Magic Leap 2 for Enterprise, Really? Plus Another $500M

Introduction

This blog has covered Magic Leap for nearly five years. I correctly predicted the key display capability of Magic Leap 1 three years before it was formally announced (See 2016 “best fit” and the iFixit Teardown I helped analyze in 2019). As predicted, Magic Leap burned through over $3 billion of investor money and offered little new over existing headsets with the Magic Leap One.

On October 11, Magic Leap released the first pictures and clues about the Magic Leap 2 (ML2), and I wanted to try my hand again at gleaning some details from little information that they have given out.

Under new Magic Leap CEO Peggy Johnson, they claim to have changed focus from the consumer to the “enterprise” market (I will try and lay out why I am skeptical). Also, as just announced, they have found investors to give them yet another $500 million at a $2 billion valuation.

First, I will piece together what Magic Leap showed on the Magic Leap 2 (ML2). Then I will go through the information and pictures and analyze it.

What We Learned from Peggy Johnson’s CNBC Interview and Her Article

Most of the source content of this article comes from an October 11, 2021 CNBC Interview with Peggy Johnson, CEO of Magic Leap,  and an “op-ed” by Peggy Johnson, “My First Year at Magic Leap and the Opportunity Ahead” on Magic Leap’s website published the same day. The information below is quoted from the CNBC interview with my bold highlights and comments (in red brackets): 

CNBC’s Julia Boorstin stated in hear introduction, “The headset also has a unique feature. It’s able to dim the real-world background to make it easier to see augmented reality elements in the foreground.”

JohnsonThe feature you mentioned [diming the real world] segmented is going to be very important, particularly for folks like surgeons in very brightly lit operating rooms. You can still in that environment show digital content.

We’ve doubled the field of view [mostly vertically].

We’ve really improved image quality color fidelity text legibility.  But maybe most importantly for many enterprise users is we’ve reduced size and weight it truly is an all-day everyday wearing device.

Tyler Mathisen, of CNBC, “What if you wear spectacles do they do they work. How do you accommodate them?”

Johnson, “We have inserts that you can put in if you do wear spectacles and what that does is it keeps that headset very close to your eyes which is very important to have the right focus [Untrue – you don’t need it that close] and so the inserts just slip right in and off you go so we’ve accommodated for that.”

It sounds like the ML2 has an electrically controllable dimming shutter to block incoming light. It has about double the FOV (interestingly, it doubles it vertically). Doubling the FOV (mostly vertically) while improving image quality and text readability (if true) to me translates into ML2 using a higher resolution and better quality LCOS display. The fact that they still are using eye correction inserts means they still don’t have enough eye relief, like Hololens 1 and 2, to support someone with glasses wearing the unit.

Each of the issues above will all be addressed in more detail later.

Addition Pictures and Information

The picture (below left) from Magic Leap shows the form factor of the ML2. For reference, I took a picture of the Magic Leap One (ML1) from approximately the same viewpoint (below right).

The figure below zooms in on a single lens of the ML2 and a corresponding view looking inside the ML2 (taken with illumination and a polarizer on the camera lens to see through the outer cover).

Some key points of difference between ML2 and ML1 based on the image:

  • The front is slightly thinner, with flat lens covers and a more squared-off look.
  • The ML2 has a hinged/tiltable headband
  • There is still a cable that looks wired into the headset, whereas the ML2 had two. A wired-in cable is a major problem as it is a snag hazard, and it also will result in damage due to the cable pulling the headset off surfaces and broken cables.
  • They have two different size cameras, likely one with a wider angle than the other, in the center of the headset on the HL1 versus two identical cameras in the temples on the ML2.
  • The hand gesture cameras that aimed slightly downward have been angled more forward.
  • A sensor receiver pair below the cameras on the ML2 is gone entirely on the ML1.
  • It looks like they are still using waveguides, most likely diffractive waveguides.
  • There are at least two glued together sandwiched flat devices that appear to have holes cut through them for the IR Eye Detection camera and indicated by the curved shapes in the glue lines. I suspect the first sandwich is the dimming shutter, and the second one is the waveguides but this is just a guess at this point.
  • The Eye detection camera moved from the bottom of the ML1 to the outer lower corner of the ML2. This change is likely due to the larger vertical FOV.

Double the FOV – Vertically

Magic Leap CEMagic Leap CEO Johnson’s article includes a picture comparing the FOV of the ML2 to the ML1. I have added to the drawing the Hololens 1 and 2 for reference (below). Unlike most other AR headsets, the ML2 has a wider vertical rather than horizontal FOV.

