Magic Leap One Teardown

https://www.ifixit.com/Teardown/Magic+Leap+One+Teardown/112245

by jrnkntl • 8 years ago

310 136 8 years ago

No matter what you think of Magic Leap generally, they did some amazing engineering here and I think that's worth appreciating. The stacked waveguide model is a great hack around the varifocal plane (accommodation) problem, but to make it work in a form factor that isn't gargatuan was a feat. Yes, it's only two planes, but it's a step forward.

Not to mention that it runs on batteries and is a fully integrated development environment - you gotta hand it to them that this was an engineering feat.

CydeWeys4• 8 years ago

Practically speaking, what does only two planes mean? Does this mean they can only overlay elements at two different apparent distances in the environment? Are these distances fixed or variable?

icegreentea2• 8 years ago

They can only project elements at two depths. I believe they are fixed since the simulated depth would be a function of the diffraction grating built into each layer.

AndrewKemendo2• 8 years ago

Does this mean they can only overlay elements at two different apparent distances in the environment?

It means the focus distance of objects is idealized at two separate focal points. The objects themselves will lie around or at the idealized focus distance with cutoff points between the two distances. I'm not sure how ML determined the cut off points, but it's probably median defocus point between the two planes.

Are these distances fixed or variable?

Fixed. The only way to very focus infinitely is with a completely different structure, namely a fiber scanning display. See our patent here for an example:

http://pdfaiw.uspto.gov/.aiw?docid=20180131926&SectionNum=1&...

Ajedi324• 8 years ago

The headset can overlay elements at any distance, but when it does there may be a mismatch between the point your eyes look at and the distance they focus at. This can cause the images you see to appear blurry and out of focus. This is called vergence-accommodation conflict, and it's a common issue in VR and AR.

Magic Leap attempts to address this by having two fixed planes where it can focus the projected images. This does help, but according to reviewers it seems like two fixed planes still isn't enough to eliminate the problem entirely. Future headsets may be able to address this by using more planes, or by allowing the plane(s) to dynamically change focus (varifocal displays).

rtkwe• 8 years ago

They can still use two eye parallax to make items appear to be at different depths the limitation is that your eye wants to focus differently on far and near objects. In VR with a simple screen your eye has to focus on a constant distance while you look at objects that should be deeper based on their parallax. It's not a huge deal but can make things blurry or increase eye strain. ML1 has two focus planes near and infinity that it will switch between depending on what object you're focusing on. You can actually see the switch between the two planes depending on what object you're looking at in the scene.

In theory this will help a bit with eyestrain and make things look a bit better but with only 2 planes it's a bit limited.

kjeetgill• 8 years ago

Thanks you. I think this thread is too ready to be dismissive. It's pretty impressive even if it's not the absolute best, best, best! We're taking a little too much joy in trying to disprove the hype.

my_usernam3• 8 years ago

I needed to hear this as I'm on the hate bandwagon too. Its unfortunate that they promised so much. Its like they told everyone they were going to create world peace, but only solved one military conflict.

mattnewport3• 8 years ago

If they hadn't been putting out unrealistic hype for years people might give them a bit more slack for a seemingly competently engineered product. They presented themselves as having some revolutionary proprietary technology. In reality almost everything here is off the shelf components being assembled by some unnamed third party manufacturer. The only unique part they make themselves is the "photonic chips" which aren't actually that revolutionary, other people had more or less the same idea and decided the benefits didn't outweigh the problems.

wilsonnb32• 8 years ago

Can anyone out there tell me if the Magic Leap One is as revolutionary compared to similar products as they want me to think it is?

I'm not that familiar with the AR/VR/HUD space and I'm having a hard time cutting through the marketing.

Ajedi324• 8 years ago

My current impression is that it's a decent, incremental step up from the Hololens. It's cheaper, more comfortable to wear, has a wider FOV, and supports displaying images at two different focal planes (which helps with realism when viewing objects at different distances).

