August 28, 2020


Light Engines for
XR Smartglasses

By Jonathan Waldern, Ph.D.

The near-term obstacle to meeting an elegant form factor for Extended Reality[1]Extended Reality (XR) is a term that encompasses Augmented Reality (AR), Mixed Reality (MR), and Virtual Reality (VR) (XR) glasses is the size of the light engine[2]Also referred to as a Picture Generation Unit or “PGU” that projects an image into the waveguide, providing a daylight-bright, wide field-of-view mobile display.

For original equipment manufacturers (OEMs) developing XR smartglasses that employ diffractive waveguide lenses, there are several light engine architectures contending for the throne. Highly transmissive daylight-bright glasses demanded by early adopting customers translate to a level of display efficiency, 2k-by-2k and up resolution plus high contrast, simply do not exist today in the required less than ~2cc (cubic centimeter) package size.

This thought piece examines both Laser and LED contenders. It becomes clear that even if MicroLED (µLED) solutions do actually emerge as forecast in the next five years, fundamentally, diffractive waveguides are not ideally paired to broadband LED illumination and so only laser based light engines, are the realistic option over the next 5+ years.

Bottom Line Up Front
  • µLED, a new emissive panel technology causing considerable excitement in the XR community, does dispense with some bulky refractive illumination optics and beam splitters, but still requires a bulky projection lens. Yet an even greater fundamental problem of µLEDs is that while bright compared with OLED, the technology falls short of the maximum & focused brightness needed for diffractive and holographic waveguides due to the fundamental inefficiencies of LED divergence. 
  • A laser diode (LD) based light engine has a pencil like beam of light which permits higher efficiency at a higher F#. This in turn results in a much improved  Liquid Crystal on Silicon (LCoS) microdisplay panel contrast when compared with traditional LED LCoS designs.
  • Looking to the future and miniaturizing still further, the projection lens itself may be integrated into the waveguide using diffractive metasurfaces containing sub-wavelength structures or “metalenses,” which are better suited to laser illumination compared to LED illumination.
  • By leveraging the benefits of lasers and its holographic switchable optics platform, DigiLens is using its extensive experience with diffractive optical waveguides to create a new class of light engines referred to as Waveguide Integrated Laser Display (WILD). The WILD category of light engines offer OEMs more immediate time to market and is a lower-risk alternative to µLED based light engines.
  • Laser coherence and phase also provide added efficiency gains in holographic thin film waveguides.
The Need for a Compact Light Engine for XR Glasses

Delivering a wide field-of-view (FoV), color, daylight-bright device in a spectacle-like form factor at an acceptable price continues to prove to be a major development hurdle for consumer XR displays. The form factor challenge is only partially met using thin diffractive and holographic waveguides. Current light engines comprising of the microdisplay beam splitter and projection lens are increasingly seen as too bulky to satisfy the aesthetic requirements of glasses.

In addition, misunderstanding also prevails with respect to outdoor contrast requirements which has an oversized impact on brightness and battery life. Here, ambient light can wash out the content and render the image unrecognizable. Under generally accepted guidelines, an Ambient Contrast Ratio (ACR) of 3:1 is required for recognizable images, 5:1 for adequate readability, and 10:1 for appealing quality.[3]Dobrowolski J A, Sullivan B T, Bajcar R C. Optical interference, contrast-enhanced electroluminescent device. Applied, Optics, 1992, 31(28): 5988–5996 DOI:10.1364/AO.31.005988; and Chen H, Tan G, … Continue reading  

Given a 90% transparent waveguide and indoor office lighting condition (150nits), the XR display target luminance level should have at least 550nits for ACR 5:1 – which is considerably less brightness than that required to overcome outdoor lighting condition of 300-3000nits (so 2-20x). It is notable that the HoloLens 2 display is 1000nits and that all current XR displays use sunshades, which longterm are unwelcome for consumers. Ultimately, even assuming a dangerously darkened room, the waveguide display contrast is likely perfectly acceptable at 30-40:1, so well short of and not to be confused with the LCD, µLED or µOLED native display panel contrast, irrelevant for an XR display application.

The low etendue and the narrow band of laser emitters makes them the ideal choice for delivering excellent color gamut and high brightness with high transmissivity that can meet the large number of outdoor use cases requiring compact form factors. However, laser emitters have traditionally brought new problems including laser speckle and other illumination non-uniformities. 

Let’s dig deeper into the leading light engine options on the market today illuminated by either Light Emitting Diodes (LED) or Laser Diode (LD).

