| HITL Home | Projects Page | VRD Home |

Optical Engineering Challenges of the Virtual Retinal Display

Joel S. Kollin and Michael Tidwell
Human Interface Technology Laboratory, Washington Technology Center
University of Washington, Box 352142
Seattle, WA 98195-2142

Copyright 1995 Society of Photo-Optical Instrumentation Engineers.

This paper was published in Proceedings of Novel Optical Systems Design and Optimization and is made available as an electronic reprint with permission of SPIE. Single print or electronic copies for personal use only are allowed. Systematic or multiple reproduction, or distribution to multiple locations through an electronic listserver or other electronic means, or duplication of any material in this paper for a fee or for commercial purposes is prohibited. By choosing to view or print this document, you agree to all the provisions of the copyright law protecting it.


  1. Introduction
  2. Light Sources
  3. Scanner
  4. Optical System
  5. Holographic Optical Elements
  6. Experimental Results
  7. Summary
  8. Acknowledgments
  9. References

Abstract

 The Virtual Retinal Display (VRD) is a unique approach to developing a high-resolution head-mounted display currently under development at the University of Washington's Human Interface Technology (HIT) Laboratory. Rather than looking at a screen though a magnifier or optical relay system, the viewer of the VRD has a scanned beam of light enter the pupil of the eye and focused to a spot on the retina. This type of optical system is subject to different design constraints than a typical HMD. With the VRD it may be possible to realize higher resolution, greater color saturation, higher brightness and larger field-of-view than a traditional LCD or CRT screen-based system. In this paper the author will present the VRD approach and how it can provide these advantages. Issues to be resolved for the VRD to reach its full potential and some of the solutions developed at the HIT lab will also be discussed.

Keywords: Retinal scanning, HMDs, portable displays, HOEs, virtual reality, augmented reality

1. Introduction

This paper presents an overview of the VRD project at the HIT lab from an optical systems designer's viewpoint, followed by details of some system components, design criteria, and specific subproject developments - primarily in optical imaging. Due to both breadth and patent considerations it is necessary to mention some aspects of the project in passing only.

1.1 System Overview

The VRD can be considered a portable system that creates the perception of an image by scanning a beam of light directly into the eye. Most displays directly address a real image plane (typically a CRT or matrix-addressed LCD) which might be relayed to form a larger, more distant image for a head-mounted display (HMD). The VRD uses a scanned, modulated light beam to treat the retina as a projection screen, much as a laser light show would use the ceiling of a planetarium. The closest previously existing device would be the scanning laser opthalmoscope (SLO) which scans the retina to examine it; the SLO is designed to capture light returning from the eye whereas the VRD is designed as a portable display. Another class of related systems are "Maxwellian-view" systems used in psychophysical research.

The VRD has several advantages over CRTs, LCD, and other addressable-screen displays :

All components in the VRD are small and light, making them ideal for use in a portable display.

1.2 History

The VRD display concept was initially conceived by Dr. Thomas A. Furness as a means of eliminating large aperture optics and expensive high-resolution addressable images sources such as CRTs. Soon after joining the HIT Lab in 1991, Joel Kollin realized a key feature about the VRD - movements of the eye would not result in perceived movement in the image. Therefore, eye tracking would not be necessary beyond that what might be needed to ensure that the light beam entered the eye. He then designed and constructed the original bench-mounted VRD, using an acousto-optic device as the horizontal scanner. Electronics largely designed and built by Bob Burstein then allowed it to be driven directly by a DEC workstation, although it was still significantly lower in both contrast and resolution than a standard SVGA display and offered an image only in uncalibrated shades of red. We subsequently began work on patenting the display and brought on board David Melville to engineer the mechanical design, especially a new scanning system. In 1993, a newly formed corporation, MicroVision Inc., licensed the VRD technology and signed a 4 year, $5.1 million development contract with the University. Rich Johnston was hired specifically to manage the VRD and other hardware products of the Lab. By forming relationships with other researchers in the College of Engineering, he has orchestrated a program to solve the challenges and bottlenecks of the project.

