Kollin, J. (1993). A Retinal Display for Virtual-Environment Applications. In Proceedings of Society for Information Display, 1993 International Symposium, Digest of Technical Papers, Vol. XXIV. (p. 827). Playa del Rey, CA: Society for Information Display.
The Virtual Retinal Display (VRD) is a unique display technology developed at
the Human Interface Technology Laboratory for Virtual Reality and other display
applications which require high-resolution in a low-profile, portable
head-mounted package. By scanning modulated light directly onto the retina of
the eye, we have eliminated the need for screens and heavy, expensive imaging
optics while allowing for higher resolution and wider field-of-view. Our
long-term (5 year) goal is a stereoscopic, full color, very high resolution (4K
x 3K pixels), 80 Hz refresh display which can be mounted on conventional
eyeglasses at a marginal cost of a few hundred dollars.
Existing virtual displays typically use a liquid crystal array, light emitting
diodes, or miniature cathode ray tube as an image source, then relay this image
via an infinity optical system to the eye. In order to match the
field-of-vision capabilities of the eye, an ideal virtual display system
should have a field-of-view of 140 degrees horizontally by 80 degrees.
Ideally, the number of picture elements should match the 1-3 minute-of-arc
dynamic acuity of the eye across the entire visual field. These
requirements necessitate the use of an image source capable of producing up to
8400 (horizontal) by 4800 (vertical) pixels. The conventional approach would
require an extremely high resolution miniature color image source to create the
original object, and large optical elements to relay the image to the eye.
Both CRT and LCD image-generation approaches generate real images, which are relayed to the eyes through an infinity optical system. The simplest optical approach is to view the image source through a simple magnifier lens. For fields-of-view greater than 30 degrees, this approach leads to a number of problems including light loss and chromatic aberrations. In addition, the optics are bulky and heavy. Virtual projection optical designs create an "aerial image" somewhere in the optical path which is then viewed as an erect virtual image. This approach increases the flexibility by which the image from the image source can be folded around the head, but large fields-of-view still require cumbersome reflective and refractive optical elements. Ideally, this optical system should be pupil-forming to gain maximum image quality; however, such an approach increases the number and size of the optical elements to provide a sufficiently large exit pupil diameter (15-20 mm diameter). Another approach is clearly needed.
The VRD works on the principle of a dynamic "Maxwellian-view optical system".
The instantaneous entrance pupil of the eye and the exit pupil of the virtual
display device are coupled so that modulated light is scanned directly on the
retina, producing the perception of a stable, erect image. The size of the
exit pupil can be smaller than the pupil of the eye, on the order of 1 mm. A
simplified ray tracing illustrates the basic principle below. An optical
layout from a working 2-axis prototype is shown in Ref. 1.
High-resolution scanners are used to deflect a beam of light both horizontally and vertically while the intensity of the light is modulated by the video signal in order to create an image. The lens of the eye focusses the light to a point on the back of the retina. As in an astronomical telescope, the angle of deflection of a collimated beam corresponds to a position of the focussed spot on the retina for any given eye position, just as if an image were at an infinite distance away from the viewer. In other words, as long as the light enters the pupil eye movement will be unimportant since it is the direction of the light which determines the apparent position of the image point. Therefore, when the viewer's eye moves, they will perceive a stationary image while looking at different parts of the scene. The lateral extent of the image is proportional to the angle of scan.
Obviously, it is not practical to place the scanners right in front of the eye. Instead, the scanner planes are relayed to the entrance aperture (pupil) of the eye, using anamorphic optics as necessary to scale the image and eye tracking to couple the exit aperture to the eye pupil if necessary. While the position of the exit aperture is not critical as long as light enters the eye, any resulting offset in the scan angle must be precisely registered with the scene imagery. If desired, the offset might be directly correlated to a positioning "index" of a 1Kx1K pixel "window" on a much larger frame buffer. Alternatively, it might be more desirable to use a number of parallel scanning systems to lower bandwidth - either by using scan conversion or taking full advantage of special rendering hardware.
In the current optical bench demonstration of the VRD, an acousto-optic
deflector is used to scan a red laser beam horizontally, while a galvanometer
is used for the vertical scan. The scanning system is driven directly by a high
resolution (RS 340) computer graphics frame buffer, such as a DEC VT 1300 or
Silicon Graphics VGX.
Our current scanners will allow for a resolution equivalent to more than 1000 x 1000 pixels. (Presently the display is about 500 x 1000 due to limitations in the laser diode driver and optical aberrations). After we reach this resolution we plan to increase the field of view from 40 degrees to over 80 and then construct a stereoscopic head-mounted version by early 1994. We will then incorporate eye tracking to increase the field-of-view and push the resolution to beyond HDTV levels. Full color will follow the expected development of suitable light sources. Eventually, we hope to use solid-state scanning technology so that the marginal costs and system size are reduced to the greatest extent possible.
The author would like to acknowledge the effort and electronic design skill of Robert A. Burstein and the inspiration and contributions of Prof. Thomas A. Furness.
1. Kollin, J. S. "The Virtual Retinal Display", to be published in the Proc. of 3Dmt '92, Montreal.
Fig.1 Ray-tracing of simplified version (1-axis) of Virtual Retinal Display