The most common format for videos today, which has driven display devices used in AR, is HDTV’s 16:9. But many experts think that something closer to a square (1:1) aspect ratio would be better for AR and VR as it more closely matches the FOV of each human eye. We can see this trend with Hololens 1 being 16:9, whereas Hololens 2 is 3:2. Vuzix’s Blade is the only other AR device I can remember that has an aspect ratio that is significantly taller than wide.

When someone does something that on the surface appears wrong, the reason is usually that “the alternative was worse. For example, it could be with the waveguide technology Magic Leap that it was easier to expand the FOV vertically than horizontally. While they could have rotated the waveguide 90 degrees, it would have put the projector in a location that would not have integrated well into the glasses-like form factor.

Likely Still Using LCOS for the Display

There is not a lot to go on with the display technology. Magic Leap’s first design used an LCOS device, and there are many better LCOS devices today that they could choose to support the bigger FOV. I don’t see why they would change, given that they are going into a waveguide.

Dimming – A Big Deal?

Both in the CNBC interview and Johnson’s op-ed made a big deal out of (assumed) electronic dimming. Quoting Johnson’s op-ed, “dimming – a first-to-market innovation that enables the headset to be used in brightly lit settings.” Unless they have some breakthrough in dimming/shutter technology I don’t know about, this seems to hyperbolic.

While electronic dimming is an article in itself, the drawbacks the issues with electronic diming include:

  • Transmissivity of the “clear” state is usually less than 40%. So they start out blocking more than 60% of the light before the dimming is applied.
  • There is often some view distortion, including color-shifting, dispersion (milky), and/or diffraction.
  • Polarization effects with polarization technologies. This can particularly be a problem when trying to read LCD devices that also use polarization.
  • Usually, non-polarization-based technologies require higher voltages (often more than 12 volts) and are much slower changing.

Maybe Magic Leap has found a new type of much improved dimming effect, but it is more likely that they decided to live with one or more of the well-known negative effects listed above.

Dimming Real-World Suggests A Brightness Issue

Back in 2018, I measured the ML1’s brightness and only got about 210 nits. That is compared to the Hololens 1 at about 320 nits and the Hololens 2 in the 500 nits range. Typical leading edge smartphones have more than 600 nits. Going further, the recent Lumus Maximus has over 3,000 nits. The ML2 use of dimming may say more about their lack of brightness than the innovation of a dimming shutter.  

What About Vergence Accommodation?

Going back as far as their 2013 investor presentation, Magic Leap made a big deal about vergence accommodation conflict (VAC). This issue manifested itself in the ML1 having two sets of waveguides, each set to a different focus distance. Supporting two focus distances cause problems both with both the view of the real world and the virtual images.

While VAC was a big deal for Magic Leap One (see, for example, my 2018 article), it wasnot mentioned at all by Ms. Johnson. Combined with all the talk about improved image quality and larger FOV, I tend to think that the dual focus distances may have been dropped.

Waveguides without a diopter adjustment will appear to focus at near infinity. Hololens 1 and 2 use a diopter lens after the waveguide (cleverly molded into the back of the shield, so you don’t normally see it) and a lens glued on the front of the waveguide (to negate the diopter effect for the real-world view).

ML1 built in a diopter adjustment into the exit grating of the waveguide as shown in the figure (right) from Bernard Kress’s “Optical Architectures for Augmented-, Virtual-, and Mixed-Reality Headsets.”

As Kress, who works for Microsoft, points out on page 164 of the book, in reference to the ML1’s use of the exit grating-based diopter adjustment:

A powered out-coupler grating might reduce the MTF of the image, especially in the direction of the lens offset (direction of TIR propagation), since the input (I-E) and output (O-E) couplers are no more perfectly symmetric (the input coupler being a linear grating in both cases, and the out-coupler an off-axis diffractive lens). Thus, the spectral spread of the image in each color band cannot be compensated perfectly and will produce LCA in the direction of the lens offset.”

In short, Kress is saying that ML1’s exit-grating diopter adjustment method tends to degrades the image quality. The dual diopter ML1 required the exit-grating diopter adjustment, but it will be interesting to see if Magic Leap keeps with that method if they only have a single focus distance.

The Need for Both Near and Far Focus

Supporting (at least) dual focus distances is not just an academic subject but a serious issue for “enterprise” applications.  Most VR and AR headsets aimed at “games” set the focus at about 2 meters. At this focus distance, the eye muscles are relatively relaxed and won’t have serious VAC problems dealing with anything that, due to the binocular disparity (stereo glasses), appear to be from about 1.5 meters or beyond.  

