304 North Cardinal St.
Dorchester Center, MA 02124
304 North Cardinal St.
Dorchester Center, MA 02124
2019-06-28: Corrected Nreal’s light throughput to be 30%
I had another busy four days a CES this year, followed by catching a lingering cold and having a business trip. So, I am just starting to write about what I saw at CES. There are about 20 companies that I visited and want to write about in future articles (some of my meetings were confidential).
Inbetween seein demos of various AR and display hardware, I was feeling around and confirming suspicions for what technology Microsoft is using in the next Hololens, which is expected to be announced soon. I have some answers to reveal in an upcoming article.
Below are summaries of some of the more interesting thing I saw at CES 2019. While I mostly concentrated on AR headsets, I did looks as some other displays technologies. I hope to circle back to provide more detail about these products in future articles.
Correction: I originally reported that Nreal’s light throughput was only 10%. Subsequent measurements have proved the Nreal optics to have ~25% light throughput (blocks 75% of the real world light)
Nreal caused a lot of buzz in the AR hall, and I have had multiple people ask me about them since. They have a very simple “birdbath optical” design with a Sony 1080p Micro-OLED display. The optical design is very similar to ODG R9 from CES 2017, but lighter and with significantly fewer reflection artifacts as far as I could tell with my brief experience.
With a former Magic Leap person founding the company and system configuration, the comparisons are inevitable. Regarding image quality, Nreal blows away Magic Leap or Hololens with much higher resolution, better contrast, and no field sequential artifacts. The one area where Nreal is lacking is regarding real-world light throughput,
which I roughly measured to be only 10% which is about 30% or about double that of Magic Leap’s 15% and a bit less than Hololens at about 40%, but well behind Lumus Vision’s 1080p and Vuzix’s Blade are about 80% transparent to the real-world.
Plessey and Lumens LED were both showing MicroLED microdisplays on the floor of CES. Many, myself included, believe that MicroLEDs are key to the future of near-eye displays and displays in general. They have many technical advantages over all other types of displays. The question becomes when they will become practical for mass use.
Vuzix had a demo with a fixed image, blue only, Plessey MicroLED display. The projector optics were impressively small and was about ½” by ½” by ¼ in (12mm x 12mm x 6mm). As the demo was quickly put together for the show and Vuzix had not had time to optimize it, I was not allowed to published pictures, but I did have permission to report I had seen it. Vuzix had grafted a tiny display engine that was only about a ½ inch on a side and a quarter inch thick onto Vuzix’s Blade waveguide. It certainly demonstrates the promise of MicroLEDs.
Lumens LEDs had working individual Red, Green, and Blue 720p microdisplays that were combined with an X-Cube to make a projector I reported on last year at CES 2018.
For CES 2019, Lumens was also demonstrating a working green (only) 1080p device. They had the device on a table in their booth showing a live video feed and was able to take a direct picture of the device.
CREAL3D was demonstrating an honest to goodness true light field near-eye display and not the fictitious marketing hype “digital light fields” by the likes of Magic Leap. CREAL3D was hidden in the back of the Swiss pavilion, and ROADTOVR has written a good article including links to some stills and videos. It’s sad we have to add the word “true” in front of “light field” to distinguish CREAL3D and others like Fovi3D (with a direct view light field display) from the charlatans, like Rony Abovitz of Magic Leap, who abuse the meaning of well-defined concepts.
I first learned about CREAL3D at CES 2018 in a private meeting. They had a fascinating but crude green-only demo to prove that it was a true light field display. It was nice to see the progress they have made with a full-color demo this year.
The size of CREAL3D’s demonstration is because it is still on an “optical breadboard” complete with a large metal plate with optics fixtures screwed into it. While I’m not sure whether CREAL3D’s technology can be reduced to a head-worn form factor that they claim and the resolution and other image quality aspects improved using new display devices, it is one of the more technically interesting things I have ever seen in display technology.
WaveOptics had a good demonstration of a series of compact diffractive waveguides using TI-DLP based displays. They showed a 40-degree FOV 720p headset with the display over the eyes and lower resolution 25-degree FOV with the displays in the temples. Both designs were very compact relative to other waveguide designs I have seen.
WaveOptics 40 degree FOV waveguides have been designed into Rokid’s Project Aurora Headset.
WayRay was demonstrating a true hologram mirror automotive HUD display. Once again, I need to emphasize the word “true” to distinguish it from Microsoft’s Hololens “marketing in name only” holograms. I was expecting it to be yet another “Pepper’s ghost” effect but was pleased to find it was a flat hologram acting as a curved color wavelength selectable mirror. Hyundai is planning on incorporating WayRay’s HUD technology into future automobiles. They use a TI-DLP illuminated by lasers to produce a narrow color bandwidth image to make the hologram work well.
Eyelights in the French Pavilion was nice enough to take me for a test drive with their aftermarket HUD for cars. Eyelights has a very bright and ruggedized flat panel display with a simple semi-mirror combiner sheet that sticks to the windshield.
Eyelights’s design is in market contrast to WayRay’s HUD. While Eyelight provides a very good and easy to see image even in most daylight conditions, it requires a large dark patch on the windshield, and which may run afoul some some state’s safety codes that limit blocking the view out the windshield. As the picture above demonstrates, it does not move the focus of the image out into the user’s far vision. WayRay’s use of holograms and a laser illuminated DLP projector supports a much more transparent view of the real world and moves the focus of the HUD image into the drivers far vision.
North’s Focals laser beam scanning glasses that I wrote about a few months back, on the other hand, managed to fall below my worst expectations. Amazon invested in North Focals which is why they were in the Amazon showcase area at CES and garnered a lot of publicity.
North’s image quality was horrible any way you would care to measure it, low resolution, poor contrast, tiny eye box, and a smudge you look through. When asked about brightness, the response was, “we don’t talk about brightness” which I translate to mean “it is not very good.” As I am fond of saying, “missing spec’s are usually ones that are not good.”
It has a tiny eyebox which means you not only have to have the glasses custom fitted, but you must wear them in an exact position or else you see a double image or no image at all. When the display is off, you are left looking through a smudge in your glasses. With a claimed 10-degree field of view, low contrast, and low resolution, a person would be infinitely better off with a smartwatch. North claims they are selling well and will be opening up more custom fit stores soon, let’s just say I am very skeptical. All I can say is, “Amazon, call me first before throwing money at dubious display technology.”
Even though it was my old company, I was surprised to see a demonstration of a true 4K LCOS microdisplay. Once again, I’m using the word “true” to distinguish it from the displays that use a lower pixel count display and then optically shift it. The display has 4K (3840 by 2048 pixel) pixel-mirrors with very small 3.2-micron pixel pitch. They were demonstrating on the floor both near-to-eye optics and in a projector.
Raontech was demonstrating their WQHD (2506×1440) field sequential color LCOS microdisplay. They also made a point to show me that they no longer had the asymmetry I discussed in my article on the Lumus 1080p engine.
Bosch (using TI DLP), ASU (using Syndiant’s LCOS), and Microvision (using their laser beam scanning – see above) were all showing Kitchen Countertop or Speaker on Table projectors. I don’t understand the use model for this type of application. It looks to me like it will only work (and not that well) in contrived demo setups.
All of these systems will have washed out pictures in good lighting and require a surface that can double as a decent projector screen. The typical tabletop or countertop hardly makes a good screen even when new, no less when it has been scratched from use.
I made a special trip to see Microvision at the ShowStoppers event at the Wynn. Unfortunately, Microvision had pulled out of the event that day. The ShowStoppers staff told me that Microvision had tried to set up their demos but claimed that the room lights caused problems and left. I don’t know if it was the amount of light or some other problem, but Microvision decided it was better not to show.
Microvision demos were supposed to be, according to their news release, for an “Interactive Display Engine” and high-resolution Lidar for depth sensing including automobiles. The interactive Display engine consists, according to their marketing material and videos, of a projector shooting onto a tabletop with time of flight (ToF) detection of very simple interaction. This application is inferior in just about every way to a touchscreen both regarding display quality and for interaction/touch-interface. They should stick a fork in this concept because it is long past being a done (see also my comments on similar concepts from Bosch and ASU below)
I don’t follow the Lidar market, so I don’t know how their unit compares to the many other companies working on Lidar with both laser scanning.
Hisense Dual LCD TV looks like it could have a big future Hisense sandwiched at 4K color LCD on top of a 1080p black and white LCD to give an LCD panel OLED like blacks while supporting higher peak brightness using Quantum Dots. While using two LCDs in-series has been theorized for years as a way to give high contrast, Hisense has figured out how to make it. I would not be surprised to see other companies, particularly Samsung with their QLEDs adopting similar methods. For more information, I would suggest the Display Daily article on the Hisense Dual LCD.
Hisense was also one of several companies showing a very short throw “Laser TV.” Compared with the image quality and relatively low cost of LCD and even OLED TVs, the short throw laser projector TV seems to me too little, too late. Hisense was using a high gain and light rejecting screen, which helps but also resulted in hot-spotting and limiting the viewing angle. It might have been a great product 10 years ago.
Lemnis is developing software for vergence/accommodation. They are hardware agnostic and have been working on using eye tracking that controls the focus adjusting hardware. They had a working demonstration in their suite. The demo demonstrates using eye tracking to detect the eye’s vergence and then control a focusing mechanism in the optics.
I plan to follow up with a more detailed articles on some of the products highlighted above. Also as stated above, I am working on a description of what I think the next Hololens is and is not going to be using for their new design expected sometime in 2019, reportedly as early as Mobile World Congress 2019).
I would like to thank Ron Padzensky for reviewing and making corrections to this article.
Karl, Thank you for sharing your observations and expert comments.
I’d like to shout out congratulations to the Syndiant team for the true 4K microdisplay.
Also, major kudos to the Vuzix and NReal teams.
Looking forward to your next posts.
Karl, thanks for you nice brief summary, hope you get well soon. Always interesting to see that Syndiant and Raontech are still making new panels, but why oh why do they insist on making such wide aspect ratios?! It makes it harder to increase vertical FOV, let alone ignores the fundamentals of projector design where a square panel is far more efficient to illuminate, after which you can recover the desired aspect ratio with anamorphic prisms if needed. The problem with the AR industry is that too many electronics engineers / none-optical people are trying to determine the components that make up the optics, with the consequences that lead to unnecessary compromises in image performance. To anyone reading this, if you’re one of them… take note! 😉
They are being driven by customers that want HDTV aspect ratios. Of all the display technologies, LCOS must be the easiest to change the aspect ratio. If their customers wanted 4:3 or some other more square ratio, like Magic Leap, they would build them. It sounds like Magic Leap got a custom size device from Omnivision.
The whole AR market for Microdisplays (thus not counting the “cell phone plus large bug-eye combiner”) must be less than 200KU per year. Hololens, for example, is said to only sell about 25K units per year and I am hard pressed to find someone selling more than them. I suspect most of the microdisplay company sales are still in small projectors more than headsets. While they hope that future sales will come from headsets, they are not seeing enough to drive them to build custom displays. They would be happy to wack off the ends of say a 1080P display and give you say 1440 by 1080 pixels if someone would pay for it.
Karl, yes you’re right, and customers with deep pockets – although often the decision makers are not optical designers… And again, even in compact projectors, the system can be made more efficient with square rather than rectangular displays… but of course its the tradeoff of cost: added prisms to recover 16:9 vs some light loss. However – with the increasing ubiquity of large area emissive display screens, projector sales are also in rapid decline, so they should keep an eye on other applications (perhaps as dynamic holographic optical elements in medical applications having the most future potential for LCOS).
There seems to be more of a desire for small projectors in the parts of Asia if the manufacturers are to be believed. But I agree, that the long term prospects for projectors are not good (note my comments about the counter-/table- top projectors.
The problem with the other applications you mention is that while they might be interesting lab products, they don’t buy many units. Without volume sales, its impossible to sustain a component business.
Have you heard about the MicroLED(5000ppi, 1 million nits) displays from the company Jade Bird Display? https://www.jb-display.com/about https://www.jb-display.com/applications
Yes, I know about Jade Bird Display. I talked with them at their booth at SID. They apparently know of my blog and were very nice to me.
They certainly appear to be in the lead pack of companies working on MicroLED. The issue is going to be how long it is going to take for the MicroLED companies to get to high enough yields to be economical. In a demo, without optics, you could have a few thousand dead pixels on a high-resolution display and nobody would know about it. It is about 10X harder to make RGB on a single device.
Thank you for your reply, Karl!
Nreal & magic leap – it seems fairly obvious how similar nreal and magic leap are in design with their processing pack and controller..considering nreal was founded by an ex magic leap engineer wouldn’t magic leap sue nreal at least to stop them from selling in the US? Surprised magic leap hasn’t taken any legal action on an ex employee making a similar headset and now trying to get into the US market (magic leap probably doesn’t care about China but seems like they should stop them from the US)?
Magic Leap has been pretty litigious against other former employees, so we will have to see. But then again, there may be nothing that special or patentable in what Magic Leap did over say Hololens. There have been a lot of AR headsets in the past that had a computer and battery separate from the headset and having a controller is obvious. What Magic Leap MIGHT be able to claim is that they used inside information to get a jump on designing the product, but I don’t know what deal they had or the timing. The optics are totally different from Magic Leap and very like ODG’s. Short of a patent that should have been hard to get, it would seem tough to stop anyone else today.
It would also be foolish take action until the infringing party was actually making money from your invention.
That is generally true with patent infringement, but less so if the issue is a trade secret.
The reason that they aggressively pursue former employee may be an indirect form of non compete. The “shoot a few prisoners and the rest will fall in line” philosophy and/or simple ego spite.
Thank you for your review! Did you take a look at DigiLens Crystal? How was Vuzix’s other closed door demos? Do WaveOptic devices have a huge eye box like they claimed?
Can’t wait to see your prediction of Hololens2
I did not get a chance to see Digilens at CES but hope to in a couple of weeks at Photonics West. I have seen them in the past and from my experience, their older designs behave similarly to other diffractive waveguides.
I didn’t have time for any quantitative checks, but from what I remember the eyebox was good/large. The image quality was better than I was expecting and the engine more compact. I’m still a bit concerned about light sources in the real world turning into flashes of color as I have seen with others diffractive waveguides.
I saw the Waveoptics device and would say only the lower FOV looked better in terms of color, but that’s to be expected, as physics tells us diffraction is an angular problem. At 40° the color was all over the place, worse than Magic leap – again thats a consequence of the expansion – yes massive eyebox, but something has to give. And yes as you say Karl, from what I could see of some external reflections, outside world diffraction is not nice. BTW the resolution also appeared less, more like 2 pixel from what I could tell. Nice try though.
Thanks for the comment. As I have written many times, it is really hard to judge display quality unless you can control the input. I often say/write “If I can choose the product or choose the demo material, I will pick the demo material.” By this I mean, that companies will choose content that does not show their known problems. Often for displays, the simplest test patterns like a simple white or gray screen will show up the worst problems such as color and intensity uniformity. Just looking through a display when it is off with light sources in front of you can pick up problems.
Karl, thanks for sharing your expert comments
The pictures “Nreal Side View” show a problem that the Semi-transflective mirror definitely reflect the facial or other images superimposed on the virtual image, in addition to doing ugly structural occlusion, is there an optical solution to solve this problem?
Thanks, I’m going to get into the reflection issue from below the 45-degree beam-splitter in a follow-up article. This reflection could be stopped by simply putting a black “cup” under the beam splitter to block light. The birdbath optics NREAL is using has to be one of the most common optical structure used. There is an alternative birdbath configuration used by Raontech and Google Glass, among others, that is a little thinner and has better light throughput to the real world.
In the alternative birdbath, there is a curved mirror (not semi-mirror) opposite the display and then you look through the beam splitter (see Google Glass figure I labeled: https://i2.wp.com/www.kguttag.com/wp-content/uploads/2017/02/Google-Glass-Birdbath-001.png?ssl=1). A problem is that this alternative birdbath requires a polarized display image for the “quarter waveplate” to change the polarization and the OLED display is unpolarized. I guess one could then polarize the OLED but this has its issues as well. Both Raontech and Google Glass were using LCOS displays that required polarized light so polarization what not an issue for them.
In the end, the thickness is caused by the beam splitter. The beam splitter is used to have the image be on-axis with the curved mirror to give an undistorted image with uniform focus across the image. There are ways to go off axis such as the Everysight Raptor, but then you want to have pre-compensation optics to correct for the image and focus distortion. To make them “thin,” then waveguides become the obvious choice but then waveguides have a different set of issues be that the common diffractive waveguides or the multi-prism Lumus LOE.
In short there is no perfect solution or even an absolute best solution, only different solutions with different pros and cons.
It will be very interesting to hear your ideas on Hololens.
There have been recent speculation that Microsoft will shift to scanning micromirrors for sensing and perhaps also for display technology. It sounds implausible (especially the latter), but they do undoubtedly have a lot of relatively recent IP in that area and they have had a big team working on micromirrors for several years and it has not been decreasing. They have also been fabricating scanning micromirrors.
My estimate would be that all this work is mainly a research exercise (albeit ambitious) and as a fallback (since they are seeing that other companies are working on scanning micromirrors, e.g. Intel did Vaunt and also invested in North), but it is still curious that they invest as heavily into it as they do.
What is your idea about this, Karl?
Overall very insightful and interesting comments. I’m aware of the Hololens IP relative to laser scanning. Microsoft has to find a way to spend all the money from the near monopoly of the PC market in a way that looks like they are working on the future.
Laser beam scanning is a pot of gold at the end of the rainbow technology (perhaps an apt analogy); no matter how far you chase it, when you get to where you think you need to get to you will find new problems based on your improved understanding that will say you are at least as far away from a solution as you started. It also sells well to the semi-technically-literate that only understand the advantages and are oblivious to the problems which are tougher to understand (and how Microvision has lost over $500M of investors money over the last ~25 years).
Hololens is an R&D project that “escaped the lab.” Reportedly they sold only 50KU over the first 2 years or 25K Units per year. If you are making components for them, they would be a nothing customer if their name was not Microsoft. Himax got both Google Glass and Hololens, wow what a deal; not enough to pay the electricity bill to make the units.
I heard (and it seems believable) that Intel’s “investment in North” was to “sell/trade” the patents to them for Stock in North. It was a fair trade IMO, worthless patents for worthless stock. Intel’s judgment in near-eye displays has been perfect, they never got anything right.
That’s pretty harsh on North! They are new to the scene, so going through the learning steps, and I suspect do not have to remain with lasers forever – for one SLEDs are almost there… and there are one or two tricks (might add a bit of bulk though). I think they’ve done well in such a short space of time, everyone else has been working on HMDs for 10-20 years.
As for the recent Microsoft IP – I would not read much into it – the attempt of trying to get an outside world image back up a waveguide without any form of lens will not work – object has to be at infinity. They have filed similar patents over past 10 years or so (e.g. eyetracking up a waveguide) which of course bore no fruit for same reason. According to the latest patent the use of the MEMs still requires polarising beam splitters which are still one of the most expensive parts of a display, so it doesn’t really lower cost as the marketing spiel would have you believe 😉
“I heard (and it seems believable) that Intel’s “investment in North” was to “sell/trade” the patents to them for Stock in North. It was a fair trade IMO, worthless patents for worthless stock. Intel’s judgment in near-eye displays has been perfect, they never got anything right.”
I think there is some truth to this, but it’s not the full story. Intel made an investment in North early on (while the name was still Thalmic labs, also long before they decided to scrap Vaunt). Probably they got a lot more stock for the IP when that deal was done. It is said that Intel had trouble selling the Vaunt IP and after not being able to sell it for the amount they desired they decided to hand it over to North, where they already had an investment.
Also, I am much more optimistic regarding North’s future. It is a risky business, of course, and the chance of their value going to nothing is far from zero. But I also think that they are acting so much smarter than all other AR startups, everything written about is positive and they seem to promise only exactly what they can deliver. I think there could very well be a relatively large niche for what they are going for (they are not trying to compete with Hololens or Magic Leap).
Have you seen the display? It’s very poor in image quality and the eyebox is tiny and if you are just a little off you get a double image. They compound this with having to have custom fitting stores and a $1,000 price tag for a display that is lower in resolution than a smartwatch. When I asked simple questions like, “what is the resolution” and “what is the brightness” the answer I got from their “technical person” is “we don’t like to talk about things like that.”
Frankly, I don’t see any advantage over a smartwatch, do you? Do you really think you are going to be inconspicuous when people see the lights from the image flashing in your glasses? As a business, they seem to have the worst of all worlds with a poor quality, very expensive, niche product and the requirement for a large amount of overhead and labor per sale. Nothing about this product picture looks good.
true. Especially for projectors in smartphones. Cool technology – but no use case, too dim projections, too low resolutions for all day use and the unanswered question: Where to project?
The kitchen projectors, e.g. from Bosch, instead may be useful in some situations because you can project onto food. In addition, many kitchens do not have extra space for fixed screens and integrating a screen into a table would be also no good idea. With projectors you may project onto scratched tables. With a build-in screen you will have a scratched or even broken screen…
Also, I think projectors in smart speakers could have a chance. You do not have to reserve space in your room and can place it everywhere. If on table, the projector may for short time project some information on it but alternatively you can move it near a wall and then you may have the projection there. You do not have to “update your furniture” for such a smart speaker. It is more flexible than a traditional screen. Of course, you will have much lower resolution and poor image quality that depends on the surface and light conditions.
For AR I am still missing a mass market use case that is acceptable for such big, heavy and uncomfortable glasses. If you refer to Microsoft it is also very true that Microsoft likely produced only 50.000 units – that are sold-out for a while. Even if they pay $ 100 per unit to a supplier of the screen and doubles the production from 25.000 to 50.000 units per year this will result only in revenues of $ 5.000.000 per year. And that are revenues. Most money – if not all – will be need to manufacture them. There will be no money left to finance a company. Even if $ 1 million is over after the costs: E.g. a company like Microvision needs approx. $ 25 million per year (I do not know why so much) for operations without profits. So, they can finance by selling that units only two weeks of operations. On the other hand – the AR customer of Microvision, which could be Microsoft, paid $ 24 million for the development and some units. Why if they do not expect a much bigger market?
Thank you very much for sharing your news and analysis’ to us that we will not find on other websites.
Thanks, but I think you really have to stretch credibility to justify the kitchen and the speaker projector.
Kitchen Projector — First, a good kitchen is designed to brightly light the countertop work surfaces. For the Bosch demo, they had a white inset in the countertop. The inset/screen can’t be shiny (specular) or seriously off-color (ala granite or another surface). In short, it has to be a very durable but good projection screen which is in the real world an oxymoron. The things you do to make it durable will tend to work against its optical properties as a screen. Do you really need to project slices on a cake and how long will it take versus the old fashion methods? Nothing in the demo Bosch gave made me think, “this is going to really help in the kitchen.” The detect the steak and bring up steak recipes was kind of neat, but you mean it would not be faster to type “steak recipes” and throw the steak on a very accurate $20 scale rather than have it guestimated? It all seemed very contrived like they had a solution looking for a problem.
Same goes for the projection speakers. You still need a decent projection surface. I would hold out more hope for a rollable OLED/MicroLED screen to roll out when you need it. Before then LCDs are just too inexpensive with much better image quality in a typical lit room. You coudl contrive a situation where a projector would work but it is not a mass market item. The LCD will be so much less expensive and “reliable” in that it will work in any room. If you are Amazon, Google, Apple, Facebook, you are looking for a mass market product not a novelty niche. There is a reason why projectors are on the decline.
You just don’t get it. I’m hearing Microsoft has spent billions of dollars to get to about $75 million in dollars of revenue (not profit) a year with Hololens (yes, I am being sarcastic). You are also right that the “supply ecosystem” is terrible. Himax “won” both Google Glass and the first Generation of Hololens. They spent several hundreds of millions in their LCOS factory to get less about 5 million in revenue over a period of about 5 years.
When it comes to Microvision, I’m always reminded of the Clevon Little’s Blazing Saddles’ scene where he gets away by threatening to shoot himself (Warning, like all of Blazing Saddles is it very non-PC – https://www.youtube.com/watch?v=Z_JOGmXpe5I). Microvision has survived by finding someone, a “fish,” at a big company that is desperate for their technology for some pet project to act like a sugar daddy. Microvision then is then able to use the NRE from the “fish” to then raise a significant multiple of the NRE amount by selling stock since it is a public company (they are a 25-year-old “startup” that went public during the dotcom boom when you didn’t need revenue or profit to go public). In the end, the “fish” realizes that while LBS solves one set of problems for them, it has a whole bunch of its own problems that are worse than the problems it solves. Time after time there is a “one and done” order (rinse and repeat, Lucy with Charlie Brown and the football). Often times the fish is an R&D group that is trying to justify their existence in a larger company, which does seem to fit Hololens. For all the reasons you outlined earlier, the numbers don’t work and whether or not Hololens uses LBS, it does not significantly change the equation in size, weight, or supply economics.
NOTE From Karl Guttag Moderating: I’m letting the comment below go through as it is civil and worth discussing, but I want to emphasize that these are the comments from iGlass which is a direct competitor to Nreal. Some of the points may be valid and some of which I disagree. Later when I have time, I will try and address the individual points in my comments
Nreal vs iGlass AR, which one will work in real life? Nreal need to improve its privacy and comfortability flaw
At CES 2019, I tried Nreal’s 85g fancy looking glasses, while it was a correct step towards the right direction, it have 6 major problems waiting to be solved.
MILPITAS, CA, USA, January 13, 2019 /EINPresswire.com/ — During CES 2019, I tried Nreal 85g fancy glasses looking gadget multiple times. While I liked its form-factor, 85g weight and fancy looking, it have a few major problems to be solved before it can work in real life.
1) Nreal’s 85g weight on nose is a fake industry design and it simply do not work in real life, as human nose can only support less than 30g daily prescription glasses (typical prescription eye-glasses weight between 15g and 30g) for long period usage. Nreal’s fancy industry design appearance was specifically designed to attract public attention, not for real life usage.
2) Nreal has fundamental privacy protection flaw, as others can directly see what you are watching from outside, from ALL angles. This is the same problem showed up on ODG R8, Lenovo Disney StarWar AR, AntAR, etc., all of these headset share the same “Birdbath” optical design.
3) Nreal’s 52 degrees FOV is too small to provide Movie Theater experience. Its screen size can compete with TV screen though, but the screen size is just not large enough to compete with Movie Theaters.
4) Nreal glasses setting on nose is too close to sensitive middle of eye area and it get noticeably hot in a 10 minutes use. Nreal need to move the thermal source out by half an inch or so to improve the comfortability, but it will change its current ID noticeably.
5) Nreal do not fit daily prescription glasses, while 40% of the world population wear their daily prescription glasses (this ratio is even higher in China). It means 40% of population either need to change their daily habit or be excluded from the customer base.
6) Nreal used the so-called “Birdbath” optics architect, which is complicated and not the simplest possible high image quality, large Field of View optical deign form. It is cheaper than the $2300 MagicLeap waveguide though, but Nreal’s $999 price tag are still out of the reach of ordinary consumer.
As a comparison, iGlass used the simplest off-axis optical architect, which use the inner surface of a reflective plastic mirror, and cheap LCD panel to achieve 80degree large FOV (30-feet giant screen size at 10-meter away) and 4K ultra-sharp image quality. iGlass puts its 130-gram weight on the forehead, with zero pressure on nose or face. It is so comfortable that even a 10-year old kid can wear iGlass for hours, like wearing a hat. And iGlass allows wearers to fit their daily prescription glasses under iGlass easily. More importantly, iGlass provide total privacy protection, only the wearer can see the 30-foot giant virtual screen in front of you, no light pollution to others. And iGlass cost is only $299, one third the cost of Nreal, within consumer affordability range.
1) Yes, it is a bit of an illusion as to the size and it is heavy for long term use. They will want to have more support on the product.
2) I think you are overblowing this issue. There is an issue with “glowing eyes.” Being distractive. None of these, including iGlass is very secure in terms of privacy.
3) Untrue. If you sit in the “prime” location (usually near the center of the theater) you typical get 50 degree FOV or less. The super wide FOV a “VR Thing.” See https://www.dolby.com/us/en/technologies/screen-size-the-impact-on-picture-and-sound-ioan-allen.pdf . A 65” diagonal TV from 6 feet (~2 meters) away has about a 48.5 degree FOV.
4) Heat management was a definite problem with the Nreal prototypes. Solving this will cause weight and size issues.
5) True, and this is a negative issue with many of today’s headsets. There is no free lunch, working around most normal glasses requires a much bulkier headset.
6) The birdbath is the most common optical architect used in headsets. It gives very good image quality for the cost. What drives the nreal headset’s cost up is the use of very high-quality OLED microdisplays. The biggest drawback is that it is very inefficient and blocks at lot of real-world light.
iGlass off-axis design has it’s own set of problem. It’s architecture is essentially the same as Meta2 Mira and many others. Because it is off-axis, it both distorts the image and the focus from top to bottom. “Digitally” correcting the distortion lowers the effective resolution significantly and you cannot digitally correct the focus, the top of the image is going to focus much closer to the eye than the bottom, which in addition to being a problem, is made worse by glasses with varifocal or bifocals that have the “far vision” at the top and “near vision” at the bottom.
While iGlass has a wide FOV it has very low angular resolution. Thus it is better for doing things like games and poor for things like web browsing. You also get “crosstalk” between the two eyes in the middle where the two domes join as there is no separation as there would be in a typical VR headset.
I’m not trying to put down iGlass, but merely point out it is a different set of trade-offs.
Very comprehensive summary as always, thanks Karl!
It seems a calm-down year for all AR companies, only NReal gains some noises in CES. Moreover, since it uses the same optical solution as ODG, which collapsed recently, do you think they can pick up what ODG left, or just another ODG which won’t end well in the future?
Also, how does Waveoptics’ image look like? How do you compare it with other similar waveguide product, such as Vuzix, Hololens, Magic Leap, Digilens, etc.?
Looking forward to your response.
ODG got a lot of buzz and you can see where that go them. In the end, the product has to be useful. The Nreal AR glasses have their pros and cons. While smaller and lighter than most AR headsets, they are still big and bulky compared to normal glasses. The image quality was very good for AR (see-through) type headsets with a decent FOV and resolution (1080p). A big downside was the 10% see-through but this could be improved somewhat. To me is a vastly better starting point than say Magic Leap, but I don’t know if it will get over the finish line even if the 6DOF and SLAM are good (don’t know about these capabilities). There is a question about the whole market for a consumer AR headset. Nreal is still expected to be in the $1,000 range and bulkier than glasses. It is still more of an indoor gaming product. While the image quality is much better than other AR headsets, it still is not as good for what a movie as a $400 4K 55″ TV you can buy at Best Buy, no less what you can get for $1,000. So to me, it seems more of a strictly gaming machine. The “AR/MR” market is someone turning into an “indoor only” gaming oriented market at present and not the “it is going to replace your cell phone.”
WaveOptics seemed to me to be in about the middle of the pack in terms of image quality compared to other diffractive waveguides. I’m guessing because I could not do rigorous testing at CES and the vendors control the demos material, they are better than Magic Leap, about the same as Vuzix and Digilens, and maybe not as good at Hololens. They are definitely worse than Lumus and NReal by a wide margin which are better than any diffractive waveguide I have seen.
Lovely write-up as always Karl, I’m glad that your new job hasn’t stopped you from updating this blog. I use your writings as (a very useful and accurate) anti-bullshit filter on the hype machines around this kind of tech. Thanks very much.
Marvelous write up.
Karl, do you believe the Hololens2, which is expected to be announced next month, will be the gold standard for AR or are these other companies closing the gap?
My expectation is that the next Hololens will have about as much impact as the 1st Hololens. I expect it will still be very expensive and sell in the few 10’s of thousands per year. I’m not expecting great improvement in FOV or image quality. I’m fond of say there are 30 major technical problems
If there is any gap closing, it is very slow. Everyone is fighting the same issues with physics. It is very tough to get a good image in a lightweight and compact form factor so close to the eye. Everyone is just making a different set of tradeoffs. Magic Leap took the battery and computer out, which reduced the weight of the headset but left you with a cord that is a snag hazard and a pain to use; they also increased the FOV, but then made some very bad choices, particularly the multiple focus planes that damaged the image quality. Nreal relatively clean and light with comparatively very good image quality (blows away ML and Hololens, and likely Hololens2), but it is still an indoor gaming-oriented product and not the kind of thing you would wear walking around outdoors.
> high-resolution Lidar for depth sensing […] time of flight (ToF) detection of very simple interaction. This application is inferior in just about every way to a touchscreen both regarding display quality and for interaction/touch-interface.
IR-pattern-stereo depth sensing (Intel RealSense) has been used for multitouch. But the >1 mm depth error was problematic. If Microvision can do better, I at least would be interested.
I’d like to write code in a setup with improved HIDs. But I’m disinclined to sacrifice existing productivity. So I start with an existing ThinkPad keyboard. Even in VR. A keyboard which would obviously be better as a full multitouch surface. But optical finger tracking has been a pain. So I’d pay for a “make anything into a good multitouch surface” solution. I can correct for fingertip orientation, to get from sub-millimeter depth to touch. Though it would be nice if IR environment wasn’t polluted. Ah well. Perhaps by combining rough depth with a mirror and individually lit keys…
Keyboards are very problematic with laser scanning and any other projection-based sensing. The obvious problems include the occlusion/shadows, knowing when to actuate, and how does the user rest their hands. They tend to end up forcing the user into a very exaggerated 1 finger typing to get it to work.
Maybe with cameras and processing, you could figure out when the finger has done a press on a surface.
Note that you don’t want to do it the unbelievably dumb way that Hololens did where you have to hold your hand up for the camera to see a gesture. After about 15 minutes your arm feels like it is made of lead.
I was unclear, sorry. Imagine depth-sensing multitouch like , but instead of a table surface, there’s a laptop keyboard surface, like . So first, you have a relatively nice laptop keyboard for typing. But the keys are smooth and flat, and the gaps between them are smallish. So fingers can glide surprisingly well over the entire face of the laptop. As if the surface of the keyboard and its surrounding laptop, were a slightly bumpy table. So second, you can do multitouch on top of the entire laptop surface. Rather than only on the small touchpad region. So you have a multitouch tablet, with full finger height/proximity and hand pose data, and a nice keyboard (now n-key rollover), as one nifty human interface device. If only the touch sensing was a bit cleaner, and thus didn’t require exaggerated touch down and up motions. (Also, getting a stylus to nicely glide over keys, while providing usable pressure data… is a problem).
In the NYU demo, they have the camera more ideally placed far overhead and even then it is tough. You can do some multi-sense but only if nothing is obscured. It is really hard to judge intent with a keystroke. With the speaker projectors, you are generally projecting at an angle which makes things harder. In the end, you are using a lot of technology to get a very poor keyboard.
Thanks for your expert comments to the products.
As you talked above, the see-through difference is much large, which could affect the virtual and reality object. is there any standard or comment to the see-through value?
I’m not sure I fully understand your question. There are no standards I know of for transparency, but I think most people accept that about 80% transparent (20% light blocking) is good enough for most purposes and that is what most “transparent” headsets try to maintain particularly for industrial use. At 50% transparent, a person will clearly perceive the difference. By 75% light blocking, you are seriously blocking the light (starting to approach wearing dark sunglasses).
Another major factor is how much the headset blocks the person’s vision. This can be an important safety concern.
Karl, ANSI z87.1 for eye protection requires >=85% transmission on indoor use. It is worth noting Microsoft uses weasil words on its website for Hololens “complies with ANSI impact requirements” because they cannot pass the full spec due to transmission. No diffractive waveguide can.
It should be noted that about 80% transmissive is about the most I have seen with any of the AR headsets.
I took a look at ANSI z87.1 2003 (I don’t have the 2015 update handy, and I don’t know if it changed regards to this) and I don’t see the >=85% requirement for indoor use. I do see that as the minimum standard to be considered “clear.” This “clear” rating may then be picked up and used in other safety standards. Maybe I am missing something. [For others] It should be noted that uncoated glass or plastic reflects about 4% of the light.
My biggest worry with diffractive waveguides even more than the transmission is the flashes of color you get when there are light sources as I showed in this article: https://www.kguttag.com/2018/09/26/magic-leap-review-part-1-the-terrible-view-through-diffraction-gratings/
Waveguides typically have better light transmission that most other “see through” optics. Most diffractive waveguides I have seen claim to have about 80% transmission before you consider other lenses. Of course, diffractive waveguide transmission will vary by a combination of the wavelengths of light and the light angle which might make spec’ing them tough. Lumus has around 80% transmission although it varies across the waveguide. Birdbath type optics are usually terrible.
Not sure I can share my copy, but theres some of 2010 online here https://support.automationdirect.com/docs/ANSI%20spec.pdf
Under 5.1.2 – “Luminous transmissin shall not be less than 85%” As far as standards go, ANSI is pretty badly laid out!
I’ve measured Lumus WGs to be ~>85% transmission (as you say, does depend on reflector position, but at max reflectivity it appears to be in this domain), and this is consistent with Daqri’s Smart glasses being ANSI certified, must be some good AR coatings on their other lenses. Hololens is ~35%, indeed some of that is the outer visor, but remember diffractive displays are usually made up by a number of stacked guides to handle each color (4 for Hololens, 6 Magic Leap…) so as you correctly point out there are the Fresnel losses, which yield ~ <70% transmission on Hololens, omitting diffractive layer scattering losses. Yes birdbath is pretty terrible!
I wonder if you have visitied LetinAR during the CES, who showed 80 degree AR demo. It was impressive for me. Do you have any idea?
I wonder if you have visitied LetinAR during the CES, who showed 80 degree AR demo. It was impressive for me. Do you have any idea?
[…] I wrote last time, Nreal caused a lot of buzz both in the AR hall in news articles. They have a very simple […]
Hi Karl nice analysis. Wondering about a few things:
1) did you think nreal was faking their demo? It seemed like their slam was being faked with the table setup
2) do you think small glasses style with limited functionality like Vuzix or focals North will be the big hit or more advanced glasses like odg or Thirdeye or nreal that have slam ?
3) what did you think of mad gaze? It seemed like nreal but without the packs
4) did you see ThirdEye’s X2? It was the smallest stand alone glasses with slam and the optics seem waveguide
5) do you think hololens 2 will be a “leap above” everyone else or just slightly better FOV and Slam?
1. I didn’t have time to check out nreal’s SLAM demo. I possibly could have “pull rank” to a demo, but my focus these days is checking out displays and optics.
2. It’s hard in either direction. The problem is that things are getting conflated with people expecting great image quality with SLAM in a sunglasses form factor.
– North Focals from what I could see is terrible; they say they are adding stores but I don’t see it.
– Vuzix, IMO, is closest to be a “consistent product” with OK but not great image quality but at least it is small and light. It has a decent size eyebox and does not require custom fitting and only blocks about 20% of the real world light. It is lower in resolution and FOV than most but still blows away North. I think it is a LOT brighter than North which will be good for outdoor use. I’m not sure if there is big market for what they are doing, but at least it is much better than North.
– The AR with SLAM (or Mixed Reality) market is more for video games, museums, and training. Basically, they are semi-see-through VR. It is a totally different market than the “walk around an Augmented real world.” This is what I call a “Segway market,” is exist, it is just not going to be that big a market.
3. I think Mad Gaze is still using the Raontech LCOS birdbath design kit optics. That design uses LCOS. It is about 40% transmissive but with lower image quality that nreal. Nreal was using 1080p OLEDs with good very good black levels. Nreal would be better for watching a movie, but Mad Gaze is more AR-Like.
4. ThirdEye’s X2 is using Lumus’s waveguide. Lumus waveguides look much better than any diffractive waveguide. It is using a 720p LCOS. The image quality is similar to Mad Gaze and Third Eye 1 (which is the same optics as Mad Gaze). The Thirdeye X2 is using the older Lumus waveguide that results in the optics blocking a lot of the user’s peripheral vision. Newer Lumus waveguides are much better but just coming to market.
5. I would have “modest expectations” for Hololens 2 in terms of image quality. Somewhat better FOV and better color uniformity but little or no improvement in resolution (just a guess based on what I think they are doing). I would expect them to keep improving the SLAM. In the end, I think it will be another “25K units a year type product.” It is hard to predict the price when then spend billions of dollars in R&D for a 75 million dollar a year market, they could price it anywhere and it won’t change the bottom line significantly.
If you are expecting only 25k a year in units for Hololens 2, does that include the recent military contract? I believe I read they will be supplying up to 100k units over the next two years. Is this a different product? You also wrote you believe Hololens is the wrong product for the military. Why do you think the military believes other wise and in your opinion what is the right product? Sorry for all the questions in one comment!
The short answer is “yes the 25k included the military.” I should also note that I work for RAVN that is developing a military headset based on the needs of the troops (RAVN’s CEO is a former Navy Seal with two tours in combat). The DNA of our company is to be trying to help the troops with what they actually need with the head of the company having lived it, and not what some Ph.D. in a lab theorized that they want.
I don’t think the military as a whole believe that Hololens is the right product, just a segment with money to spend. The US military wants to be “high tech” and keep ahead of the enemy, but they are often misled by theoretical advantages that are impractical in actual combat.
Keywords in the military contract are “up to.” There is a lot that goes on in military contracts that have nothing to do with what is best for the troops. At some point, they will have to pass field qualification where they can’t handwave around issues like how much the equipment blocks the light.
Right off the bat, Hololens blocks too much light. There are specs out there for all kinds of eyewear that have to block less than 15% of the light and Hololens blocks much more than that. Military documents talk about the danger of blocking too much light in case troops have to go from outside to indoors quickly (see for example: https://phc.amedd.army.mil/PHC%20Resource%20Library/Safety_Glasses_and_Tinted_Lenses_63-001-1013.pdf). Then you have all the issues with diffractive waveguides and the flashes of colors they cause. Unless Hololens changes everything about what they are doing for optics including dropping the diffractive waveguides they are likely continuing to use in Hololens 2, I don’t think it will get deployed in mass numbers to the troops. Part of the 25K units a year Hololens already sells goes into military training, but that is a very different requirement than deploying it with troops in the field.
It’s hard for me to see Hololens as a competitor other than their soaking up money and maybe putting the military off of high tech. I have a saying, “if I am a small company, I want the big company spending their time and effort on the wrong thing.” There are lots of different groups in the military and many of them are not bought off on Hololens.
I really want to see a display that uses active lenses to always keep the point the display in focus. It could then simulate depth of field by blurring the “out of focus” regions of the display. You’d need good eye-tracking to see the convergence point of the eyes. As a bonus, you wouldn’t need to worry so much about the sweet spot of the lenses since you could adjust focus based on what region of the display is being looked at.
For the sake of the men and women in our military, I hope they get the right product. As a competitor I would expect nothing less than what you said, but I think you are being a little too dismissive of Microsoft.
If you do not see Hololens as a competitor, who do you see as a competitor?
One additional question. It has been 4 years since Hololens was released. Are your assumptions based on this version of Hololens or are things you may have learned about the upcoming release?
Hi Karl, thank you very much for bring us these useful information.
Regarding the tri-laser TV provided by Hisense, I am curious about the speckle level of it. Usually, green laser generates most obvious speckle and that is one reason why people only use single/dual lasers in projection light source before (another reason would be the price of green laser is way more expensive).
I haven’t seen anyone else talking about the speckle level of it. Since you saw the product, what would you rate the speckle level? Like, completely speckle free, minor speckle, moderate speckle or still very obvious?
I did not notice any speckle. Most “laser projectors” today are actually made with a blue laser for blue that is also converted to green with a phosphor wheel. Red is sometimes an LED, sometimes a red laser, and sometimes a blue laser converted with phosphors to RED. So if the phosphor convert green there will be no speckle in green. Another simple trick is to vibrate the screen (this is what Dolby does in the Laser Cinemas.” It does not take much vibration for a major speckle reduction and it is the only thing I have seen really work eliminate any noticeable speckle with true laser light. They are likely using multiple (many) blue lasers and this will reduce speckle in blue.
Hi Carl. I want to know if you had the opportunity of test the Realmax Qian(100 AR) in CES 2019? They claim the FOV is 100 degrees. Some reviews say in youtube that the image is very clear and the FOV is the biggest of all AR headsets. Everything is integrated is the headset. Is a very interesting AR headset.
Hi. Carl. Did you have the opportunity of check Realmax Qian in CES 2019? They claim 100 degrees FOV and very bright image.
what is your opinion about the new 230 pages long Microsoft Hololens patent that shows (only) a Syndiant LCOS projector in the assembly and in the specifications (page 28 in the PDF):
The patent shows the complete assembly with BOM and all parts (including the Syndiant projector with all parts). Is this also your information that Microsoft switched to Syndiant for Hololens?
Specifications form the patent:
“Optic Characteristics Value
virtual display field ~25-30 degrees (equivalent to the
of view (Diagonal) FOV of a 24″ monitor viewed at
1 m distance)
see-through field of view more than 80 degrees
eye clearance more than 18 mm
Material zeonex optical plastic
weight approx 15 grams
Wave Guide dimensions 60 × 30 × 10 mm (or 9)
Size 15.5 mm (diagonal)
Material PMMA (optical plastics)
FOV 53.5° (diagonal)
Active display area 12.7 mm × 9.0 mm
Resolution 800 × 600 pixels
VIRTUAL IMAGING SYSTEM
Type Folded FFS prism
Effective focal length 15 mm
Exit pupil diameter 8 mm
Eye relief 18.25 mm
Number of free form surfaces 2-3
AUGMENTED VIEWING SYSTEM
Type Free form Lens
Number of free form surfaces 2
Wavelength 656.3-486.1 nm
Field of view 45° H × 32° V
Vignetting 0.15 for the top and bottom fields
Distortion 10% at 30 lp/mm
In an embodiment, the Projector Characteristics may be as follows:
Projector Characteristics Value
Brightness Adjustable, .25-2 Lumens
Voltage 3.6 VDC
Illumination Red, Green and Blue LEDs
Display SVGA 800 × 600 dpi Syndiant LCOS
Power Consumption Adjustable, 50 to 250 mw
Target MPE Dimensions Approximately 24 mm × 12 mm × 6 mm
Optics Housing 6061-T6 Aluminum and Glass-filled
Weight 5 gms
RGB Engine Adjustable Color Output
2x 1 GHZ processor cores
633 MHZ DSPs
30M polygons/sec DC graphics
This patent APPLICATION is one acquired by Microsoft from ODG. I don’t think it has much to do with Microsoft’s direction with Hololens in the future, if that is what you are asking.
I’m catching up on my VR news and saw the comment you dropped on RoadToVR about the CREAL3D a while back on it being “not done like Nvidia’s” lightfields.
I’ve only recently managed to wrap my head around lightfield optics so I gotta ask, how is CREAL3D different? Does it not use a microlens array to reconstruct the lightfield image? Or perhaps it uses many dedicated microdisplays per lens instead of just one display?
The Nvidia light field display generates small sub-images simultaneously. Thus the resolution drops with the number of subimage. CREAL3D generates a series of full resolution images time sequentially and used the illumination angle to generate the various. How it is done is explained here: http://lightfield-forum.com/2019/01/patent-how-creal3d-light-field-display-works-without-a-microlens-array/
Karl, With the anticipated release of an updated version of Hololens at MWC, will you be communicating your thoughts on display technologies you expect to be incorporated? LCOS or otherwise? Enjoy your work! Thanks
I hope to write about it if I have time. I have a lot of things going on right now both personally and with work.
Thanks Karl. Look forward to your insight if schedule allows.
I’ll like to thank you for such in-depth research on smartglasses display as a whole.
I’m particularly interested in WaveOptics and Digilens, both companies are extremely similar and focus a lot on manufacturing waveguide lenses for the use of smartglasses.
I’ve spoken to both companies and WaveOptics mentioned that their waveguides are more on surface-relief while Digilens is on holography waveguide. Could you explain the pros and cons between the two?
Both work by diffraction. A surface relief waveguide, as the name suggests essentially adds a diffraction grating surface on a substrate. I’m not sure how Waveoptics does theirs, but the grating can be more or less printed or using some form of deposition using photographic masks. With Digilens, they are using a liquid crystal photo-polymer that is “frozen” by exposure.
Both types claim advantages in manufacturing cost but there is so little public information that it is hard to substantiate the claims. A lot of the issue will come down to yield. There seem to be more companies use surface relief (Hololens, Magic Leap, Waveoptics, and Vuzix come to mind) with fewer doing Holographic/Photochemical processes (Digilens being one of a few). In terms of image quality, they seem to have similar issues.
[…] AM Glass version shows a “light shield” below the optics. As I noted in my 2019 article on Nreal, a significant amount of light is reflected up (see lower right picture above). This is because […]
[…] Microvision has probably been the longest-lasting promoter of the down projector touch screen concept, starting with development kits in 2017 and following up with full-blown design concepts at CES 2019 and 2020. Microvision used their LBS projector technology, but others have also shown the DLP and LCOS concept at CES 2019 and 2020. I wrote about the Microvision, Bosch (using DLP), and ASU (Syndiant’s LCOS) projectors at CES …. […]
I need buy rdp501h where?
I would think you would have to contact RaonTech directly. (http://developer.raon-tech.com/index.php/contact/)