dlppleading

Pleading for a low-cost 1-chip-DLP
active-stereoscopic-3D projector
by
Christoph Bungert
stereo3d.com
page introduced July 20, 2003
last update March 31, 2005

Content:
Why a single 1-chip 3D-DLP projector plus 3D-shutterglasses could be superior...
Available products
Time-sequential technology
DLP speed
Conclusion
Doing a low-cost 1-chip DLP-3D-projector - How to
webboard-thread for this page

March 31, 2005: The waiting is over. A low-cost 1-chip-DLP-stereo projector is now available: www.DepthQ.com

Why a single 1-chip 3D-DLP projector used with 3D-LCD-shutterglasses could be superior compared to all other available 3D solutions
Advantages:
- virtually no ghosting opposed to polarization-solutions, CRT's and many autostereo displays
- multiple viewers watching from multiple angles and distances - opposed to monitors, (autostereo-)monitors, HMD's and boom mounted displays
- high refresh rates possible
- low cost - compared to most 3-chip 3D-DLP projectors, dual-projectors, film-projectors, fast phospor CRT-projectors, hi-res HMD's and autostereo displays
- low weight, high mobility and easy setup - opposed to dual-projector setups, CRT's, plasma-autostereo-displays and even HMD's
- cheap, light, comfortable, wireless shutterglasses are already available
- no expensive, external de-multiplexers required - opposed to dual-projector setups
- no non-depolarizing screen required - opposed to dual-projector setups
- back-projection is possible
- small back-projection monitors for labs, simulators, offices and operating rooms are possible - opposed to larger 3-chip setups
- no isolation and no lack of orientation - opposed to closed HMD's
- more comfortable than HMD's
- full resolution - opposed to autostereo monitors
- excellent image quality
- large field of view without eye-strain - opposed to monitors and most HMD's
- perfect channel-separation at any angle, audience can tilt their heads - opposed to polarizer-solutions
- 3D-DLP-projector works as a 100% standard 2D-projector as well - no compromises opposed to autostereo monitors
- works for video and VGA sources
- works with existing stereo-software
- better software compatibility than volumetric displays and multichannel-autostereo-displays
- a set composed of a low-cost 3D-DLP-projector with wireless shutterglasses would make a 'sexy' product - since professionals and consumers are currently hot for digital projection
Disadvantages:
- requires dark room
- requires glasses
- potentially less brightness than (autostereo-)monitors, dual-projectors or 3-chip-3D-DLP-projectors
- shutterglasses are more expensive and heavier than polarization-glasses - and they require emitters, batteries, maintenance and even disinfection after public use (in IMAX3D theaters all polarization- and shutterglasses are disinfected)
- no head-tracking
- less immersion than HMD's
- inferior 'black-level', less contrast and less frequency-variability than CRT's
Conclusion: I believe the combination of a single 1-chip-3D-DLP-projector and 3D-shutterglasses could be the most economical, most desireable stereo3D-solution available to date. The absence of ghosting, the ease of use, the low cost and the possibility to present to larger audiences is what would set this solution apart. The backdraws are acceptable.

Barco Galaxy 3-chip 3D-DLP-projector
Available products
Standard DLP projectors usually use a native frequency of 60 Hz. The buffered, internal frequency may be out of sync with the source input signal. Models with advanced line-doubling may mix images in video mode destroying the 3D-information. Therefore standard DLP-projectors have a flicker problem, a compatibility problem and often a sync-problem when used with 3D-shutterglasses.
Special 3D-DLP projectors by Barco and ChristieDigital use 3-chip DLP configurations and are expensive (Christie Mirage series: 50-150 k$).
Synchronized dual DLP-projector setup + demultiplexer + shutterglasses by digital image

The situation isn't satisfying for consumers, universities and professionals on low-budget.

Time-sequential technology
DLP in general and 1-chip DLP projection in particular are time-sequential technologies and therefore are related to time-sequential stereoscopy (i.e. shutterglasses).
The micro-mirros in a DLP system are very fast ( >50 kHz). All the mirrors are able to is switching a pixel on or off. There are no grey scales and no colors. Grey scales and colors are faked by fast switching. The electronics in a DLP system are designed to do all kind of time-sequential tricks.
So the marriage between 1-chip DLP projection and shutterglasses would be a very natural one.
By the way Texas Instruments just announced to go for a new technology where the color wheel doesn't have distinct color-windows, but a color-Archimedes-spiral on it (SCR - sequential color recapture). Don't know yet what the implications for stereo3d will be.
However the spiral-trick, where 3 different colors are projected on the DMD simultaneously, is another example for the kind of delicate time-sequential-tricks one can play on the DMD. Compared to these twists and turns doing a 120Hz-shutterglasses-mode with a conventional color-wheel would just be a snap.

