Depth Perception for DCS Mirrors (VR)

[Magic Zach]

Quite straight forwards.  Mirrors for all modules (to my knowledge) do not have depth for VR players.  It gives the mirrors an appearance of being essentially a small flat-screen television in your cockpit.  Mirrors as they are now only compute the reflection per the center of the head, like TrackIR, when in VR it should render different for and to both eyes.  Ofc if you look in mirrors in reality, you know that you can see depth and distance in them.

 

Just a little request for VR, is all.

[sirrah]

51 minutes ago, Magic Zach said:

Quite straight forwards.  Mirrors for all modules (to my knowledge) do not have depth for VR players.  It gives the mirrors an appearance of being essentially a small flat-screen television in your cockpit.  Mirrors as they are now only compute the reflection per the center of the head, like TrackIR, when in VR it should render different for and to both eyes.  Ofc if you look in mirrors in reality, you know that you can see depth and distance in them.

 

Just a little request for VR, is all.

As much as I agree and would love to see the mirrors properly implemented, I fear it could have high impact on performance

(that is however not based on any knowledge of programming, so maybe someone can take away that fear with a good argument)

[cfrag]

15 hours ago, Magic Zach said:

Quite straight forwards.  Mirrors for all modules (to my knowledge) do not have depth for VR players.  It gives the mirrors an appearance of being essentially a small flat-screen television in your cockpit.  Mirrors as they are now only compute the reflection per the center of the head, like TrackIR, when in VR it should render different for and to both eyes.  Ofc if you look in mirrors in reality, you know that you can see depth and distance in them.

 

Just a little request for VR, is all.

 

As someone who has written code to do this (mirrors in a VR game) let me tell you that 'real' (depth perception) mirrors are some of the hardest (performance-wise) things to do in non-raytraced (i.e. DCS and 99% of all other games that use the 'painter's algorithm' of z-ordering pixels that requires an order of magnitude less performance) environments. Usually, mirrors are simulated using a back-facing camera (meaning, yes, the entire scene is re-calculated for each mirror), and the resulting image looks convincing in 2D, but loses all depth perception and 'reflects strangely' in VR. This is because the mirror produces an image centered in a (horizontally inverted) camera that is situated behind the mirror, and it is only drawn once, not once per eye as required for VR. When you move your head to look at the mirror, the image you see isn't a reflection of the rays emanating from your eyes, to the mirror, and then into the world, but from the camera's position behind the mirror, and that camera sees the world from a slightly different angle than your eyes. That is why the image looks like a 2D photograph, you have that unsettling parallax error and the image moves so strangely when you shift your eyes (DCS seems to move the camera somewhat according to the HMD's position; that movement seems linked to the HMD's center, not your eye(s)). To save performance, each mirror is only calculated once (not once per eye), and the resulting image is simply simulated with an x-inverted camera, positioned at an interpolated position behind the mirror that skews your parallax angle (move your head closer to the mirror, and you will see that the reflected image of your helmet in the mirror recedes instead of getting bigger). Yet even in this heavily restricted form, mirrors are performance killers (as said before, they need to re-draw the entire scene per mirror). Switching to a raytrace engine will make mirrors trivial (they'd simply be reflective surfaces), but will incur a phenomenally higher performance tax on everything else (hence the 'RTX' moniker of some hardware vendor that promises to deliver a ray-trace image at the same fps you previously got from z-ordering.

 

So, that little request is understandable (and seconded, as I hate the way the mirrors work in VR too), but today's technical limitations prevents this until we get a raytraced engine. Even then, the question remains if we should blow all this performance on prettier mirrors, or stick with z-ordering to draw a better overall image and put up with shoddy mirrors. 

 

Edited August 18, 2021 by cfrag

[lefuneste01]

1 hour ago, cfrag said:

So, that little request is understandable (and seconded, as I hate the way the mirrors work in VR too), but today's technical limitations prevents this until we get a raytraced engine. Even then, the question remains if we should blow all this performance on prettier mirrors, or stick with z-ordering to draw a better overall image and put up with shoddy mirrors.

Hello, just for your information, mirror are working well in VR in another WWII sim, which is not based on a raytraced engine and provides globally better fps than  DCS...Even if not so usefull than in real life because of HMD resolution, I uses them in P47D razorback, hs129, P38, P39,...

So you may just add to your elements that they are limited to DCS...

Edited August 18, 2021 by lefuneste01

[cfrag]

11 minutes ago, lefuneste01 said:

Hello, just for your information, mirror are working well in VR in another WWII sim, which is not based on a raytraced engine and provides globally better fps than  DCS...

 

Does 'working well' really mean full 3d support with correct parallax? Or are those mirrors also flat? DCS mirrors work, but they aren't truly 3D, merely 2D images pasted on the mirror's surface. I'll check asap (if you are referring to that sim that also also has the German word for 'storm' as part of its name), as VR mirrors have been some obsession of mine ever since they posed such a challenge to me. Thanks for the tip !

 

 

[lefuneste01]

27 minutes ago, cfrag said:

 

Does 'working well' really mean full 3d support with correct parallax? Or are those mirrors also flat? DCS mirrors work, but they aren't truly 3D, merely 2D images pasted on the mirror's surface. I'll check asap (if you are referring to that sim that also also has the German word for 'storm' as part of its name), as VR mirrors have been some obsession of mine ever since they posed such a challenge to me. Thanks for the tip !

 

 

They are not flat, they provide depth perception, and the sim is a matter of great battles...

[cfrag]

It took me some time to try it out (I didn't fly that particular sim in ages), and indeed, those mirrors allow for depth perception! The mirrors (one per plane as far as I can tell, I only have three planes that allow for a mirror) are small, but they do allow you to correctly approach them (growing the reflected image correctly) and they dynamically re-calculate the mirror's FOV (larger the closer you are) on 'Complex' setting and they exhibit no parallax shift. That means that thissimwhosenameweshallnotmention does go the extra mile to correctly calculate individual mirror images for each eye, and the individual camera positions are correctly derived from the HMD's PD (i.e. they use each eye's position, and then reflect that off the mirror's surface to calculate the correct virtual camera position behind the mirror's surface before they render the scene)! There's also next to no perceptible scaling / quality artifact, meaning that the mirror texture isn't a (heavily) scaled representation of a small fixed-size render texture, it seemingly sets the quality dynamically from the mirror's size in your view. Very, very nice work indeed! Thanks for the headsup, @lefuneste01. 

 

If I have the time, I'll try and see how much performance these mirrors gobble up, and what the lesser mirror settings are.

Edited August 18, 2021 by cfrag

[CMDRennie]

On 8/18/2021 at 8:41 AM, lefuneste01 said:

They are not flat, they provide depth perception, and the sim is a matter of great battles...

 

I wish DCS would impliment mirror tech akin to that in IL-2, especially for those using VR or head tracking it is clear that the mirrors in DCS do not behave as thay are supposed to.

[Magic Zach]

(cough)

[Camble]

This is one of my biggest gripes with DCS at the moment, but it's not just a VR issue. It's just more apparent in VR and completely disorienting and immersion-breaking.

I'd say this is less about depth perception and more about the way mirrors are implemented in DCS. I presume this is because because originally, the pilot's head was stationary, but you can witness a similar perspective issue with the mirrors in 2D by moving the pilot's head.

And as others have said already, it's not impossible to implement without a performance impact. The existing mirrors are a throwback from Lock-On and desperately need replaced.

 

[Capt_Haddock]

Bump!

With all the new improvements to the engine it's surprising that the mirrors have been left untouched for years.
Just make them like IL2. You can tweak performance by deciding how much they draw in terms of distance and content, and they are properly 3D, with different images per eye. Worth a try in VR if you've never seen them.

[Camble]

Not exactly untouched, because work had to be done to get them working in MT. So why not address the issues then instead of bringing over an old implementation that isn't relevant for the current state of the engine? Fine if you never move your head, but almost everyone has 6DOF head tracking or VR these days. These mirrors are completely wrong and immersion breaking.

[draconus]

2 hours ago, Camble said:

These mirrors are completely wrong and immersion breaking.

Yeah, not consistent among the modules, obvious bugs as for the mirror scale and angles, no angle change with moving canopies...

At least we got adjustable mirrors in F-15E and F-14 recently.

[Gunfreak]

1 hour ago, draconus said:

Yeah, not consistent among the modules, obvious bugs as for the mirror scale and angles, no angle change with moving canopies...

At least we got adjustable mirrors in F-15E and F-14 recently.

 

[Camble]

Bump. This doesn't just affect VR.

[TORC]

Bump, because even though it doesn’t just affect VR, it badly affects VR!

[Raven (Elysian Angel)]

On 10/14/2023 at 2:52 PM, Camble said:

I'd say this is less about depth perception and more about the way mirrors are implemented in DCS.

I agree: it bothers me more that they work in reverse - leaning in zooms out, leaning back zooms in.

[average_pilot]

On 1/31/2024 at 12:25 AM, Raven (Elysian Angel) said:

I agree: it bothers me more that they work in reverse - leaning in zooms out, leaning back zooms in.

It bothers me even more that there is no zoom happening in real mirrors 

Edited February 12, 2024 by average_pilot

[Camble]

Bump again. New Fulcrum announced but still having to put up with janky mirrors.

[Qcumber]

Yes please. Flat mirrors breaks the immersion in VR. 

[Camble]

It would be nice if someone from ED at least acknowledged this is an issue.

[Hiob]

6 minutes ago, Camble said:

It would be nice if someone from ED at least acknowledged this is an issue.

I think the main problem with a proper mirror implementation is, that it will probably tank performance. I assume that is, why they rather not touch in right now.

I don't see how you wanna improve in a meaningful way (especially in VR) without proper 3D-Rendering for the mirrors.

[Camble]

19 hours ago, Hiob said:

I think the main problem with a proper mirror implementation is, that it will probably tank performance. I assume that is, why they rather not touch in right now.

I don't see how you wanna improve in a meaningful way (especially in VR) without proper 3D-Rendering for the mirrors.

The current implementation is more of a performance hit than a more modern implementation could be.

And nobody is expecting 3D mirrors, because mirrors don't work like that. See the video I linked above for an example of the problem.

Edited April 11, 2024 by Camble

[cfrag]

12 hours ago, Camble said:

And nobody is expecting 3D mirrors

Doesn't this thread's title

Quote

Depth Perception for DCS Mirrors (VR)

literally suggest the opposite?

[Hiob]

I'm a little confused as well. Why don't mirrors "work this way"?