aprile@ghost.unimi.it (Walter Aprile) (05/15/91)
I hear you have some available info on polarized light 3D systems. I'd
be really grateful to you if you let me knoz something about this.
Enclosed is the posting I planned to put on nn.
Thanks, Walter A. Aprile
p.s. we have ftp.buckland@ucs.ubc.ca (Tony Buckland) (05/16/91)
In article <1991May15.232536.27134@milton.u.washington.edu> aprile@ghost. unimi.it (Walter Aprile) writes: >I hear you have some available info on polarized light 3D systems. I'd >be really grateful to you if you let me knoz something about this. >Enclosed is the posting I planned to put on nn. The traditional polarized-light 3-D system is appropriate for movies, but not for systems involving viewing computer displays. In the case of movies, two separate streams of light are directed at the screen, one containing the right-eye image and one the left-eye image. They pass through separate polarizing filters. Each audience member's goggles contain correspondingly-oriented polarizing filters so that only the right-eye image reaches the right eye, and only the left-eye image the left eye, even though both photon streams are bouncing off the same place on the theatre screen. The key is the polarization of the photon stream which each eye sees. Achieving this with a computer display would involve is to view. To make a similar scheme work with left-eye and right-eye images coming from the same display would involve having a mechanical device rotating 90 degrees in front of the display every 1/60th second or faster, while the displayto ows alternating views for each eye. This is a big mechanical What I have seen that works very well is a display that shows alternate-eye views at a 120-Hz rate, while each viewer's goggles, synchonized via infrared emitted from a small device on top of the display, obscure alternate eyes with a switched LCD filter; the equivalent of a small round laptop screen in front of each eye, with all the pixels switching at once. This is an application of polarization very locally within the goggle "lens"; is that what you had in mind? If so, the system comes from Silicon Graphics [for whom I do not work, and who do not bribe me, and whose stock I do not own, etc]. Ask about their VGX models; but I'm told the stereo capability can be added to vanilla SGI workstations as well.
webber@csd.uwo.ca (Robert E. Webber) (05/17/91)
In article <1991May17.030937.9450@milton.u.washington.edu> buckland@ucs. ubc.ca (Tony Buckland) writes: . .. . What I have seen that works very well is a display that shows . alternate-eye views at a 120-Hz rate, while each viewer's . goggles, synchonized via infrared emitted from a small device . on top of the display, obscure alternate eyes with a switched . LCD filter; the equivalent of a small round laptop screen in . front of each eye, with all the pixels switching at once. . This is an application of polarization very locally within . the goggle "lens"; is that what you had in mind? If so, the . system comes from Silicon Graphics [for whom I do not work, . and who do not bribe me, and whose stock I do not own, etc]. . Ask about their VGX models; but I'm told the stereo capability . can be added to vanilla SGI workstations as well. Actually, the system you are describing is CrystalEyes which is manufactured by StereoGraphics (of San Rafael, California; 1-800-24STEREO) and is resold by SGI for people who want a Stereo-ready SGI box (if you want to retrofit a non-Stereo-ready SGI box, then SGI refers you back to StereoGraphics). They also make equipment that will connect to other workstations and even to PC's. --- BOB (webber@csd.uwo.ca) [MODERATOR'S NOTE: CrystalEye's inventor and vendor, Lhary Meyer, can be reached via email to The WELL: lmeyer@well.sf.ca.us -- Bob Jacobson]
brucec%phoebus.labs.tek.com@RELAY.CS.NET (Bruce Cohen) (05/18/91)
In article <1991May17.030937.9450@milton.u.washington.edu> buckland@ucs.ubc.ca ( Tony Buckland) writes: >>Enclosed is the posting I planned to put on nn. > > The traditional polarized-light 3-D system is appropriate for > movies, but not for systems involving viewing computer displays. > In the case of movies, two separate streams of light are directed > at the screen, one containing the right-eye image and one the > left-eye image. They pass through separate polarizing filters. > Each audience member's goggles contain correspondingly-oriented > polarizing filters so that only the right-eye image reaches the > right eye, and only the left-eye image the left eye, even > though both photon streams are bouncing off the same place on > the theatre screen. > > The key is the polarization of the photon stream which each eye > sees. Achieving this with a computer display would involve [ lines missing ? ] > is to view. To make a similar scheme work with left-eye and > right-eye images coming from the same display would involve > having a mechanical device rotating 90 degrees in front of the > display every 1/60th second or faster, while the displayto ows > alternating views for each eye. This is a big mechanical [ lines missing ...] There has been at least one 3D viewing system for computer displays on the market which does very well with the simple glasses and which does *not* use a mechanical shutter. Rather, there's an LCD shutter, as you describe for the synchronized glasses system. This shutter system was developed by Tektronix and marketed on their 3D terminals adnd workstations. Now that Tek has introduced a Pex terminal I expect that they'll be putting the shutter on it some day. Note that though I work for Tek, I no longer work in the graphics division, and have no official or unofficial knowledge of product plans; I'm just speculating on my own hook. The displays the system was used on put up 60 frames/sec, alternate frames for each eye with the polarization rotated by the shutter. I've used the shutter on several different displays (I worked on some of the internal display firmware for 3D interaction using tablets and dialboxes and such), and the stereo effect seemed quite good to me. At least I was able to do reasonably accurate tracking of a cursor in 3D without needing other depth cues like depth shading or color desaturation. -- ------------------------------------------------------------------------ Speaker-to-managers, aka Bruce Cohen, Computer Research Lab email: brucec@rl.labs.tek.com Tektronix Laboratories, Tektronix, Inc. phone: (503)627-5241 M/S 50-662, P.O. Box 500, Beaverton, OR 97077
jkollin@milton.u.washington.edu (Joel S. Kollin) (05/19/91)
In article <1991May17.030937.9450@milton.u.washington.edu> buckland@ucs.ubc.ca ( Tony Buckland) writes: > The key is the polarization of the photon stream which each eye > sees. Achieving this with a computer display would involve... > having a mechanical device rotating 90 degrees in front of the > display every 1/60th second or faster, while the display to ows > alternating views for each eye. This is a big mechanical... WRONG! Not only is it possible to electronically change the polarization of light at screen refresh rates, but Tektronix makes a device which does just that (using, I believe, liquid crystals). Don't assume that because you can't figure out a neat way of doing something that it is impossible. It's bad enough when 'experts' make this mistake, but you really should either do some research or put in a caveat when you post... Actually, most VR/'cyberpunk' people think things are much EASIER than they really are. This is a refreshing change... joel
lmeyer@uunet.UU.NET (lhary meyer) (05/25/91)
Thank you for posting the info on the CrystalEyes system...the correct number is 415-459-4500 (FAX: 415-459-3020) or EMail as above... If your interst is in movie projection schems, I highly recommend "FOundations of the Stereoscopic Cinema" available thru REEL 3D in Duarte,CA. (this best suppler of 3D "stuff" around) or from Stereographics.
testarne@athena.mit.edu (Thad E Starner) (05/27/91)
Here is a quick evaluation of the Tektronics SGS610 Stereoscopic
System. Note that I do not give any warranty to the accuracy of this
info.
Basic Principle:
The idea is to flip rapidly between images created for the left and
right eye. These images should be created by taking a snapshot of the
world and then translating ~6.5cm left (or right...arbitrary depending
on which eye you are rendering for) recentering the view frame (NOT
rotating the camera) and taking another snapshot. These images are
then flashed on the screen synchronously with the flipping of a
shutter. This shutter (in the Tektronics case) is a LCD screen which
fits over the computer screen. The shutter switches between acting as
a right circularly polarized filter and a left circularly polarized
filter at rates of up to 120 Hz (60 Hz per eye). The user wears
passive glass which are right and left circularly polarized over the
right and left eye respectively. Note that
these glasses are passive and look like normal sunglasses. Thus they
are light and non-electical with no tethers. Provided the images are
in sync with the shutter and the appropriate image is displayed for
each case, the user should get the proper effect: the left image
only seen by the left eye and the right image only seen by the right
eye. More information on circular polarized filter can be found in
most advanced physics textbooks dealing with light.
