Several liquid-crystal technology goals must be considered for 3-D flat-panel-display implementations to achieve high visual performance.
by Philip J. Bos and Achintya K. Bhowmik
WITH high-quality 2-D liquid-crystal displays having become commonplace, interest is now shifting toward the development of 3-D displays. It is clear from the large number of papers published in the leading display technology journals and conference proceedings that research and development efforts are increasingly being dedicated to 3-D displays, both in industry and academia. As a result, a number of consumer-electronics companies are now offering systems based on 3-D displays in the marketplace.1 These are primarily high-definition televisions (HDTVs) in addition to some early models of notebook computers and handheld devices. However, the viewing quality or the visual performance of currently available 3-D displays lags far behind their state-of-the-art 2-D counterparts, which indicates that a large number of issues have yet to be solved.2 In this article, we will provide an overview of the requirements for achieving a “true” 3-D display, and then point to the liquid-crystal technology goals that must be considered for high-quality 3-D flat-panel-display implementations.
The requirements for the optical system of a true 3-D display can be understood by thinking of the display system as a window. Imagine the window divided into small area patches – so small that if we were to block off the entire window except one of the patches, we would see only one set of color and intensity because the patch would be too tiny to allow us to see image detail. In the depiction of the display system below, each of these window patches corresponds to a pixel. Figure 1 shows light passing through a very small patch of a window near its center.
The color and intensity of light that comes through the small patch depend on the light ray’s angle. From the angle shown in Fig. 1, the viewer sees the yellow color of the walls, but at other angles the viewer would see the red color of the roof or the blue color of one of the windows. So, from each patch on the window, there are a bundle of rays emerging that is characterized by an angle, color, and intensity. What is different about 2-D and 3-D displays is the angular dependence of the pixel information.
Fig. 1: The light rays passing through one “pixel” of a window come from different points of the scene. To the viewer, in the position shown, the pixel will appear yellow.
This angular-dependent information conveys three aspects of the 3-D scene:
1. Relative motion of objects (an eye sees a different view as the viewer’s position is changed).
2. Stereopsis (each eye sees a different view).
3. Focus (the angular spread of rays, intercepted by the pupil from a point in the scene, is determined by its distance from the viewer).
In adding these characteristics to a typical 2-D display to make it a 3-D display, it is also important that the resolution of the display remains high, near the limiting resolution of the eye, because the textural cues are important in the perception of depth and “realism” of the image. If we consider the “ultimate” 3-D display as one that can emulate a window as described above, we need to have a display with a high density of pixels that change color and intensity for different directions of view with very high angular resolution. If we say the eye collects a cone of rays spanning several tenths of a degree, and we would like to have an adequate angular resolution to achieve a proper focus, it is likely that we will need an angular resolution of 1°. And then if we would like the window to be viewed over a 100° field of regard, we will need each pixel to provide 1 million rays of light, each with a defined color and intensity. A 3-D display with these specifications would have its bandwidth increased by a factor of 1 million over a 2-D display with the same size and resolution, which is well beyond the current state of liquid-crystal technology.
So, we need to think of a more limited solution. One method of doing that is used in theaters equipped to show 3-D films, where the scene is far away from the viewer and the viewer is sitting still. In that case, we can ignore the relative motion and focus cues of the required optical system and only provide the stereopsis cue. But this simplification may not work for scenes close to the viewer, such as would be portrayed on a desktop or a mobile display. In this case, both the cues of relative motion and focus could be significant, especially the relative motion cue (for an example of how important, check out this video from Johnny Lee, then at Carnegie Mellon University’s Human–Computer Interaction Institute:http://www.youtube.com/ watch?v=Jd3-eiid-Uw). Including the relative motion cue for multiple viewers, but not attempting to take into account the focus cue, reduces the requirement on the angular resolution to that which is required to produce stereopsis and smooth motion. The angular resolution in this case might be on the order of 1° and can further be considered to be limited to only the horizontal direction. In this case, the bandwidth of a 3-D display is increased by a factor of “only” 100 over a 2-D version.
