When we set out to create the Index HMD, one of our primary goals was to improve the overall fidelity of the VR experience, including visuals, audio, ergonomics, tracking quality, and more. Clearly, providing a wide field of view is a critical part of visual fidelity: it increases immersion, makes wearing an HMD more comfortable, and, depending on the experience, can enhance satisfaction around gameplay and interaction.
Before we dive in, we should mention a couple of aspects of visual fidelity that are intertwined with FOV.
Angular pixel resolution. Measured in pixels per degree (ppd), angular resolution is a major factor in the sharpness and realism of your virtual world. In terms of HMD design, angular resolution is driven by the display resolution and the FOV. Unfortunately, providing a big FOV directly decreases the angular resolution as the available pixels are spread out over the large viewing area. Obviously this is a critical trade-off in HMD design as both visual clarity and FOV are important for great VR. The complete visual clarity story involves many factors beyond just pixels per degree, like subpixel layout, fill factor, optics, and even ergonomics. So that's a big topic for another day.
Display refresh rate and display illumination time ("persistence"). Some of the benefits of higher refresh rate are well understood in the Desktop PC space. But in VR, where the display is attached to your head, frequent update and lower persistence are both key to reducing motion blur. Similar to increasing resolution, reducing motion blur helps to improve the perceived sharpness of the system. But it also provides a couple of things that increasing pixels per degrees alone cannot: it improves the sense of physical permanence of virtual objects and at the same time improves the overall stability of the virtual environment. Due to the physiology of human vision, these qualities increase in importance as FOV increases.
We plan to cover these two in greater detail in future posts, but no discussion around FOV would be complete without mentioning them.
What is VR FOV?
The field of optics has a well-established set of terms and conventions around FOV. VR is a different animal, and in many ways has unconventional requirements, so we tend to use the term a bit differently. For non-VR optical systems, pupil* locations are fixed and panel or sensor sizes are the defining FOV limiter for a given lens. In those cases FOV is easily described in horizontal, vertical and diagonal terms. Those measurements are derived from the edges of the sensor or panel through the pupil of the optical system. However, in VR optical systems, the pupil is the combination of the human pupil location (which includes eye relief and IPD position), the HMD lens aperture (typically not circular for ergonomic reasons), focal length of the lens, display size, and binocular relationship between the two eyes. Thus measuring VR optical FOV gets a lot more complicated. (We use the term eye relief to refer to the distance between the front of the lens and the closest point of your eye, typically the front of your cornea.)
Every HMD has a maximum possible single-eye FOV determined by its design, independent of the user. This is what people commonly talk about, try to measure, etc. But from a VR product design standpoint we are primarily concerned about what each individual user actually sees. Surveying where HMD designs were when we started designing Index, we had observed that it was common for a user to get less (even much less) than the theoretical maximum FOV due to the fit of the headset and their individual facial geometry. For example, basic trigonometry will tell you that if your eye is too far back from the lens relative to the lens's diameter, the entire lens won't cover a very wide angle and you can’t possibly see a big FOV. In this lens-limited situation, for HMD designs like Index which aim to provide FOV higher than 90 degrees, even a single millimeter of excess eye relief reduces your FOV by about 3 degrees.
Wearing an Index HMD, you can directly see the high sensitivity of FOV to eye relief by rolling the eye relief knob all the way back and forth and observing the large effect on FOV from the seemingly small amount of travel. Facial geometry differences between individuals can fairly easily cause the eye relief distance to vary by +/- 6mm.
Understanding the geometry of the situation gets even more complex when you think about other factors that affect your eye’s position relative to the lens. Consider what happens when your eye is looking straight ahead, then begins rotating away, moving the pupil farther away from the lens and closer to one edge. Similarly, if the hardware IPD of the headset is not adjusted correctly, that can also limit FOV. One common example is if the hardware IPD is too small, it limits the FOV toward the outside.
Fit matters as well. You might adjust the headset a little tighter or looser, or have it positioned a little off-center. All of these subtract from FOV. More complexity is introduced with prescription eye wear, which also tends to optically change the effective eye relief value. Beyond the physical uncertainties, there are additional complications on the software side, such as the offset frustums of the projection matrix and the shape of the masked render target not being circular. Those impact FOV by making it asymmetric. Also, compositor panel masking, which is software that controls stray light and chromatic fringing, is dynamic and dependent on/responsive to the re-projection system... Meaning that FOV for any modern HMD is not even completely static from frame to frame.
And all of the above is still considering FOV for only a single eye. Considering binocular FOV adds another layer of complexity, introducing stereo overlap and pushing further into the realm of individual subjective perception.