DigiLens Visualize Design v1

LIGHT EMITTING DIODES

µLEDs Engines

µLED is a thin emissive panel technology that dispenses with beam splitting and illumination optics, thereby offering higher brightness and much smaller size. There is consequently great interest in this technology. Yole, a leading industry analyst, believes that µLED technology could match or exceed OLEDs for most key display attributes, especially brightness and efficiency.[4]µLED Displays 2018, Yole Development, July 2018 However, as Yole points out, there is no sign of a color µLED light engine for XR applications. The reasons for this can be attributed to both basic device physics and technological obstacles to achieving process maturity.

As µLEDs get small enough to satisfy the 2k-by-2k resolution ideal for XR, the pixels suffer from surface recombination, etching- related defects on the side walls of the chip, current crowding, and other thermal effects which all impact the external quantum efficiency.

Driving the LED pixels in the panel brighter to overcome low efficiency is limited by higher power consumption, current drop and heat dissipation requirements. Solutions to the efficiency problem have been identified (e.g. thermal annealing, side wall passivation and improvements to LED structures and materials). Reducing the sheet resistance of the current-spreading layer of the µLED chip has also been shown to give a uniform current distribution and alleviate current crowding.

Another effect that appears as the pixel size is reduced, is a gradual increase in the light emission through the side wall leading to a far field deviation from the normal Lambertian characteristic of the µLED. This side emission depends on the semiconductor material index and the device structure, but can certainly be a problem in color displays as different materials are used for different colors. The combined side emissions can lead to an angular color shift that can also distort the color balance of the final image, resulting in a subpar experience for enterprise and consumer use cases.


Plessey and Compound Photonics’
0.26”diagonal display module

Efficiency is not the only challenge for µLEDs. An integration technology for active matrix driving, technologies for full-color realization, and defective pixel control all still remain major development hurdles. Driving µLEDs is more complex than OLEDs and using standard low temperature polysilicon or oxide TFT backplanes might not be as straightforward as expected. Full color also requires integration of phosphors or QDs to the µLEDs, a process that is in its infancy.

On a positive note, Yole reports that leaders are achieving 99% yields, and small die efficiency is approaching or exceeding that of OLEDs.[5]µLED Displays 2018, Yole Development, July 2018 MOCVD reactor suppliers also have credible roadmaps to deliver cost-effective tooling. Despite these advances, there are still significant challenges to developing a consumer µLED based light engine to match diffractive waveguide requirements.

Current Commercial Activity and Status of µLED Displays

In recent years the market has seen significant µLED business development, with Facebook (Oculus) taking a lead by acquiring two University spinoffs: Irish µLED developer InfiLED (2016) and a UK company mLED (2017). Apple is also reportedly considering investing a further $330 million into a Taiwanese factory to manufacture both LED and µLED displays for future iPhones, iPads, MacBooks, and other devices, so µLED display technology must be seen in the context as being enabling to a broad range of display product categories.

Additionally, Facebook has demonstrated a green (520 nm) nHD (640×360) µLED array with 6-micron pixels on a 20-micron pitch.[6]www.mled-ltd.com Pixel sizes down to 2-4 microns may be in the cards. Improvements to circuit complexity, efficiency drop at low pixel sizes and wall plug efficiency resulting from reduced specific contact resistivity (<10-4 ohm/cm2) and sheet resistance of ~100k ohms square are also claimed.

Facebook and Plessey

A partnership between Facebook and Plessey announced in March 2020 will leverage Plessey’s LED manufacturing expertise to help Facebook prototype and develop new technologies.

Plessey is increasing investment in their fab to improve volume and yield. It is notable that with Oculus, Facebook can use µLED panels in both VR headsets and AR transparent glasses, so they are likely taking a long term view into several sub-categories of XR displays, both handheld and head mounted, with VR likely being earlier to market. 

Earlier this year, Compound Photonics and Plessey Semiconductors Ltd announced the first fully addressable µLED XR display modules resulting from a partnership to develop GaN-on-Silicon µLED displays for XR. Other significant movements by Plessey in 2020 included a partnership with Axus Technology, a leading global provider of semiconductor processes, to commercialize high-performance GaN-on-Silicon monolithic µLED technology and another partnership with WaveOptics to develop next generation smartglasses. 

