Update 12-19-2019 – Added missing hyperlink for one of the photographs to the text.

In editing the document, an important hyperlink that pointed to where I got one of the photos was accidentally lost. The link was included label of the photograph but was missing from the text. In subsequent online discussions, it appears that this error has caused people to assume I had gotten the photo from a different online forum that cited the same photographs. I added the link to the section “Hololens 2 Problems are Now an Open Secret.”

Introduction

Someone sent me one of the pictures in this article a few days ago, but I was waiting for permission and getting some more and better pictures. But since pictures were published in the Reddit discussion, Hololens 2 waveguides in the wild. Color banding is very close to in headset experience. today, I decided to write about it today.

I hope to get some much better pictures in the near future, but it could be after CES in January 2019. From the pictures available, you can tell the horrible color uniformity problems, but they are not of high quality and with the right input to show the resolution issues. I also hope to prove/show some of the other issues like flicker and the disappearing lines temporal artifact that will be discussed below.

I wrote this article in a hurry to be timely, so there may be more than the usual typos and grammatical errors.

Hololens 2 Problems are Now an Open Secret

It is now an open secret that Microsoft is having serious problems making the Hololens 2. There are threads on various groups discussing the problems. In addition to the thread with the pictures, the Hololens Sub-Reddit discussion “The elephant in the room: HL2 display” cites and Hololens 2 waveguides in the wild [edit: this second link was included in the photograph but was accidentally left out]. Color banding is very close to in headset experience. confirms the information presented this blog’s articles Hololens 2 is Likely Using Laser Beam Scanning Display: Bad Combined with Worse and Hololens 2 (HL2): “Scan Lines” Making Text Hard to Read and Quality Issues with Waveguides.

Prior to the Reddit threads, there was a Facebook thread reported in by this blog in Hololens 2 (HL2): “Scan Lines” Making Text Hard to Read and Quality Issues with Waveguides. That thread was quickly deleted after I publish, but I had already the content in the article. The new pictures confirm what the drawing showed.

The Pictures – Yes, It Really is This Bad

Below are the two pictures from the Reddit thread (the second one was originally published on Twitter). Both public and private information sources are telling me that the first picture is “typical” of what people see with the Hololens 2. There are units that are somewhat better and those that are considerably worse (as described on the Facebook post above). Yes, Microsoft is really shipping products this bad.  

When Smart People Do Something that Looks Dumb, the Alternative Seemed Worse

I’m often asked why otherwise smart people did something that looks dumb, and my response is “because doing the alternative seemed worse.”  In this case, Microsoft had to choose between shipping obviously defective devices or it becoming clear they could not ship.

Based on all the evidence, Microsoft felt pressure to announce the Hololens 2 (HL2) before they had figured out how to manufacture it. Worse yet, they were having not just one but at least two serious yield problems, namely the laser scanning engine and the butterfly waveguide (see Hololens 2 is Likely Using Laser Beam Scanning Display: Bad Combined with Worse).

Likely a Hiearchy of Customers and Binning

I would suspect there is a hierarchy where very important customers (ex., the US Army, and Toyota) are getting the best units. That “less important” companies are getting the units from the lower “bins.” These lower binned units may put up an image but have poorer image quality.

The yields are so low and the costs are so high that they don’t want to trash headsets were the laser display engine works and so are “functional.” Perhaps the waveguide problem is showing up late in the assembly process, say at the point of attachment to the rest of the optical assembly. Once it is glued together (as all optics like this must be glued together), there is no chance to rework. They are left to choose between shipping bad looking units that “function” or shipping nothing at all.

All this put together, I suspect that Microsoft is shipping a higher percentage to bad looking, but functional units to “lesser” companies. But it is also possible that a large percentage of units have serious image quality problems. We don’t currently have a large sample size of published images.  

Background Information on Hololens 2 quality issues

I’ve known for over a year that they were having serious problems with the laser scanning engine. It sounds like they may have even had some infant mortality problems, which is the worst kind of problem as it means units could go dead after glues expensive waveguides to them.

Another open secret that I have discussed many times is that Microsoft got the Laser Beam Scanning (LBS) technology from Microvision. Strangely in spite of this being an open secret, Microsoft keeps claiming to have invented the LBS engine (see  Hololens 2 Video with Microvision “Easter Egg” Plus Some Hololens and Magic Leap Rumors), in spite of claiming to have invented it, is using Microvision’s laser beam scanning technology.  

I have multiple sources that have told me that Microvision was having trouble making the fast scanning mirror. It sounds like Microsoft thought it could solve the problems, but apparently, the problems were bigger than Microsoft could readily solve.

Until recently, I was less aware that they were having serious yield/manufacturing problems with the diffractive waveguide as well.

Hololens Diffractive Waveguides, Bad or Really Bad

Even “good” diffractive waveguides are notoriously bad (it is baked into the physics) in terms of image uniformity with colors shifting across the field

The question with Hololens diffractive waveguides is not whether it produces a good image. Realistically, the best Hololens (1 or 2) looks very bad compared to the cheapest TV you could buy today. You would return a TV as defective if it looked at bad as the Hololens 1 or 2.

So we are already grading on a curve when it comes to Hololens’ image quality. What a “dumb consumer” would call a bad looking image is considered “good” by Hololens standards.  

As nothing looks very good, what is considered bad by Hololens’ “standards” becomes subjective. A person with multiple units can judge that some look worse than others.

Reports of Another Temporal Artifact, Disappearing Lines

I have reported for some time that the HL2 has flicker caused by too slow a refreshed rate, and I continue to get confirmation. Another temporal artifact issue has been reported, that of “disappearing lines.”

In talking with someone that has been using the HL2, they said that in addition to having problems reading the text, occasionally, every other line disappears. I believe this to be caused by the Interlaced scanning process combined with eye movement. The human visual system blanks out vision while the eyes are moving (proven in many studies). If the eyes happen to be moving when one of the interlaced fields is presented, the human visual system will see momentarily two of the same field and miss the lines from the other field.

I expect there are other artifacts yet to be reported cause by the Interlaced scanning process. As I first reported in Hololens 2 First Impressions: Good Ergonomics, But The LBS Resolution Math Fails!, the scanning rate of the Hololens 2 is too slow for the claimed resolution not to cause problems.

Some Good News for Hololens 2 – Ergonomically Great Compares to Magic Leap

The one comment I keep hearing and reading about the HL2 is that ergonomically, it is great compared to the Magic Leap One and better than the Hololens 1. There are developers that can look past the image quality problems and see uses for the Hololens 2. They particularly like how it can be worn with glasses and that they are fairly open.

Magic Leap has recently announced a pivot to the enterprise market. The problem is that the Magic Leap One has a terribly ergonomically design compared to even the Hololens 1 no less the better ergonomically designed HL2. Magic Leap went for steampunk styling and marketing hype and made a complete mess of utility. I don’t see people taking the Magic Leap One’s pivot to enterprise seriously.

Conclusions and What to Expect Going Forward

It is hard to say how long it will take for Microsoft to solve their manufacturing problems. Based on announcing it in February only to ship very limited and yet still pretty bad units 10 months later, they are obviously having much bigger problems than they expected.

I would expect that eventually, they will reduce the worse case problems with the waveguides. But they will still be diffractive waveguides and still have serious color uniformity problems as they did with the Hololens 1 or perhaps worse due to the wider FOV. By consumer standards, they will still look bad, but they could be useful for industrial purposes where image quality is less important.

The refresh rate is still too low by a factor of 2 to reduce flicker and a factor of 4 to both reduce flicker and support the resolution that Microsoft has claimed for the HL2. I think this will be a problem that will plague the product in some applications. There are some people that are very adversely affected by flicker (including causing headaches and nausea) and this problem is built-in by design.

I am also concerned about the long term effects of shining a laser beam into the eye for very long periods of time. The HL2 is claiming to display about 2560 x 1440 = ~3.7 million pixels (not all of which are visible at the same time due to digital IPD adjustment). Assuming they are displaying a 500 nits average image, it means at any instant in time the laser beam has ~3.7 million x 500 nits = ~1.8 Billion nits or brighter (since the spot is smaller than a pixel) than looking directly at the sun. The only thing that keeps it from blinding you is that the beam is rapidly moving.  I’m sure they have circuitry to shut down the laser if it stops (assuming it works 100% of the time), but that is the obvious problem. I have not seen any study on the long term effects of scanning a laser into the eye for many hours a day, 300 or so days a year. As an analogy, smoking a cigarette for a short period of time may not give you lung cancer; it is the long term effect that causes the problem.