Jump to content

Eye box

From VR & AR Wiki

The eye box (also written eyebox or eye-box) of a near-eye display is the region of space in front of the optics within which the user's pupil can sit and still see the entire virtual image. Xiong, Hsiang, He, Zhan and Wu, in their 2021 review of AR and VR displays, define it as "the region within which the whole image FoV can be viewed without vignetting"; it is a three-dimensional volume whose size depends strongly on the exit pupil of the optical system.[1] The IEC eyewear-display measurement standard calls the same quantity the "qualified viewing space": the three-dimensional space within which users place their eye so as to properly see the entire virtual image without moving the head or making any adjustment other than the natural rotation of the eye.[2]

A large eye box lets a head-mounted display tolerate differences in interpupillary distance (IPD) between users and small shifts of the headset during use.[1] A small eye box forces the user to hold the device in one exact position, and in the extreme case of Maxwellian-view and holographic displays the image disappears entirely when the pupil moves a small distance off the viewpoint.[1][3] Because the product of eye box area and field of view is bounded by the etendue of the optics, eye box size is one of the central trade-offs in the design of both VR headsets and AR glasses.[1][4]

Reviewed 20 September 2026. Definitions, quoted figures, eye box and field of view values, IPD statistics, product examples and all 19 sources checked against the cited papers, data sheets and reviews. About review dates.

Definition

In a near-eye display, light from the microdisplay or panel passes through an eyepiece or combiner and converges toward the eye. The area where the ray bundles from all field angles overlap is the exit pupil of the optical system. The eye box is the related but distinct volume in which the eye's own pupil must be placed to receive every one of those bundles at once; Xiong et al. note that its volume "is strongly dependent on the exit pupil size".[1] In their 2021 review of waveguide combiners, Bernard Kress and Ishan Chatterjee define the eye box as "the volume that the user's pupil can sit in and view the entire virtual image field-of-view", and add that it is usually not a rectangular prism but a frustum, since the region over which all fields overlap shrinks or grows with distance from the optic.[4]

Two related terms are used with the eye box. Eye relief is the distance from the user's corneal surface to the last surface of the display optic.[4] The eye point is the design location at which the entrance pupil of the eye should be placed for best performance; the IEC standard treats the eye box as a volume around that point and asks manufacturers to specify either the eye point position or the eye relief.[2] A display's eye box is therefore quoted at a stated eye relief: moving the eye closer or farther changes the usable region, which is why the box is a frustum rather than a slab.[4]

The human pupil is small, roughly 3 to 5 mm in diameter, and its size changes with ambient illumination.[3][1] This matters for displays that build their eye box from several discrete viewpoints: the viewpoints must be spaced farther apart than the pupil diameter so that only one enters the eye at a time, but not so far apart that gaps open between them.[1]

Distortion that changes as the pupil moves around the eye box is perceived as pupil swim, the apparent warp and shift of virtual objects during eye and head motion.[4]

Etendue and the trade-off with field of view

For displays built from conventional geometric optics, the field of view and the eye box are linked by the conservation of etendue (also called the optical invariant). Enlarging one without shrinking the other requires larger optics, which in turn compromises the form factor of the device.[1] Conventional VR headsets use a transmissive lens with a focal length of about 4 cm to obtain both a large field of view and a large eye box, and this focal distance is the main reason the headsets are much bulkier than eyeglasses.[1]

The same limit applies in a different form to holographic displays, where the etendue is set by the spatial light modulator (SLM): the maximum diffraction angle of the SLM multiplied by its size gives the system etendue, so a wide field of view can only be obtained at the cost of a small exit pupil, and vice versa.[1] Xia et al. state the relation directly: the etendue of a holographic display "is the product of the exit pupil size and the FOV and is linked to the resolution of the SLM", and "as etendue is always preserved, when a large FOV is achieved for holographic display, the size of the exit pupil at each viewing position will be traded off".[3] Jang, Bang, Li and Lee describe the same constraint as "an inherent trade-off relationship between the field of view and eye-box size" imposed by the limited bandwidth of the SLM.[5] A 2019 study by Ratnam, Konrad, Lanman and Zannoli summarised the practical consequence: "state-of-the-art near-eye displays often compromise on eye box size to maintain a wide field of view", which is why pupil steering is needed to keep the box aligned with a moving eye.[6]

