A Backlight for View-Sequential Autostereo 3-D

A backlight that emits collimated light whose direction can be scanned through 16° has been demonstrated. Combined with a high-frame-rate LCD, this could enable a stereo 3-D display that does not require glasses.

by Adrian Travis, Neil Emerton, Tim Large, Steven Bathiche, and Bernie Rihn

THE recent launch of 3-D displays has been thrilling, but let us not pretend that people actually want the glasses. No, what they really want is a 3-D image that they can see by doing nothing other than glancing at the screen. Providing this is quite a challenge, so many of the attempted solutions have been somewhat radical, requiring, for example, carefully registered lenticular arrays and high-resolution panels. What would be ideal would be to make minimal changes to technology that already exists. This article explains how we might get a 3-D LCD to work without glasses by altering the shape of the light guide in the backlight.

It is easy to forget that we see a picture on a display because rays of light travel from that display to our eyes. The glasses used with a classic stereo 3-D display block the rays trav-eling to one eye at a time so that the display can control what each eye sees. An alternative strategy is not to send the rays of light to that eye in the first place.1 This would require a backlight that is like that of an overhead projec-tor in the sense that it has the area of a liquid-crystal panel, but concentrates rays to a point that for our purposes is one eye of the observer. We must also be able to switch the direction in which rays are concentrated so that we can shine them into the other eye. Automobile headlights do this when they are dipped by switching between light bulbs in the focal plane of a curved mirror. However, overhead projectors and automobile headlights are rather bulky, so this article describes how to perform the same trick in a slim light guide.

The wave guide behind a conventional liquid-crystal panel is usually a wedge embossed with structures that make its emission uniform and diffuse. The passage of rays is predictable by tracing them through a stack of replicas of the wedge2 and the emission can be made partially collimated by letting fan-out take place in the same wedge as that from which light emerges.3 View-sequential 3-D, however, demands precise collimation, which requires that the surfaces be smooth and that the geometry of wedge replicates alone be used to achieve uniform intensity.4 We explain here how the direction of emitted light alters with the point of input in the manner needed for view-sequential 3-D.


Rays leave a light guide only upon reaching the critical angle, so we can trace rays in parallel at this angle backward from the surface at which they are to emerge. From within, any light guide appears like a kaleidoscope – as if the rays were traveling in straight lines through multiple reflections of the guide, as shown in the cross-section in Fig. 1.



Fig. 1: Rays appear to travel straight through a stack of light guides.


The authors arranged for the rays to reflect off the thick end before they emerge and curved the thick end so that the multiple reflections of Fig. 1 stack into a curve of constant radius, as shown in Fig. 2.



Fig. 2: The thick ends stack into a curve, which would focus rays to a point if they remained guided.


When parallel rays reflect off a curve, they converge toward a point of focus, but the rays in Fig. 2 will reach the critical angle and cease to be guided before they reach this point. So, we embossed the thick end with facets sloped to reduce ray angle and truncated the wedge at the point of focus, which is halfway to the center of curvature from the thick end.

The facets should ideally swivel the point of focus to a position where the central ray (the thick ray in Fig. 3) is reflected parallel to the plane of the wedge because the ray bundle will then be symmetric, which maximizes the ability of the light guide to collect light.



Fig. 3: The thick end is curved and faceted so that ray paths can be concentrated to a point at the input.


However, mirror images of the facets are formed at the interface between adjacent wedges, so the embossed structure must be symmetric, i.e., a zig-zag, which means that one-half of all backward-traveling rays are lost to the system.

In reality, the rays are traveling forward from the thin end to the thick end and because the situation they encounter is symmetric, no rays are lost: rays hitting upward sloping facets emerge from the upper surface of Fig. 1, while rays hitting downward sloping facets emerge from the lower surface. It was then a simple matter to add a mirror to one surface of the wedge so that all rays finally emerge from the same side.

The direction of rays resolved in the plane of the light guide must also be made parallel, which was done by giving the thick end of the light guide the same curvature as shown in Fig. 1; i.e., its surface (ignoring the facets) is spherical. Lastly, we added the usual prismatic film so that the rays that emerge into the surround at a shallow angle to the plane of the light guide were turned to the perpendicular.

With the uniform, collimated illumination that emerges, it is a simple matter to add a Fresnel lens that concentrates rays into the eye of a user. It is possible to switch between different points of concentration in the horizontal plane by switching between different LEDs at the input to the light guide. Two LEDs will be sufficient if there is only one user who is prepared to hold the screen perpendicular. Otherwise many LEDs and head-tracking may be necessary.


The Wedge backlight is an acrylic slab that tapers from a thickness of 6.2 mm to 10.8 mm over a distance of 320 mm and is 195 mm wide. At ±30 mm from the center of the thin end, three red, three green, and three blue light-emitting diodes were placed from right to left. Initially, only the red and blue light-emitting diodes were switched on, and they formed the image shown in Fig. 4 on a white surface placed 2 m in front of the backlight.



Fig. 4: Shown is a projection onto a screen 2 m from a wedge backlight with 60 mm between red and blue sources at the input.


Next, a Fresnel lens was placed over the surface of the light guide, whereupon there formed on a distant screen the image shown in Fig. 5.



Fig. 5: Here, a projection from a wedge backlight travels through a Fresnel lens onto a screen.