In late 1993 and 1994, Mike Tidwell redesigned the VRD to maximize the resolution possible with the A-O scanner while David Melville designed a new Mechanical Resonant Scanner (MRS) which would be capable of the high rates of horizontal scanning without the costs and other limitations of the A-O devices. The MRS was then utilized in full-color inclusive and "see-through" systems.

1.3 Subsystems

In the VRD a light source is modulated with image information, either by direct power ("internal") modulation or by an external modulator. The light is passed through an x-y scanning system, currently the MRS and a galvanometer. Light from the scanner pair enters an optical system, which in present implementations of the VRD forms an aerial image and then uses and eyepiece to magnify and relay this image to infinity.

2. Light sources

2.1 Coherence and Source Size

According to conventional additive color theory, any color can be represented as a mixture of three appropriately chosen primaries. The three ideal VRD light sources would be monochromatic for maximum possible color saturation. The ideal colors would be near the corners of the CIE chromaticity chart but not too far in the blue or red as the nominal gains in addressable color space would be more than offset by sensitivity decrease in the eye. Spatial coherence is also important - larger source spots will correspond to larger spots on the retina, decreasing resolution. The primary cause of the real (if sometimes exaggerated) hazards of laser light are the result of spatially coherent light focusing to a small area on the retina, causing highly localized heating and ablation of tissue. In the VRD the spot is traveling in two directions and even when stationary is not at a power level that would cause damage. We are working with ophthalmologists and will publish a definitive article on this in the near future. Incidentally, polychromatic sources can be shown to form spots comparable to monochromatic ones of the same spatial extent. Therefore spatial coherence is responsible for the small spot size which leads to both high resolution and (given enough power) retinal hazard.

To achieve the desired resolution, all current VRD prototypes have used lasers for their superior spatial coherence characteristics. In order to use a point source such as an LED, the image of the source should be smaller than the diffraction limit of the scanner (Fig. 1). Using the lens magnification, one can determine the maximum source size that can be used before degrading the diffraction limited spot size at the image plane. The angular divergence of the source is effectively limited by treating the scanner as a stop. Light which does not hit the mirror does not contribute to the image plane spot size. From this geometric argument we can derive an equivalent point source size between 4 and 5 microns for a VGA resolution image in our current system. For a system where the scanner is illuminated with a collimated Gaussian beam, similar arguments can be made to determine the required divergence and beam waist from the equations for image plane spot size.

(Figure 1)

Fig. 1 Relationship between source size, objective focal length, scanner aperture & image spot size

2.2 Devices

Red laser diodes have more than adequate modulation for our current needs but suffer from contrast problems due to effects of extinguishing and re-establishing the lasing action. When the current drops below the lasing threshold the laser acts like an LED, exhibiting reduced optical power and coherence but still glowing. This can be seen as a bend in the current vs. optical power curve. We presently adjust the range on this curve to obtain satisfactory image contrast. External A-O devices are used to modulate the gas lasers currently used as blue and green sources. We intend to switch to solid-state lasers in the green and blue in the very near future as they become commercially available.

Work is proceeding on the use of LEDs as the VRD light source. LEDs can be directly modulated, are inexpensive, and are available in a wide range of colors. The spatial coherence and modulation bandwidth of available visible LEDs is below the requirements for high-resolution displays. We are working with LED manufacturers to eliminate these issues.

3. Scanners

The horizontal scanning mechanism of the VRD must be capable of both relatively high scan rates (15 kHz-90+ kHz) and high resolution (500-2000+ pixels) for NTSC to HDTV formats, respectively. To date we have built SVGA format systems (80 kHz) in monochrome/greyscale using a A-O scanner and 30 kHz in full-color with a mechanical resonant one.

3.1 Acousto-Optic Scanners

High-speed A-O scanners require a highly anamorphic beam and expensive drive electronics. In addition, they suffer from light loss, small deflection angles, variations in diffraction efficiency and non-linear chromatic aberration. For these reasons we now use the more commercially viable mechanical devices.

