Magic Leap Fiber Scanning Display (FSD) – “The Big Con” at the “Core”

Intro to “The Big FSD Con”

Of all the things Magic Leap has done, I consider Fiber Scanning Displays (FSD) to the “The Big Con”. The reason why is that Magic Leap made such a big deal about it being one of their “core technologies” early on, so many people today are that it will work, and yet is it so easy, with 4th-grade math, to prove it is impossible. Fiber Scanning Display” is still showing up in 33 Magic Leap patent applications published in 2017

That anybody would promote that FSD could support the resolution required was incredibly naive/ignorant and/or incredibly corrupt. It also would seem to show, hard as it is to believe with all big money involved, that nobody was looking at Magic Leap in serious technical detail. Maybe they were just “big picture” people who didn’t sully themselves with 4th-grade math.

A “Good” Con Should At Least Look Believable

It has said that “a good con must be believable.” In the case of FSD, the University Washington HID Lab in the mid early to mid-2000’s had shown they could make a low resolution highly distorted FSD projector with a fiber projector and they had papers and patents being published ever since.

So the con goes something like, “it is just a matter of improvements, “Moore’s Law”, and scaling up right?” . . . Wrong, as you scale up the resolution, the fiber has to move impossibly fast. As I showed in a previous article, even at lower resolution/speeds is the image highly distorted and non-very non-uniform in intensity.

One Web Calculation and 4th Grade Math

There is a one piece of semi-complex math that is solved by doing an online search for a “roll length calculator,” of which there are many, such as on Handymath. Working from Ivan Yeoh’s 2015 Ph.D. dissertation (the best and most recently available information on FSD), he had 180 spirals which would nominally be 360 pixels across the circle at the equator of the image with only a 50Hz frame rate. This still ignores the Nyquist re-sampling from a grid to the spiral.

Starting with 360 pixels across in 1 mm or 1mm/36 = ~0.00278mm per loop. Now scale this up to 1080p (1920 by 1080 pixels), requires 1080 loops to give an inscribed 1920 x 1080 pixels image.

Using the coiled rope web calculation (inputs and results show below), plugging in  0.00278 mm which gives a ~6mm (=0.00278x1080x2) diameter coil to give 1080 turns. The path length (of the tip) works out to about 10.17 meters (see below) whereas for the Yeoh paper with only 180 loops and a nominal 360 pixel across the circle, the length was only ~0.26 meters and still they dropped down to a flickering 50Hz just to do that.

To prevent flicker with a zero persistence display, you really should have a refresh rate of about 120Hz (and not 50Hz or even 60Hz). And since there is one out and one back for each complete “scan” that hits the outer rim only once, this means you it must go the path length of the spiral 240 times a second to “hit” the center and outer most pixel 120 times a second.

Thus the fiber needs go 10.17 meters X 240/sec = 2,400 meters/sec. For reference, the speed of sound is a mere ~343 meters per second (varies with temp and air density) so the fiber has to average about 7 times the speed of sound and because it is accelerating and decelerating back to zero at each end, it has to have a peak velocity of greater than 14 times the speed of sound!!!!!

If this is not insane enough, they need an optical (clear) fiber that will move at the resonate frequency required, while driving it all contained in a near perfect vacuum so whatever it is won’t spontaneously combust due to air friction along with the sonic boom destroying the users hearing. Not to mention turning the user’s brain to Jell-O from the vibration. And if you find this miraculous material, it must also follow an exact uniform path particularly at the center as is passes through zero and reverses or else the pixels will not be in the right place and/or wiggle.

You can play with my assumptions above and plug in different numbers, but you can’t come up with any rational way to make it even come close. There is nothing in any of Magic Leap’s or the University of Washington’s papers or patents that show how it could be scaled up as much as 720p not less 1080p effective resolution.

Now For The Sad Part of this Story – Were Is the “Smart Money?”

What gets me is that with all the “very smart” people and those with all that money that reviewed the Magic Leap investment, this “little problem” should have been quickly discovered. Magic Leap’s whole credibility should have been gone back then.

Just about every person at Magic Leap with math, engineering, and/or science degree must know by now that FSD is a crock. And still, the myth is still being perpetuated that FSD will work and Magic Leap is still filing patents like FSD will exist (to support the myth or fool investors?). Of course, not of Magic Leap’s patents or papers deals with the speed of the “magic” fiber.

And this is not the only things that are downright silly in their patents. The whole depth planes concept also falls apart as I discussed last time; to support the six focus planes the dream about in their patents, they would need a stack of 6 x 3 = 18 waveguides with even more insane optical routing.


