Optical combiner

An optical combiner is the optical element in an optical see-through augmented reality (AR) display that merges the user's direct view of the real world with the virtual image produced by a microdisplay or light engine. It is the defining component of any optical see-through Head-mounted display, because it lets the eye see through to the physical scene while a second, synthetic image is overlaid on top of it. The combiner is what makes the display "see-through" rather than fully immersive: a virtual reality headset blocks the outside world and needs no combiner, whereas an AR headset must add light to the scene without taking the real world away.

Because the combiner has to be partly transparent and partly reflective at the same time, it sits at the heart of the hardest trade-offs in near-eye display design. The same element must transmit enough ambient light for a natural view of the world, redirect enough projected light to make the virtual image bright, do so across the full color range and the whole Field of view, keep both images sharp, and still fit in a wearable form factor. Improving any one of these usually costs another, so every combiner architecture is a compromise.

What an optical combiner does

A see-through AR display has two light paths that must end up overlaid on the retina. The first is the see-through path: light from the real scene passes through the combiner to the eye. The second is the virtual path: light from a small image source (typically an LCOS, micro-OLED or laser microprojector) is relayed and bent by the combiner so that it also reaches the eye, appearing as a magnified virtual image at a comfortable focus distance. The combiner is the single surface or substrate where these two paths join.

The physical conflict is fundamental: a combiner has to change the direction of the projected light while leaving the real-world light passing straight through it as undisturbed as possible. A surface that reflects more of the virtual image also tends to reflect (and therefore block or distort) more of the real scene, so designers must split the available light between the two paths. This is why combiner design is dominated by figures such as how much real-world light is transmitted, how much projected light reaches the eye, and what unwanted reflections or color errors the element introduces.

Key trade-offs

Combiner architectures are usually judged on a recurring set of properties, most of which pull against each other.

• Field of view (FOV). The angular size of the virtual image. A wider field is more immersive but generally needs a larger or more complex combiner, and in waveguide designs it is capped by physics (see below).
• Eye box. The volume in front of the eye within which the whole image stays visible. A larger eye box tolerates different face shapes, eye movement and a range of interpupillary distances; some combiners exist mainly to enlarge it by replicating the exit pupil.
• See-through transparency. The fraction of real-world light that reaches the eye. High transparency gives a natural view but leaves less of the surface available to redirect the virtual image.
• Optical efficiency. How much of the light engine's output actually reaches the eye. Low efficiency forces a brighter (and more power-hungry) display to stay visible in daylight.
• Image quality. Sharpness, color uniformity, contrast, and freedom from ghost images or rainbow artifacts.
• Form factor. Thickness, weight and how close the design comes to ordinary eyeglasses.

A deep constraint underlying several of these is the conservation of etendue, the product of the eye box area and the field-of-view solid angle. For a given image source, a compact combiner cannot freely enlarge both the field of view and the eye box at once: pushing one up tends to push the other down. This is one reason wide-FOV AR optics so often have small eye boxes, and vice versa.

Combiner types

Optical combiners fall into two broad families. Free-space (or "off-axis") combiners use macro-optical elements such as flat or curved half-mirrors and prisms placed in the open light path. Waveguide (or lightguide) combiners trap the projected light inside a thin transparent substrate by total internal reflection and release it in front of the eye using micro- or nano-scale structures. Free-space designs tend to be optically efficient and easy to make but bulky; waveguides tend to be thin and glasses-like but harder to manufacture and less efficient.

Simple beam splitter (half-mirror)

The most basic combiner is a single flat partially reflective plate, a beam splitter or half-mirror, tilted in front of the eye. The projected image reflects off the plate toward the eye while real-world light passes through it. The approach is simple, cheap and optically efficient, and because reflection from a coated surface is wavelength-independent it has good color fidelity. Its weakness is that a flat plate has no focusing power, so a wide field of view needs a large plate placed well in front of the face, which becomes bulky. Many older or low-cost AR headsets and most automotive head-up displays use a tilted-plate combiner, and the principle is the starting point for the birdbath design below.