Conceptually, this could be accomplished by removing the typical backlight assembly of an LCD and putting a highly collimated “searchlight” in the form of a line that sweeps from –50 to +50° while the image on the display changes at a rate of once every 1° of sweep. But, of course, this would need to happen at a rate faster than the flicker rate, so we would need an LCD that can update a full screen in about 100 μsec. Alternatively, it could be accomplished with a parallax-barrier or lenticular-lens approach, in which we have 100 images behind each lens-let. But this would require a display that has over 100 Mpixels and image resolution would be a major challenge.
Neither of the above options is outside of the range of consideration for liquid-crystal technology, and both parallax-barrier- and lenticular-lens-based devices using a more limited number of views have shown good performance over a more limited field of regard.
Another approach that we can consider in order to reduce the required bandwidth of the system is limiting the number of viewers. A head-tracking system can then be used to determine the location of viewers’ eyes, and each pixel then only needs to be able to provide light of a color and intensity corresponding to the angle of a ray leaving the pixel and headed for each eye. For a single viewer, we will need to increase our system bandwidth by only a factor of 2 (if the focus cue is left out of consideration).
A relatively easy way to do this is to use glasses on the viewers’ eyes that ensure that each eye perceives each pixel on the display as having the color and intensity corresponding to its angle with respect to the pixel. The standard approaches are divided into the classes of active and passive glasses and are described in the article in this issue, “Tutorial on 3-D Technologies for Home LCD TVs” by Seonki Kim and a previous article by Jeong Hyun Kim.3
For the best brightness, a segmented active shutter can be considered. Segmenting is useful because if there are no shutter segments, as in the case of active glasses, the display must be black for half of the frame time, resulting in a maximum display duty cycle of about 50% (again, see the article by Kim). However, if we consider a shutter as having many segments, then the crosstalk or brightness is limited by the slower switching time of the display, as shown in Fig. 2. The duty cycle will be (FT–ST)/FT, where FT is the frame time andST is the display switching time. If we assume that FT is 8 msec, then to achieve a duty cycle of 90% will require a switching time of about 0.8 msec. If a 75% duty cycle is acceptable, then a switching time of 2 msec is needed, which is more easily available.
Fig. 2: With the display scanning from top to bottom, the changing image is displayed from the blue left-eye view to the green right-eye view. In front of the display is a polarization-state controller with about 20 segments, the state of which is synchronized with the scanning of the display panel. The green-colored segments are providing the polarization state that is transmitted by the right lens of the viewers’ glasses, and the blue color corresponds to that of the left eye. The transition region shown in the display represents the area of the display at the instant the picture is considered where the image will exhibit cross-talk, unless this region is blanked. Assuming that the display switches more slowly than the polarization-state controller, it is this transition time that limits the performance of the system.
Therefore, we can see that a main issue for the liquid-crystal technology is to provide a fast-switching display, as well as a fast-polarization-state controller to ensure that the correct image gets to the correct eye and a high duty cycle is maintained.
There are different factors that can be addressed to increase the speed of a liquid-crystal device. These might be divided into factors related to the material and device design. It is well established that the material factors desired for fast response are low viscosity, high birefringence, high elastic constants, and high dielectric anisotropy. Related to the device design factors, there are three that come to mind (there could be more!).
The first is to design the device so that the material flow in the cell does not slow the relaxation of the device. In electrically compensated bend (ECB) type devices, a design that accomplishes this is the pi-cell4 and for the twisted-nematic (TN) case it is the –3π/2 device.5 The typical switching times for these devices are between 1 and 2 msec and have been used in 3-D systems with success.
The second device design factor is to use a pretilt that is not too close to being parallel to the surface (for positive-birefringence materials) or perpendicular to the surface (for negative-birefringence materials). The reason is that the torque upon application of an electric field is zero if the director is fully perpendicular or parallel to the applied field. This factor has recently been addressed in an overview article presented at the 2011 SID Symposium.6
Another factor is to enhance the stored elastic energy in the field applied state so that when the field is removed, the director field “relaxes” to its zero-field configuration quickly. This approach has been used in polymer-stabilized devices called “stressed liquid crystals,” 7 flexolelectric devices,8 and in blue-phase devices.9 This approach has a drawback, however, in that it necessarily means that the torque needed to bring the director to a given field-on state goes up, and so this approach leads to a requirement for high voltages (given that the material has already been picked that has the highest practical dielectric anisotropy).
Yet another factor is to use ferroelectric liquid crystals that can be “driven” to both desired states and have a strong coupling to an applied electric field. These types of materials have been considered for commercial applications by Micron (formerly Displaytech) in high-frame-rate field-sequential displays.10 While these devices use binary switching between two states and thus require very high frame rates for gray-scale images, analog SmC* devices are also being considered.11
So, there are liquid-crystal technology-based approaches that can provide acceptable response times for two field-sequential devices needed for single users, and improvements in the future will allow for faster refresh rates for multiple users.
However, for a “true” 3-D display, we may want to consider including the focus cue. If the focus cue is incorrect, it causes asthenopia (visual fatigue) after a person has been viewing the display for a short while because a spatial discrepancy exists between the accommodation plane and convergence point as illustrated in Fig. 3. It has been demonstrated that using a corrective lens of appropriate refractive power effectively reduces this discrepancy and the fatigue.12
Fig. 3: The display panel is at the distance the unaided eye would focus at, whereas the convergence plane is at the distance desired for the eyes to focus to minimize eye fatigue. The tunable lens shown can correct the discrepancy.
Therefore, if we could determine the power of the needed corrective lens, and then provide a tunable lens of that power, it might be possible to solve the last major problem for single-user 3-D systems using a flat-panel display and glasses. The first issue of determining the power of the needed corrective lens might be accomplished by having a sensor or a micro-camera on the inside of the frames of the glasses to detect the “toe-in” of the viewer’s pupils. This information will allow the com-puter to know the convergence point of the eyes, which is the location of the object in the 3-D scene being considered. We would like the power of the eye’s lenses to be adjusted as they would be if focusing at that distance and not at the distance to the physical display surface.
By using the lens makers’ formula, we can determine the required power of the corrective lens. If we call the distances from the eye’s lens to the panel dp, to the convergence pointdc, and to the retina dr, then we would like the power of the eye’s lens to be Pe = 1/dr + 1/dc for it to be focused at the convergence plane. But for the image to actually be focused on the retina, we need the power of the electronic lens Pl to be determined by Pe + Pl = 1/dr + 1/dp. This leads to Pl = 1/dp – 1/do. If we consider the value of dp and dc to be 50 and 67 cm, the power of the corrective lens, Pl, is 0.5D, where D is the diopter.
However, we need an electrically controllable lens for this system to work. So, liquid-crystal-based lenses are another area of liquid-crystal technology that could be important toward realizing the future “true” 3-D systems that meet all of the cues required that we listed earlier (relative motion, stereopsis, and focus). Recent developments in this area were described at the 2011 SID Symposium.13
The above system has two drawbacks that might be considered. One is that the viewer needs to wear glasses; the other is that the focus cue is not determined directly from the 3-D display, but in fact by the user’s eye convergence point. So rather than being an independent cue, it is derived from the stereopsis cue and the viewer’s response to it. Correcting both of these problems will require an autostereoscopic display that comes closer to emulating the window described at the beginning of this article, where the angular resolution of the rays is high enough so that multiple rays of light will be intercepted by the pupil of each eye, evoking the focus response to form a sharp image on the retina.14
Approaches that have been considered to yield a display of this type include holographic15 and integral-imaging-based devices.16 But due to the high information content required, flat-panel implementations of these approaches that can maintain all of the good characteristics of existing 2-D flat-panel displays have not currently been shown.
