Finally some content to broaden your own horizons. Thank you @Heliosurge I missed that here last year. Real informations used to be one of the strengths of this forum. Too bad that many qualified experts left the forum because a wall of nonsens.
1 Like
Well as mentioned the forum was much better prior to the original KS. During and after folks got stuck on support issues and the unfortunate somewhat consistent poor communication ie release dates etc…
It would be great if we could get back to those fine roots.
In the p4k days a lot of ppl complimented this forum as being one of the most engaged places discussing vr. Back then a lot if simmers had a very keep interest in pimax due the p4k.
The forum has been stuck under a dark cloud for too long.
1 Like
Now this I agree with! This is something one would expect any HMD maker to have done, when creating their distortion correction profiles, and performing calibrations.
Do read doc_ok’s papers, as suggested! Among them, he’s got some old pictures from his setting up a calibrated camera, and shooting test patterns in the original Rift CV1, and HTC Vive.
Allow me to postulate a few things:
-
The whole point of a VR headset, is to provide your eyes a head-tracked view, that is as close to 1:1 with your real-world view as possible. This makes it look to you as were you inside the virtual world.
-
Contrast the above against the purpose of a pair of binoculars, which is to let you see things far away – they blow up e.g. 5° to 45°, so that you can better make out that tiny thing on the horizon. Panning naturally becomes a really rapid optical flow.
-
If our eyes could comfortably focus on something only a few centimetres away, VR headsets would not have lenses; They would not be needed, and we could have stared directly at the display panels.
-
The distortion caused by the lenses is meticulously counteracted by the VR runtime, though warping the frame images equally much in the opposite direction, so that when they have gone through the lenses, and reach your retinas, they appear as is the lenses hadn’t been there at all, as far as distortion goes.
-
…but they do appear farther away, specifically so that each eye can focus on them, as previously mentioned. -Instead of a tiny screen just in front of your eyes, it looks like a huge screen something like two metres away (EDIT: focus-wise - not stereo vision wise), BUT, crucially, the relative scale of that screen is such at it fills out the full possible view of the physical display panel you can see through the lenses in the real world.
-
The HMD maker determines the FOV games are told to render for their HMD (these are the numbers Risa’s tool expose), and can do so arbitrarilý.
-
…but 99% of the time, they want to meet postulate 1: Something held 30° out to the side from the user’s eye should appear 30° out to the side in the virtual world, too. This is an immersive view - not a telescope one.
-
They want the view to encompass the farthest possible point on the display screen that you can see through the lenses, to maximise the FOV of their device, in real world angles, between your eye and the lens. Therefore the rendered FOV needs to be at least this large.
-
They do not, however, want to render anything past that point, because that is wasted work, on imagery that you can not see anyway, and that work could be better put to use rendering more detail, or acheiving shorter frame times.
-
…so optimally, the rendered FOV is made to match visible FOV.
-
You can live with some zooming and distortions, and even greater abberrations, as eyeglass wearers do daily, because our brains are highly adaptable, but even with neuroplasticity accounted for, unnecessary incorrect bits are less than desireable.
7 Likes
Haha just wanted to write this. The whole point of VR is indeed to trick your brain into thinking you’re looking at reality. The better the headset, the better it does that. It’s really THE goal of any VR headset. Rendering a FoV that does not match the actual FoV, will not result in a believable virtual image and thus will result in a bad headset.
Of course a manufacturer can render 200 FoV into a 100 Fov headset and risa’s tool will report 200, since it simply hooks the steamVR API so it just reports the FoV that the manufacturer tells SteamVR to render. But this will really be a horrible headset to use, nobody wants that. Because if you just want big FoV, why not simply render it into a monitor plane?
6 Likes
@philpw99 Everything was already said, so I will just add an analogy in the non-VR world. Imagine playing a FPS game on the normal monitor - for example Quake. This game engine renders the game world at fixed horizontal FOV (90° by default), so basically you see 90° of the virtual world (on the screen).
Now, regardless how close or far away you put your eye to/from the monitor, or regardless of the size of the monitor, the displayed image will still convey those 90° of the game world. Even if you could see it from an arbitrary distance and thus at an arbitrary angle. And if you put your eye close enough you may even see it at larger angle than 90°.
But, no matter what you do, there still will be only 90° of the game world. If you want to see more (at once), you need to change the game options to render larger FOV. This will pronounce more the geometrical distortion, but some pros do it anyway, to get better situational awareness.
This is what rendered FOV means, and why it is important to people - it defines how big portion of the virtual world we can see.
Now, as already explained by @jojon and @Djonko (and others), it is only half of the whole story, as in the VR headset (contrary to the flat monitor) the rendered image is further obscured and distorted by the lens. This is why it is important to also do the real world check and by using some other tool (e.g. ROV tool) to confirm the visible FOV.
As you noted, however, the real visible FOV is individual and alone would not tell the whole story either. So we combine both to judge the headset design. Because the rendered FOV is independent from the viewer it is a good benchmark and this is the reason why there is a database for them.
