Doc Ok new Blog Post about 8k


"This assumption is wrong in Pimax 8k’s case, which causes two problems. The big one is that SteamVR’s internal HMD representation does not support HMDs with a total field of view of 180° or more. Consider Figure 2: the FoV for each eye is formed by the triangle defined by that eye’s pupil position and the left and right edges of its associated screen. The closer each pupil is to its screen, the bigger the FoV. At the limit, when the pupil touches the screen, FoV approaches 180°. As there is no way to push the pupil through the screen, FoV can not be made any larger. If per-eye FoV is limited to less than 180°, and the two screens are parallel, combined FoV must also be less than 180°. This in turn means that SteamVR software can not render the full FoV required by the HMD’s screen layout, and since there is also no way for an HMD to tell SteamVR to only render to part of its screens, SteamVR will take the reduced FoV and stretch it across the real FoV, which causes visible distortion not only in the part of periphery that extends beyond 180°, but everywhere.

Second, even if Pimax 8k had a total FoV of less than 180° (or were able to tell SteamVR to ignore parts of its displays), and SteamVR were told to render to the correct FoV, there would still be distortion, specifically, keystone distortion, because SteamVR’s projection would be based on parallel screens, while Pimax 8k’s screens are rotated with respect to each other.

The proper way to fix both of these problems is for Valve to add screen rotating angle fields to SteamVR’s HMD specification protocol, so that SteamVR can set up correct projections. This would most probably not even require changes to existing SteamVR applications, as those are only receiving general projection matrices, which can already represent rotated screens.

Until then, there is a work-around Pimax could use to at least confine geometric distortion to the outer edges of the periphery, beyond 170° or so. I have been ignoring lenses and lens distortion thus far, but now they come in handy. SteamVR already has a very flexible way for HMDs to communicate the optical properties of their lenses, concretely, the formulae needed to pre-distort rendered images to correct for the lenses’ non-linear geometric distortions and chromatic aberrations, to the SteamVR run-time. The stretching resulting from FoV truncation, and the keystone distortion resulting from wrong screen orientation, while not in fact caused by the lenses, can still be baked into the lens distortion correction formulae and fixed that way."


1 Like

Let’s not start multiple threads about the FoV issue. Please keep it central in the current thread. Pimax already replied there. I’m closing this thread.