Brightness is part of the same budget. For a fixed input luminous flux, the output luminance of an AR combiner falls as the field of view or the exit pupil grows; Xiong et al. calculate that a lossless combiner with a 60 degree diagonal field of view and a 10 mm square exit pupil could deliver at most about 17,000 nit per lumen, and note that a larger eye box or field of view "usually decreases the image brightness".[1] Kress and Chatterjee show that for a waveguide combiner the area of the out-coupler grating is dictated solely by the field of view and the eye box, so a larger box also means a larger grating in the lens.[4]

How large an eye box needs to be

The eye box has to cover two kinds of misalignment: the spread of IPD across the user population, and the movement of the headset relative to the head during use.[1] A 2004 survey of anthropometric data by Neil Dodgson found that mean adult IPD is around 63 mm, that the vast majority of adults fall between 50 and 75 mm, and that a range of 45 to 80 mm is likely to include almost all adults; the minimum for children down to five years old is around 40 mm.[7] A headset with a mechanical IPD adjustment can centre its optics on the user's eyes, so its eye box only has to absorb the residual error; a device without adjustment has to accommodate the IPD spread within the eye box itself.[1]

Kress and Chatterjee list a large eye box among the basic visual comfort features of a mixed reality headset, and note that while consumer optics could in principle ship in small, medium and large IPD versions, enterprise headsets shared between employees have to accommodate a wide IPD range in a single design.[4] For AR waveguides they give concrete targets: covering the 95th or 98th percentile of the IPD distribution, including different facial types, "requires a large horizontal EB, typically 10-15 mm", and because of fit and nose-pad variation a vertical eye box of 8 to 12 mm is also desirable.[4] Xiong et al. give roughly the same figure, describing the eye box of diffractive waveguide combiners as "generally large enough (~10 mm) to accommodate different user's IPD and alignment shift during operation".[1]

Eye box in VR headsets

VR headsets place the eye behind a magnifying eyepiece a few centimetres from the panel, and their eye box is the region where the full field of view stays sharp and unvignetted. Users experience its edge as the point where the image blurs, dims or shows artifacts, and reviewers usually call the region inside it the "sweet spot".[8] According to the optics firm HyperVision, pancake lenses, which fold the light path with polarisation optics, generally provide a bigger eye box than aspheric or Fresnel lenses.[9] Xiong et al. note the cost: with two passes through a half mirror, a polarisation-folded pancake lens has a maximum efficiency of 25 percent for polarised input light.[1]

Meta's Meta Quest 3 is an example of this generation. In its review, UploadVR reported that Meta claims an eye box of plus or minus 5 mm for the Quest 3 pancake lenses, an upgrade over the Quest Pro, and found that the pancake lenses were "far less sensitive to your eyes being properly aligned" than the Fresnel lenses of the Quest 2, so that the image stayed sharp even with the IPD set incorrectly. The headset also offers a stepless IPD adjustment from 58 mm to 70 mm and four eye-relief positions.[10]

The opposite design choice appears in the Bigscreen Beyond, which uses small pancake lenses in front of micro-OLED panels. In its review, Road to VR wrote that "the eyebox (the optimal optical position relative to the lenses) is so tight that even small deviations can amplify artifacts and reduce the field-of-view", and that swapping between two facepads that differed by only a few millimetres changed the clarity noticeably. Bigscreen makes this workable by manufacturing each facepad from a 3D face scan and building each unit for a fixed IPD in the 53 to 74 mm range, so the company knows where the customer's pupils will sit; the reviewer measured a field of view of 98 degrees horizontal by 90 degrees vertical.[8] HyperVision's analysis of the Apple Vision Pro makes the general point: for pancake lenses designed around micro-OLED panels with very small pixels (7.4 micrometres in the Vision Pro) the magnification is so strong that "there is not enough resource for the eyebox", so the visual and optical axes must be aligned well.[9]

Research VR optics show how far the trade-off can be pushed. Xiong et al. cite a curved lenslet-array VR design with a 180 degree diagonal field of view and an eye box of 19 by 12 mm, and a compact system combining pancake optics with a Fresnel lenslet array that reaches a 102 degree horizontal field of view with an 8 mm eye box, at the cost of some image discontinuity and crosstalk.[1]

Eye box in AR displays

Free-space combiners

AR displays based on traditional geometric optics (birdbath and freeform combiners) have a relatively simple design with a field of view of about 60 degrees and an eye box of about 8 mm, together with reasonable efficiency; shrinking them further runs into the etendue limit, and the placement of the half mirror leaves the form factor relatively bulky.[1] For reflective designs with a single curved extractor, the field of view is proportional to the size of the reflector, so widening the view means a larger waveguide; multilayer coatings and embedded polarised reflectors are used to enlarge the eye box.[11]