The width of the projected image was 160 mm, the result of which would be a 3-D image with a field of view of 16°.

If we exchanged the distant screen for a head, its left eye would see rays only from, say, one of the red LEDs and its right eye would see rays only from one of the green LEDs. Switch-ing between a pair of white LEDs instead, and adding an LCD panel fast enough for stereo, would provide the desired 3-D image.

Lastly, the central green LEDs were switched on alone and white diffusive paper was placed over the surface of the wedge in order to show uniformity. A photograph of the result is shown in Fig. 6, and more rigorous measurements at various points across the surface indicated a non-uniformity of less than ±10%.



Fig. 6: This wedge backlight example uses three green LEDs within a 30 mm width at input and with a white paper diffuser over the surface.


In order to assess the performance when illumination is off-perpendicular, the red LEDs were switched on instead, but there was no perceptible shadowing or vignetting.


A nascent market for autostereo 3-D exists in portable devices, for which 3-D glasses are almost unacceptable. A portable display typically has a single user, whose natural tendency is to hold it square-on, so head-tracking is unnecessary. The backlight described here would also help to reduce power consumption by not wasting light to wide angles, and the portrait orientation typical of a portable device works well with the optics of the light guide. The success of 3-D based on glasses may stem from its limited aims, and perhaps the lesson for 3-D is at first to be content with cracking one application at a time.

To the extent that it collimates light, the new light guide can be thought of as a flattened lens with a quasi one-dimensional focal plane at the thin end. Like all lenses, it has aberrations that limit performance both at large angles to the perpendicular and at short focal lengths, i.e., when the display is much wider than it is high. There exist more radical approaches, based, for example, on electro-wetting5 and virtual imaging,6 but if we want a wide field of view, we will need an LCD with the high frame rate needed to display more views or head-tracking or both.

Ferroelectric liquid crystals and polysilicon transistors have long had the switching times required to enable high frame rates, but the display industry has instead developed nematic liquid crystals and amorphous-silicon transistors, which switch too slowly for the display of many views. However, work on stereo 3-D and color-sequential displays has prompted the development of liquid crystals7-9 that switch on and off in less than 1 msec yet have the gray scale lacked by classic ferroelectrics. These effects typically require undesirably high switching voltages of more than 50 V, but LCDs are nevertheless following a trend of a rising frame rate.10Simultaneously, advances in head-tracking technology have led to a reduction in the number of views needed and therefore to a reduction in the frame rate required of the LCD. The authors therefore see a bright future for the view-sequential approach.


1A. Travis, "Autostereoscopic 3-D Display," Applied Optics 29, 4341-4342 (1990).

2A. Travis, J. R. Moore, and J. J. Zhong, "Opti-cal design of a flat panel projection wedge display," Proc. IDW '09, paper FMC6-3 (2002).

3K. Käläntär, S.F. Matsumoto, T. Katoh, and T. Mizuno, "Backlight unit with double-surface light emission using a single micro-structured lightguide plate," J. Soc. Info. Display 12, 379-387 (2004).

4A. Travis, T. Large, N. Emerton, and S. Bathiche, "Collimated light from a wave-guide for a display backlight," Optics Express 17, 19714-19719 (2009).

5A. Schwerdtner, N. Leister, R. Häussler, S. Reichelt, G. Fütterer, and A. Schwerdtner, "Eye-Tracking Solutions for Real-Time Holographic 3-D Display," SID Symposium Digest 39, 345-347 (2008).

6C. Lee and A. Travis, "Flat-panel backlight for view-sequential 3-D display," IEEE Proc. Optoelectronics 151, 486-489 (2004).

7N. Koshida, Y. Dogen, E. Imaizumi, A. Nakano, and A. Mochizuki, "An over 500-Hz Frame-Rate Drivable PSS-LCD: Its Basic Performance," SID Symposium Digest 40, 669-672 (2009).

8F. Castles, S. M. Morris, and H. J. Coles, "Fast-Switching Flexoelectric Display Device with High Contrast," SID Symposium Digest 40, 582-585 (2009).

9Y. Shimbo, Y. Takanishi, K. Ishikawa, E. Gorecka, D. Pociecha, J. Mieczkowski, K. Gomola, and H. Takezoe, "Ideal Liquid Crystal Display Mode Using Achiral Banana-Shaped Liquid Crystals," Jpn. J. Appl. Phys. 45, L282-L284 (2006).

10S. S. Kim, B. H. You, H. Choi, B. H. Berkeley, D. G. Kim, and N. D. Kim, "World's First 240-Hz TFT-LCD Technology for Full-HD LCD TV and Its Application to 3-D Display," SID Symposium Digest 40, 424-427 (2009). •


The authors are all with the Applied Sciences Group at Microsoft, which is led by Steve Bathiche. He can be reached at stevieb@ microsoft.com. The concepts stem from work performed at Cambridge University by Adrian Travis who is currently a visiting professor at the Université de Paris Sud. He can be reached at adriant@microsoft.comTim Large and Neil Emerton are physics researchers formerly with CamFPD who have performed the detailed design and made the light guides. They can be reached attlarge@ microsoft.com and emerton@microsoft.com, respectively. Bernie Rihn has recently graduated from Stanford University and he thinks spectacles (glasses) are pretty cool. He can be reached at bernier@microsoft.com.