3.2 Resonant Scanners

Mechanical resonant scanners are free from chromatic aberration, extremely efficient, small and inexpensive to produce. The MRS will be discussed at length in a future paper, but certain issues involved with all resonant scanners might be noted here as they directly affect the optical design. By their nature, resonant scanners are very efficient as they convert potential energy to kinetic and back again, in a manner similar to a pendulum. Because of their sinusoidal scan, it is highly desirable to use resonant devices as bi-directional scanners . At the current time the MRS is capable of roughly +/- 12 degrees at 3 mm clear aperture at 15kHz bi-directional (i.e. 30 kHz effective scan rate.) We expect this will soon improve to greater than +/- 20 degrees with the next generation of driver circuitry. Larger scan angles mean less magnification is necessary, leading to larger exit pupils and a smaller system.

We are still researching other scanning technologies, especially those which might have advantages that outweigh the low cost and high efficiency of the MRS. Currently we are using a galvanometer for the vertical scanner while a less expensive solution is being developed.

4. Optical System

4.1 Basic Constraints

Aperture size and scan angle are directly related by the Optical (LaGrange) Invariant which states that the product of the field angle (tangent) and aperture is constant. Aperture size and resolution are proportional, and are usually related by the Rayleigh Criterion:

h = 2.44 l (f/#)

where h is the resolvable spot size (diameter) and f/# (f-number) is the imaging element diameter divided by the distance to focus.

In a typical retinal scanning system, a telescope is used to form a reduced "image" of the scanner at the eye. This is not the image seen by the user, but an exit pupil or aperture stop of the system in image space. By forming an exit pupil smaller than the scanner aperture, the scan angle is effectively amplified by a proportional amount (Fig. 2). In other words, the ratio of the size of the aperture stop image (exit pupil) to the original scanner is inversely proportional to the magnification of the aerial image, in accordance with the Optical Invariant. However, this smaller exit pupil can lead to two problems: difficulty in making sure the light still enters the eye, and reduced resolution due to diffractive spreading as given by the above equation.

(Figure 2)

Fig. 2 Angular magnification in a pupil-forming system

Wide FOV polychromatic eyepiece design is non-trivial - adding light weight and form factor constraints makes it even more challenging. Here the small instantaneous ray bundle size helps considerably. For instance, with the fast (small f-number) eyepiece only the aberrations across one beam width will degrade the image, rather than over the entire aperture.

4.2 System Design Considerations

For a simple two-lens system, the objective/eyepiece pair will ÔscaleÕ the scan by the ratio of the lenses:

Tan Qeye / Tan Qscan = fobj / feye

where fobj and feye are the focal lengths of the objective and the eyepiece, respectively. The f/#'s of the lenses will be similarly scaled. The critical imaging component in this system would therefore be the eyepiece, whose f/# can limit the field-of-view of the system. Low f/# or "fast" lenses tend to be heavy and complex, especially for polychromatic systems. The first polychromatic system used a telescope eyepiece. The eyepiece was subsequently replaced with a mirror in a modified Schmidt camera configuration. Mirrors are simple, rugged, have no chromatic aberration and can be very light for their power. They can also be made partially reflective for a "see-through" system. The mirrorsÕ flaws and aberrations should be limited to approximately one-quarter wave across the any pupil-sized area on its surface, which is a much easier target to reach than quarter-wave accuracy over the entire aperture. Another advantage of mirrors used in a Schmidt configuration is that post-objective scanning can be used. Such systems are simpler and works well with the resulting curved aerial image field. One drawback of the classical Schmidt design is the field-of-view limitation due to the fold mirror keeping the viewer away from the mirror.

An important limit on field-of-view for any pupil-forming system is eye relief. The FOV can be defined by the (half-)angle formed by the chief or principle ray with the optical axis. This implies that its tangent is inversely proportional to eye relief, defined as the distance from the final element to the exit pupil, for an eyepiece of given focal length. Therefore a wider FOV system will less eye relief. Note that inadequate eye relief (less than 20 mm) may prohibit the use of eyeglasses, and even more eye relief may be needed for a wide FOV system. This is because the exit pupil must be closer to the center of the eye's rotation to permit viewing the edges of the image.

Pupil distortion is another concern. If the center of the exit pupil is farther from the eyepiece than the edges of the pupil, the eye must be moved closer to the eyepiece when viewing the periphery of the image. This is clearly unacceptable for a head-mounted display.