Karl Guttag
Karl Guttag
Articles: 243


  1. If you think magic leap is a con, you’re assuming Google and whoever the big investors are are idiots and buy into 4thgrade disprovable math.

    Sounds to me like you don’t know anything. You see some patents but don’t know which are being used in the device.
    You assumed you were going to see a magic elephant in your hand with out glasses too right?

    Since you are so smart, don’t you think you should write an email to Google and warn them of this “con” u have discovered? Nothing gets past you, does it sherlock? Magic leap will soon have the chance to put the speculation to rest. Then you can say ha! I told you so!

    • A lot of what Magic Leap is doing the average person would have a hard time understanding, but I have laid out the simple math of the spiral scanning display. Why don’t you argue with my math?

      I can’t explain “why”, but clearly something is wrong/fishy. There appears to be something of a herd mentality where nobody asked/asks even simple technical questions. You might want to watch the movie “The Big Short.”

      • The working device as you see in the press photos was demonstrated and I personally know developers who work with the hardware daily.just wait for the developer units to ship. This isn’t vapor ware, no con. We will all see the comprises that were made in order to create the end product.

        Havevyou seen the video hyper reality? Its not there butvthats the kind of world they are envisioniing

        Someone close to me works there, trust me they have something pretty crazy 😉

      • I haven’t written that is it “vaporware”. They are doing some real things that are going to be a lot closer the Hololens than anything that “magical.” The Fiber Scanning Display is a con; something they were using to sell what they were doing that was not possible and they either did or should have known it.

        I get the world they are envisioning, but that is not what they are actually doing.

      • A vacuum wouldn’t exactly be exotic given the tiny volume involved. But even without a vacuum, the speed of sound in Helium (not exotic, harddrive use it) is 3X faster. At 180°f (easy to insulate and thermally isolate given the tiny volume) it is 4000km/h. Now we’re 40% of the way to 2400m/s and beyond 720p.

        And why would it have to be exactly a 16:9 aspect? A much better pixel measure could be achieved with a circular display output. Between this and putting some portion of the display of each eye outside of the stereo overlap of the other you could probably hit a pixel total nearing 1080p (not counting any of the overlapping region as being doubled).

      • It is not just the burning up of the fiber, it is the mechanical stresses of accelerating an optical fiber to this speed. As I wrote, you could put it in a vacuum to keep it from burning up, but you would still have the stress on the fiber and the vibrations to deal with, not to mention absolutely accurate/repeatable control of a wiggling fiber to make a display. The velocity must past through zero at the center a region where the image becomes unstable (see )

        While Fiber Scanning Displays mentioned in Magic Leap patents in 124 out of 305 patent application (, they never get into any details like you have suggested in how to make the fiber actually work. There is nothing like say putting a magnetic coating on it so you could help correct the path, no mention of Helium as you suggested, nothing, nada. They have not even addressed how to get a much lower resolution FSD to work better than the Yeoh paper of 2015.

        Their only suggestion is to put many of the fibers in parallel which is yet another deception for the casual observer. The fundamental problem with multiple fibers is that you can simultaneously get the light rays collimated (necessary for a waveguide and other near eye optics) and overlapped to create a continuous image. Once again, they never address this issue in the patents. Then we can come back to the issue that they can’t control even low resolution FSD well at the center.

        If Magic Leap was serious about FSD, you would expect they would be trying to patent HOW to make it work in practice. What you see is just the assumption that it works and it can be used for all these optical systems. It looks to me like they are just trying to deceive someone.

      • Your math says that you have to make 1080 coils to produce a 1080p image. In one coil, the fiber will intersect a given vertical line in two places, not one. You need HALF as many coils as you calculated.

      • Maybe I was not clear. I wrote, “Now scale this up to 1080p (1920 by 1080 pixels), requires 1080 loops to give an inscribed 1920 x 1080 pixels image.” The key word here being “inscribe.” The image is 1920 wide they there will generally be some of the spiral cut off in fitting a rectangle. At a minimum, you need 1920/2 or 960 and to get to full rectangle you need about 1080 loops.

        Even if it were half as you wish, the fiber still has to move impossibly fast.

    • I have worked with waveguide optics for 30 years, and I came to the same conclusions as Karl did. That was before I even knew who Karl was. After I passed my opinions onto a friend, he sent me a link to Karl’s site.
      The idea that a large number of intelligent people could make some very stupid investment decisions is hardly a new one. Greed is a very strong motivator. With few new groundbreaking technologies being developed, there is a strong desire to not miss out on the “next big thing”. A few years ago, 3D printing was supposed to change the world. The reality turned out quite different, and the predicted 3D printers on everyone’s desktop never materialized. My guess is that AR will follow the same trajectory.