Waveguides and exit pupil expansion

Waveguide combiners change the picture because they can replicate the exit pupil. Light from a small projector is coupled into a thin glass plate, travels by total internal reflection, and is coupled out a little at a time on each bounce, so the same image leaves the plate through many adjacent exit pupils. Xiong et al. describe this exit pupil expansion (EPE) as the distinctive feature of waveguides: it "effectively enlarges the system etendue", and the effective eye box is enlarged with it.[1] Kress and Chatterjee reproduce a 1991 patent for a waveguide-embedded partial-mirror combiner with exit pupil replication as an early example, and note that Lumus multiplies the extractor mirrors in its reflective light-guide optical element for the same purpose.[4]

Whether expansion is needed depends on the field of view. For small fields (under about 20 degrees diagonal), as in many smart glasses, a single exit pupil can suffice, which simplifies the design and even allows a curved waveguide; once the field of view exceeds about 20 degrees, especially in a binocular design, one- or two-dimensional pupil replication is required.[4] One-dimensional expansion along the horizontal axis is often enough, since the horizontal eye box is the one that must absorb the IPD spread, but it then requires the projector to produce an input pupil that is already tall in the unexpanded direction, larger than the replicated pupil in the other axis. Two-dimensional expansion is used when the projector cannot form such a tall pupil because of its own etendue and size limits.[4][1] Expansion is not free: each replicated pupil carries only a fraction of the light, and a typical diffractive waveguide delivers around 50 to 200 nit per lumen, far below a free-space combiner.[1]

Manufacturer data sheets show the trade-off inside one product family. Dispelix's DPX 30-50 single-layer diffractive waveguide family is specified as follows:[12]

Diagonal field of view Eye relief Eye box Efficiency (waveguide and projector)
30 degrees 20 mm 12.5 x 8 mm 220 nits per lumen
40 degrees 16 mm 12 x 6 mm 110 nits per lumen
50 degrees 16 mm 10 x 6 mm 75 nits per lumen

As the field of view grows from 30 to 50 degrees, the eye box shrinks from 12.5 x 8 mm to 10 x 6 mm, the quoted eye relief drops from 20 mm to 16 mm, and the efficiency falls by about two thirds, all with the same 5 mm input pupil.[12] The theoretical upper limit on a waveguide's field of view is set by its refractive index;[1] Meta states that the silicon carbide waveguides in its Meta Orion prototype, with a refractive index of 2.7, are what allow a field of view of approximately 70 degrees without stacking plates.[13]

Maxwellian-view and holographic displays

Retinal projection displays based on the Maxwellian view focus a point source into the eye pupil so that the image is always in focus on the retina regardless of the eye's accommodation. Their weakness is "the tiny exit pupil, or eyebox": a small displacement of the pupil from the viewpoint makes the image disappear completely, and expanding the eye box is regarded as one of the most important challenges for this class of display.[1] Xia et al. report the same for holographic near-eye displays, which can reach a wide field of view (80 degrees horizontal in a Microsoft Research prototype by Maimone, Georgiou and Kollin[14]) in an eyeglass form factor but have an eye box of less than about 1 mm, so that "a user needs to place his eye very accurately within the display's tiny eyebox, and thus such NEDs do not allow any eye movement".[3] Karl Guttag observed the practical result on the Focals by North laser glasses at CES 2019: a tiny eye box meant the glasses had to be custom fitted and worn in an exact position, "or else you see a double image or no image at all".[15]

Methods to expand the eye box of these systems fall into two families. Pupil duplication generates several viewpoints at once to cover a larger area, dividing the light between them; the viewpoint spacing must exceed the pupil diameter so that only one viewpoint enters the eye per frame, which avoids ghost images but reduces efficiency. Pupil steering produces a single viewpoint and shifts it to follow the pupil, which is more light-efficient but requires a real-time eye-tracking module; Xiong et al. note that a perfect steering system would need five degrees of freedom once eyeball rotation is included.[1] In a steered system the eye box for any single viewpoint is only the size of the eye pupil, and it is the steering or duplication that expands the total system eye box.[1]

Representative research systems include:

System Approach Reported field of view Reported eye box
Jang, Bang, Li and Lee (2018) Holographic display; exit pupil shifted with pupil tracking through a pupil-shifting holographic optical element, described by the authors as the first practical eye-box expansion for a holographic near-eye display[5] Not reported in the abstract Expanded by exit-pupil steering
Xia et al. (2020) Holographic display with a lenslet-array holographic combiner that creates a static array of focal points, no moving parts[3] 45 degrees About 10 mm (10 x 10 mm demonstrated)
Kim et al., "Holographic Glasses" (2022) Pupil-replicating waveguide, SLM and geometric phase lens in a 2.5 mm thick optical stack weighing 60 g without its driving board[16] 22.8 degrees diagonal 2.3 mm static, dynamic with beam steering
Chae, Bang, Yoo and Jeong (2023) Sparse eye box: a lens-array eyepiece rearranges a dense holographic eye box into a sparse one so the limited SLM etendue is not spent on redundant area[17] Not reported in the abstract Sparse eye box, size not reported in the abstract
Wang et al. (2025) Freeform and holographic optics projecting rays onto the retina from many independently controllable viewpoints that track the eye without mechanical steering[18] 50 degrees 10 x 10 mm, with a depth of field from 0.25 to 10 m

Ratnam and colleagues published a taxonomy of pupil-steered architectures and an optical model with a wide-field schematic eye to predict how imperfect steering degrades the retinal image, as a basis for choosing which designs deserve psychophysical testing.[6]

Measurement

IEC 63145-20-10:2019, the fundamental optical measurement standard for eyewear displays, defines the eye box as the qualified viewing space and adds that "able to properly see" means the displayed image meets every requirement in the product specification, so the reported eye box depends on the thresholds written into that specification.[2] Its clause 6.8, "Eye-box based on luminance", sets out a measurement procedure and report, and the reported quantities are the width and height of the eye box in millimetres, written W box and H box. The standard places the eye point (the design location of the eye's entrance pupil) at the origin of the measurement coordinate system and requires the manufacturer to specify either the eye point position or the eye relief; the light measuring device is meant to stand in for the eye, with an entrance pupil between 2 mm and 5 mm that is smaller than the display's output light field. An informative annex describes estimating the eye point from full-field luminance or from Michelson contrast.[2]

Head-up displays

The term is also standard in aviation head-up displays (HUDs), where the eye box is the volume of cockpit space within which the pilot must keep at least one eye to see the whole display. Modern aircraft HUDs allow head movement within an eye box of roughly 5 inches laterally by 3 inches vertically by 6 inches longitudinally; outside it the image is cut off, unclear or distorted.[19] In a near-eye display the same concept applies at millimetre scale.