Overall it's evolutionary, not revolutionary.

yohann3054• 8 years ago

So was the iPhone when it came out. It was an evolution over the available phones at the time. BlackBerry Razor if i recall correctly was the hottest phone at the time. Actually the dominance of the iPhone was all but certain at the time. I remember a female friend sticking to Blackberry because it was easier for her to type on a physical keyboard than a touch-screen with her long nails.

Good job to Magic Leap team for getting us closer to the last stroke in MR that will trickle down a plethora of innovations

k__3• 8 years ago

I saw a review and it said that the ML-1 is still far away from "good".

Better than the Hololens, yes, but still too blurry most of the time.

DmenshunlAnlsis3• 8 years ago

Comfort to wear is still unknown frankly, and while it has marginally improved FOV and 2 focal planes it also has a much worse view of the real world because of the 6 layers of waveguides. It feels like an expensive dead end, especially when you take the corddd “Lightpack” into account.

reggieband• 8 years ago

Worth watching Tested's review on youtube https://www.youtube.com/watch?v=Vrq2akzdFq8

I was expecting typical astroturfing and there is a bit of that. But they are also tempering expectations.

Notorious_BLT• 8 years ago

I'm fairly familiar with the space, and my office picked up a Magic Leap to tinker with it. It feels revolutionary if you haven't tried its predecessors; if you have then it feels like the logical next step. If you've tried the Hololens, it's a nice step up in both the control department (the controller is miles better than the crummy'stare-at-it-and-pinch-fingers' Hololens interaction) and the field-of-view department (It isn't perfect, but its a far step beyond Hololens' "tiny viewport if you look straight ahead". In general it feels more polished, and at a lower price point as well, I can't think of any reason you'd waste your time with Hololens now.

It is, however, over-=hyped. I hear the engineering team at Magic Leap hates the marketing team for hyping it like it's literal magic. The hardware team is supposedly steaming ahead, and compared to the one they are currently working on, the released version is pretty dated, but what are you gonna do when they want to finally release a product after all this hype?

oflannabhra• 8 years ago

One big takeaway is the confirmation that ML-1 is using focus plane sequential LCOS SLM with waveguide optics. Perennial Magic Leap critic KGonTech correctly called this in 2016[0] based on Magic Leap patents.

[0] https://www.kguttag.com/2016/11/20/magic-leap-separating-mag...

lhl• 8 years ago

Just as an FYI, Karl posted an update referencing the iFixit teardown today: https://www.kguttag.com/2018/08/23/ifixits-magic-leap-one-te...

For more technical information about the Magic Leap rendering stack and their dual-focal plane approach, here's the deck from a SIGGRAPH talk: https://drive.google.com/file/d/1h36TJRkK4KteRUVcoXAnbHzTFfu...

DmenshunlAnlsis3• 8 years ago

Damn, that’s someone who really knows where their towel is. He nailed it two years ahead of time, through a morass of hype, based on patents alone and got it all right. Looking over more of his site he correctly predicted why the view of the real world would be so dark, that they’d be limited to only 2 focus planes, and pretty much everything else.

Wow.

yohann3054• 8 years ago

In this article, Karl mentions he's busy working on RAVN. Anyone knows what he is referring to?

russdill2• 8 years ago

Do the dual focal planes actually work well? What happens when something is moving on the boundary? Does it pop back and forth between them?

AstralStorm• 8 years ago

The fun question is the lifetime of lasers used in this thing. I'd bet on 10000 hours or so to 50%. Not that long at all.

CydeWeys4• 8 years ago

10,000 hours is five full work years. That's a really long time. The device itself only costs a few thousand dollars, and it's going to be technologically obsolete long before the lasers wear out.

I can't think of anything I own in my life that lasts for 10,000 hours of use, except for housing. Hell, cars don't last nearly that long, not without major maintenance. 10,000 hours of highway driving is a million kilometers.

LeifCarrotson• 8 years ago

The lithium-ion battery in this thing will be toast long before the lasers dim, but 10,000 hours is a relatively low bar. I'm still using a 5-year-old laptop, with no component replacements (other than the battery, and an SSD I installed shortly after buying it) and it still works fine.