What's included (paraphrased from 1988 User Manual):
19" (610), 16" (410), or 12" (310) liquid crystal Stereoscopic
Modulator which is attached to the user-supplied monitor
Stereoscopic Modulator Driver with AC power supply
Four pairs of viewing glasses
Modulator Interconnect Cable
Velcro Mounting strips (used for attaching modulator to display)
User's Manual
What you do:
First you must have a graphics board which can switch images at least
as fast as 60 Hz (30 each eye). In my case this is a Sun TAAC
accelerator in a Sun 4/360 (the images are ~500x500). These boards
synch using the synch pulse sent to the monitor. The modulator driver
also sits on this synch line so that it can keep the modulator in
synch. The velcro strips are used to hold the modulator on the
computer screen. You render the appropriate views as described above,
start the stereo mode on the graphics board, plug in the modulator
driver, put on your glasses, and suddenly you have 3D! Note that
depth reversal may occur due to the right image being given to the
left eye, etc. This can be fixed by flipping the depth reversal
switch on the modulator driver. Also, the image will appear to follow
you. This is due to the lack of motion parallax in the system. While
you are moving, the same images are being presented to your eyes.
Thus, the objects seem to follow you.
More Info:
There are six different modes for giving synch pulses to the driver.
Control inputs are TTL levels; composite synch levles are >= .2V neg.
synch and (<5V p-p video and sync).
These are vector direct (from one input (BNC connector) : on = right,
off = left), vector latched (first true on input 1 latches right eye
on, first true on input 2 latches left eye on), vector gated (at true
edge of input 1, the data at input 2 will be used....if it is true
then right eye else left eye), raster frame direct (raster mode...on
true edge from input 1, the right eye is turned on, the right eye is
turned of and the left eye on at the appropriate interval
afterwards...delays for modulator switching and phosphor decays of
1.5ms or less are included), raster composite sync (hard to
explain...right thing for a Sun monitor model HM-4119-S-AA-O
however...again compensates for phosphor decay and switching), and
raster field direct (again hard to describe but compensates for
phosphor decay and switching). We use raster composite sync. There
are also options for termination or feed through of sync input and 30
or 60 Hz rates. We use the 30 Hz rate without termination or
feedthrough (last two did not effect performance).
Interesting Manufacturer's Specs (again from 1988 User's Manual):
Warmup time 60 sec MAX (ours about 10-20 secs)
Right eye turn-on time 0.35 mS MAX (switching from left to right)
Left eye turn-on time 3.2 mS MAX (switching from right to left)
Average light transmission 12% (????? seemed much better than
that)
Ave extinction ration (on image/off image)
left red 14/1 14/1
green 9/1 10/1
blue 5/1 8/1
right red 20/1 20/1
green 15/1 15/1
blue 10/1 10/1
The display used needs to have fast phosphors (~1.5 msec for decay)
for this system to work well.
Equipment Used:
This is primarily a rehash, but:
Sun 4/360 with TAAC graphics accelerator board
Sun monitor HM-4119-S-AA-O
Tektronics SGS 610 19" Sterescopic 3D Display Kit, 120/60Hz (running
in 60Hz mode)
Software Environment: ThingWorld 3D Modeling System (A.
Pentland and a host of others in the Vision & Modeling Group, MIT
Media Lab). Runs primarily under X.
Evaluation:
The binocular disparity provided by this system causes a rather
striking depth effect (as exclaimed by several viewers). Unfortunately, this ef
fect was limited to when only red objects were presented (white and green object
s were also tried, a black background was used in all cases). Otherwise, ghosti
ng was prevalent. Ghosting is when the image from one view persists on the scree
n while the other eye is being addressed. The advantage of red objects is
probably due to the good extinction ratio of red as opposed to green
and blue as shown in the manufacturer's specs. There are several ways
to address this issue. The first is to get a display with faster
phosphor decay. It is possible that the display used has a decay rate
of > 1.5msec. Another solution may be to increase the frame rate. If
the manufacturer's specs are correct, this should reduce the effect
somewhat. Unfortunately, our graphics board is not set up for the
higher speed. Another solution may be to adjust the timing of the
presentation of the views. However, this may be difficult depending
if internal adjustments can be made to the modulator driver or the
accelerator board. Another drawback was the flicker observed in our
system. Again, the higher frame rate could be used to reduce this
effect. There are several major advantages that have been noticed with
this system. They are:
1) Portability of the stereoscopic system to other displays and
systems (as long as the graphics and sync requirements are met).
2) Passive glasses. This is a major advantage.
a) There are no tethers to the system. The glasses feel and
look (unless viewed through another pair) like normal sunglasses.
b) Multiple users can view a screen at once (although
they will get different effects from not all being at the
optimal viewing distance).
c) The user can tilt his head in any direction and still
receive left and right views due to the circular polarization
method.
3) Lack of mechanical parts (except for the LCD crystals)
I can think of one other inherit disadvantage besides the ones already
given. This is the interference the modulator can generate. When
using the system with 2 Polhemi, strong interference occured.
However, this was rectified by simply moving the Polhemi sources
farther from the modulator.
In the next few weeks I will be experimenting more with this system.
One of my goals is to reduce ghosting. If anyone is
interested in the results or clarification of the above, I can be
reached at testarne@media-lab.media.mit.edu much quicker than the
athena account. Also, my S.B. thesis (for which this equipment was used)
gives a brief overview of the various stereoscopic and 3D imaging methods
available from the Wheatstone stereoscope through holography if anyone is intere
sted.
Standard Disclaimers:
Note that I do not work for Tektronics and do not give any warranty
for my information. Also, while I am affiliated with the Vision &
Modeling Group, MIT Media Lab, my opinions and ideas do not
necessarily reflect those of the group or lab.
Thad Starnertestarne@athena.mit.edu (Thad E Starner) (05/27/91)
I should have mentioned that the Tektronics System was tried out in the drumset reality I posted (or tried to post) several months ago. The results were pretty good. Hitting the drumheads was much easier with the extra depth information. For those who don't know (or couldn't guess given the number of people who've played with this idea) this reality involves a rendered drumset (2 snares, 2 toms, bass, 2 cymbals, and a "cow" bell...it goes moo). Two drumsticks with Polhemi sensors attached two them are used to interact with the system. As the two drumsticks are moved, rendered versions of them are displayed real time on the screen. Predictive filtering is used to compensate for delay. The drums make the appropriate sound when given an impulse with the drumsticks. Csound is used to generate (or in some cases play back samples) sounds for the system. The sounds have in the past been tied to the size and shape of the object (though nothing complex). All this is runnin g on a Sun 4/360 with a TAAC board.
rcd@ico.isc.com (Dick Dunn) (05/29/91)
Thanks to testarne@athena.mit.edu (Thad E Starner) for info on the Tektronix stereoscopic system. A few comments and questions... > ...The shutter switches between acting as > a right circularly polarized filter and a left circularly polarized > filter at rates of up to 120 Hz (60 Hz per eye)... Later in the description, there are various figures of merit for turn-on time and extinction ratios, all of which are better for the right eye than the left. Can you explain this? Is there some mechanism such as the shutter being actively driven for one polarization but allowed to "relax" for the other polarization? I'm curious about the mechanism; I'm also curious whether the response asymmetry will be noticeable (probably not overtly in the short term, but perhaps over the longer term or at a sub- liminal level?). > Interesting Manufacturer's Specs (again from 1988 User's Manual): ... > Average light transmission 12% (????? seemed much better than > that) Light levels can be deceptive...12% is about 3 f-stops. > The display used needs to have fast phosphors (~1.5 msec for decay) > for this system to work well. This is where I start getting very interested in how it feels to use the display. Phosphors for typical color monitors seem to be selected without a lot of attention to matching persistence among the colors. The differ- ences aren't dramatic, but they can be made noticeable. This could add to (or partially compensate for) the effect of the "shutter" itself, which Thad notes: > ...The advantage of red objects is > probably due to the good extinction ratio of red as opposed to green > and blue as shown in the manufacturer's specs... Actually, I'm confused about whether the color-dependent variation in the extinction ratio is all due to the shutter itself (polarizers tend to be more effective at longer wavelengths), or if some phosphor characteristic is also factored in. >...There are several ways > to address this issue. The first is to get a display with faster > phosphor decay. It is possible that the display used has a decay rate > of > 1.5msec... My concern with that would be flicker. An interlaced monitor at 60 Hz will be objectionable to a lot of people. (I once tried a monitor at 56 Hz; I started feeling physically ill after a few minutes.) Longer-persistence phosphors help cover the flicker problems of existing monitors. So, are you on the horns of a dilemma here, where faster phosphors will just trade ghosting for flicker? Thad already mentioned that >...Another drawback was the flicker observed in our > system. >...Another solution may be to increase the frame rate... That gets you out of the ghosting/flicker dilemma. It does come back to my earlier question about the effect of the asymmetry of switching time (about a factor of 10) for the shutter...if you double the switching rate, does it become noticeable? -- Dick Dunn rcd@ico.isc.com -or- ico!rcd Boulder, CO (303)449-2870 ...Simpler is better.
testarne@media-lab.media.mit.edu.MEDIA.MIT.EDU (Thad Starner) (05/31/91)
The only response I can give to Dick until I run a few more tests is
to give some more info:
According to the Tektronics Stereoscopic manual (1988), there is "a
limit [I assume of time "on" per iteration] of about 20msec for the
left eye state, which limits the lower rate at which the Modulator can
be used to about 20 Hz."
Also:
"The Pi-cell is the fastest large area liquid crystal effect switching
from non-driven to driven in .2 msec and from driven to non-driven in
approximately 2 msec. To enable a 120 Hz vertical refresh rate, the
cell is electrically split into top and bottom segments. Multiplexed
2:1 from a PROM it is possible to achieve an acceptable contrast ratio
with flicker free operation."
This info supports your driven/relax hypothesis.
As far as the color-dependent variation in the extinction ratios, I
assumed that this included phosphor characteristics since the header
to the stats I gave was
"Optical Performance (P4 phosphor and Tek Stereo glasses):"
and there were other mentions of assumptions of 1.5msec decay
phosphor. The extinction ratios seem to be the most dependent on the
phosphor values, so I believe that this is where the assumption comes
in (though ave light transmission is also related).
As far as the pure shutter effects on the extinction ratios, I'm
curious myself.
Can anyone think of a good fast way of testing this, or does anyone know?
About asymmetry effects - it is possible that Tektronics made their
driver so that this is taken into account. If I have time I'll put a
scope on it. If not the direct modes on the modulator should allow you
to design a circuit to compensate.
Flicker- I didn't notice it after a while, but of course as soon as
someone brought it to my attention I would see it. I did get a
monstrous headache while getting the system up, but that was mainly my
own fault due to problems with depth reversal and incorrect views.
The final system I used for an extended period of time with no ill
effects except an increased sensitivity to depth information ("Wow!
I've been sitting in front of the stereoscopic for so long I see
everything in 3D!").
I'll post more info as I do more tests.
If people are really curious and are in the area, call me up and I'll
show it to you.
