304 North Cardinal St.
Dorchester Center, MA 02124
304 North Cardinal St.
Dorchester Center, MA 02124
I was making a list of “what we know about Magic Leap One” when I thought back to Sarah Kimberly Eusche’s analysis on her blog SAKIE and her article where she calculated the Field of view based on the Rolling Stone Reveil of Magic Leap One (MLO). In a second article, she added dimensions to U.S. Design Patent D797,735 which looks to be very close to the Magic Leap One design in the front. Sarah used a table of the range of feature of the human head to help add dimension to the dimensionless patent figure.
Lookin at Sarah’s work plus the second of the Magic Leap Recode video interview that I analyzed, I found I liked her overall approach but it was missing some details. I think, based on the videos and pictures of people wearing the MLO, she had the MLO positioned too far forward. So I decided to do an analysis building on what Sarah had done.
In Magic Leap One Video – Diffractive Waveguides Confirmed, this blog identified the glint of light reflecting off the diffraction grating of the exit pupil diffraction grating (see below right).
Then using Photoshop®, I scaled and corrected the perspective distortion to fit the outer lens with the blue glint into a scale drawing on Sakie. I then measured the size of the blue glint, and it came out to about 24mm wide by 14mm tall.
The next step was to fit the goggles onto a human head which is tricky because head sizes and shapes vary considerably. Additionally, Magic Leap has said they have multiple sizes of MLO. There are also some clues from the various pictures put out by Magic Leap and the video with Shaq. Shown on the right is the overall fitting on a “typical” head and below left is just the region around the eye in more detail with some key measurements.
The outer lenses have about 12 degrees of “pantoscopic tilt” (more on this at the end) from the vertical, and this angle seems to carry through the front of the headset. In the drawing on the left, I have indicated the approximate location of the exit grating based on the blue glint and the shape of the front of the goggles. The exit grating works out to be about 21.5mm from the eye.
For reference, the distance of the back of a pair of glasses to the eye is typically 13mm +/- ~1mm. The MLO’s outer lenses are about 34mm from the eye our about 2.6x further from the eye than a typical pair of glasses. This distance has a big impact on restricting the view out of MLO. Each eye is restricted to seeing only about 73.6 degrees total (or +/- 36.8 degrees). This result means the MLO block off almost all of the person’ mid- and far peripheral vision of the real world.
For comparison, a typical pair of glasses are about 50mm wide by 34mm tall, but they are typically only about 13mm from the eye. The roughly 73.6-degree view out of MLO translated to a vertex of 13mm gives viewport circle of only~18.5mm in diameter (smaller than a U.S. penny). In short, the MLO wearer will have a severe case of tunnel-vision.
With the rough size and location of the exit grating as being about 21.5mm, it is a simple matter to calculate the maximum possible field of view (FOV) of the display. The actual image will be inset somewhat within the exit grating. I get about 50.1 degrees horizontal and about 30.4 degrees vertically and about 57.5 degrees diagonally.
Assuming they have about 5 to 10% of the boarder keep-out area, this would indicate that it could support a 45-50-degree diagonal FOV or about in line with other peoples estimates and assumptions, but not much more. For reference, a 65-inch TV from about 6 feet (1.8 meters) away has a diagonal FOV of about 48-degrees.
The physical size of the exit grating on MLO is smaller than Microsoft’s Hololens, but then the grating on Hololens is much further from the eye which allows people to wear most normal eyeglasses. Rony Abovitz, CEO of Magic Leap and a wearer of glasses, has said that Magic Leap will have a solution for people with glasses in the future. It is not clear how far in the future or whether this includes the MLO generation. Magic Leap’s US design patent D795,952 patents show using a (prescription) insert which is similar to what Vuzix uses with their Blade product, but it is not obvious that MLO supports a custom insert or not. You want the insert lenses vertex distance to be about 13mm and looking at the drawing above, the MLO googles would interfere, but then this drawing is not of the final design. None of the pictures release so far, nor the Shaq video shows the insides of the goggles.
Supporting enough eye relief for regular glasses requires about 30mm plus about 150mm or more of left to right opening. You end up with something the size of Hololens.
I keep coming back to MLO looking to me being much more like VR with a small FOV. According to my prior estimates based on the Shaq video, Magic Leap One blocks about 85% of the light. The location and size of the outer lenses tunnel the wearer’s vision to about a 74-degree circle. But then it apparently has between a 45- and 50-degree diagonal display FOV.
I’m also expecting the MLO to have the usual issues with image quality inherent in diffraction waveguides and using microdisplays. I think people were expecting the image to look anywhere near as good as cheap LCD TV are going to be disappointed.
The pantoscopic tilt and location of the outer lenses suggest that they are expecting the user to be looking down more than straight out. The looking down view seems to be confirmed in the Shaq Recode video starting at about 6:25 when Shaq says, “I put on the pair of these glasses, and I watched a full-court game right here. Not flat, LeBron was right here.” He is looking down at a tabletop and indicating the size of LeBron Jame his hand (captured at 6:34). Shaq seems to be indicating that LeBron James looked like he was (only) about 4-inches tall from about arm’s length away.
Everything about MLO screams to me “VR Headset Game Player.” But one that will cost 5 to 10 times as much, have lower image quality, and with less than half the FOV of a typical VR headset.
Blocking so much light and giving a person tunnel-vision, it does not seem to be very good for mixing in the real world as expected with AR/MR. Not the vision everyone else is talking about for AR/MR. I can’t see someone buying seats to an NBA game and wanting to watch the game through them as Rony Abovitz talked about in the Recode video; the user will be taking them off to watch the Jumbotron first.
