING implied about the companies mentioned here: just an evaluation from my POV. Discussion invited. -Dave Stampe