Received: from watserv1.uwaterloo.ca (watserv1.waterloo.edu) by karazm.math.UH.EDU with SMTP id AA29130 (5.65c/IDA-1.4.4 for ); Wed, 23 Oct 1991 19:21:12 -0500 Received: by watserv1.uwaterloo.ca id ; Wed, 23 Oct 91 20:16:58 -0400 Date: Wed, 23 Oct 91 20:16:58 -0400 From: Dave Stampe-Psy+Eng Message-Id: <9110240016.AA08943@watserv1.uwaterloo.ca> To: glove-list@karazm.math.uh.edu Dave Wargo (dwargo@cs.ucsd.EDU) replies: >> >>I would be interested to hear descriptions of any homebrew eyephone >>projects that have been completed. What diplays are you using? >>Monochrome or color? Binocular or monocular? Field of view? >>Optics? Are you monitoring the user's head positon? Weight? >>I'm trying to get an idea of the level of sophistication that >>is currently being used, for use in future postings. >> >>Please post to the list, as I'm sure others care too! >> >>- Dave Stampe >I am in the process of trying to put somthing together. I have tryed at >first to use a couple of lcd tvs from radio shack and they just would not >work for what I wanted to do. > >Right now we are trying a couple of sharp displays. I think the real trick >is going to be the optics. This is not a trival task and I would be >interested in hearing what can be done short of spending $800.00 for a pair >of LEEP optics. > >Please pass along what you find out. > >Do you think that this could be the begining of a mail list for >eyephone-list? > >Thanks > >Dave Wargo > What model of the Sharp displays are you using? I recently did a (non-VR) project using the 5" color NTSC/RGB displays (forget the model). What did you pay? I can't seem to find smaller ones for less than $800 (maybe it's my supplier though...) and the really decent 5" models are over $1000. From what I've seem recently, a lot of the newer pocket TV LCD display have crummy resolution-- it's cheaper for them to digitally preprocess the image than to use a decent LCD panel, I guess. I guess the LEEP optics cost *is* dropping-- last time I talked to Eric Howlett (designer of LEEP systems) they were still $3500 each. Guess he decided the big markup on 6 cheap plastic lenses and a case wasn't helping sales (B_{) or maybe the production volume went up. From what I know of the LEEP optics, though, it would be difficult to add LCD panels, as the display's active areas must actually TOUCH at the center to get a full display. The monchrome LEEP units that I worked with actually had part of the PCB cut away on each display, and bridged with 40 or 50 wires, and the right panel was flipped over. The original NASA design used one 640x240 LCD panel to drive both eyes (not a bad idea, if you can find one that has the right dimensions). As for optics, I think that we should try to stick with single-lens systems. More lenses will either add distortion and chromatic abberation or will reduce the field of view. The LEEP system actually USES the distortion and chromatic effects, and compensates for them at the camera side. The way to get a big FOV is to bring the displays close to the eye. But then the user can't focus properly, so a converging lens must be added. I've checked out these types: - Biconvex: Forget it , the distortion is TOO GREAT (unless you want it...) - Planoconvex: Less distortion. Fits nicely against glasses. - Meniscus: Least distortion, but can interact with glasses. - Aspherical prescription lenses: These have NO distortion over a fairly large FOV. I think +10 or +12 Diopter is the strongest available with a decent FOV. There is some interaction with glasses. If you can get them, try out cataract surgery lenses. Prescription lenses are special, as the back surface is ground to work properly with the eye "rolling" behind them. A rig that Telepresence Systems in Vancouver put together abuot 3 years ago (no they're not currently doing much on VR, concentrating instead on 3D video for telerobotics) used a pair of +10D glasses and color LCD panels from handheld TVs. There was a lot of other problems with the system, but the optics worked (if you didn't need strong glasses). Some adjustment was provided in the spacing between the displays and the glasses, for about +/- 4 diopters for those who needed prescripton eyewear. It is probably better to mount the lenses on a frame than to use glasses, because of the convenience factor. Also, if the displays (mounted on the headband) move with relation to the glasses on the head, it causes a LOT of display shifting. Some of the wholesalers for prescription eyewear stock preground plastic lens blanks and may be willing to cut them to the desired shape. I got a set of +10D blanks cut at Imperial Optical in Toronto for $35 each (or was it for the set? Lost the invoice). Distortion may not be too great a consideration-- UNLESS you're using 2 displays for stereo! Then the distortions cause a reverse-spherical distortion in depth except at the center of the display, and loss of stereo depth toward the display corners, due to vertical misalignment. Look at a sheet of grid paper thru the lenses you propose to use to judge the amount of distortion. To reduce the effect of distortion, set up the displays and the 3D software for a 1 m convergence distance, as this will minimuze disparity between the displays for a stereo range of 40 cm to infinity. Another consideration with stereo displays is how to make the 2 displays overlap in the visual field. The average distance between a person's pupils is about 6 cm, so unless you have small displays you can't just hang them in front of the eyes-- they'll be too far apart. One way to fix this is to tilt the displays away from each other and use prisms to "bend" the images back in front of the eyes. This can also be achieved by offseting the lenses (Brewster stereoscope). Problem is, this causes a WEIRD warping of the picture (vertical lines are bent), more color problems and blurring, and greater distortion at the display edges. The BEST method is to mount the displays SIDEWAYS and use 45 degree mirrors to put the image in front of the eyes. The picture gets flipped left-to-right, though, but this isn't a problem if you can define new fonts on the computer. If you're using CRT-type displays, just exchange the leads to the horizontal deflection yoke in the display, and everything's back to normal. Interestingly, this limits you to 54 degrees of horizontal FOV, because the display must be as far from the lens as it is wide (draw a picture and you'll see). If you're REALLY picky, use front-surface mirrors to prevent double reflections of bright lines on dark backgrounds. Front surface mirrors are pretty heavy, though, as they are usually 3/8" thick. Edmund Scientific (Efton in Canada) stocks then in various sizes. If you want lighter weight or cheaper mirrors, just get standard mirror glass. You might also try flipping the LCD panels over, but that is difficult with color LCD modules, because the electronics are then on the wrong side. Also, the vertical view angle of the panels then gets flipped, causing contrast mismatches between the left and right eye displays. Now, how to decide on lens powers, display distances, etc? The formula for lens diopter is 1/D = 1/F - 1/d; where D is the lens diopter required, F is the lens-to-display spacing, and d is the desired "apparant" distance (I recommend 0.5 to 1.0 meters for comfort). All distances are in meters. Be aware that distortion increases with lens power! Nondistorting prescription glasses have smaller FOV's with increasing power, too. For FOV, use tan(FOV/2) = s/d/2, where s is the pertinant dimension of the display, and d is the lens-to-display distance. Most display sizes are given diagonally, with hsize 4/5 * diagonal, and vsize = 3/5 diagonal. A factor of 1.1 or 1.2 increase in effective FOV _MAY_ happen because of interactions between strong lenses and eye movements. There are some subtle problems with LCD displays. First, the vertical viewing angle range of these things is less than +/- 15 degrees, so if you have a vertical FOV of 30 degrees or greater, the top and bottom of the scene will be washed out or darkened. This is why the new color LEEP optics use a diffuser plate: the viewing angle increases at the cost of resolution. The other problem with LCD color displays is the visible RGB pixel bands. These may be a problem with wide-angle displays, due to the high magnification. And, in some cases, they cut down the usable resolution. If a display's specs say "288 pixels H resolution, RGB stripe" then you KNOW that there are 96 red pixels, 96 blue... etc: so if you want pure white, your horizontal resolution is 96 pixels! (This type of display gives good apparant resolution for small FOV's, but not for large ones...) What do I recommend for FOV and pixel size? Well... I think that we should try for AT LEAST 54 degrees horizontally and 40 degrees vertically (about 2 1/2 by 3 1/2 feet at arm's length). This is sort of a happy medium between psychophysical effects and the limits of single-lens technology. Pixel size should be LESS THAN 1/3 degree plus a diffuser (actually, 1/10 degree or less is best, but let's not kid ourselves). Using 320x200 resolution with the 54x40 degree FOV, we get 0.2x0.16 degree pixels... hmm, not bad. BTW, these numbers based on the human contrast sensitivity function, which peaks at 10 pixels per degree. This means that bigger pixels will be tend to be seen as _pixels_ rather than a continuous image. A diffuser acts as a spatial lowpass filter, blurring the pixels together. Myself, I am _considering_ using some 4" portable B&W TV's because of their low cast, rlatively large screen size and high resolution. Also, pixel blurring can be done by adjusting the focus. The only problem is the weight-- they would have to be counterweighted, so the inertial load on the head is pretty big. Also no color... Have to think about it some more. DISCLAIMER: Some of the above is based on work with a 35x25 degree system, some is knowledge of the technology and experience with an (older) LEEP system I'm not the most knowledgable person about optics, so maybe someone with more experience can design a cheap 2-lens system with less distortion and the full FOV needed. Myself, I subscribe to the KISS philosopy, but it would be nice to have a higher lens power so smaller displays can be used. -------------------------------------------------------------------------- | My life is Hardware, | | | my destiny is Software, | Dave Stampe | | my CPU is Wetware... | | | Anybody got a SDB I can borrow? | dstamp@watserv1.uwaterloo.ca | __________________________________________________________________________