      • I think it a “matter of physics” that diffraction waveguides will have poor image quality. Lumus’s multi-mirror looks much better than any diffractive waveguide, but I have not had a chance to seriously evaluate it. Waveguides requirement for collimated light puts some limitations on them when it comes to VAC.

        I’m not sure it is a much “greed” as it is people desperately looking for a “thin glasses look.” The end up rationalizing that it is OK for the image quality to look worse so long as is thinner. Most of the solutions that look better are big and bulky by comparison. There is a great deal of Fear Of Missing Out (FOMO) driving the AR market. Researchers like to “play” with interesting technologies. I think we can agree that the expectations for AR are out of line with the reality of what can be done.

        3-D printers are nice and useful but there they can be overblown. They are particularly helpful to people that like making things. But the idea that you will just print many end products at home is going too far.

  2. Why would Google be fooled by this if it’s nonsesnse as you claim? Are they that gullible? The device was demonstrated recently and I can confirm the press photos are real not renderings. Anyways u deleted my first comment I doubt this will pass the gate keeper of posts ?

    • I have laid out the math, why don’t you argue with the math, it is pretty simple. I can’t explain why other than maybe a herd mentality and FoMO (Fear of Missing Out) or the looked past the basics and focused on the “big picture.” What I don’t get is people that treat Magic Leap like it is some kind of religion that must be venerated and should not be questioned. I’m not making mindless attacks, I am showing the evidence.

      I have had over 4,000 “users” go to my site since yesterday and not one has argued with my math. Start there and ask why is Magic Leap still pushing FSD?

      I’m not saying that Magic Leap is doing nothing (pardon the double negative). What I am saying is that all the evidence suggests in terms of display (I have not analyzed the tracking/slam) that is it going to be pretty much like what Hololens, Vuzix, DigiLens, and others are doing.

      The only comments that get filter out are those that are spam, abusive, or personal attacks. I welcome dissenting point of view, but not personal attacks.

    • the FSD was confirmed by the information article(the only one that is not under NDA) that it was set aside for RnD. karll never mentioned magic leap as a con, its there FSD tech because they still trying to fortray to the public that they are using it.

      “Why would Google be fooled ” ? you dont know how this works. the one who handles the investments are not technical people. recently facebook and apple was even got scammed 100million of investments through email communication without visiting the actual company

  3. Ok I see my previous comment awaiting moderation. Interesting article but I think the speculation I’ve read online is out control. Let’s see what it is when released. It’s definitely a game changer I know people who work there. It’s cutting edge.

    Nothing better than beating low expectations.

    • Magic Leap has so hyped what they are doing and some of their followers expectations are so high, that it is going to be impossible to beat them. If you won’t take the time to understand what I have written and just assume “smart people know” then to you are just believing in “magic.” I suggest you watch the movie “The Big Short”, while fictionalized, it does show how herd mentality and the conspiracy of silence works.

  4. u missed the e in …..4th grade math ….in the first paragraph

    if they expressly promised to deliver an fsd which might very well be the case and they had a demo unit with rgb fibers…….well at least the so called con worked…..they got the money and made a bit better hololens …..but the company might not be a success in the long term

    • Thanks for the correction. I’m working on getting a proofreader.

      I think FSD does goes to their whole credibility. They are willing to say FSD was a “core” core technology which is not true either cognitively or out of ignorance. They are still putting FSD into their patents. It is simply amazing that they were able to raise so much money went their “core” technology is worthless. They never answer anything other than softball questions, and with only one known exception (The Information) that was not under NDA. We don’t know that it is better than Hololens in terms of image quality and won’t know until they allow non-NDA demos with people that understand displays in uncontrolled environments.

      • yeah
        I meant it might be better as far as ….being slightly bigger fov ….maybe similar picture quality ….apparently somewhat good resolution ….and some depth to it
        and u get a much smaller lighter unit on your head in comparison and a 6dof controller
        if it works smoothly and as intended as far as tracking and graphics power which is pretty tricky ….it will be pretty nice to use…..but we’ll see how it will compare to the next hololens
        I never expected something magical from them….I always called them Good Enough Leap…..but as far as their presentations and promises….yeah they r falling short …..maybe as far as a con

  5. Me again, the non scientific person. I know from ‘experiments’ that the human eye/brain is somewhat flexible in terms of resolution when pixels are non square, more granular to the likes of celluloid film. We have worked with microLED units that had more hex pixels. The eye perceived them slightly fuzzy but still more homogenous and realistic then tradition lower res imagers.
    With that if the FSD does not intent to cover the full field of view, it could get away with lower resolutions.
    I don’t know if its possible to steer the fiber with much more center focused resolution and taper off to the edges, this would sort of intuitive as the radius increases anyway. Wouldn’t this simulate a sort of foveation if paired with eye tracking to keep the projected image at the point where the highest resolution of human vision lays ?
    Just thinking outside of the box (spiral).