In all of the above cases taken separately, the effect might be on the order of a millimeter or two, but taken together it means that A) you need significant margin and/or adjustment built into the headset design in order to deliver your FOV to the user, and B) it is difficult (really impossible) to make a single objective, quantitative measurement of the FOV that predicts what any individual user is actually going to get. That makes us hesitant to talk about FOV using a single number, as it has never been anything close to the clarifier we've wanted it to be. So let's expand on the first...
Designing For FOV
In order to make it possible for users on the “close“ side of eye relief (and users with glasses) to use a headset at all, headsets without good eye relief adjustment and careful attention to comfort have to be designed with the eye relief distance biased well outward. And that in turn leaves two unpleasant design options: either design the maximum FOV to be pretty low for everyone, or design the maximum to be reasonably high but leave many users with a truncated FOV and wasted angular resolution.
For the Index HMD, we instead kept the theoretical maximum FOV fairly close to the high end of previous-generation headsets and then focused on delivering the full FOV to every user. This was done with a combination of a number of design elements which together add up to a big difference in effective FOV and comfort:
First and foremost, we implemented physical eye relief and IPD adjustments in order to provide optimal nominal eye position, and thus maximal comfort and FOV, to as many users as we could. In addition to being easier to adjust, the Index eye relief mechanism allows the display assembly to sit closer to the eye compared to headsets of the previous generation. This means that much more of the complete image rendered by the GPU is delivered to the eyes of most users. The Valve Index HMD's physical design also sets up the optical subsystem to work as well as it can and helps constrain the trade-off space it must operate within.
Canted Eye Tubes:
Second, we canted each lens/display assembly by 5 degrees to optimize inner vs. outer FOV and also improve available interior IPD range. The first benefit of the canted approach is simple: It nudges a few more degrees of FOV towards the outer sides, at the expense of the inner sides of each eye where stereo overlap is at play. Stereo overlap is still vitally important, of course. The canting simply provides a way to keep angular resolution of the system high while still striving for the higher overall binocular FOV that we were hoping to provide.
The main downside of canting is that both the existing software content library and the field of GPU rendering hardware are all typically optimized for parallel eyes. Fortunately, this may be readily compensated for in software using the re-projection techniques we already depend on for maintaining a constant frame rate. We just need to do a tiny bit every frame.... This way, apps past, present, and future may continue rendering in parallel as they always have, and they will "just work" for HMDs with mild amounts of cant angles.
Third, we made the front surface of the lens much flatter. This allows the eye to comfortably get closer to the lens, particularly for people using eyewear. While this effect is on the order of a few millimeters, we've seen how every little bit helps. Also, while it is technically possible to approach this from the other direction by increasing the clear aperture, there's an obvious practical limit to the outer diameter of the lens that will fit into the HMD and still provide enough IPD range to give a wide distribution of users a good experience.
Beside the main three, there are a few other aspects of HMD design that impact FOV worth mentioning, and that we had to consider when designing Index.
- Lens Diameter: We maintained a large lens diameter of 50 millimeters so that the eye can maintain a comfortable distance and still receive a high, geometrically stable FOV. This has been our approach because smaller diameter lenses reduce the effective light aperture to the eye, and can quickly become a limiting factor for FOV.
- Edge to Edge Clarity: The new lens design for the Index allows for more even clarity across the whole of the optical field. If it isn't of high enough quality, additional FOV may not be nearly as beneficial as we'd like.
- Geometric Stability: As FOV increases, it becomes harder and harder to accommodate distortion and keep the image geometrically stable. Instability comes from many factors, but is most clearly manifested as a wobbliness in the world, where things that should appear solid instead undulate like gelatin as you move your head. We believe that providing geometric stability is a critical aspect to long term comfort and sustained growth of VR usage.
So there are dozens of considerations that impact FOV, and all must be designed together in order to maximize the actual, delivered FOV across the entire user base.
- The Index HMD maximizes FOV by allowing the lenses to sit much closer to your eyes, even with full gasket foam.
- The FOV rendered by the GPU for Index is similar to that of a Vive or Vive Pro, but more of that view is delivered to most users.
- Index’s careful design delivers a higher effective FOV to the individual user without sacrificing pixels per degree.
- The canted eye tubes effectively shift a small amount of horizontal FOV from the inside to the outside, making them more balanced.
- It's really hard to use a single number to effectively describe the FOV of an HMD.
*pupil is referring to the stop of the system while accounting for any refraction or reflection that may occur in the system. It can include the stop, entrance pupil, or exit pupil of the system.