I know that batteries wear out in use, and expect that. Not everyone does. I did not know that laser diodes wear out that fast, and I'm glad to have the information.

soneca• 8 years ago

My takeaway: Magic Leap is definitely not vaporware. It is a solid first product.

I agree with Ajedi32 comment that Overall it's evolutionary, not revolutionary, which is great per se.

The hype was unacheivable, as their marketing videos are (were?) misleading to say the least. Which I think is a strategic mistake that can hurt its adoption.

If they have a chance to return all of their investment, it depends of what they will be delivering in two or three years. If it continue its evolution agressively or this product is all they have.

TrainedMonkey3• 8 years ago

If the choice comes to "be an optimist, over-hype, get funding, and survive until engineering investments pay off" or "be realist, get less funding, deliver even more mediocre product (because less funding = less engineering talent), and die" which one would you chose?

In the end what matters is runway and culture. With enough runway and solid engineering management they have a pretty high chance to eventually release revolutionary product.

jonny_eh2• 8 years ago

Has the former option ever worked?

russdill2• 8 years ago

The device promoted by their marketing videos still does not exist.

ericd• 8 years ago

Unless you have some inside knowledge, that should probably read "The device promoted by their marketing videos still has not been released".

They may very well have a version that's like that that just can't yet be miniaturized within the constraints of a consumer device.

pome• 8 years ago

Still no real demos? :-) [0] https://www.theregister.co.uk/2018/06/06/magic_leap_revealed... [1] https://www.theregister.co.uk/2016/12/09/magic_leap_neither_...

kalleboo2• 8 years ago

The Tested review has actual footage shot through one eye of the goggle https://youtu.be/Vrq2akzdFq8?t=958

maxxxxx• 8 years ago

I am kind of glad that they actually have a real product. IT still feels like they took in way too much investment but at least they are not a complete fraud like I started to suspect.

modeless2• 8 years ago

Yeah, it may have taken a billion dollars but a billion dollars doesn't guarantee success. The product is impressive, even if Hololens stole their thunder two years ago. The engineers should be proud.

I don't see a market for it though, so it's hard for me to see how the company can survive until the technology matures enough for the mainstream. I think it will be more than 5 years, possibly 10. Timing is so important for a startup and Magic Leap is too early.

garmaine3• 8 years ago

If it was $100 it would revolutionize the world. At $1000 we should dust off our Magic Leap obituaries.

nradov• 8 years ago

A large consumer market seems unlikely at this stage, however I'll bet they could find some lucrative industrial and military niches. Those customers will be more tolerant of limitations as long as they see operational benefits.

derekdahmer• 8 years ago

I suspect if the finances get tight they can make back their money on their patent portfolio alone. If this tech is the future of AR then Magic Leap will have the rest of the manufacturers up against the wall when they want to build their own versions.

h4b4n3r0• 8 years ago

Proud of what, though? It's basically a 2-piece hololens that doesn't have any software. This stuff won't take off until field of view is fully immersive. In Magic Leap (and hololens) it's not.

PascLeRasc• 8 years ago

Ifixit is such a treasure trove these days. Their guides are so well-made, accessible to all audiences, and they don't seem to have any biases towards any one platform/technical agenda. I really hope they're able to stay independent and not get bought out by Microsoft or Facebook.

craftyguy• 8 years ago

I've used them a number of times to take apart various devices I've come across for repairing/replacing things. I would love to meet the ultra-super patient folks who create these guides.

saagarjha• 8 years ago

Plus they have these great high-quality images. There used to be a bunch of these guides with information of similar detail, but often they have with the most horrible pictures…

sspencer• 8 years ago

Wow! Anyone else reminded of the gargoyle rig from Snow Crash?

I went from "probably vaporware" to "I should look into buying one of these" just from looking through this teardown.

portlander12345• 8 years ago

I absolutely love that they went with a 90s cyberpunk look.