The VAC issues occur when someone is on something with their hands if the focus is about 2 meters. When working on something with their hands, a person will be looking at about 0.3 meters away, which will cause VAC issues such as headaches and eye strain with prolonged use. One of the examples Ms. Johnson gave in the CNBC interview was doctors, which is a well-known case where you would want the closer effective focus distance as doctors are working with their hands for prolong periods.  It will be interesting to see if:

  1. Magic Leap supports the dual diopter and suffers image quality losses
  2. Has both a “near use” and a “far use” designs
  3. Ignores the issue in favor of better image quality

Interpupillary distance IPD adjustment

The ML1 had two different models, and while many thought it was due to head sizes, it was due to the very wide range of distances between human pupils, known as interpupillary distance or IPD. While having two models adjusting for a different range becomes a logistical hassle having the right number of headsets of the various types. Supporting a single design requires giving up some amount of horizontal FOV for dedication to IPD adjustment. There is no indication I have seen if ML2 will have more than one IPD model.

“For Enterprise” — Really?

Magic Leap, from the beginning, focused on consumer entertainment and gaming, and this focus seeped deep into their hardware and headset decisions. By 2019 ML1 was failing, and Magic Leap was running out of money when they said they were pivoting to the enterprise market. They took the ML1 and deemed it an enterprise device with zero changes to the design, and they announced the ML2 would target the enterprise market.

Either the ML2 was too far into the design, or Magic Leap didn’t have the money to design a product for the enterprise market. While the Hololens 2 image quality is terrible, as this blog has gone into great detail to explain (see these articles), they at least improved on basic ergonomic issues.

Importantly compared to the HL2, the ML2 fails on the following:

  • The ML2 lacks enough eye relief to support most glasses, and each headset must be customized for people that need eyewear before it can be used. It eliminates applications for “occasional use” such as auto showrooms, museums, and similar applications.
  • Restricted peripheral vision with the ML2 and the ML1 (see above right) should be a safety concern in many enterprise applications. In many environments, it is dangerous to block the user’s peripheral vision. The issue of not blocking peripheral vision has been discussed on this blog many times, including comparisons between ML1 and Hololens 1.
  • The cable is a dangerous snag hazard in many environments, and it will also cause an asymmetric pull on the headset due to the weight of the cable. With the ML1 and ML2 designs, not only are the glasses close to the eyes, but the temples and cameras block off much of ones peripheral vision.
  • Of secondary concern is that the ML2 does not appear to be supported well enough. Note how the HL2 has a support over the forehead and (not shown above) has an optional over the head strap. There may be added straps or support not shown in the ML2’s pictures shown to date.

The one feature that is “enterprise-like” versus consumer may be the very large vertical FOV.  

Magic Leap Patents Appear to still be in hock to Citibank rather than J.P. Morgan

As this blog reported on November 10, 2019, apparently all of Magic Leap’s I.P. was put up as collateral. It turns out that on November 07, 2019, the assignment of all the I.P. was then sold to Citibank. As of October 12, 2021, the US Patent office is still showing Magic Leap patents as assigned to Citibank (see the screen capture from USPTO Public Pair – click on it to see it full size).

It appears that the new $500m investment may not include securing the rights to the I.P. as it is still held by Citibank. I have not seen the terms of the agreement, so I don’t know what it will take to get the I.P. back. Assuming Magic Leap secured another $500m with a $2b valuation without they I.P. combined with Magic Leap’s track record says something about the parties involved.

Conclusion – “An Enterprise Sticker On a Consumer Product”

So far, what I have seen of the ML2 seems to confirm what I thought it would be when Rony Abovitz was still CEO and said the ML2 would be aimed at the enterprise market. It seems the shift was mostly in marketing and not so much the design

The cable problem might have been relatively easily solved with the “puck on the back” approach like the HL2 (and Nreal Enterprise Edition). But the eye relief and allow much better peripheral vision would have required a major redesign of the entire optical system from scratch.

I suspect they didn’t have the time and money to do a complete redesign and opted to slap a “make for enterprise” sticker on an obvious consumer-type design as they did in December 2019 for the Magic Leap One.

Appendix: Some P.T. Barnum Fun

At this point might be fun to watch the video of Michael Crawford singing “A sucker born ev’ry minute from “Barnum.” I saw Crawford as Barnum at London’s Victoria Palace in 1985, which made me think to look it up 😊.

Default image
Karl Guttag
Articles: 195

5 Comments

  1. Hi Karl – excellent write-up, but from what I can tell your conclusion about “An Enterprise Sticker On a Consumer Product” is a good thing, yes?

    Shouldn’t this mean that there is room for consumer content, on ML2 and in any case I heard that these will be devkits as well as enterprise solutions to XR in the field. So if anything it’s just adding to what was already there. Isn’t this what we want?

Leave a Reply

%d bloggers like this: