304 North Cardinal St.
Dorchester Center, MA 02124
304 North Cardinal St.
Dorchester Center, MA 02124
Last week I was in San Jose for a quick trip, and it so happened that as I was about to leave, Digilens invited me to see their latest developments. I found several different aspects interesting, including their brightness, efficiency, modularity, manufacturing, and plans for plastic waveguides later this year.
DigiLens was founded in 1997 and has developed a reputation of being a company that lived on grants and funding but never seemed to make it into a volume product. However, many optical experts in AR have met over the years used to work for DigiLens. Perhaps the most well-known DigiLens alumnus is Microsoft HoloLens’s, Bernard Kress.
DigiLens has a new management team working to move Digilen’s technology from the R&D labs into products. It is probably worth taking a fresh look.
The meeting with DigiLens started with them greeting us outside with a demo on a sunny, nearly cloudless day. DigiLens wanted to show off their waveguide display against a bright sky. A blue sky on a cloudless day has about 3,000 to 7,000 nits (depending on the angle of the sun and other factors). So the picture on the right gives you some idea as to the brightness. The image being visibility green against the sky proves they were outputting about the >3,000 nits DigiLens claims.
The waveguide was their Crystal 30-G, 30-degree FOV green only, waveguide driven by a DLP microdisplay. As you can see, the display in portrait orientation with a 15° vertical and 26° horizontal FOV.
Below is a gallery with a few more pictures I shot with an Olympus D5iii with a 17mm lens (The images have all been cropped). I think they fairly represent what the eye sees looking through the waveguide.
The pictures below show the waveguide shot from the outside (left) and the inside (right). Even though the projector is outputting about >3,000 nits, there is relatively little front projection (Digilens says less than 11%). Also, notice the transparency of the waveguide, which DigiLens specs as being 95% transparent, and the picture shows it to be very transparent.
The waveguide is made on Gorilla-Glass and survived my accidental drop test onto a metal mesh seat (seen right) as I was holding it for a picture of the waveguide.
As you can see in the top view (below) of the waveguide while being driven, the waveguide physically independent of the optics driving it. This decoupling can be advantageous as it supports the waveguide being easily replaced independently from the rest of the optics. You might also notice the path of light from the DLP projector to the waveguide in this view that there appears to be a substantial amount of “dump light” (unused light) after the waveguide entrance. I suspect the amount of light is due to the DLP projector outputting unpolarized light and the waveguide only using polarized light.
As shown earlier with the outside view of the waveguide above, DigiLens has a significantly less front projection (eye glow) of light than other diffractive waveguides. The comparison below is taken from a DigiLens presentation. Most diffractive waveguides project between 50% and 100% of the light forward they project toward the eye. Not only is this inefficient, but it gives the glowing eyes cyborg effect.
They use laser illumination of an LCOS display on the far right example above, what DigiLens calls a Waveguide Integrated Laser Display (WILD). To get the front projection down to 1%, they use a “notch filter” coating tuned to the wavelength of the laser.
I want to note that while the forward projection is very good for a diffractive waveguide, Lumus expects to have only 1% forward projection with their Maximus reflective-base waveguide as discussed in this blog’s Exclusive Article on the Maximus.
We moved indoors for the demonstration of the Crystal30 color. While a similar view, it is now on the other side of a heavily tinted (likely >80%) window.
The Crystal30 uses two waveguide layers, one Red-Green and the other Green-Blue. As you can see in the picture, the color uniformity falls off significantly toward the top and bottom ends of the image.
Even at 30°, color uniformity is a weak point for DigiLen’s waveguides. Still, it is not as bad as the Hololens 2, which may say something about the importance, or lack thereof, of image quality in some enterprise AR applications. Lumus has the best waveguide uniformity I have seen, and Hololens about the worst (see below), and both of these have a much wider FOV (a little over 50°).
DigiLens has developed a novel development platform they call the Design v1. I got a chance to see the same demo shown in their true through-the-lens video. I added the word “true” because so many companies fake these videos.
The v1 headset was configured with modular dual 1280×720 DLP projectors and snap-in interchangeable dual Crystal50 waveguides. The Crystal50 uses 3 waveguide lasers (red, green, and blue). With the 3 layers of waveguides, the light throughput with the Crystal50 spec’ed at >80%.
The v1 is a development vehicle that will let designers try out different configurations without designing everything. It is not a product concept, although it does demonstrate the ability to have interchangeable/replaceable waveguides, something I have not seen before from any other company. Some technologies may also be amenable to replaceable waveguides, whereas others require the projector, optics, and waveguides to be permanently glued together.
The ability to replace the waveguide can be extremely important for some applications, particularly those in rugged use where lenses can get scratched or broken. It could also lead to some unique applications such as what happened with Mira for their use in theme parks (see: Mira – From <$20 Fishbowl To New Mario Kart Ride at Universal Japan (In About 4 years).
OK, this section is a bit of a rant because when I ask a simple question, I receive complicated answers requiring many assumptions. This is not a criticism of DigiLens or any other company, and I think it has some to do with it being hard to give a simple, fair answer. It could sometimes also have to do with wanting to hide their not-so-good specs.
What counts the most to the end-user is “how many nits go to the eye for how many Watts-in” for a given FOV and eye box (both FOV and eye box are factors). So far, only Lumus has given me a simple answer (one watt of LED for >3,000 nits going to >4,500 nits soon). From this blog’s teardown of Nreal, I calculated that Nreal gets about 120 nits-out for about 0.85 Watts-in for a similar FOV. It appears that Lumus is getting about 30 times the nits/Watt of Nreal. I know Lumus is doing something better than the others because they are getting so many nits out of a small set of LEDs.
As I wrote in the Lumus Maximus article about efficiency, producing image light efficiently is extremely important to AR, and becomes critically important as brightness has to be very high for outdoor use. Unfortunately, measuring efficiency is extremely difficult. Another issue is that an image with a lot of haze (poor contrast) may be outputting light that “counts” in terms of nits, but is making the image harder to see. Haze should not only not be counted, but it really should count against the “net-nits.”
AR headsets are not in high enough volume to warrant the type of extensive, objective tests such as done for cell phones by Displaymate. Worse yet, there are more variables and things to test for in AR than for a cell phone display. Still worse, there are no uniformly agreed ways to test things and avoid “counting” things like a haze contributing to brightness. As I often say/write, this blog is about the only one calling out Hololens 2 on their falsehood about resolution, and that is simple to measure. So if companies will “lie” about easy-to-measure things, you can only imagine what they will do more complex specs.
DigiLens is claiming that the Crystal50 will have >350 nit/lumen for “polarized light” (and ~175 nits/lumen for unpolarized light) for a 50° FOV and 12x10mm eye box. WaveOptics has a spec of only 50 nits/lumen for a 56° waveguide and a 12mm x 7mm eye box. The WaveOptis system uses an unpolarized DLP, so maybe they are about 1/2 as efficient as DigiLens with the same assumptions (very approximate). We can see that the DigiLens does a much better job of steering light to the eye at the exit grating, as evidenced by the lack of eye-glow (see above), so it looks like they have at least one significant efficiency advantage over WaveOptics. Lumus is saying they will have >650nits/lumen (polarized nits?) with the Maximus. But as I caution, these are unverified numbers, and there are likely different assumptions in each of them.
It looks like DigiLens is more efficient than WaveOptics by about 2X to 4X, and Lumus is more efficient than DigiLens by another 2X to 4X. Sorry, but very rough guesses are about the best I can do with the available information.
Waveguide developers have been telling me they will have plastic waveguides for many years. I have even seen an R&D sample or two, but never a projection device.
Think waveguides require the image light to make massive numbers of TIR bounces. Any imperfections in the surface of the waveguide are multiplicative. Also, any stress in the medium will cause birefringence that can disturb the image. Thus the waveguide has to be a very high-quality medium, with highly parallel and nearly perfect surfaces. So far, nobody has achieved a waveguide that works well in plastic. Several companies tell me they see Mistubishi Chemicals as having developed what may be a suitable plastic for waveguides.
At least as far back as SID 2019 Display Week (Display Daily Article), DigiLens has been saying they are developing plastic waveguides. DigiLens is working with their partner and investor, Mistubishi Chemicals, to release a plastic waveguide this year.
Glass waveguides pose a safety concern for many industrial and military applications. The current solution is to surround the thin waveguides with thick plastic to protect both the waveguide and the user’s eye, adding bulk and weight. And still, there is the concern for the glass breaking if the waveguide is dropped onto a hard surface.
DigiLens manufacturing process has a much faster cycle time than surface relief gratings. Whereas it can take many days to manufacture a surface relief waveguide, DigiLens can go from glass to waveguide in one hour. This fast turnaround means that Digilens has less work in progress (WIP) that could be messed up if there is a problem in manufacturing. It also means they can run many more cycles to improve their process.
DigiLens also requires less sophisticated and expensive equipment. DigiLens plans for their customers to be able to have their own in-house manufacturing of the waveguides.
DigiLens fundamental waveguide formation method is photographic which can be changed almost at will. Most other waveguides use a microscopic printing or semiconductor like deposition to form a surface relief graiting. DigiLens can better tailor their waveguide characteristics for different requirements.
Still, the theoretical manufacturing advantages have yet to be proven. It is hard to judge all the manufacturing plusses and minuses until companies get into volume manufacturing.
There are many pros and cons to the DigiLens waveguide, and I have decided to summarize them in the lists below. First the pros, then the cons, and finally “My Takeaways,” what I think is most important.