4.3 Image Distortion

The scanned image through a conventional eyepiece suffers from Ôf-QÕ distortion, where the image size is proportional to the tangent of the scan angle rather than the angle itself, i.e.

X = tan Q

where X is the displacement of the spot relative to the center as a function of scan angle. Normally this is solved by using a special 'f-Q' or scan lens that follows the condition X = Q instead. In addition, with the MRS we also must deal with the sinusoidal scan pattern:

Q = A sin (Kt)

where A is the maximum amplitude, K is a proportionality constant related to the frequency and t is time. If we combine the equations above, we have

X = tan [A sin (Kt)]

Where A is the maximum half-angle excursion of the sinusoidal scanner. If A = 1 radian, we find that X is a nearly linear function of t over most of its range. Unfortunately, with the +/- 12 degrees range of the current MRS scanner the function is dominated by the sin function. Currently we are implementing a frame buffer to remap the image to compensate for this distortion. This will also allow for more flexibility in the optical design by allowing us to electronically correct distortion while optimizing other constraints such as form factor and resolution.

5. Holographic Optical Elements

5.1 Definition

Holographic Optical Elements (HOEs) are diffractive optical elements that are constructed holographically i.e. by interfering two beams of light. Typically one beam resembles the "playback" beam that will illuminate the HOE in the final system. The second beam corresponds to the "image" beam desired to exit the HOE upon playback (Fig. 3). In practice, the construction beams can differ from this paradigm since there is considerable variation in the playback beam, especially in a wide field scanning system. In any case, most HOEs can be approximated as holograms of point sources or collimated beams of light, such that light from one source is imaged into the other.

(Figure 3)

Fig. 3 Recording and Playing a HOE (Off-Axis Parabola Shown)

5.2 Characteristics

The diffractive power of the HOE leads to several interesting possibilities as well as a new set of constraints. For instance, "curved" mirrors can be made on a flat substrate by interfering expanding and contracting beams of light from different sides of the HOE plate. This can be accomplished by simply placing the HOE plate between a spatial filter (beam expander) aperture and a concave mirror, with the beam reflecting from the mirror converging back to the aperture. The interference fringes thus recorded will resemble an actual mirror, as the fringes formed are normal to the intersection between the crossed construction beams. More specifically, they will be the same "speed" (f/#) as the construction mirror but with a focal length given by the distance between the HOE and the aperture (Fig. 4)

(Figure 4)

Fig. 4 Construction of a non-conformal reflection HOE using a spatial filter beam expander and concave mirror. Width of HOE and fringes are greatly exaggerated.

The HOE in Fig. 4 would be considered a non-conformal reflection HOE, meaning the fringes are not conformal or parallel to the surface. The playback beam will reflect off the HOE since the construction beams impinge from different sides, causing the interference fringes to be formed roughly perpendicular to the propagation directions. If the HOE is recorded on a curved substrate so that the fringes conform to the surface, it would act as a multi-layer coated mirror. A HOE of this type can reflect nearly all of the light whose propagation vector component k was twice that of the fringe spacing - this is known as the Bragg condition (Fig. 5). The Bragg condition is the combination of wavelength and angle that causes the impinging and diffracted light of the HOE to be in phase with each other. Not surprisingly, the wavelength and angle of the construction beams will determine the Bragg condition of the HOE fringes, although this can be modified after recording by chemical or physical distortion of the fringes. For a diffuse and/or polychromatic illumination source, those rays meeting the Bragg condition will be diffracted much more strongly - this is known as Bragg selection. We say that holograms with Bragg selection are "thick", meaning they have significant structure inside the plane of the hologram, which may be less than 7 microns but still contain 10 or more fringes to form a "multilayer stack". Reflection holograms are by definition thick, whereas transmission holograms can be thick or thin depending on the exposure geometry and choice of material. Embossed or surface relief holograms, such as those found on credit cards, are thin holograms and therefore have little or no Bragg selection. They are colorful because they are "rainbow" or Benton holograms, which are vertically dispersive transmission holograms mounted in front of a mirror. Thick reflection and transmission holograms can approach 100% diffraction efficiency (at least for a point-source hologram), whereas thin transmission holograms are limited to 33.9% maximum theoretical efficiency.

(Figure 5)

Fig. 5 Satisfying the Bragg condition for first-order diffraction off of a plane mirror

Besides a decrease in the bulk and weight, HOEs have several other advantages over conventional elements due to Bragg selection. For a see-through system, the wavelength used by the display can be selectively reflected from the HOE which also tends to reject outside light of the same color. This means higher contrast relative to a broadband beamsplitter for a given brightness and/or transmissivity for the system as a whole, although the outside light may be color shifted as a result. Also, the HOE can multiplex functions by acting differently for beams impinging from different directions or with different wavefronts. For example, light might pass through the HOE virtually undisturbed, be reflected from a beamsplitter, and then be magnified by the HOE. This is a system being tested by HITL as described below. A more complex system might act as a prism for one wavefront and a lens for another.

A reflection hologram for a full-color system can be accurately recorded by three separate exposures at chosen red, green, and blue laser wavelengths. Each set of fringes is Bragg selective for a narrow bandwidth - usually around the exposure wavelength, although this can be shifted and the bandwidth broadened. Since each color is independently exposed, it is possible to modify the exposure geometry to minimize chromatic aberrations or to combine spatially separate color images. Each exposure uses up some potential refractive index modulation, but this limitation can be circumvented by copying the HOE. This is often accomplished by placing the copy hologram in contact with the master and exposing it to the playback beam. The resulting image beam from the master interferes with the playback beam to form nearly identical fringes in the copy hologram. Computer generated diffractive optical elements can also be "contact copied" in this manner.

In general, HOEs also allow greater freedom in an optical system since diffractive optical elements cause chromatic dispersion opposite in sign to refractive elements. Diffractive power can be added to a refractive and reflective element by incorporating diffractive features on a lens or mirror surface. Thin holograms or other diffractive devices can be pressed onto the surface of molded or cast parts. Thick holograms can be placed on curved reflectors or lenses by coating the surface with photopolymer or gelatin and exposing. Conformal exposure increases expenses quite a bit - from perhaps $5 to $50, whereas complex lenses and reflectors can be molded for under $5 in large quantity.

Thick HOEs have some other disadvantages besides cost. Bragg selection can limit the field of the HOE, with efficiency and FOV decreasing away from the Bragg condition. At the same time, insufficient Bragg selection can lead to significant dispersion in a non-monochromatic source such as an LED, and even cross-talk between the different color exposures. Also, flare from bright ambient light could be diffracted into the eye by a reflection HOE acting as a transmission hologram. This is more of a problem for non-conformal holograms, where the fringes can be perpendicular to the plane of the hologram at the surface, allowing for significant diffractive transmission. Proper design and the use of baffles can help minimize flare. Also, non-conformal holograms can lead to distortion, since the HOE might intersect the scanned beam at a different point a curved mirror would.

5.3 Materials

While the first holograms and HOEs were recorded on silver halide material, (aerial photography plates and films) the high efficiencies mentioned above require nearly perfect phase materials. Silver halide holograms can be "bleached" to change their amplitude variation into phase retardance, but this is usually at the expense of increasing the film grain noise inherent in the material. Dichromated gelatin (DCG) is the traditional recording medium for HOEs as it is nearly noiseless and very efficient. Unfortunately, DCG is also toxic, fragile, usually mounted on glass, and requires tight process control and a perfect seal against moisture and humidity. More recently, photopolymers have become available as a more durable and easier alternative. DuPont has produced a family of products (Omnidex) which require no wet processing, only flooding with white or UV light after exposure and heating to raise the diffraction efficiency. While Omnidex does not match the high index of refraction variation and efficiency of dichromated gelatin, it does meet the requirements of many HOE applications and is relatively impervious to environmental stress including heat and water. Polaroid also has a family of materials which offer some interesting variations in reflection bandwidth and refractive index modulation. It should be noted that all of these materials require significantly more light to expose than silver halide products. Because conventional silver-based films and plates have a grain structure to amplify the exposure, since each grain is activated as a unit. While silver halide is more sensitive (since it takes fewer photons to expose an area of film), this grain amplification effect adds noise and scatter, making silver materials less desirable for technical applications.