  6. Karl, your evaluations have been quite helpful in my trying to sort out some reality with Magic Leap given the CEO’s tendency to wander into fanciful wordplay. However, I’d suggest the patents do not necessarily reflect what is going on here.

    I’ve been part of a high profile effort where we had to crank patents as part of the process and you do your best early on to layout what you think you will be creating, and it is definitely speculating on development direction. It does not surprise me at all that what they have built is not what they started out thinking. In fact, I would suspect that a non-trivial portion of what they have done is not in the patents and may be just considered trade secrets. Plus there are probably a bunch of more realistic patents of the actual hardware coming — I’ve been through that as well.

    The fiber scanning display essentially violating physics is definitely poor form and I would assume the engineers behind that know so, but I would also think that the patents were not created with intent to deceive. It seems like they were pressured to get as much in place as early as possible to get a good negotiation position for financing. It’s not uncommon and with this being one of the highest profile hardware investments in years and its directly competing with Microsoft, it’s quite believable.

    I used to think that Magic Leap was going to get Segway’d — they had built up so much expectation that they could not possibly fulfill that no matter what they delivered it would be considered a flop. However, because we’ve known the product area for some time now and had multiple examples of things heading in that direction, most notably the Hololens, which has shown just how hard this area is to produce a market appropriate device/system, I think the press is going to give them a bit of a pass here as long as it is better by a reasonable amount than the other AR devices/systems. Segway had the problem that they hid the actual product in pretty all its details and built the hype — no one had any idea it would be a silly balancing thingy. The disconnect of the hype to reality was so profound that people were dumbfounded.

    Magic Leap is going to benefit from the context of the others already making hard fought progress in the same area. But they could still get hammered if they don’t continue to tone it down, as they are already doing, as long as the CEO doesn’t go into “we are all going to become energy beings” word salad mode too much.

    • Thanks,

      I’m looking at a combination of their patents, statements and what is possible based on other work in the field. You can see in the public evidence where Magic Leap started and where they are heading and so far it seems to check. You can see them tacking from something that was impossible to something much more like Hololens.

      They have filed over 300 patent in the last few years. While it is possible, it would seem to me unlikely that they would file on their bad stuff and risk not filing on their good stuff.

      I can’t come up with a good explanation why they would keep filing patents talking about Fiber Scanning Display. The smart people at Magic Leap should know better.

      The problem of coming up with a mass market headset is very tough. I can see something like what they showed going to the mass market with big cables going to a half-in-half-out the pocket computer. I think Segway and/or Hololens is a reasonable analogy. They are never going to live up to the expectations they raised. This is a test

      I don’t see how they will benefit from the work of others if they don’t have a mass-market headset.

      It was a long comment and I have a lot going on today, so hopefully, I address your issues

      • What makes you think that the spiral pattern is the one they would use? If you’re going by ML’s patent filings, have you considered that they might be using their funds to license other patent holders? I’m not at liberty to say which patents they have licensed, but the total length of the path does not come anywhere close to what you’ve calculated on some of the patterns.

      • There are two most likely paths to consider. One would be the spiral and the other would be a Lissajous pattern similar to mirror scanning laser beams. The spiral only requires one start and stop per trip whereas the Lissajous pattern requires a stop and stop per “line.” Either method has it severe drawbacks. The path length for a Lissajous pattern is going to be worst to give the same coverage. A fundamental problem is that they must make a wiggling fiber pass through zero velocity where it is unstable. It is also a nightmare to control the intensity.

        I have laid out the math I used. Please feel free to show where it is in error and I will issue a correction. Fundamentally you must follow the path and the path gets longer by a square law as the resolution increases.

  7. No, the lissajous pattern never stops. It’s moving on a continuous path. There’s no point where it’s velocity is zero.

    I don’t have a problem with the math you laid out, I just think you’re looking at the wrong path and making assumptions about acceleration that aren’t necessary.

    • I’m not sure what point you are trying to make but you are dead wrong on a point of fact about the Lissajous pattern. The Lissajous pattern has to reverse direction for each horizontal sweep. It has to pass through zero velocity to reverse direction. Please explain if you think otherwise. The whole point of the spiral scan is that there is only one reversal.