See also

References

  1. 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.20 1.21 1.22 1.23 Xiong J, Hsiang E-L, He Z, Zhan T, Wu S-T (2021). "Augmented reality and virtual reality displays: emerging technologies and future perspectives". Light: Science & Applications, vol. 10. pp. 216. doi:10.1038/s41377-021-00658-8. https://www.nature.com/articles/s41377-021-00658-8.
  2. 2.0 2.1 2.2 2.3 "IEC 63145-20-10:2019 Eyewear display - Part 20-10: Fundamental measurement methods - Optical properties (Edition 1.0, preview sample)". iTeh Standards. International Electrotechnical Commission. 2019-08. https://cdn.standards.iteh.ai/samples/101828/8ac7b0875e974aa085e18a1d6cb73b49/IEC-63145-20-10-2019.pdf. Retrieved 2026-09-20.
  3. 3.0 3.1 3.2 3.3 3.4 Xia X, Guan Y, State A, Chakravarthula P, Cham T-J, Fuchs H (2020). "Towards Eyeglass-style Holographic Near-eye Displays with Statically Expanded Eyebox". 2020 IEEE International Symposium on Mixed and Augmented Reality (ISMAR). https://www.cs.unc.edu/~cpk/data/papers/eyebox-expansion_ismar2020.pdf.
  4. 4.00 4.01 4.02 4.03 4.04 4.05 4.06 4.07 4.08 4.09 4.10 Kress B C, Chatterjee I (2021). "Waveguide combiners for mixed reality headsets: a nanophotonics design perspective". Nanophotonics, vol. 10, no. 1. pp. 41-74. doi:10.1515/nanoph-2020-0410. https://www.degruyterbrill.com/document/doi/10.1515/nanoph-2020-0410/html.
  5. 5.0 5.1 Jang C, Bang K, Li G, Lee B (2018). "Holographic near-eye display with expanded eye-box". ACM Transactions on Graphics, vol. 37. pp. 1-14. doi:10.1145/3272127.3275069. https://dl.acm.org/doi/10.1145/3272127.3275069.
  6. 6.0 6.1 Ratnam K, Konrad R, Lanman D, Zannoli M (2019). "Retinal image quality in near-eye pupil-steered systems". Optics Express, vol. 27, no. 26. pp. 38289-38311. doi:10.1364/OE.27.038289. https://opg.optica.org/oe/fulltext.cfm?uri=oe-27-26-38289.
  7. Dodgson N A (2004). "Variation and extrema of human interpupillary distance". Stereoscopic Displays and Virtual Reality Systems XI, Proc. SPIE 5291. pp. 36-46. doi:10.1117/12.529999. http://www.neildodgson.com/pubs/EI5291A-05.pdf.
  8. 8.0 8.1 Ben Lang (2023-09-26). "Bigscreen Beyond - Promising but Incomplete, Just Like This Review". Road to VR. https://roadtovr.com/bigscreen-beyond-review-pc-vr-headset/. Retrieved 2026-09-20.
  9. 9.0 9.1 "Apple Vision Pro Optics insights". HyperVision. 2023-06. https://www.hypervision.ai/tech-research/apple-vp-optics-insights. Retrieved 2026-09-20.
  10. David Heaney (2023-10-09). "Quest 3 Review: Excellent VR With Limited Mixed Reality". UploadVR. https://www.uploadvr.com/quest-3-review/. Retrieved 2026-09-20.
  11. Xia X, Guan F Y, Cai Y, Magnenat Thalmann N (2022). "Challenges and Advancements for AR Optical See-Through Near-Eye Displays: A Review". Frontiers in Virtual Reality, vol. 3. pp. 838237. doi:10.3389/frvir.2022.838237. https://www.frontiersin.org/articles/10.3389/frvir.2022.838237/full.
  12. 12.0 12.1 "The DPX 30-50 degree waveguide (specification sheet)". Dispelix. 2020-10. https://assets.ctfassets.net/lgro3zg4tx7x/1houHw1hPeYANMLB8ClF6i/4c6747d3df43fb1eda1de406eda113b0/Dispelix_DPX30-50_WG-specs_202010.pdf. Retrieved 2026-09-20.
  13. "Crystal Clear: Our Silicon Carbide Waveguides & the Path to Orion's Large FoV". Meta. Meta Platforms. 2025-03-06. https://www.meta.com/blog/orion-silicon-carbide-waveguides-ar-glasses-large-field-of-view/. Retrieved 2026-09-20.
  14. Maimone A, Georgiou A, Kollin J S (2017). "Holographic near-eye displays for virtual and augmented reality". ACM Transactions on Graphics, vol. 36, no. 4. doi:10.1145/3072959.3073624. https://www.microsoft.com/en-us/research/publication/holographic-near-eye-displays-virtual-augmented-reality/.
  15. Karl Guttag (2019-01-24). "CES 2019 - AR and Other Interesting Display Technology". KGOnTech. https://kguttag.com/2019/01/24/ces-2019-ar-and-other-interesting-display-technology/. Retrieved 2026-09-20.
  16. Kim J, Gopakumar M, Choi S, Peng Y, Lopes W, Wetzstein G (2022). "Holographic Glasses for Virtual Reality". ACM SIGGRAPH 2022 Conference Proceedings. doi:10.1145/3528233.3530739. https://dl.acm.org/doi/10.1145/3528233.3530739.
  17. Chae M, Bang K, Yoo D, Jeong Y (2023). "Etendue Expansion in Holographic Near Eye Displays through Sparse Eye-box Generation Using Lens Array Eyepiece". ACM Transactions on Graphics, vol. 42. pp. 1-13. doi:10.1145/3592441. https://dl.acm.org/doi/10.1145/3592441.
  18. Wang Y-D, Yang T, Lyu X, Cheng D, Wang Y-T (2025). "Large Depth-of-Field, Large Eyebox, and Wide Field-of-View Freeform-Holographic Augmented Reality Near-Eye Display". Advanced Science, vol. 12, no. 37. pp. e08773. doi:10.1002/advs.202508773. https://pmc.ncbi.nlm.nih.gov/articles/PMC12499390/.
  19. Radiant Vision Systems (2021-01-13). "Quality Considerations for Aviation Head-up Displays (HUDs)". Aviation Today. https://www.aviationtoday.com/2021/01/13/quality-considerations-aviation-head-displays-huds/. Retrieved 2026-09-20.