110 degrees Fov stretched to 200 degrees? EDIT: All the games render in a native wide FOV through the engine

The projection is correct, but the tested review claimed there was warping at the edges. Is it possible to take a photograph of the full lenses displaying an image to see how it looks in reality?

I have told @PimaxVR to get in touch with Doc-Ok to get this fixed.

What is it @deletedpimaxrep1? Do you guys not trust that we want to help you?

All the tested guys recognized warping. This needs to be fixed, otherwise that huge FOV goes to waste.

1 Like

If the game is unaware of the angle between the 2 display panels, then the projection is very incorrect and would cause overstretching. Compare the length of the 2 yellow lines in the pic I posted above.

This would explain what Tested described as stretching.

I don’t know if this is the cause though, maybe there’s an angle setting in the API somewhere. Still the game would need to process that. Maybe this is solved by SMP ? Would love the hear what Pimax says @yangjiudan

@PimaxVR @deletedpimaxrep1 Could we get a video taken through the lens of the prototype?

@PimaxVR Robin Weng Please pay attention to this thread

yes, that’s the point, most games built on openvr sdk knows that you can find the API in the openv githup pages, the provide driver api and aplication api for that. and others games we have done some special hacks to make it right, and this I can’t tell much detail here.

2 Likes

SMP is used with Mult-Res shading tech in VR Application Rendering. it is for improve the performance of VR Application Rendering.

Ok so good to hear you guys are working on this. Would this maybe explain what the Tested guys experienced as ‘stretching’ (due to your ‘hack’) ? Or what’s your personal opinion, do you have an explanation for the fact they felt they were looking at stretched images ? Thanks for your input, very valuable !

An alternative explanation could maybe lie in the lens distortion parameters ? And then there’s the IPD issue that I could see as a difficult thing to get right too, since you guys are not moving the panels. So IPD then directly influences distortion.

1 Like

The issue with high FOV rendering you see here: Breaking the FOV limit in Quake - YouTube

The Pimax currently has the same issue, above ~140° FOV everything in the corners gets stretched and everything in the center gets smaller. Here is an explanation of the issue:

The solution for this is to render multiple views per eye. If the Pimax devs design the drivers and the lenses for just one planar projected view, it will always look badly stretched at the edges (that’s what the guys from Tested saw).

4 Likes

thanks for your understanding.
we already have done the hack, and make as many games as we know works good.
the only reason may cause the Tested guys experienced “stretching” is distortion issues, and the IPD mismatch is one reason, maybe they did not adjusted the IPD much well in a short time.

1 Like

So just to be sure, you guys ARE changing the distortion parameters according to the IPD setting, right ?

yes, we have algorithms to do that.

2 Likes

Difficult stuff to grasp. I kind of get what he’s saying. But look at that last picture, how would this then project onto 1 screen ? Or take the drawn picture that was posted before, how then would this look if you’d render it like a cylinder ?

Thanks, I’m really happy you guys are participating here. It takes away the worries to hear things like this :slight_smile:

Take a look at the video I linked. You see standard 90° projection, that looks fine. Then you see standard 120° which still looks fine. Standard 150° already looks a bit stretched, and standard 170° looks horrible. A bit later you see “Panini 200” where you see how a “correct” 200° FOV would look like, thats done with rendering multiple views. You can read more about it here: GitHub - shaunlebron/blinky: Exploring peripheral vision in games (using Quake)

Heres a comparison:

On the left you see what the Pimax headset is currently using.

The issue with the “correct” view is that you have to render a lot of different views per eye and thats more expensive to do:

6 Likes

4 Likes

Thanks ! I’m going to think some more about this. I have a feeling this indeed is the core of the problem.

@John_Alcatraz If you’re right (and I have a feeling you are), then we’re basically f*cked, right ? None of the games can render like this. It would need huge overhaul in game render engines. I highly doubt any game developer is interested in doing that, just because of a few thousand pimax 8k users out there.

EDIT although … in the ‘globe’ example, you’d just need 6 viewports and have the Pimax render those into the lens projection… I wonder if you can do that with SMP.

The games can render like this, its just more expensive. The important thing is that, and I’m not fully sure about that, the Pimax devs would have to design the lenses in a way that a correct image can be used. That would mean that the games would have to specifically support the correct rendering, so you couldn’t just play any existing SteamVR game with it.

The Pimax devs probably decided to do it the “wrong” way because they wanted to have all current games be compatible and they considered that more important than having a correctly looking game.

It’s not really an issue for game developers to support it since its just the engine that has to support it. Almost all games you see are either done with Unreal Engine or Unity, so someone could just release a “Pimax Plugin” for the engine (or the engine could just directly add support for it) and then all games made with that engine could just use it. The developer of the game has to add that though, so at the beginning most games would be incompatible.

1 Like

THIS IS WHAT IVE BEEN SAYING FOR SEVERAL POSTS NOW. @deletedpimaxrep1 @PimaxVR