5.4 HITL HOE Prototype

After constructing several promising HOEs in DuPont photopolymer to test Bragg selection and image quality, a VRD system incorporating a HOE was designed and built as shown in Fig. 6. In this system, light from the MRS/objective system is transmitted undisturbed at a shallow angle through a reflective HOE and is reflected from a beamsplitter. (Light transmitted through the beamsplitter on this first pass simply passes off to the side of the viewer's eye.) The reflected light now impinges on the HOE with the optical axis of the HOE co-linear with the center scan beam (i.e. on-axis). Light reflected from the HOE is now collimated and the aerial image is relayed to infinity. For this particular implementation we tilt the HOE vertically so that the specular reflection of the scanner off of the HOE substrate misses the eye. This leads to a small amount of keystoning in the image, and might be avoided by the use of anti-reflective (AR) substrate and/or polarizing elements instead of tilting.

(Figure 6)

Fig. 6 See-through VRD system incorporating HOE eyepiece (not to scale)

6. Experimental Results

Our first VRD system incorporating HOEs was designed using ZEMAX and exposed onto DuPont OmniDex 706 film. We started by constructing simple, nearly on-axis non-conformal 35mm focal length, f/0.8 mirrors by two different methods. At first we used a two-beam exposure at 532nm with back to back singlets to converge an expanded, spatially filtered beam to a point near the pinhole of another spatial filter. This method allows for more flexibility for future experiments with off-axis systems, but has a few drawbacks. First there is distortion introduced by the singlet pair - we cannot be measure this or predict it using ZEMAX directly but it is fairly obvious barrel distortion which accentuates rather than correcting the sinusoidal distortion. Also we noticed that there were several "false foci" recorded on the HOE - these were from stray surface reflections and manifested themselves as extra exit pupils with distorted images. These will presumably be reduced or eliminated by anti-reflective coatings on the optics and plate.

We then made a simple one-beam "Denisyuk" HOE by using a mirror to reconverge light to the spatial filter, as shown in Fig. 4. As in the previous setup, Bragg angle can be altered by varying the conjugates from the nominal 1:1 ration. Stray reflections were reduced, with only the specular highlight from the scanner being apparent. This system also makes it easier to construct multiple color HOEs by sequential exposure in different wavelengths. So far the results obtained for a two-color HOE recorded on DuPont 700HD 15-20 full-color reflection film have been disappointing, but we have just begun experimenting with this material. Exposure times are very long - 10-12 minutes for red (633nm He-Ne) followed by 3-4 minutes for the much stronger green (532nm doubled Nd-YAG). Once we have obtained better results and have optimized recording conditions, we plan to try three-color exposures with blue for a full-color VRD system.

A testing system now under construction will enable us to quantitatively measure system MTF, spot size, resolution and contrast. We can therefore only offer preliminary results based on test patterns which are observed visually. For the single color (green) HOEs we were able to resolve the full resolution of 640 by 480 pixels except for extreme left and right edges of the display, which are compressed by the sinusoidal distortion. This was determined by displaying gratings of alternating off and on pixels in the vertical and horizontal directions. Cosmetically, the holograms show surface defects incurred during lamination and striated variations in diffraction efficiency. The two-color HOEs were much dimmer and noisier than the green ones in the VRD setup, possibly due to errors in the Bragg selection, long exposure times necessary because of limited He-Ne power, and mechanical properties of the material.

