Magic Leap One Teardown
ifixit.com
ifixit.com
[0] - https://www.kguttag.com/2016/11/20/magic-leap-separating-mag...
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...
Wow.
The only thing Ravn needs to care about is awareness within the defence industry.
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.
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&...
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).
So basically it can only correctly overlay elements at two distances, then? Any other distances are going to be out of focus to varying degrees, which will start looking increasingly incorrect the farther they are from the correct focal distance?
Does the hyperfocal distance come into play here? Is there some range around each focal length that is acceptable, or do the eyes work differently from camera lenses here?
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.
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.
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.
What did whatsapp over-hype?
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.
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.
At $100 people would be willing to use this for various new applications. Education comes to mind as a particularly good application. Maybe sports commentary? Maybe pop-up AR booths in malls. That would be enough to jump-start an ecosystem, while the hard technical work is done.
we dont scoff at $1,000 phones hardly at all these days.
this is the first step, we should be thinking about where tech will be in 20 years at all times.
Hololens is focusing on industry - live monitoring, architecture planning, 3d modelling and visualization. I don't know how successful they are.
Engineers who participate in this sort of activity need to take a long look in the mirror and ask themselves if the paycheck is worth it.
I'm not that familiar with the AR/VR/HUD space and I'm having a hard time cutting through the marketing.
Overall it's evolutionary, not revolutionary.
Good job to Magic Leap team for getting us closer to the last stroke in MR that will trickle down a plethora of innovations
The original iPod was not all that much different from other MP3 players at the time, it simply had great design (both from a UX and aesthetic perspective). Similarly, the iPod Touch or early iPhones were technologically not much more impressive than a Blackberry. Touchscreens were nothing new at the time. But I agree with you that the way you interacted with the device was a new paradigm. Sometimes little evolutions make a big difference, especially when they make something easier to use, which is probably the most important factor for mass adoption.
Equivalent statement:
"When the iPod came out, it felt like someone had handed you a little bit of Star Trek."
In both cases, once you started using it you had the same reaction "this is REALLY cool but has major issues, I can't' wait for the next version".
I remember my boss gleefully showing me his new iPod, and then him cursing from his office that afternoon (and many after that) as the thing crashed, and nuked his music library again, and again.
It had similar features, but not useful. iPhone was very useful.
The iPhone was a massive upgrade to anyone using a flip phone still, but an evolutionary upgrade to people that had been using Palm Pilots, Blackberry's, and other devices since the late 90's.
We can debate about whether that's "evolutionary" vs. "revolutionary," but if it was evolutionary, it's the kind of evolutionary success that leads to epochal shifts.
To bring this around to the actual OP, I'm dubious that Magic Leap is going to pull that same thing off, but that's because I'm dubious that AR in the form of "things you stick on your head" is ever going to be a mass-market phenomenon.
AR is the same thing, it takes the dopamine hit and puts it on your face. Subsequently you get a lot more dopamine hits.
for me, being in tech my whole life and growing up solidly a nerd and in the cyberpunk culture of the 80s, when the iphone hit, it was more of a sigh of relief of "ok finally we are making the steps into the cyberpunk future weve been fantasizing and thus building for decades"
it wasnt until the snowden stuff hit and all the current tech problems we have that the sentiment is "oh crap, the dystopian cyberpunk future is a heck of a lot more of a slippry slope than we were ready to deal with as a society"
I think you're correct that the smartphone was a phenomenon because it put the internet in your pocket, but we already knew the use case for the internet. People use their smartphones to do everything computers do, and sometimes more: browse the web, read books, find restaurants, make reservations, get directions, use social media platforms, pay for purchases with NFC, listen to music and podcasts, manage their calendars, control their home entertainment systems, check the weather, (...deep breath)
Anyway: "Great dopamine hits, man" falls, I think, into the "necessary but not sufficient" bucket. AR needs to do that, but it needs to do more, too. I agree with "the smartphone was a phenomenon because it put the internet in your pocket," but I think we're still looking for the correct X and Y values of "AR will be a phenomenon because it puts [X] in your [Y]."
The iPhone was completely different. The design looked different mechanically and had a totally different ui paradigm with the touch screen.
It was revolutionary as evidenced by media coverage at the time, growth in mobile web, and massive market share still today.
(Well, at least in my recollection.. :)
Only if you ignored all the other smartdevices on the market at the time. Palm and WindowsMobile phones had touchscreens before the iPhone came out. The only big change with the iPhone from an interaction perspective was a switch to capacitative screens, which made it easier to use with fingers but far worse to use in any sort of weather or industrial situations.
I suspect that by the time Apple starts signaling that they've got a headset and announce it at WWDCXX its going to feel like it fell straight out of the year 3000.
I think ML will eventually be part of FB, MS, Google, or Apple, or even Amazon within 5 years. It'll be another 5 before AR is mainstreamed.
Google seems to have had some decent experience lately with hardware, Amazon still feels like it flounders a little in this area with failed fireos phones, and alexa/echos' and maybe firesticks are the only major things they've really come out with that stuck, I guess kindles too a bit, but not great.
Samsung I guess could also be another contender to buy them, they put out some great devices as well... I don't know who it'll be - but I think ML will be acqui-hired.
With the arguable exception of MacOS X, I don't think Apple has ever just outright bought major products.
https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitio...
I mean, I feel ML team is definitely worth more than Whatsapp, but they don't have a product that's ready for mainstream yet... having the full backing of one of the bigger companies could bring mass AR to mainstream audiences faster.
The Magic Leap One costs $2.2k, which is cheaper than the competition but still pretty expensive for a product that hasn't yet established its usefulness to the extent PCs or smartphones have.
If Magic Leap (or a competitor) can cut the price to a third, bump the specs a bit further, and get developers on board creating compelling experiences that just might do it, but it will take a couple more evolutionary iterations before we get to that point.
When the iPhone came out it was a true revolution and the plans for Android phones at the time were entirely redone.
Better than the Hololens, yes, but still too blurry most of the time.
That, and the documentation is a lot better. HoloLens wasn't all that bad, but Magic Leap has really put a ton of effort into their tooling and docs.
How did you get in HoloLens development?
The display may be 1080p but I would love to see a measurement of the modulation transfer function of the display system... i.e. display vertically alternating bars, and horizontally alternating bars... then take a picture with a good camera
Edit: I think you're talking about Hololens, which does put all of the batteries on your head. Yes, I think you've correctly identified another company taking a different approach. And you've correctly identified the usability trafeoff associated with the belt-mounted battery. I maintain that the tradeoffs are such that we will see many companies using non-head-mounted batteries in order to support longer use sessions.
Hololens is notably not good enough, graphically. Neither is Magic Leap. I don't contest that you can do a not-good-enough head mounted computer
I was expecting typical astroturfing and there is a bit of that. But they are also tempering expectations.
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?
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.
> ...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.)
I went from "probably vaporware" to "I should look into buying one of these" just from looking through this teardown.
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.
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.