As first steps toward practical implementation of the above devices, lenticular-lens-based solutions are being considered.17 Two-view devices may not be acceptable in the long term, but advances are being made. In the short term, because of the tradeoff of image quality required for 3-D, a current aspect of liquid-crystal technology development has been switchable lenses, to allow for displays to switch between a 3-D and 2-D mode.18 As such, a device consisting of an autostereoscopic 3-D display, such as a notebook computer, could utilize the 3-D mode while displaying 3-D content such as 3-D movies or graphics applications and switch to 2-D mode while displaying 2-D content such as conventional productivity applications. Lenticular-lens-based autostereoscopic devices are making rapid progress toward having a wider field of regard while maintaining reasonable resolution. Especially notable are multi-view devices as demonstrated by Toshiba.19 Further improvements in auto-stereoscopic systems are achievable by limiting the number of viewers, as discussed earlier, and designing the system with knowledge of viewer’s location to be able to add a focus cue.20 Early commercial implementations of this approach in notebook computers are making their way into the marketplace,21 in which an eye-tracking software utilizing the built-in webcam locates and tracks the viewer’s eye positions such that the images corresponding to the left and right eyes are directed accordingly. So, while the challenges of autostereoscopic systems are great, we can expect that rapid progress will continue to be made.
In summary, the realization of a “true” 3-D display that mimics the real world to give an impression that the viewer is looking through a “window” requires the display system to provide all the important visual cues: relative motion, stereopsis, and focus. In this article, we have reviewed the electro-optical system design considerations to achieve such a system. The key requirements for liquid-crystal technology to yield high-quality 3-D displays appear to be within reach, if the implementations are constrained to a limited number of viewers who are willing to wear glasses. Technology advances are being made toward autostereoscopic display systems, especially for single-user devices in the near-term and ultimately the multi-user devices of the future.
4P. Bos and K. Koehler-Beran, “The pi-cell: A fast liquid-crystal optical switching device,” Mol. Cyst. Liq. Cryst. 113, 329 (1984); P. Bos and J. Rahman, “An optically ‘self-compensating’ electro-optical effect with wide angle of view,” SID Symposium Digest 24, 273 (1993).
7J. West, G. Zhang, A. Glushchenko, and B.-S. Ban, “Stressed Liquid Crystal Materials for Fast Display Application,” SID Symposium Digest 37, 761 (2006); J. Sun, Y. Chen, S-T. Wu, and R. Ramsey, “Submillisecond Response Sheared Polymer Network Liquid Crystals for 3D displays,” SID Symposium Digest 42, 86 (2011).
9H. Kikuchi and H. Higuchi, “Fast electro-optical switching in polymer-stablilized liquid crystalline blue phases for display application,” SID Symposium Digest 38, 1737-1740 (2007); H. Coles and M. Pivnenko, “Liquid crystal ‘blue phases’ with a wide temperature range,” Nature 436, 997-1000 (August 2005); S-T. Wu, “Blue Phase LCDs,” SID ’11 Seminar Lecture Notes (2011 ); H. Lee et al., “The world’s first blue phase liquid crystal display,” SID Symposium Digest 42, 21 (2011).
11M. J. O’Callaghan, M. Wand, C. Walker, W. Thurmes, and K. More, “High-Tilt, High-PS, de Vries FLCs for Analog Electro-Optic Phase Modulation,” Ferroelectrics 343, 201-207 (2006); M. Reznikov, L. Lopatina, M. O’Callaghan, and P. Bos, “The effect of surface alignment on analog control of director rotation in polarization stiffened SmC* devices,” J. Appl. Phys. 109, 054108 (2011).
12T. Shibata et al., “Stereoscopic 3D display with optical correction for the reduction of the discrepancy between accommodation and convergence,” J. Soc. Info. Display 13, No. 8, 665 ( 2005); T. Shibata et al., “The zone of comfort: Predicting visual discomfort with stereo displays,” J. Vision 11, No. 8, 1 (2011); M. Banks et al., “Consequences of Incorrect Focus Cues in Stereo Displays,” Information Display 7, 10 (2008); Hoffman et al., “Vergence-accomodation conflicts hinder performance and cause visual fatique,” J. Vision 8, 1 (2008).
13G. Love et al., “High Speed Switchable lens enables the development of a volumetric stereoscopic display,” Optics Express 17, 15716 (2009); See, Session 3, “Liquid Crystal Lenses for 3D displays,” papers in SID Symposium Digest 42 (2011).
15H. Stolle and R. Haussler, “A New Approach to Electro-Holography: Can This Move Holography into the Mainstream?,” Information Display 7, 22 (2008); S. Tay and N. Peyghambarian, “Refreshable Holographic 3D displays,” Information Display 7, 16 (2008).