Update (Oct. 19th, 2024)

While the general premise of this article is that Meta Orion is using similar waveguide technology to Snap (Wave Optics) and that Magic Leap 2 is correct, it turns out that a number of assumptions about the specifics of what the various companies actually used in their products were incorrect. One of my readers (who wishes to remain anonymous) with deep knowledge of waveguides responded to my request for more information on the various waveguides. This person had both a theoretical knowledge of waveguides and what Meta Orion, Wave Optics (now Snap), Magic Leap Two, and Hololen 2 used.

My main error about the nature of waveguide “grating” structures was a bias toward linear gratings, with which I was more familiar. I overlooked the possibility that Wave Optics was using a set of “pillar” gratings that act like a 2D set of linear gratings.

A summary of the corrections:

1. Hololens 2 had a two-sided waveguide. The left and right expansion gratings are on opposite sides of the waveguide.
2. Prior Wave Optics (Snap) waveguides use a pillar-type 2-D diffraction grating on one side. There is a single waveguide for full color. The new Snap Spectacles 5 is likely (not 100% sure) using linear diffraction gratings on both sides of a single waveguide full color, as shown in this article.
3. Magic Leap Two uses linear diffraction gratings on both sides of the waveguide. It does use three waveguides.

The above corrections indicate that Meta Orion, Snap Spectacles 5 (Wave Optics), and Magic Leap all have overlapping linear gratings on both sides. Meta Orion and Snap likely use a single waveguide for full color, whereas the Magic Leap 2 has separate waveguides for the three primary colors.

I’m working on an article that will go into more detail and should appear soon, but I wanted to get this update out quickly.

Introduction and Background

After my last article, Meta Orion AR Glasses (Pt. 1 Waveguides), I got to thinking that the only other diffractive grating waveguide I have seen with a 2-D (X-Y) expansion and exit gratings, used in Meta’s Orion, was from Wave Optics (purchased by Snap in May 2021)

The unique look of Wave Optics waveguides is how I easily identified that Snap was using them before it was announced that Snap had bought Wave Optics in 2021 (see Exclusive: Snap Spectacles Appears to Be Using WaveOptics and [an LCOS] a DLP Display).

I then wondered what Magic Leap Two (ML2) did to achieve its 70-degree FOV and uncovered some more interesting information about Meta’s Orion. The more I researched ML2, the more similarities I found with Meta’s Orion. What started as a short observation that Meta Orion’s waveguide appears to share commonality with Snap (Wave Optics) waveguides ballooned up when I discovered/rediscovered the ML2 information.

Included in this article is some “background” information from prior articles to help compare and contrast what has been done before with what Meta’s Orion, Snap/Wave Optics, and Magie Leap Two are doing.

Diffractive Waveguide Background

I hadn’t looked at in any detail how Wave Optics diffraction gratings worked differently before. All other diffraction (I don’t know about holographic) grating waveguides I had seen before used three (or four) separate gratings on the same surface of the glass. There was an Entrance Grating, a first expansion and turning grating, and then a second expansion and exit grating. The location and whether the first expansion grating was horizontal or vertical varied with different waveguides.

Hololens 2 had a variation with left and right horizontal expansion and turning gratings and a single exit grating to increase the field of view. Still, all the gratings were on the same side of the waveguide.

Diffraction gratings bend light based on wavelength, similar to a prism. But unlike a prism, a grating will bend the light in a series of “orders.” With a diffractive waveguide, only the light from one of these orders is used, and the rest of the light is not only wasted but can cause problems, including “eye glow” and reduce the contrast of the overall system

Because diffraction is wavelength-based, it bends different colors/wavelengths in different amounts. This causes issues when sending more than one color through a single waveguide/diffraction grating. These problems are compounded as the size of the exit grating and FOV increases. Several diffraction waveguide companies have one (full color), or two (red+blue and blue+green) waveguides for smaller FOVs and then use three waveguides for wider FOVs.

For more information, Quick Background on Diffraction Waveguides, MicroLEDs and Waveguides: Millions of Nits-In to Thousands of Nits-Out with Waveguides, and Magic Leap, HoloLens, and Lumus Resolution “Shootout” (ML1 review part 3).

Meta Orion’s and Wave Optics Waveguides

I want to start with a quick summary of Orion’s waveguide, as the information and figures will be helpful in comparing it to that of Wave Optics (owned by Snap and in Snap’s Spectacles AR Glasses) and the ML2.

Summary of Orion’s waveguide from the last article

Orion’s waveguide appears to be using a waveguide substrate with one entrance grating per primary color and then two expansion and exit/output gratings. The two (crossed) output gratings are on opposite sides of the Silicon Carbide (SiC) substrate, whereas most diffractive waveguides use glass, and all the gratings are on one side.

Another interesting feature shown in the patents and discussed by Meta CTO Bosworth in some of his video interviews about Orion is “Disparity Correction,” which has an extra grating used by other optics and circuitry to detect if the waveguides are misaligned. This feature is not supported in Orion, but Bosworth says it will be included in future iterations that will move the input grating to the “eye side” of the waveguide. As shown in the figure below, and apparently in Orion, light enters the waveguide from the opposite side of the eyes. Since the projectors are on the eye side (in the temples), they require some extra optics, which, according to Bosworth, make the Orion frames thicker.

Wave Optics (Snap) Dual-Sided 2D Expanding Waveguide

Wave Optics US patent application 2018/0210205 is based on the first Wave Optics patent from the international application WO/2016/020643, first filed in 2014. FIG 3 (below) shows a 3-D representation of diffraction grating with an input grating (H0) and cross gratings (H1 and H2) on opposite sides of a single waveguide substrate.

The patent also shows that the cross gratings (H1 and H2) are on opposite sides of a single waveguide (FIG. 15B above) or one side of two waveguides (FIG. 15A above). I don’t know if Wave Optics (Snap) uses single- or double-sided waveguides in its current designs, but I would suspect it is double-sided.

While on the subject of Wave Optics waveguide design, I happen to have a picture of a Wave Optics 300mm glass wafer with 24 waveguides (right). I took the picture in the Schott booth at AR/VR/MR 2020. In the inset, I added Meta’s picture of the Orion 100mm SiC wafer, roughly to scale, with just four waveguides.

By the way, in my May 2021 article Exclusive: Snap Spectacles Appears to Be Using WaveOptics and [an LCOS] a DLP Display, I assumed that Spectacles would be using LCOS in 2021 since WaveOptics was in the process of moving to LCOS when they were acquired. I was a bit premature, as it took until 2024 for Spectacles to use LCOS.

In my hurry in putting together information and digging for connection, it was looking to me that WaveOptics would be using an LCOS microdisplay. As I pointed out, WaveOptics had been moving away from DLP to LCOS with their newer designs. Subsequent information suggests that WaveOptics was still using their much older DLP design. It is still likely that future versions will use LCOS, but the current version apparently does not.

Magic Leap

Magic Leap One (ML1) “Typical” Three Grating Waveguide

This blog’s first significant article about Magic Leap was in November 2016 (Magic Leap: “A Riddle Wrapped in an Enigma”). Since then, Magic Leap has been discussed in about 90 articles. Most other waveguide companies coaxially input all colors from a single projector. However, even though the ML1 had a single field sequential color LCOS device and projector, the LED illumination sources are spatially arranged so that the image from each color output is sent to a separate input grating. ML1 had six waveguides, three for each of the two focus planes, resulting in 6 LEDs (two sets of R, G, & B) and six entrance gratings (see: Magic Leap House of Cards – FSD, Waveguides, and Focus Planes).

Below is a diagram that iFixit developed jointly with this blog. It shows a side view of the ML1 optical path. The inset picture in the lower right shows the six entrance gratings of the six stacked waveguides.

Below left is a picture of the (stack of six) ML1 waveguides showing the six entrance gratings, the large expansion and turning gratings, and the exit gratings. Other than having spatially separate entrance gratings, the general design of the waveguides is the same as most other diffractive gratings, including the Hololens 1 shown in the introduction. The expansion gratings are mostly hidden in the ML1’s upper body (below right). The large expansion and turning grating can be seen as a major problem in fitting a “typical” diffractive waveguide into an eyeglass form factor, which is what drove Meta to find an alternative that goes beyond the ML1’s 50-degree FOV.

Figure 18 from US application 2018/0052276 diagrams the ML1’s construction. This diagram is very close to the ML1’s construction down to the shape of the waveguide and even the various diffraction grating shapes.

Magic Leap Two (ML2)

The ML1 failed so badly that very few were interested in the ML2 compared to the ML1. There is much less public information about the second-generation device, and I didn’t buy an ML2 for testing. I have covered many of the technical aspects of ML2, but I haven’t studied the waveguide before. With the ML2 having a 70-degree FOV compared to the ML1’s 50-degree FOV, I became curious about how they got it to fit.