As I see it, it fails to come close to meeting expectations for AR or VR. It is in its own category, but not a good one.
small mistake Karl — “I get about 50.1 degrees horizontal and about 30.4 degrees horizontal”
Thanks, it has been fixed.
I did this calculation using the Sakie methodology on the day you wrote the ‘Diffractive Waveguides Confirmed” piece (pixel counting the diffractive element from Shaq, which is scaled from the ergonomic drawings put together by Sakie). I wrote up a long comment, but, annoyingly, it got lost in the ether of WordPress/non-updated browser. I didn’t save it and it was long and couldn’t be bothered to re-write it. It less precise than what you’ve done here, but satisfied for me the idea that the diagonal FOW would be larger than the Hololens. Here’s a summary of what I calculated:
Diffractive element size: 27.4 x 18 mm (H x V), or diagonal 32.8 mm
Estimated distance from diffractive element to eye: between 20-30 mm
Therefore, diagonal FOV: between 57-79 degrees.
This assumes they will use the entire size of the diffractive element, which they won’t. In any case, this is quite a lot larger than the Hololens, which has a diagonal FOV of around 32 degrees.
My conclusion was also that, owing to the darkness of the lenses and its FOV, the ML1 is more like VR than it is truly AR.
what about vhs at half arm’s length…..that puts it at way less than 50 horizontal fov
That was a person using their memory and judgment which is not very precise. The calculations, however, only give the upper bound on the diagonal FOV.
The weight of evidence suggests the diagonal FOV will be greater than 35 and less than 55-degrees. I think it is likely going to be between 40 and 50 degrees. Note these numbers depends on the person and how the wear the glasses as the eye to lens distance will vary from person to person.
he said he had plenty of time to figure out how to discribe it….and I took that to mean that he had time while he was using it….thinking about the experience later would be pretty pointless….. but whatever
The Diffractive element you used is quite a bit bigger than what I found and the distance from the eye has a wider range.
All the evidence, including the comments by the Rolling Stone reporter, is that the FOV is going to be bigger than Hololens, but certainly nowhere near 79 degrees or even 57 degrees. Everything seems to suggest it will be between 40 and 50 degrees.
I think both the darkness and the tunnel vision effect will rule out what most people think of as “AR.” I think the image quality other than center resolution is going to be worst than a good VR headset.
You are assuming no eyebox then… 😉
“The pantoscopic tilt and location of the outer lenses suggest that they are expecting the user to be looking down more than straight out.”
Rony Mentioned many times in the recode Video that you will have many screens
If that’s so I think it will also be used for straight out as well.
But everything else in the article looks good
Rony says a lot of things that are only in his imagination. You could have multiple virtual screens, they would just below the centerline if you look straight out.
I’m not saying you cannot look straight out but there appears to be a bias in the optics for looking slightly down (about 18 degrees from vertical). You can see this in the two lens barrels at the front of the goggles. Note also, the exit grating is (blue glint) is below center. All this is pointing to the vertical FOV biased below the centerline if you looked straight out.
With a tilt like that, what you makes you think it’s not a birdbath / freeform element? I’m not entirely convinced that there is enough information to conclude that the output surface is diffractive – and if there was a diffractive element it could equally be an intermediate relay for pupil expansion. Dielectric coatings, particularly for narrow bandpass applications (which might be the case for optimizing transmission from an OLED or other weak source) which reflect well at other wavelengths and angles. Having said that, coming back to eyebox, for it to be reasonable – 15mm say, would bring the FOV closer to 30° – so not far off Hololens, and in the realm of where diffractive optics is just about ok – but 18° tilt would lose more than half the vertical field.
I’m certain that they are using waveguides. Not only do they have a patent trail including recent applications, there is evidence in the videos (see my article: https://www.kguttag.com/2018/02/14/magic-leap-one-video-diffractive-waveguides-confirmed/). OLEDs have notoriously wide spectrums which makes them lousy candidates for dielectric coatings.
In the this the FOV and Tunnel vision article, I should all the geometries and measurements. The front part of the headset in the video appears to agree with the design patent. The configuration appears to be biased in favor of looking down like through the reading part of Varilux type glasses. Thus you have Shaq gesturing like people were running around on a table top and not out in front of him.
Based on all the images seen so far, it is highly unlikely to be a waveguide, but a holographic beamsplitter using some power on reflection from the outer “lens” – built some prototypes of similar schemes in the past. A waveguide viewed from a distance will not give a near-perfect rectangular field, you always get noticeable trapezoidal roll-off in luminance due to the different optical path length losses for different angles. 😉
And still nobody yet talks about eyebox… and trade-off with FOV… where do they think non-mechanical adjustable IPD comes from?
[…] I am calling “The Magic Leap One (ML1) experience.” After writing my earlier article Magic Leap One – FOV and Tunnel Vision, I realized that there was enough information to build a 3-D model that would simulate the view of […]
Didn’t want to be so blunt before in case it might give them some ideas 😉
[…] It blocks 85% of the real world light (as measured and as predicted in my analysis of the Shaq video back in February 2018) […]
[…] of which you’ll examine somewhere else. (One position to get started: Karl Guttag‘s Magic Leap One – FOV and Tunnel Vision.) I’ve discovered sufficient, thus far, to make two […]
[…] blocking out of the real world is something I faulted the ML1 with in “Magic Leap One – FOV and Tunnel Vision,” and also in “Magic Leap Review Part 1 – The Terrible View Through Diffraction Gratings”. […]