Magic Leap: When Reality Hits the Fan

Largely A Summary With Some New Information

ml-slm-beam-splitter-lcos-type-optics-colorI have covered a lot of material and even then only glossed at the surface of what I have learned about Magic Leap (ML). By combining the information available (patent applications, articles, and my sources), I have a fairly accurate picture of what Magic Leap is actually doing based on feedback I have received from multiple sources.

This blog has covered a lot of different topics and some conclusions have changed slightly as I discovered more information and with feedback from some of my sources. Additionally, many people just want “the answer.” So I thought it would be helpful to summarize some of the key results including some more up to date information.

What Magic Leap Is Not Doing In The Product

Between what I have learned and feedback from sources I can say conclusively that ML is not doing the following:

  1. Light Fields – These would requires a ridiculously large and expensive display system for even moderate resolution.
  2. Fiber Scan Displays – They have demonstrated low resolution versions of these and may have used them to convince investors that they had a way to break through the limitations of pixel size of Spatial Light Modulators (SLM) like LCOS, DLP, and OLEDs. Its not clear how much they improved the technology over what the University of Washington had done, but they have given up on these being competitive in resolution and cost with SLMs anytime soon. It appears to have been channeled into being a long term R&D effort and to keep the dream alive with investors.
  3. Laser Beam Scanning (LBS) by Microvision or anyone else – I only put this on the list because of an incredibly ill-informed new release by Technavio stating “Magic Leap is yet to release its product, and the product is likely to adopt MicroVision’s VRD technology.” Based on this, I would give the entire report they are marketing zero credibility; I think they are basing their reports on reading fan-person blogs about Microvision.
  4. OLED Microdisplays – They were using these in their demos and likely in the video they made, but OLED are incompatible optically with there use of a diffractive waveguide (= ML’s Photonic Chip).
Prototypes that Magic Leap Has Shown
  1. FSD – Very low resolution/crude green only fiber scanned display. This is what Rachel Metz described (with my emphasis added) in her MIT Technology Review March/April 2015 article, “It includes a projector, built into a black wire, that’s smaller than a grain of rice and channels light toward a single see-through lens. Peering through the lens, I spy a crude green version of the same four-armed monster that earlier seemed to stomp around on my palm.
  2. ml-495-applicationTI DLP with a conventional combiner and  a “variable focus element” (VFE). They use the DLP to generate a series of focus planes time sequentially and change the VFE between the sequential focus planes. Based on what I have heard, this is their most impressive demo visually and they have been using this for over a year, but the system is huge.
  3. OLED with a conventional combiner (not a waveguide/”Photonics Chip”). This is likely the version they used to shoot their “Through Magic Leap Technology” videos that I analyzed in my Nov. 9th, 2016 blog post. In that article I though that Micro-OLED might be used in the final product, but I have revised this opinion. OLEDs output very wide bandwidth light that is incompatible with waveguides, so it would be incompatible with working with Photonics Chip ML makes such a big deal about.

What is curious is that none of these prototypes, with the possible exception of #1, the single color low resolution FSD, are using a “waveguide.” Waveguides are largely incompatible with OLEDs and having a variable focus element is also problematical.  Also none of these are using LCOS, the most likely technology in the final product.

What Magic Leap Is Trying to Do In Their First “Product”

I’m going to piece together below what I believe based on the information available from both public information and some private conversations (but none of it is based on NDA’ed information as far as I am aware).

  1. ml-slm-beam-splitter-lcos-type-optics-colorLCOS Microdisplay – All the evidence including Business Insider’s October 27, 2016 points to ML using LCOS. They need a technology that will work well with waveguides using narrow band (likely LED) light sources that they can make as bright as necessary and control the angle of the light illumination. LCOS is less expensive, more optically compact, and requires less power than DLP for near eye systems. All these reason are same as why Hololens is using LCOS. Note, I’m not 100% sure on them using LCOS, but it by far the most likely technology they will be using. They could also be using DLP but I would put that at less than a 10% chance. I’m now ruling out Micro-OLED because it would not work in a waveguide.
  2. Two (2) sequential focus planes are supported – The LCOS microdisplay is likely only able to support about 120 full color frames per second which is only enough to support 2 sequential focus planes per 1/60th of a second of a moving image. Supporting more planes at a slower rate would result in serious image breakup when things move. The other big issue is the amount of processing required. Having even two focus planes greatly increase the computation that have to be done. To make it work correctly, they will need to track the person’s pupils and factor that into their processing and deal with things like occlusion. Also with the limited number of focus planes they will have to figure out how to “fake” or deal with a wider range of focus.
  3. Variable Focus – What I don’t know is how they are supporting the change in focus between the sequential focus planes. They could be using some form of electrically alterable lens but it is problematical to have non-collimated light entering a waveguide. It would therefore seem more consistent for them to be using the technique shown in their patent application US 2016/0327789 that I discussed before.
  4. Photmagic-leap-combiner-croponics Chip (= Diffractive Waveguide) – ML has made a big deal about their Photonic’s Chip, what everyone else would call a “waveguide.” The Photonics Chip likely works similar to the one Hololens uses (for more information on waveguides, see my Oct 27th, 2016 post). The reports are that Hololens has suffered low yields with their Waveguides and Magic Leaps will have more to do optically to support focus planes.

Overall, I think it it is very clear that what they will actually make is only a fraction of he vision they have portrayed to the press. They may have wanted to do 50 megapixel equivalent foveated displays, use FSD as their display device, have 6 focus planes, or even (from Fortune July 12, 2016) ““light-field” technology essentially mimics the brain’s visual-perception mechanisms to create objects and even people who look and behave just the way they would in the real world, and interact with that world seamlessly.” But then, they have to build something that actually works and that people can afford to buy. Reality then hits the fan

Karl Guttag
Karl Guttag
Articles: 244


    • Also note the authors comment that although the Epson has a smaller FOV it has a better picture quality than the LUMUS model he tried which reportedly uses a LCOS display .

      “Though other companies have larger transparent displays – like the Lumus headset I tried back at CES in January – they fall short of the picture quality of the Moverio set. “

      • Take heart :-). It is not clear to me that “diffractive waveguides” won’t be the Wankel engine of display technologies. They enable thin optics which are good for the egos of the designers and stylist, but the image quality suffers when being bent at such tight angles. The diffraction gratings have some error in them that hurts image quality and they are expensive to make. They really need (expensive) laser illumination and holograms to do a good job.

        As you many know ODG is planning on using Micro-OLEDs in their new design. ODG uses simple beam splitting mirrors. Any display viewed in a simple beam splitter is going to look better than the best waveguides I have seen. The glasses just may not look as high tech.

    • Epson is NOT using a waveguide. Epson is using a simple prism beam splitter. A waveguide would be thinner and be using total internal reflections (TIR).

      But I should have been clearer and say “diffractive waveguide”. The is a chance that OLEDs might be able to use a prismatic type “waveguide” such as by Lumus.

      The problem for OLED is that they put out broad spectrum colors (sort of like phosphors) and these become a total mess with a waveguide. It looks like Magic Leap is using a think waveguide with a “flat injection” like hololens and this would requires diffractive or holographic elements which in turn would requires narrow spectrum/bandwidth colors from either a LED or laser.

      Another problem for OLEDs is that the light is very diffused which might be problematical for getting collimated to make the TIR work well. With DLP and LCOS you can illuminate with collimated light.

      • Do you have an opinion on how well inorganic LED (micro-LED) displays will work with “diffractive waveguides” ?

        Apple is pursuing Micro-LED and purchased LuxVue in 2014

        Oculus purchased InfiniLED recently .

        eMagin also has a patent app on the tech .

      • What, me express an opinion :-)?

        Only the Seeking Alpha had any technical detail. They talk about them only have “blue.” The fundamental issue is that the different colors require different semiconductors and/or crystal structures. So you can’t get small DIRECT red, green, and blue on the same semiconductor chip easily and small. Likely what they will do is use the blue LEDs to stimulate red and green phosphors (this is what LG OLED TVs do as well). The phosphors will have broad spectrum and very diffuse light. So I would think it would be as bad or worse for waveguides. It may be more efficient than OLEDs because blue LEDs can be very efficient and phosphor conversion is about 80% efficient.

      • Thanks ,

        The SA article also linked to some VerLASE tech which is an alternative to phosphor conversion that sounds very interesting .

        VerLASE Technologies announced today that US Patent No. 9,019,595 has issued on its revolutionary Chromover™ wavelength conversion technology. The technology introduces an entirely new approach for efficiently converting colors from inexpensive, widely available blue/violet light sources such as LEDs or laser diodes to any color in the visible range; for example, to the optimal greens and reds needed for the ideal projection of images, which have been otherwise proven very difficult to create. Significantly, not only can color be tailored but also its spectral characteristics, such that the output can be designed to be a laser, LED, or in-between, in a single color or a RGB projector light source, or as a multi-color micro-array, ideally suited for integrating with microLED displays for the next generation of Near Eye Displays, including holographic 3D variants.

        Do you think when they mention tailoring the “spectral characteristics” they are referring to the ability to make a “narrow” spectrum/bandwidth as would be required for use in ” diffractive waveguides” ?

      • Thanks, I checked a bit on this. It sounds like more of a concept than reality. They are saying that they think they can get the “pixellated microarray down to resolution significantly less than 10 microns with RGB colors per pixel as needed.” LCOS and DLP are already in the 4 micron range or about 16/100th the area. They will have to be way below 10 microns to have pixels small enough. This pixel size might be good for something light smartwatch, but seems huge for a “microdisplay.”

        If they could make it into a laser, one could argue it would be a way to drive an array of Fiber Scanning Displays. They are saying that they can control the wavelength to be as narrow as desired.

        I did not see where they had demonstrated it anywhere. It sounds a lot like it is still a concept. Have you seen anything else about it? I would also wonder about the yield if it could be made; if you have 1 defect in 2 million pixels you don’t have a good display if you are building a 1080p device.

  1. LCoS makes a lot more sense than OLED I agree, but the fact that they raised their last round with the public justification of buying a MEMS fab (and have been hiring people for their “MEMS spacial light modulator” — i.e. DLP clone) suggests to me that that is the approach they are taking.

  2. Dear Karl, thank you. Interesting article.

    Your opinion – Oled with simple beam splitter is better solution?

    As you many know ODG is planning on using Micro-OLEDs in their new design. ODG uses simple beam splitting mirrors. Any display viewed in a simple beam splitter is going to look better than the best waveguides I have seen. The glasses just may not look as high tech.

    LCOS is less expensive, more optically compact, and requires less power than DLP for near eye systems.

    Why LCOS is more optically compact and requires less power?

    By my opinion DLP as illumination source is much better for final picture than LCOS.
    Your opinion? Which illumination source DLP, OLED or LCOS is better for final quality picture with with simple beam splitter (not a waveguide combiner).

    Thank you for your answers.

    • OLED and a simple combiner, either ~45 degree or a curved semi-mirror will blow away any waveguide on image quality that I have seen. There will have to be some optics prior to the combiner to move the focus into the far vision (the curved combiner will require less). DLP and LCOS can also be very good in this configuration. Designs more often use LCOS because it is lower power and less expensive (considerably so) than OLED. Both LCOS and DLP have a major brightness advantage which is important if the system is going to be used outdoors.

      Lots of reasons LCOS is lower power. To being with it runs on lower voltages that DLP and power tends to go a Voltage-Squared. It generally transition less so you have and fxV-squared issue. The DLP requires a lot more formatting with its ASICs and therefore uses more power in the ASIC and the memory support it. So the DLP BEFORE you factor in the LEDs takes a lot more power. The DLP is more LED light efficient so for front projectors over about 50 lumens it tends to win (at least with LED illumination), but near eye devices typically only need about 0.5 lumens (more or less depending on a lot of factors). So the power of the display system tends to dominate.

      OLED has the advantage and disadvantage of being self illuminated. This simplifies and improves the optical design. But you are limited with the color spectrum, brightness, and diffuse light of the LEDs. LCOS and DLP requires a more complex path to route the light into and out, but you can control the light source and get very bright and long lifetimes.

      DLP requires “of axis illumination” which tends to be bigger. LCOS can be very compact and even use a curved beamsplitter to be almost flat.

      It is easier to get DLP to high contrast than LCOS. LCOS can also get to high contrast but it is harder. I have heard the DLP’s contrast is not as good as the older DLP on contrast with the new Tilt-n-roll due to stray light and diffraction, but I have not personally evaluated it nor know how it would compare to LCOS.

      DLP has a higher field rate which is why it would seem to be a better solution for sequential focus plane.

      The net is that LCOS seems to dominate the near eye designs these days. There are some notable exceptions. From what I have seen lately, it seems to be mostly LCOS with some Micro-OLED and a few DLP. Nothing is in seriously high (ala consumer) volumes yet so this is more based on the number of designs and not volume.

      Better is a tough question. Right now they spend a lot more for OLED, but if you put that money into an LCOS design it can look pretty good as well. Avegant chose DLP because they liked the image quality they could get.

Leave a Reply