DLP speed
An earlier argument against 1-chip-DLP-projection and stereoscopy was lack of speed. Now 5-speed and even 6-speed DLP chips are entering the mid-range and even low-end market.
"The way in which the DMD is driven has also been enhanced. The surrounding electronics have been
further developed, and new algorithms have been designed to improve performance in a range of
areas. The color filter wheel’s effective speed is now 5-6x the speed of the original color
wheels,..."
Ian McMurray, Texas Instruments

I've made a list of possible 1-chip DLP setups and the fields per second the DMD has to deliver accordingly, plus some examples of existing projector models:
DLP-spd.   wheel-r. wheel-segments              fields/s. eff.img.frequ.
1x-speed = 60 rps x 3 color-segments (RGB)     = 180 fps ( 60Hz)(1st generation projectors)
1x-speed = 60 rps x 4 color-segments (RGBW)    = 240 fps ( 60Hz)(1st generation projectors)
2x-speed = 120 rps x 3 color-segments (RGB)     = 360 fps (120Hz)
2x-speed = 120 rps x 4 color-segments (RGBW)    = 480 fps (120Hz)(NEC LT240/LT260)
4x-speed = 120 rps x 6 color-segments (RGBRGB) = 720 fps (240Hz)(NEC HT1000)
4x-speed = 120 rps x 7 color-segments (RGBWRGB) = 840 fps (240Hz)
4x-speed = 120 rps x 8 color-segments (RGBWRGBW)= 960 fps (240Hz)
5x-speed = 150 rps x 6 color-segments (RGBRGB) = 900 fps (300Hz)(Toshiba TDPM8U)
6x-speed = 180 rps x 6 color-segments (RGBRGB) =1080 fps (360Hz)(Sharp Xv-Z9000 ?)
?x-speed = 240 rps x ? color-segments (?)                        (Toshiba TDP-MT500)
SCR-wheel= 11 rps x 126 color-segm. (RGB-spiral)               (prototype)
rps = revolutions per second of color wheel
fps = fields per second on the DMD (monochromatic)
RGBW = red, green, blue, white
DMD = digital micro mirror device
DLP = digital light processing (consists of the DMD and the required electronics for control of the DMD)
SCR = sequential color recapture
I fear that the frames produced at high speed, like the 900 fields per second, on the Toshiba TDPM8U aren't complete images. Only a combination of all fields which belong to one frame may give the correct grey-scale and color-mix for the complete image. The effective frame rate - for distinct frames - may still be 60 Hz. However it can be assumed that current 1-chip-DLP's are fast enough to do 120 distinct frames for 120 Hz stereo with acceptable grey-scales and color-depth.

Conclusion:
1) Low cost 1-chip DLP packages which are fast enough for active, flicker-free stereoscopic 3D projection are already available
2) The available 1-chip DLP packages and frame buffers seem to be very flexible since various versions of color-wheel-segmenation, color-wheel-speeds and input signals are already supported. Therefore it should be very easy for developers, either at Texas Instruments or even at the projector-manufacturers, to introduce a native 120Hz mode for stereoscopic 3D.
3) The extra cost for R&D and a VESA-DIN-3 jack should be small.

Doing a low-cost 1-chip DLP-3D-projector
Goal:
- low cost 1-chip DLP projector with native 120 Hz full-res frame-sequential 3D-mode for all input formats
- internal frame-sync of projector MUST be made available to trigger 3D-shutterglasses
- VGA connector should at least accept 60 and 120 Hz frame sequential 3D content
- DVI-port should at least accept 60 Hz frame sequential 3D content
- video/s-video connector should accept 60 Hz frame sequential 3D content (50 Hz PAL mode would be nice, but is harder to achieve)
- conversion for over-under and line-sequential 3D-sources would be desirable

How it could be done:
1) Instead of taking a 50-120Hz input signal and bring it down to 60Hz and then multiply it again to 120, 150, 240 or whatever internal frequency it should be a simple software change to add a 120Hz input mode. The buffer could be used to stabilize the gap between vertical frequency of the input-signal and the native color-wheel frequency. If the input signal comes slightly too fast (i.e. 121 Hz or something) a complete stereo-pair could be dropped from time to time. If input comes slightly too slow (i.e. 119 Hz) a stereo-pair could be doubled. This may not even require new hardware, just a change in the image processing/buffer firmware, since the buffers do such kind of tricks already.
2) For NTSC/PAL 3D video input the half-frames must not be mixed! The lines of each individual half-frame should be doubled. Some DLP projectors do this already in 60 Hz. For transfer to 120 Hz the frames must be doubled from an AB to an ABAB sequence.
3) The sync from the color wheel should be used as trigger for a VESA DIN-3 stereo output on the projector, which controls the 3D-shutterglasses. There should be a stereo-reverse button.
4) Optional: In addition to giving the projector a VESA DIN-3 stereo output one could give it a VESA DIN-3 input as well. This input could be used to sync the color wheel and the buffer to it (that's the BARCO approach I believe). Problem is - this may require delicate speed control of the color-wheel, resulting in higher cost than an output-sync-only solution. (Mechanical solutions are more expensive than electronical solutions.) The only reason why Barco has a DIN-3-input may be to keep the stereo-orientation stable between the source and the output.
I don't know if the DLP projectors on the market may already have a color-wheel which allows speed-fine-tuning. If so it would be great, but I doubt it! Here's my theory, please correct me if I'm wrong: The color wheel runs at a semi-fixed speed. Since the motor of the wheel may not run 100% stable the wheel-status is given to the buffer-electronics which processes the input signal in a way to fit the output-signal (which drives the actual mirrors) into the 'color-wheel-schedule'. At a certain point the buffer is no longer able to compensate the time gaps, resulting in a 'tearing effect' in the image or a frame-drop.

Christoph Bungert

If you have any thoughts on this issue please post to this webboard-thread.

Brand and product names are trademarks or registered trademarks of their respective holders.
The author can not guarantee the accuracy of the information given on this page.
Christoph Bungert, Germany