Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Apple Vision Pro (Part 4) – Hypervision Pancake Optics Analysis
Introduction
Hypervision, a company making a name for itself by developing very wide field of view VR pancake optics, just released a short article analyzing the Apple Vision Pro’s pancake on their website titled, First Insights about Apple Vision Pro Optics. I found the article very interesting from a company that designs pancake optics. I will give a few highlights and key points from Hypervision’s article, but I recommend going to their website for more information.
Hypervision has demonstrated a single pancake 140° VR and an innovative 240° dual pancake per eye optical design. I will briefly discuss Hypervision’s designs after the Apple Vision Pro optics information.
Apple Vision Pro’s Pancake Optical Design
Hypervision’s article starts with a brief description of the basics of pancake optics (this blog also discussed how pancake optics work as part of the article Meta (aka Facebook) Cambria Electrically Controllable LC Lens for VAC?).
Hypervision points out that an important difference in the Apple Pancake optics shown in the WWDC 2023 video and other pancake optics, such as the Meta Quest Pro, is that the Quarter Waveplate (QWP) retarder 2, as shown above, must be curved. Hypervision shows both Meta (Facebook) and Apple patent applications showing pancake optics with a curved QWP. Below are Figs 8 and 9 from Apple’s patent application and Hypervision’s translation into some solid optics.
Hypervision’s Field of View Analysis
Hypervision has also made a detailed field-of-view analysis. They discuss how VR experts who have seen the AVP say they think the AVP FOV is about 110°. Hypervision’s analysis suggests APV’s FOV “wishfully” could be as high as 120°. Either value is probably within the margin of error due to assumptions. Below is a set of diagrams from Hypervisions analysis.
Pixels Per Degree (ppd)
Hypervision’s analysis shows 34 pixels per Degree (ppd) on the lower end. The lower PPD comes from Hypervision’s slightly wider FOV calculations. Hypervision notes that this calculation is rough and may vary across the field of view as the optics may have some non-linear magnification.
I have roughly measured the Meta Quest Pro’s (MQP) ppd in the center and come up with about 22 ppd. Adjusting for about 1.8X more pixels linearly and the difference of 106 FOV for the MQP, and 110 for the AVP results, I get an estimate of about 39 ppd. Once again, with my estimate, there are a lot of assumptions. Considering everything, depending on the combination of high and low estimates, the AVP has between 34 ppd and 39 ppd.
Eye Box
Hypervision makes the point that due to using a smaller pixels size that thus requires more magnification, the eye box (and thus the sweet spot) of the AVP is likely to be smaller than some other headsets that use pancake optics with LCDs rather than the AVP’s use of Micro-OLEDs.
Hypervision
Hypervision clearly has some serious optical design knowledge. I first saw them in 2022, but as their optics have been aimed at VR, I have not previously written about them. But at AR/VR/MR 2023, they showed a vastly improved optical quality design using pancake optics to support 140° with a single pancake optics and 240° with what I call a dual pancake (per eye) design. I took more notice of pancake optics becoming all the rage in VR headsets with MR passthrough.
AR/VR/MR 2022 with Dual Fused Fresnel Lenses and 270°
I first saw Hypervision at AR/VR/MR in January 2022. At the time, they were demonstrating a 270° headset based on what I call a fused dual Fresnel optical design using two LCDs. I took some pictures (below), but I was not covering much about VR at the time unless it was related to passthrough mixed reality. While the field of view was very impressive, there were the usual problems with Fresnel optics and the seam between the dual Fresnel lenses was pretty evident.
AR/VR/MR 2023 Pancake Optics
Below are pictures I took at AR/VR/MR 2023 of Hypervision’s 140° single pancake and 240° dual pancake designs. The pancake designs were optically much better than their earlier Fresnel-based designs. The “seam” with the dual pancakes seemed barely noticeable (Brad Lynch also reported a barely invisible seam in his video). Hypervision has some serious optical design expertise.
I mentioned Brad Lynch of SadlyItsBradley and who covers VR in more detail about Hypervision. Brad had the chance to see them at Display Week 2023 and recorded a video discussing them. Brad said that multiple companies, including Lynx, were impressed by Hypervision.
Closing
Hypervision is a company with impressive optical design expertise, and they demonstrated that they understand pancake optics with their designs. I appreciate that they contacted me to let me know they had analyzed the Apple Vision Pro. It is one thing for me, with an MSEE who picked up some optics through my industry exposure, to try and figure out what is going on with a given optical design; it is something else to have the analysis from a company that has designed that type of optics. So once again, I would recommend reading the whole article on Hypervision’s site.
What is your take comparing AVP to Varjo current offerings?
Its very hard to compare without a massive side-by-side effort, but a few random thoughts:
1) In the long run I don’t see a future in foveated displays over just making the display higher resolution. I think Varjo has recognized this with their Aero product.
2) Varjo has a much better directed effort with support for their product category. But Apple will get massive more developers.
3) Varjo separate compute platform and “light houses” seems to be the way to go for the high end part of the market they are addressing.
4) Assuming Apple will support external computing and light houses, it could be that someday Varjo could develop software for it.
Super interesting post (as usual!). I’ve tried using a range of consumer HMDs for traditional 2D work tasks from time to time (browser-based email, docs, etc.) and it was always a struggle.
During your AWE talk, and in part 1 of this series of posts, you mentioned the need at for at least 40 pixels per degree for the replacement monitor use case. Do you have supporting published reference for that? Or perhaps that based on experience trying a number of devices over the years?
Good question and the answer is complicated. I’m working on an article discussing this very issue.
The 40 ppd is based on “typical” 1080p to 1440p monitors view at 0.5 meters. Additionally, it includes my personal experience My main monitor is a 22:9 34-inch 1440 line monitor which works out to about 40ppd at from 0.5 meters. Standard ergonomic guides suggest sitting 0.5 to 1 meter away, but I think most people sit close enough to touch the screen which is going to be about 0.5 to 0.6 meters. Personally, I can’t use/read my monitor from 1 meter away. If I concentrate and lean in a bit, I can see the pixel grid (I don’t have the best eyesight even with glasses – I think I correct to about 20/25 with glasses).
From a practical matter, with antialiasing and what Microsoft calls “ClearType” that varies intensity with the color subpixels, it gets pretty hard to tell for most applications at 40ppd. The 60ppd, what Apple called “retinal resolution” has been a pretty common print standard of 300 pixels per inch at “reading distance” (about 12-inches). If you have extremely high contrast edges, humans with the best vision can tell the difference up to 80 to 120 ppd if you put of a black and white diagonal line with no antialiasing (intensity edge smoothing/blurring).
So the 40 ppd is not a hard and fast rule, but when you get much under 40 ppd, people with even moderately good eyesight will start seeing screen door effects. When considering optical AR, with all the issues with whatever you are looking at being the the “black,” 40 ppd is good enough. With VR with a good black level/contrast, 40 ppd might be on the edge and you might want a bit more.
Karl
Thanks so much for that additional context and detail. It’s good to have a “typical” desktop setup as a reference. An article on such a topic would be very interesting for a lot of readers, since “multiple virtual displays for traditional productivity tasks” is a use case suggested in the marketing materials for several HMDs.
I look forward to trying the Vision Pro at some point. I’ve also tried some AR HMDs for work purposes (using browser-based productivity apps), and an additional challenge for me was related to focal plane, e.g., on HoloLens the virtual content was typically further away (> 1m), and in my normal workflow I often switch between looking at a virtual display and realworld content on my desk (notebook, sketches, etc.)