11 Likes
Totally agree with what you said, as in the previous post you taught me the correct way to calculate the FOV. I do appreciate that!
I kept raising this FOV issue, because I do want people here to understand the possibility that Pimax might have incorrectly calibrated the picture. They uses 103 degree FOV in 3D space, but after the lens distortion, it shows a 115 or 125 degree image.
Just like your example of 90 degree screen, the question is, should we judge a monitor of being inferior because a game is showing only 90 degree?
Basically what I said is not much different from you. Of course you said it in a much more professional and precise way, and I had to rely on the binocular analogy.
I am not trying to cover up or defend Pimax. It’s just from this forum thread I got an impression that people are thinking Risa2000 number is equal to the final visible FOV, and I have to step up and state the obvious. I guess now everyone is fully educated, so I rest my case.
Risa’s. Tool dies indeed equal the final possible FoV as requested by the hmd driver to render. That being said if the lenses and the rendered FoV do not properly align you will have distortions/artifacts in the visuals. ie the FoV seems to be stretched much like a pan&scan 4:3 TV image being stretched distorted to fit a Widescreen 16:9 screen.(This was the initial description of the 8k during the KS; which was explained by pimax that they were still refining the distortion profile iirc). It is possible they can adjust the requested rendered FoV and adjust the distortion profiles accordingly; since this is a beta test they may indeed adjust the rendered FoV profile. As we know atm there are 2 modes for the 42ppd lenses that are doing something like this for testing small and big? iirc was the ss where Kevin explained this setting was for testing while they refine things for the 42ppd lenses. 103 for the 35ppd lenses atm is a hard fact until the software/firmware is changed. This is what is called to render as the final potential max HFoV.
The fudging of FoV values is unfortunately common ie the Vive Pro2 the ceo clearly stated the horizontal FoV is 120 and iirc it ended up being something like 116?
Now as Ben from RoadToVR has stated there is not a clear standard on how from my understanding how FoV is being measured ny the manufacturer. They could for example be using optimal pupil to lense at a given ipd. It is also possible that manufacturing marketing departments are using tool like the FoV calculator. To which then yes the advertised specs are legit to the tool being used even though it doesn’t align with the software.
@PimaxUSA has endorsed Risa’s tool as being accurate and it aligns with @knob2001 's RoV test suite as used by the m1 testers. However the Marketing team is still measuring/advertising the large FoV as 170 wide even though tests confirm it to be 160 to iirc 165 perceived. However Knob2001 has said this test has a degree of subjectivity based on variances of the end user and other possible factors.
The one simple truth is diagonal FoV shouldn’t be used.
Vrgineers with there RoboVR hmds had a range for the Horizontal FoV iirc based on eye relief and possibly ipd? These were the 2 previous hmds prior to the Xtal series.
I don’t believe hmd manufacturers are necessarily trying to deceive end users deliberatively jist that department A is not likely using the same tools causing marketing boo boos. I am pretty sure HTC Ceo didn’t like finding out he was perhaps given incorrect information that made him & the company look bad. Cross communication does indeed play a problem in corporate structures even in today’s age of technology that should make things easier and often doesn’t. Due to a wide variety of issues.
Another factor is the Advertised FoV being calculated without taking HAM into consideration? As it could also bump the FoV values even though anything outside of the viewport is unseen.
I think the whole reason this thread even exists, is because everybody here understands that there’s always the software component that sets the rendered FoV. That’s why people want to hear from Pimax what’s going on. Even Pitool has an option to change the rendered FoV (small, large etc). Also the example of Varjo has been given, where they increased the rendered FoV with a firmware update, a while after release.
I think much more likely than what you propose (that the engineers simply made an error with incorrect scaling) I think that IF the optics are able to show more FoV it could have been the case that:
- They felt that the outer edges of the lens were not good enough, maybe the image was too blurry there or they had problems correcting the distortion there, so they just limited the FoV
- The outer edges might be invisible due to the gasket, so they felt it would only waste GPU power anyway (I think think is why Varjo initially rendered less FoV).
Anyway, this is why people say Pimax should step in and clarify. Maybe they can give two modes, just like with the other Pimax headsets, small/large FoV and people can decide for themselves which they want to use.
Personally I always think it’s better to explain than to just keep quiet about it and just watch people getting their pitchforks out.
2 Likes
Personally I am in this boat, or at least I hope that’s the case. I’m hoping that we might get a larger FOV enabled as the distortion is improved.
1 Like
Well with the swappable lenses they can also create a wider FoV as well. We just need to be patient and see what tge end result of the beta yields.
1 Like
A lot of talk for what exactly? If the FOV is really drastically less than advertised then most of the people at the roadshows would have complained about it. Actually most of them stated „around Index FOV“, some said even bigger than that. I do really think that it is a „non issue“, also because of the way Pimax did - not - respond. Maybe it is just me being overly optimistic, who knows. Nevertheless i will enjoy my Crystal (with Index-ish FOV) and most of you will, too.
3 Likes