VectorLock2• 8 years ago

I'm not seeing it other than the cameras-sprouting-everywhere thing. What about this makes you think 90s cyberpunk look? The dark gray casing?

raziel2701• 8 years ago

Neal Stephenson is involved in Magic Leap as Chief Futurist: https://www.nealstephenson.com/magic-leap.html

kalleboo2• 8 years ago

I found the Tested review was very fair about the limitations https://youtu.be/Vrq2akzdFq8

reilly3000• 8 years ago

I became really rapt in the early days of Magic Leap PR ~14-15 ' on the basis that Neil Stephenson was joining their team in a creative capacity, and they had lots of job openings for writers and other creatives. It seemed that they may be trying to build a true virtual world with neat hardware to access it. I'm not certain that dream is dead, but the emphasis has been squarely on the hardware thus far. I'm really pleased to know that it isn't vaporware.

stefan_• 8 years ago

Feels like that Juicero thing all over again, billions of dollars, many many years spent on a freakishly expensive, outrageously complicated hardware design that is all over the place. While out there in the real world, it was Microsoft of all companies that got the MVP done and out there, and is actively learning and improving.

kbumsik• 8 years ago

Tegra X2, Myriad 2, Intel MAX 10 FPGA and a lot of other vision-related chips. etc. Wow a lot of expensive chips. As a dev reference kit, the BOM and the hardware development cost can justify the $2000. Glad to see they at least made a real hardware.

syntaxing• 8 years ago

iFixit videos have becoming better and better. I really like this teardown where they explain the technological motivation behind the placement of parts. One thing I do not get is why the control board seems so weird. It's as if different teams worked on different parts and just decided to slap things together on a board without any sort of optimization. Have anyone seen something like this before?

defterGoose• 8 years ago

As someone (without a huge amount of experience) designing some hardware in the space, me too. The two usb controllers was the first thing that made me double-take. Maybe one is only capable of doing back-forth on the external line and one is dedicated to control for the camera/light hardware? My huge takeaway was, "yep, charging 2k for a bunch of last year's good phone hardware means you can stuff A LOT of it in without worrying about your BOM too much".

exikyut2• 8 years ago

I jaw-dropped at something not yet mentioned here.

Taking a look at Step 9 (https://www.ifixit.com/Teardown/Magic+Leap+One+Teardown/1122...) where the LCOS display is revealed, I took a look at the linked PDF.

That PDF says the pixel size is 4.5um. 1000/4.5 is 222, which I think means a single row or column has approximately 222 pixels in it.

I ran the display through https://www.sven.de/dpi/ (specifying a display size of 0.4 inches), and it decided the display has 5507.27 PPI.

The PDF says that the active area is 8.64mm x 4.86mm, so that PPI rating isn't perfectly accurate - but _wow_, 1080p in less than 1cm x 0.5cm. Ha.

The LCOS module is shown 2nd from right in Step 15. https://www.ifixit.com/Teardown/Magic+Leap+One+Teardown/1122...

I'm not sure how to properly interpret "4.5um pixel size" within the context of a non-square module. The pixel dimensions (square? slightly rectangular?) aren't in the PDF. I'm also not sure how to compute how many 4.5um pixels there are within 1 sq mm, which I would very much like to do.

exikyut2• 8 years ago

Wow, I was really distracted when I wrote the above. Specifically

...which I think means a single row or column has approximately 222 pixels in it.

No no. This was supposed to say that, at 4.5um, 1 millimeter of space has 222 pixels in it per row/column! (So given any 4.5um column, a 1mm width/height span will have 222 pixels in it.)

newphoneguy• 8 years ago

Why does anyone think this is a viable first product? What plan is there to turn the discreet focus levels into a continuous variable field of focus? What plan is there to make objects appear opaque instead of clear? What do they say about increasing the fov to beyond a tiny patch? This is indeed vaporware. When conventional VR headsets mature, cameras can be mounted to them and they will achieve the exact same thing as the ML but be infinitely better.

pnathan• 8 years ago

Be interesting to hook up a jtag to the SoC and fpgas to pull out what they do. that's going to be the secret sauce for all the super speedy stuff.

Magic Leap One Teardown

Related Stories