To start with, the ML2 eliminated the ML1’s support for two focus planes. This cut the waveguides in half and meant that the exit grating of the waveguide didn’t need to change the focus of the virtual image (for more on this subject, see: Single Waveguide Set with Front and Back “Lens Assemblies”).

Looking through the Magic Leap patent applications, I turned up US 2018/0052276 to Magic Leap, which shows a 2-D combined exit grating. US 2018/0052276 is what is commonly referred to in the patent field as an “omnibus patent application,” which combines a massive number of concepts (the application has 272 pages) in a single application. The application starts with concepts in the ML1 (including the just prior FIG 18) and goes on to concepts in the ML2.

This application, loosely speaking, shows how to take the Wave Optics concept of two crossed diffraction gratings on different sides of a waveguide and integrate them onto the same side of the waveguide.

Magic Leap patent application 2020/0158942 describes in detail how the two crossed output gratings are made. It shows the “prior art” (Wave Optics and Meta Orion-like) method of two gratings on opposite sides of a waveguide in FIG. 1 (below). The application then shows how the two crossed gratings can be integrated into a single grating structure. The patent even includes scanning electron microscope photos of the structures Magic Leap had made (ex., FIG 5), which demonstrates that Magic Leap had gone far beyond the concept stage by the time of the application’s filing in Nov. 2018.

Magic Leap Two Disparity Correction

I then went back to pictures I took of Magic Leap’s 2022 AR/VR/MR conference presentation (see also Magic Leap 2 at SPIE AR/VR/MR 2022) on the ML2. I realized that the concept of a 2D OPE+EPE (crossed diffraction gratings) was hiding in plain sight as part of another figure, thus confirming that ML2 was using the concept. The main topic of this figure is “Online display calibration,” which appears to be the same concept as Orion’s “disparity correction” shown earlier.

The next issue is whether the ML2 used a single input grating for all colors and whether it used more than one waveguide. It turns out that these are both answered in another figure from Magic Leap’s 2022 AR/VR/MR presentation shown below. Magic Leap developed a very compact projector engine that illuminates and LCOS panel through the (clear) part of the waveguides. Like the ML1, the red, green, and blue illumination LEDs are spatially separated, which, in turn, causes the light out of the projector lens to be spatially separated. There are then three spatially separate input gratings on three waveguides, as shown.

Based on the ML2’s three waveguides, I assumed it was too difficult or impossible to support the “crossed” diffraction grating effect while supporting full color in a single wide FOV waveguide.

Summary: Orion, ML2, & Wave Optics Waveguide Concepts

Orion, ML2, and Wave Optics have some form of two-dimensional pupil expansion using overlapping diffraction gratings. By overlapping gratings, they reduce the size of the waveguide considerably over the more conventional approach, with three diffraction gratings spatially separate on a single surface.

To summarize:

• Meta Orion – “Crossed” diffraction gratings on both sides of a single SiC waveguide for full color.
• Snap/Wave Optics – “Crossed” diffraction gratings on both sides of a single glass waveguide for full color. Alternatively, “crossed” diffraction waveguides on two glass waveguides for full color (I just put a request into Snap to try and clarify).
• Magic Leap Two – A single diffraction grating that acts like a crossed diffraction grating on high index (~2.0) glass with three waveguides (one per primary color).

The above is based on the currently available public information. If you have additional information or analysis, please share it in the comments, or if you don’t want to share it publicly, you can send a private email to newsinfo@kgontech.com. To be clear, I don’t want stolen information or any violation of NDAs, but I am sure there are waveguide experts who know more about this subject.

What about Meta Orion’s Image Quality?

I have not had the opportunity to look through Meta’s Orion or Snap Spectacles 5 and have only seen ML2 in a canned demo. Unfortunately, I was not invited to demo Meta’s Orion, no less have access to one for evaluation (if you can help me gain (legal) access, contact me at newsinfo@kgontech.com).

I have tried the ML2 a few times. However, I have never had the opportunity to take pictures through the optics or use my test patterns. From my limited experience with the ML2, it is much better in terms of image quality than the ML1 (which was abysmal – see Magic Leap Review Part 1 – The Terrible View Through Diffraction Gratings), it still has significant issues with color uniformity like other wide (>40-degree) FOV diffractive waveguides. If someone has a ML2 that I can borrow for evaluation, please get in touch with me at newsinfo@kgontech.com.

I have been following Wave Optics (now Snap) for many years and have a 2020-era Titan DLP-based 40-degree FOV Wave Optics evaluation unit (through the optics picture below). Wave Optics Titan, I would consider a “middle of the pack” (I had seen better and worse) diffractive waveguide at that time. I have seen what seem to be better diffractive waveguides before and since, but it is hard to compare them objectively as they have different FOVs, and I was not able to use my content but rather curated demo content. Wave Optics seemed to be showing better waveguides at shows before being acquired by Snap 2021, but once again, that was with their demo content with short views at shows. I am working on getting a Spectacles 5 to do a more in-depth evaluation and see how it has improved.

Without the ability to test, compare, and contrast, I can only speculate about Meta Orion’s image quality based on my experience with diffractive waveguides. The higher index of refraction of SiC helps as there are fewer TIR bounces, which degrades image quality, but it is far from a volume production-ready technology. I’m concerned about image uniformity with a large FOV and even more so with a single set of diffraction gratings as diffraction is based on wavelength (color).

Lumus Reflective Waveguide Rumors

In Meta Orion AR Glasses: The first DEEP DIVE into the optical architecture, it stated:

There were rumors before that Meta would launch new glasses with a 2D reflective (array) waveguide optical solution and LCoS optical engine in 2024-2025. With the announcement of Orion, I personally think this possibility has not disappeared and still exists.

The “reflective waveguide” would most likely be a reference to Lumus’s reflective waveguides. I have seen a few “Lumus clone” reflective waveguides from Chinese companies, but their image quality is very poor compared to Lumus. In the comment section of my last article, Ding, on October 8, 2024, wrote:

There’s indeed rumor that Meta is planning an actual product in 2025 based on LCOS and Lumus waveguide. 

Lumus has demonstrated impressive image quality in a glasses-like form factor (see my 2021 article: Exclusive: Lumus Maximus 2K x 2K Per Eye, >3000 Nits, 50° FOV with Through-the-Optics Pictures). Since the 2021 Maximus, they have been shrinking the form factor and improving support for prescription lens integration with their new “Z-lens” technology. Lumus claims its Z-Lens technology should be able to support greater than a 70-degree FoV in glass. Lumus also says because their waveguides support a larger input pupil, they should have a 5x to 10x efficiency advantage.

The market question about Lumus is whether they can make their waveguide cost-effectively in mass production. In the past, I have asked their manufacturing partner, Schott, who says they can make it, but I have yet to see a consumer product around the Z-Lens. It would be interesting to see if a company like Meta had put the kind of money they invested into complex Silicon Carbide waveguides into reflective waveguides.

While diffractive waveguides are not inexpensive, they are considered less expensive at present (except, of course, for Meta Orion’s SiC waveguides). Perhaps an attractive proposition to researchers and propriety companies is that diffraction waveguides can be customized more easily (at least on glass).

Not Addressing Who Invented What First

I want to be clear: this article does not in any way make assertions about who invented what first or whether anyone is infringing on anyone else’s invention. Making that determination would require a massive amount of work, lawyers, and the courts. The reason I cite patents and patent applications is that they are public records that are easily searched and often document technical details that are missing from published presentations and articles.

Conclusions

There seems to be a surprising amount of commonality between Meta’s Orion, the Snap/Wave Optics, and the Magic Leap Two waveguides. They all avoided the “conventional” three diffraction gratings on one side of a waveguide to support a wider FOV in an eyeglass form factor. Rediscovering that the ML2 supported “dispersion correction,” as Meta refers to it, was a bit of a bonus.

As I wrote last time, Meta’s Orion seems like a strange mix of technology to make a big deal about at Meta Connect. They combined a ridiculously expensive waveguide with a very low-resolution display. The two-sided diffraction grating Silicon Carbide waveguides seem to be more than a decade away from practical volume production. It’s not clear to me that even if they could be made cost-effective, they would have as good a view out and the image quality of reflective waveguides, particularly at wider FOVs.

Meta could have put together a headset with technology that was within three years of being ready for production. As it is, it seemed like more of a stunt in response to the Apple Vision Pro. In that regard, the stunt seems to have worked in the sense that some reviewers were reminded of seeing the real world directly with optical AR/MR beats, looking at it through camera and display.