Birdbath

The birdbath combiner adds focusing power by pairing a flat beam splitter with a curved (usually spherical) partially reflective mirror; the curved element resembles a garden birdbath, which gives the design its name. Light from the display first reflects off (or passes through) the beam splitter, strikes the curved combiner which magnifies it and collimates it, and is then routed to the eye, while ambient light passes through both elements to give a see-through view.

The birdbath gives a relatively wide field of view (commonly 40 to 55 degrees in shipping glasses), good sharpness and accurate color, and it is comparatively inexpensive, which is why it is the most common architecture in recent consumer AR display glasses. Its core drawback is light efficiency. Because the projected image makes two passes at the beam splitter, the throughput is the product of the two pass efficiencies: a nominal 50/50 splitter transmits only about 23 percent of the display light (roughly 0.48 x 0.48), and the see-through path is dimmed as well. To keep the virtual image bright enough, many birdbath glasses block most incoming light; some designs transmit only on the order of 25 to 30 percent of real-world light, and reflections "rattling around" between the display, splitter and mirror can produce faint ghost images. Birdbath optics also add noticeable depth in front of the eye, so they look more like goggles than eyeglasses. Google Glass used a compact prism-based variant of this idea, and the architecture is widely used in tethered display glasses such as those built around Sony micro-OLED panels.

Freeform prism

A freeform prism combiner replaces the flat plate and separate mirror with a single molded block of plastic or glass whose surfaces are non-rotationally-symmetric "freeform" shapes. Inside the prism the projected light is steered by a combination of refraction, total internal reflection and one partially reflective surface, so that the block both magnifies the image and folds the path into a compact volume. To keep an undistorted see-through view, the prism is usually cemented to a matching freeform corrector lens that cancels the prism's deflection of real-world light.

Freeform prisms can deliver a fairly wide field of view (designs around 50 degrees diagonal and beyond have been reported) in a relatively compact, efficient package, and they were used in products such as the Meta 2 developer headset and Epson Moverio-style optics. The trade-offs are that the molded freeform surfaces are demanding to manufacture and align, the prism adds thickness and weight, and the optical power is fixed, which (as with most combiners) leaves the vergence-accommodation conflict unsolved.

Diffractive waveguide

Diffractive waveguides are the basis of most high-end mixed-reality headsets. The projected image is coupled into a thin, flat, transparent substrate by an input diffraction grating, propagates across the lens by total internal reflection, and is coupled back out toward the eye by an output grating. The gratings are periodic nano-scale structures (surface-relief or volume gratings) etched or imprinted into the substrate. Crucially, the output grating replicates the exit pupil many times across the lens area, which expands the eye box, and the resulting element is thin and highly transparent, close to an ordinary lens.

The chief limitations are field of view, efficiency and color. The maximum field of view is set by the substrate's refractive index and the critical angle for total internal reflection, so single-substrate diffractive waveguides have historically been limited in FOV, although dual-channel pupil-expansion designs have reached around 70 degrees diagonal. Because diffraction is wavelength-dependent, the different colors spread differently inside the guide, producing color non-uniformity and a visible "rainbow" effect; full color usually requires stacking two or three substrates for the red, green and blue channels. Efficiency is low, because light is lost at the input coupler and a little more energy leaks out at every pupil replication. Microsoft HoloLens (both the original, which stacks three waveguides, and HoloLens 2, which uses a two-layer scheme) and Magic Leap (the Magic Leap 2 reaches a 70-degree diagonal field of view) are the best-known diffractive-waveguide products.