Diffraction efficiency is probably also reduced by the current lasers used in the VRD - 543nm He-Ne for the green and 650nm (nominal) red laser diode. Baking the holograms seems to make them brighter in white light but not in the system, leading us to believe that efficiency is not improving much for our choice of off-peak wavelengths. Successful use of HOEs will require careful control over the peak Bragg wavelength and/or broad spectral bandwidths to match the source. LEDs are fairly well-matched to the bandwidth exhibited by DuPont photopolymer holograms if the peak wavelengths are close. In the next year we hope to add a wet darkroom and environmental controls to our optics facility which will make it much easier to try other materials such as those made by Polaroid. Recently we have also constructed a "VRD simulator" (Fig. 7) which allows us to examine optical subsystems (eyepieces) of the VRD without requiring access to the scanner and being limited in optical power or wavelength. The basic concept is to take light from a slide transparency and project it aerially as a real image in front of a apertured projection lens. The aperture of the projection lens corresponds to the stop in the system, simulating a scanner aperture and resulting pupil size. The distance from the projection lens to the image and the focal length determines the size and field angle of the aerial image, which is then re-imaged by the tested system which also forms an image of the projection aperture to the eye. The image can be made as bright as desired to test inefficient systems prior to optimization, and the chromatic response to a white-light source or various filtered colors can also be tested. This is especially useful for testing HOEs with Bragg condition problems. It is not currently capable of simulating distortion other than what is introduced by the projection optics.

(Figure 7)

Fig. 7 VRD Simulator

7. Summary

The Virtual Retinal Display has significant performance and ergonomic advantages, especially for portable displays. Subsystems and issues currently being addressed at the HIT Lab include: light source development, scanning technology, optical system engineering and compact eyepieces. In particular, holographic optical elements are a promising means of providing see-through eyepieces in a small form factor with unique advantages.

8. Acknowledgments

This work was performed under contract with MicroVision, Inc., 1420 Fifth Ave., Suite 2200, Seattle WA 98101.

The authors wish to acknowledge the critical contributions of the rest of the VRD team, who presently include David Melville, Rich Johnston, Prof. Thomas A. Furness, Bob Burstein, Heather Patrick, Steve White, Prof. Kelin Kuhn, Prof. Thomas Pearsall, Dan Bertolet, Carrie Cornish, Phillip Allison and Archie Gonzales. We also wish to thank Rich Rallison of the Ralcon Corportion for his helpful suggestions and Prof. Shawn Brixey of University of Washington Cross-Disciplinary Art program for his generous loan of equipment.

9. References

R. H. Webb, G. W. Hughes, and F. C. Delori, "Confocal scanning laser ophthalmoscope," Appl. Opt., Vol. 26, No. 8, pp. 1492-1499, 1987.

R. J. Jacobs, I. L. Bailey, and M.A. Bullimore, "Artificial pupils and Maxwellian view ," Appl. Opt., Vol. 31, No. 19, pp. 3668-3677, 1992.

R. Johnston and S. Willey, "Development of a commercial retinal scanning display," Helmet- and Head-Mounted Displays and Symbology Design Requirements, SPIE Vol. 2465, 1995.

J. S. Kollin, "The Virtual Retinal Display," Proc. 3Dmt '92, Montreal.

J. S. Kollin, "A retinal display for virtual environment applications," SID IntÕl Symp. Vol. XXIV, pg, 827, 1993.

M. Tidwell, R.S. Johnston, D. Melville, and T.A. Furness, "The virtual retinal display - a retinal scanning imaging system," Virtual Reality World '95 Conference Documentation, IDG Conferences. pp. 325-333, 1995.

M. Tidwell, "A virtual retinal display for augmenting ambient visual environments," unpublished Master's Thesis, Department of Electrical Engineering, University of Washington, 1995.

J. W. Goodman, Introduction to Fourier Optics, McGraw-Hill, New York, 1968.

W. J. Smith, Modern Optical Engineering, 2nd Ed.,McGraw-Hill, New York, 1990.

G. F. Marshall, Optical Scanning, Marcel Dekker, New York, 1991. Smith, 1990.

R.S. Johnston, unpublished HIT Lab Technical Report 1994. Tidwell, 1995.

R. Kinglake, Optical System Design, Academic Press, Orlando, 1983.

P. Hariharan, Optical Holography, Cambridge University Press, Cambridge, 1984.

W. Gambogi, K. Steijn et. al., "HOE imaging in DuPont holographic photopolymers." Diffractive and Holographic Optics Technology, SPIE Vol. 2152, 1994.

A.M. Weber, W.K. Smothers, T.J. Trout, and D.J. Mickish, "Hologram recording in DuPont's new photopolymer materials," Practical Holography IV, SPIE Vol. 2152, 1990.

Human Interface Technology Lab