Thadkilian@poplar.cray.com (Alan Kilian) (05/31/91)
Well I guess I'll take a shot at explaining the LCD polarizing filter 3D shutter system. First we need to plot what happens when in the CRT/LCD system. If you are doing 120 frames per second, one frame is 8.3ms long (120^-1) O.K now all that time will not be used for displaying the image. The CRT takes some time to get the electron beam back to the top of the screen at the end of a frame. So we'll say it takes about 1ms. It really spends .5ms above the top of the frame and .5ms below the bottom of the frame but you don't care about that. A few other numbers from Thad Starners post of May 27: 1.5ms Maximum time for phosphors to decay from 100% brightness to 0% .35ms Maximum time for switching from left eye to right eye. 3.2ms Maximum time for switching from right eye to left eye. So here is a time line. Each row is .5ms with things that don't line up listed with their "real" times. Time in ms Video action LCD action ---------- ------------ ---------- 0.0 Gun at top of CRT 0.5 First raster line being drawn 1.0 This is the right eye image 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 Last raster line being drawn (7.8ms) 8.0 Vertical sync (8.3ms) Start to switch to left eye 8.5 First raster line being drawn (8.8ms) 9.0 This is the left eye image 9.5 10.0 10.5 11.0 11.5 top 40% of screen drawn Left eye fully transparent 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 Last raster line (16.1ms) 16.5 Vertical sync (16.6ms) Start to switch to right eye 17.0 Right eye fully transparent 17.5 First raster line (17.1ms) 18.0 This is the right eye image 18.5 19.0 19.5 20.0 20.5 21.0 21.5 22.0 22.5 23.0 23.5 24.0 24.5 Last raster line (24.4ms) 25.0 Vertical sync (24.9ms) Start to switch to left eye O.K. so what does all of this mean? Two things to look at: 1) When the left eye becomes fully transparent we have already drawn 40% of the left eye image. If the phosphors decay at the minimum rate (1.5ms) The upper 17% of the left eye image has already decayed to black. So the upper 17% of the screen is not useful to the left eye. 2) When the right eye becomes transparent we are before the first raster line of the right eye, BUT the lower 9% of the CRT still contains the left eye image so the lower 9% of the screen is not useful to the right eye. So that leaves 74% of the screen useful (100% - 9% - 17%) If your phosphors take longer than 1.5ms to decay then less of the left eye is useful and more of the right eye is useful. If your phosphors take less time to decay then more of the left eye is useful and less of the right eye is useful. These problems do not have to do so much with the switching time of the LCD but due to the difference in switching times left to right and right to left. I use a pair of Sega(tm) LCD glasses with my system. They have equal switching times and I find that I cannot use about the lower 10% of the CRT because ofthe phosphor decay time. The Sega(tm) LCD glasses switch very fast. About .5ms to go from dark to clear. The bad thing about these is that they don't go very dark so you have to use a filter over them to get good blocking and this makes the image dim. If you turn the CRT brightness up you start to see the other image through the dark lens. So with the Sega(tm) glasses I use the top 90% of the CRT and all is fine. I hope this help everyone evaluate LCD swiching for 3D display. -Alan Kilian kilian@cray.com 612.683.5499 Cray Research, Inc. | If god had meant us to use the metric system 655 F Lone Oak Drive | he would have given us ten finger and ten Eagan MN, 55121 | toes. Judith Stone _Lighter Elements_ [MODERATOR'S NOTE: Thank you, Alan. Very well done. -- Bob Jacobson]
lance@motcsd.csd.mot.com (lance.norskog) (06/04/91)
[kilian@poplar.cray.com (Alan Kilian) writes an excellent analysis] 1) The Extinction Ratio refers to the ratio between the number of photons that go through in the dark:clear states. The Haitex goggles (Nintendo goggles which were never marketed outside Japan) have an extinction ratio of 1:100, as does CrystalEyes' low end product. Their high-end product has an extinction ratio of 1:300. You reported that your Sega Goggles "don't go very dark". Do you have an extinction ratio rating for them? 2) The naive scheme is to switch between states based on the vertical retrace. It seems that you may want to decouple the square wave to each lens from being alternate portions of one square wave, and overlap them a bit, initiating the clear phase just before the vertical retrace. So, the circuit inputs should be the vertical and horizontal retrace, and a horizontal counter. The circuit counts horizontal flybacks and retraces just before the bottom of the screen. Perhaps the off phase should be initiated right around the bottom of the screen, so as to cut off dying phosphors because the colors decay at different rates. 3) The right solution for large screens (> 500 lines) would seem to be bifocal shutters. The top half matches the top half of the screen, and the bottom shows the bottom of the screen. This gives tighter control in matching the scan rate of the screen. 4) A hacker at Vision Research Group (another LCD shutter vendor) told me that the control waveforms should come out of a ROM instead of an analog circuit; this gives better performance somehow. The next generation of LCD shutter gear should be designed based on these conclusions: Separate the duty cycle for the two lenses. Adjust the duty cycles of the lenses for the switching times of the lens and the phosphor decay times of the colors of the monitor. For big-screen work, try bifocals. Lance Norskog lance@motcsd.csd.mot.com