Reflective (geometric) waveguide

A reflective, or geometric, waveguide guides light by ordinary reflection rather than diffraction. The projected image enters a thin transparent slab and travels by total internal reflection until it meets a cascade of embedded, partially reflective mirror facets (often built by stacking coated plates and slicing them), which progressively couple the light out toward the eye and replicate the pupil to widen the eye box. Because reflection is wavelength-independent, reflective waveguides avoid the rainbow color errors of diffractive designs and give excellent color uniformity, and they are markedly more efficient: a single set of mirrors handles all colors at once. The vendor Lumus, which specializes in this approach, has reported luminance efficiency on the order of several thousand nits per watt and up to roughly ten times the efficiency of competing diffractive waveguides. The drawbacks are manufacturing: embedding many semi-reflective facets uniformly and without visual defects across a lens is difficult and costly, and older one-dimensional designs grew thicker as the field of view increased. Reflective waveguides are common in enterprise and defense AR, and Lumus reflective waveguides have been adopted by partners building industrial and military headsets.

Holographic optical element

A holographic optical element (HOE) combiner is a thin film of holographic material in which an interference pattern has been recorded so that the film behaves as a wavelength- and angle-selective lens or mirror. For the narrow band of light it is tuned to, the HOE acts like a curved mirror and redirects the projected image to the eye; for all other light it is nearly clear, so the real world passes through almost unaffected. This lets a HOE combiner be very thin and highly transparent, in some cases little more than a coating on an otherwise ordinary lens. HOEs are used both as the diffractive structure inside holographic waveguides and as free-space curved combiners.

The trade-offs mirror those of diffractive optics. Because each hologram is recorded for a specific wavelength, full color generally needs three multiplexed or stacked films (red, green, blue), which can introduce color crosstalk, and the angular selectivity that makes the film transparent also tends to limit the field of view and eye box. HOE combiners have appeared in lightweight smart glasses: North's Focals used a curved holographic film as a free-space combiner with a laser microprojector to place a small (about 15-degree) image in front of one eye, and Sony and Konica Minolta have built HOE-based see-through displays.

Comparison

The table summarizes the main optical combiner types, their characteristic trade-offs and representative devices. Figures are approximate and depend heavily on the specific implementation.

Combiner type	Operating principle	Typical strengths	Typical weaknesses	Example devices
Flat beam splitter (half-mirror)	Tilted partially reflective plate reflects the image and transmits the real world	Simple, cheap, efficient, good color	No focusing power, so wide FOV needs a large bulky plate	Many head-up displays; early plate-combiner glasses
Birdbath	Flat beam splitter plus a curved partial mirror that magnifies and collimates	Wide FOV (about 40 to 55 degrees), sharp, accurate color, low cost	Low efficiency (double pass at the splitter), dimmed see-through, possible ghosting, bulky in front of the eye	Google Glass (prism variant); many Sony micro-OLED display glasses
Freeform prism	Molded freeform block steers light by refraction, TIR and one partial mirror; paired with a corrector lens	Compact and efficient for its FOV (around 50 degrees and up)	Hard to mold and align; adds thickness; fixed focus	Meta 2; Epson Moverio-style optics
Diffractive waveguide	Input and output gratings couple light into and out of a thin substrate by diffraction	Very thin and transparent; replicated pupil enlarges eye box	FOV capped by refractive index; rainbow color non-uniformity; low efficiency; often needs stacked RGB layers	Microsoft HoloLens 1 and 2; Magic Leap 2
Reflective (geometric) waveguide	Embedded cascade of partial mirrors couples light out of a thin substrate	High efficiency; excellent color uniformity (wavelength-independent)	Hard and costly to manufacture uniformly; older designs thicken with FOV	Lumus-based enterprise and defense headsets
Holographic optical element	Recorded holographic film acts as a wavelength-selective curved mirror or grating	Can be extremely thin and highly transparent	Color crosstalk (needs RGB films); limited FOV and eye box	North Focals; Sony and Konica Minolta HOE displays

See also

• Augmented reality
• Waveguide display
• Near-eye display
• Head-mounted display
• Microdisplay
• Field of view
• Mixed reality
• Vergence-accommodation conflict