Oculus Concept for a waveguide-based VR headset
Image courtesy Facebook

Jade Bird
Monochrome (green only) Jade Bird waveguide eyepiece

The Shanghai based company Jade Bird is currently sampling µLED panels delivering 3M nits at 530 nm and 150k nits at 455 nm with contrast at 10,000:1, pixel sizes down to 2.5 micron and 360Hz refresh rate. Of course bichromatic and trichromatic versions are in development. The arrays use refractive micro-lens arrays, bonded onto the µLED array for beam shaping to improve efficiency.

As shown in the specifications below, the blue output is significantly less than green, even accounting for the human visual efficacy difference. Also, our understanding is that the red material set is even more of a challenge. Jade Bird is working to remedy the fundamental device physics and manufacturing issues as part of their roadmap, but no production date as been announced yet. 

µLED brightness falls dramatically as size reduces
Ostendo

The Southern California based company Ostendo is marketing a different µLED approach called RGB QPI®, that’s also receiving a lot of attention. The company’s patents describe the QPI as a “…3D-IC semiconductor device comprising a high density array of digitally addressable micro-LED pixels,” with its simultaneous color emission, so the RGB pixels share the same optical aperture per pixel but with a constrained emission profile.[7]El-Ghoroury, H. (2014/ December 4). “Quantum Photonic Imager (QPI): A New Display Technology and Its Applications“.

It is highly unlikely the LED emission physics are fundamentally different to other µLED devices, so the Ostendo QPI® offering will likely still suffer from the same luminance shortfall for waveguide based eyeglass solutions. Ostendo is reported to be making its own eyeglass optic so perhaps lessening the requirement of a diffractive waveguide design, currently proferred by OEM’s.

Apple, Facebook, Plessey, Jade Bird and Ostendo are not the only lively players in the µLED field. Yole points out that Samsung AOU, Innolux and LG are all very active,[8]µLED Displays 2018, Yole Development, July 2018 with Samsung pursuing three separate architectural efforts (one in particular: its QNED nanorod “ink” which could be a major disruption).

According to Yole, it appears Apple leads the µLED race when factoring in the Luxvue acquisition and internal developments. Yole anticipates a 30% µLED based XR penetration by 2027 and they highlight that several start-ups have already collectively raised close to $0.8B and are likely to add another $100M in 2020. It should also be noted that these dollar amounts exclude the new Apple $300M investment rumors.

LASER DIODES

Laser Based Scanning (LBS) Engines

Two notable head mounted displays have both developed laser diode driven LBS systems. One has been implemented in the HoloLens 2 (HL2) and the other in the North[9]Purchased by Google – https://www.bynorth.com Focals smartglasses. However, there remain fundamental challenges[10]Detailed tradeoff analysis by Karl Guttag at https://www.kguttag.com  to sufficiently scale either of these solutions for the future high resolutions required for wide FoV XR, especially the vertical FoV, which is lacking in the HL2.

Put simply, the limitation for future LBS XR is the laser pixel modulation requirement to pulse modulate (PWM) the brightness for each pixel location. These physics are almost saturated today with 16:9 aspect ratio in HL2’s 2048×1080 resolution. The limit is with the laser driver chip.

As waveguide FoVs increase from 52° today to a more immersive 70° or 90° this decade (in both horizontal and vertical axis), the 2k-by-2k going to 4k-by-4k resolution will demand a scanning and modulation speed that is impossible for a single RGB LBS light engine design. A tiled laser fiber virtual retinal display (VRD) advocated by Magic Leap is theoretically possible[11]US20150268415A1 Ultra-high resolution scanning fiber display – Magic Leap, but ultimately too complex and costly as it would need an array of tiled diodes and drivers to achieve the modulation bandwidth. 

Other arguments in favor of scanned laser for XR is that they are always in focus and hence solves the vergence-accommodation conflict (VAC), however new Maxwellian VRD architectures[12]Accommodation-Free Head Mounted Display with Comfortable 3D Perception and an Enlarged Eye-box https://doi.org/10.34133/2019/9273723 which use Liquid Crystal on Silicon (LCoS) spatial light modulators are also always in focus, so common to all laser based light engine architectures and a critically important feature for XR applications.

The North Focals provide a good example of an LBS niche as XR content is sparse

For low-resolution consumer market applications implementing “watch type” notification companion displays, like the North Focals, LBS can fit that niche, although the small eyebox in these non-waveguide displays will likely always limit consumer enthusiasm. This is especially the case where XR content is sparse, but it must be noted that users have expressed skepticism for that use case at scale.[13]For example the lack of sales for GoogleGlass and Vuzix devices, and even push back on special notifications with the Echo Frames described in detail here: … Continue reading

Whilst LBS based light engines come with the benefit of small volumetric packaging, efficiency and high brightness, their role in XR is likely diminished due to resolution limitations, scanning uniformity, mechanical reliability and safety. It is likely the niche for LBS will be limited to low resolution, simple notification smartglasses. 

Laser Illuminated Liquid Crystal on Silicon (LCoS) Light Engines

For the last 20 years, LCoS has been a poor second to DLP based light engines, despite LCoS panels having a distinct cost advantage and the critical smaller pixel size needed for high resolution small aperture size. As LED light has a wide cone to collect a sufficiency of light, as this light cone is mirrored to reflect off the LCoS panel, overall image contrast suffers. Liquid Crystal (LC) panels cannot modulate light (light to dark) at a high angle.

Additionally, LED light is unpolarized and like all LC displays, only polarized light can be used. DLP with LED always wins as they both use mirrors, so they have high contrast with high illumination angles and are not polarization dependent, so it can use all the LED light. When compared with LCoS, DLP with LED is twice as bright and efficient. Lasers are a perfect marriage with LCoS, as laser light is polarized and thus all the light can be collected and shaped onto the LCoS panel with a low angle, giving very high contrast.

So why do we not see more Laser/LCoS light engines for XR? The problem has previously been that the use of laser illumination comes with an unwelcome artifact: Speckle. 

Easily recognizable as a sparkly or granular structure around uniformly illuminated rough surfaces, speckle arises from the high spatial and temporal coherence of lasers. The resulting viewer distraction and loss of image sharpness has been a major obstacle to commercialization of laser Illuminated light engines. 

However, in a feat of optical origami, DigiLens has created a highly innovative Waveguide Integrated Laser Display (WILD) light engine which uses waveguide optics to integrate despeckling, homogenization, beam shaping, beam splitting and high contrast image projection from an LCoS, but in a breakthrough tiny package. This alternative approach to µLED leverages DigiLens’ electrically switchable diffractive waveguide optics to achieve high brightness, contrast and color gamut, whilst overcoming speckle. 

The WILD light engine contains a tiny array of switchable diffuser pixels in the light path, which are patterned and switched rapidly to generate a multiplicity of different speckle patterns, which in turn eliminates laser speckle from the projected image into the combiner waveguide. The fundamental methodology and architecture is a breakthrough feature of WILD and will be covered more fully in a forthcoming light engine article. WILD opens the door for Laser diodes to directly compete and surpass in many cases there µLED counterparts.

Later this year, initial demos of WILD will be showcased in a ~3cc package with a development pathway to ~1cc total light engine size (diodes, imaging waveguides, panel and coupling lens). LCoS panels already have a mature supply chain with multiple manufacturers, 2k-by-2k, with 3µm pixel panels and are available now. They also have a clear path to 1.5μm pixels for a near term 3k-by-3k tiny package. As such, WILD offers a lower-risk and overall superior performance solution for OEMs developing XR solutions starting in 2021.

Notional WILD pico-projector module
Additional Design Evolution Benefits

Given WILD has used diffractive waveguides to integrate optical functionalities, even the projection lens may benefit from advanced techniques to eliminate bulky refractive optics. In the current WILD reference design, a miniature projection lens employs a five (5) lens element, and 15mm projection lens module. Moving to a “pure” all diffractive WILD design, whilst capitalizing on the WILD londitudnal form factor, it is now conceivable to integrate even the projection lens into an even smaller volume. Narrow band laser illumination is key to enabling diffractive optics technology and thus a time to market enabler. 

Metalenses

Metalenses are a recent addition to the diffractive optics family, where the nanoscale optical structures used in metalenses can allow more subtle control of polarization, phase and amplitude down to the diffraction limit. The most significant feature in terms of imaging is the use of surface features including height which is typically ~ 500 nm, combined with at least one other dimension <<λ, (typically ~10s of nm).

A recent example developed by Metalenz (a Harvard spin-off) uses complex patterns of titanium oxide nanofins on a glass substrate. The lens eliminates chromatic aberration over the visible band by optimizing the shape, width, distance, and height of the nanofins, where a single mask semiconductor production process is used for fabrication.

Metalens designs currently being reviewed also provide a collimator and display waveguide in-coupler. Metalenses can also assist with laser aberration correction, color correction and beam shaping.

DigiLens’ analysis suggests that metalenses are highly promising candidates for a thin flat lens that can encode all the optical functionality of a multi-element projection lens, significantly reducing WILD’s overall volume by half to below 1cc. Laser illumination also allows the use of metalenses to be developed to complement DigiLens’ flat holographically recorded lenses, already proven in DigiLens’ AutoHUD. So broadband μLED therefore leaves OEMs at a disadvantage, as they are unable to harness the next step in diffractive miniaturization.

APPLES TO APPLES – 
DAYLIGHT BRIGHTNESS, XR LIGHT ENGINE REQUIREMENTS

WILD vs. µLED

A theoretical comparison was made of a full color WILD based XR smartglass display vs. an equivalent µLED powered version, given a display specification of 5,000 nits; 30° diagonal FoV, 12x10mm eyebox, assuming a waveguide efficiency of 1%, and from a laser illuminated LCoS (8.4×5.93 mm) to match the same aperture as the Jade Bird µLED panel previously referenced. Using identical F-numbers and losses, our model predicts the RGB µLED panel luminance needed to hit the display specification would be approximately 300k/nits, 750k/nits and 60k/nits respectively, or 1.1M/nits at D65 color point.  

Individual R and B µLED panels are currently a fraction of this brightness and when mounted around a dichroic beam cube, are too bulky. As mentioned earlier, there’s no sign of a trichromatic µLED panel with the required RGB efficacy. This together with the diversity of device technologies needed for RGB, fundamental physics-related device efficiency problems, and electronics integration challenges, leaves µLED at a significant disadvantage relative to WILD.

DigiLens WILD and µLED Comparison

Also, while such high luminance values have been demonstrated for monochrome µLED devices having larger pixel size, that has not been the case for the tiny 2-3um pixel size needed for XR glasses.

One obvious solution to compensate for the lack of µLED brightness is simply to increase waveguide efficiency. Yet this would entail an order of magnitude increase, which is not feasible in the short term. Even if one could make an efficient RGB µLED panel, there is no short term (next 3-5 years) match with µLED panel capability needed for an daylight use XR eyeglass display. 

What About DLP?

nother MEMS microdisplay that can withstand both high LED or LD flux is DLP, who’s light engines, have been commercially available from a range of licensed manufacturers for over twenty years. The leading 0.2” WVGA miniature light engine is used in the DigiLens smartglasses which ergonomically performs well for its resolution specification.

Fundamentally it is unlikely this or any DLP light engine will serve XR light engine requirements, given the LED size and thermal limitations compounded with DLP chip size inherent to the relatively large MEMS pixel size. When compared to the benefits of WILD, DLP based engines are unable to deliver the high resolution and small form factor XR light engine requirement.

Conclusion

By leveraging the benefits of laser illumination and diffractive/holographic switchable optics, DigiLens is bringing a range of laser diode based light engines to market as an alternative to LED and µLED based light engines with an equally dramatic form factor reduction. 

Once integrated in industrial HUDs and lightweight smartglasses, the benefit of narrow band laser combined with LCoS, is high polarized efficiency and contrast. The benefit of laser diode combined with diffractive optics in both the light engine and waveguide eyeglass, is high efficiency, improved color uniformity, planar construction, miniature size and low cost. Finally, the benefit of laser combined with meta projection lens solutions, are faster time to market, as use of narrow band laser line, leads to a simpler smaller design that is more efficient. 

Companies looking to build smartglasses and other forms of HUDs should be conscious of making the correct near and long-term choice about which light engine technology will scale and grow with their future generations of products. Otherwise they risk misallocating a significant amount of precious resources and will ultimately delay their XR product(s) time to market.

On behalf of the team at DigiLens, I offer this thought piece to posit that for OEMs, a WILD based light engine solution offers the most optimal path to realize a low cost, high resolution/efficiency miniature light engine and thereby maximize the likelihood of success in the emerging XR head worn category, over the next decade.


additional references
Legal Information and Disclosures

This thought piece expresses the views of the author(s) as of the date indicated and such views are subject to change without notice. DigiLens has no duty or obligation to update the information contained herein. Further, DigiLens makes no representation, and it should not be assumed, that the head worn hardware market will develop in a manner envisioned by DigiLens.

This thought piece is being made available for educational purposes only and should not be used for any other purpose. The information contained herein does not constitute and should not be construed as an offering of advisory services. Certain information contained herein concerning trends and performance is based on or derived from information provided by independent third-party sources. Digilens, Inc. (“DigiLens”) believes that the sources from which such information has been obtained are reliable; however, it cannot guarantee the accuracy of such information and has not independently verified the accuracy or completeness of such information or the assumptions on which such information is based.

References[+]

Scroll to Top