      • Technically speaking you are both correct, but Karl is more correct. The Lissajous will reach zero (angular and spatial) velocity along a given scan axis at the end of each scan. However, except for some unorthodox modes, the orthogonal axis will still have some scan velocity. This means that for each scan “line” there is a stop and return, just like a person on a swing set. Technically the overall velocity is not zero, but the line scan velocity is.
        As for the impact on scan velocity of the “lost time” due to the sinusoidal velocity, it ends up being a little more complex. You have to throw away a portion at the end of scan (for practical, not physics reasons), and can calculate the difference in accessible pixels as a function of the Lissajous pattern. It is not the average and it is not the minimum.
        There are a number of other impacts of the differential velocity. For example, the pixel time means the perceived brightness at the edges is larger due to more photons, so you have to correct for that. The pixels in the center are more smeared due to the angular velocity maximum in the middle, so you have to correct for that. And there are a lot more issues such as pixel smear due to fiber angle.
        Correlating to what Karl says, the biggest problem is that you end up with geometrically increasing constraints on the fiber as a mechanical device as you go up in resolution and then have to multiply this by the optical train considerations and the stability.
        Keeping the fiber in a vacuum package is not that big of a deal mechanically speaking and this lets you get high frequency and high Q at the resonant frequency. But, it impacts the optical design, FOV, achromaticity and much more.
        Seibel’s scanner works. But, it becomes one of those engineering problems that gets geometrically more challenging as you try to reach real world performance specs.

  8. What a nice blog post, I came here after reading the teardown reviews of the ML1 and what can I say?
    KarlG > Due diligence teams @ Google, Qualcomm, Alibaba, AT&T, JPMorgan, Fidelity, etc. etc..

    Now that the veil has been lifted and the prototypes leading up to the ML1 have also been displayed publicly you have to wonder how multiple Silicon Valley titans invested hundreds of millions of dollars each in a company that had zero new technology demo-able, and could not even implement existing technology well (e.g. magnetic tracking).

    All of the reviews I read said the 3D surround sound system was the most immersive feature of the ML1..the sound shouldn’t be the highlight of a AR headset with a development cost in excess of a billion dollars.

  9. […] さらに詳しく説明すると、Magic Leap Oneは6つの導波路を使っています。それぞれお互いの上に重なり、RGB3層がそれぞれ2つの異なる焦点面に割り当てられています。Magic Leap Oneは2焦点面(Bi-Focal)のディスプレイと考えられます。つまり、ディスプレイの焦点をアイトラッキングに基づいて2つの異なる位置の間で動かせる物です。ただしOculusのHalf-DomeやNVIDIAの本物のライトフィールドディスプレイのように、可変的に焦点を動かす(vari-focal)ことはできません。私はまだ正確な数値を把握していませんが、近距離は0.75メートル、遠距離は5メートルに焦点が合っているようです。もしMagic Leapがこの技術を使い続けるのであれば(そして実際、同社が他の技術、特に散々宣伝されたファイバーディスプレイを使えると示すものを見たことがありません)、他の焦点を設定するためにはさらに多くの導波路と、実行不可能なレベルの高いフレームレートが必要です(逆算すると、焦点面1つに対して最低60hz必要になります)。納得のいく重さ、画質を実現し、さらにコストの制約がある中で、この仕組みが実現するとは思えません。 […]

  10. […] さらに詳しく説明すると、Magic Leap Oneは6つの導波路を使っています。それぞれお互いの上に重なり、RGB3層がそれぞれ2つの異なる焦点面に割り当てられています。Magic Leap Oneは2焦点面(Bi-Focal)のディスプレイと考えられます。つまり、ディスプレイの焦点をアイトラッキングに基づいて2つの異なる位置の間で動かせる物です。ただしOculusのHalf-DomeやNVIDIAの本物のライトフィールドディスプレイのように、可変的に焦点を動かす(vari-focal)ことはできません。私はまだ正確な数値を把握していませんが、近距離は0.75メートル、遠距離は5メートルに焦点が合っているようです。もしMagic Leapがこの技術を使い続けるのであれば(そして実際、同社が他の技術、特に散々宣伝されたファイバーディスプレイを使えると示すものを見たことがありません)、他の焦点を設定するためにはさらに多くの導波路と、実行不可能なレベルの高いフレームレートが必要です(逆算すると、焦点面1つに対して最低60hz必要になります)。納得のいく重さ、画質を実現し、さらにコストの制約がある中で、この仕組みが実現するとは思えません。 […]

Leave a Reply

%d bloggers like this: