Eyestrain, Software IPD experiment

Pimax 8K Series
1

For this experiment I’ve used a picture posted by @risa2000

pimax_default
pimax_default2560×1440 23.8 KB

I’ve added to semi transparent squares, one blue and one red, they are right in the middle of the picture. An a vertical green line, right in the middle too.

First replace it as “pimax_default.jpg” in C:\Program Files\Pimax\Runtime\resource and restart service

For the experiment, (I have 64 IPD) I start looking at one point in my room at about 2m distance, and with software IPD at 0, I put the headset on, and I see 4 colored squares instead of two, and two green lines instead of two. Very bad.

I repeat the process, take off the headset I set the software IPD at -1,5, look at 2meters distance, and I put the headset on again, now I see three colored squares.

I repeat the process with software IPD at -3, when I put the headset on I see two colored squares, and one green line, but still not perfect.

Finally, with software IPD at -3.5 the picture is perfect, two squares, one green line and the numbers are perfectly aligned, and theoretically my eyes are aligned as if they were in the real world so it shouldn’t give me eyestrain, right?. (I’ve never tried such a lower software IPD, and I always set the physical IPD a bit lower than mine) I start a game and yes, everything look right.

What concerns me though, is that the green line should be right in the center of the picture, but it’s not. It’s displaced two squares to the left, I mean, if I close the right eye I can count 13 squares from the green line to the center of the lens, and if I close the left eye I can count only 11 squares. I can clearly see the line displaced to the left. So the whole world is not well aligned.
I thought it was my eyes that were not symmetrical in my face, but my wife have tried it and it’s the same for her.
I’ve tried it in virtual desktop attaching the screen to the hmd with the picture in full screen, and its the same there, so it’s not a problem of Pitool default picture. The view of both screens are not well aligned.

So, maybe my unit is bad calibrated, maybe it’s our eyes or maybe all the screens are bad calibrated. Is it possible that it’s causing eyestrain? Maybe some of you can try it and give some feedback.

Would be nice if in new Pitool was implemented a screen horizontal offset value per eye, then with a simple test like this we could get the world perfectly aligned, and without eyestrain. Hopefully

EDIT; This test is useful to check if your screens are well aligned, the green line must be right in the middle of your view.
You can check it by closing one eye, and counting the number of squares from the line to the center of the lens, both eyes must see the same number of squares.
If it is, better use a 3d object to calibrate the IPD

5 Likes
2

Thx, will test it and se how it works for me.

1 Like
3

Shouldn’t the distance be around 3m instead?

1 Like
4

@Mlkconcept I am not sure this method is a good one for the IPD calibration as I have a suspicion that the static image projection is not correct (see the Observation 1 and Observation 2 paragraphs on the original post here - https://community.openmr.ai/t/pimax-native-projection-and-pre-lens-warp-transformation/15775).

In short, what I believe is wrong about this is that the images are projected as if you were focusing on an infinite distance, while they should be projected at some (comfortable) finite distance (e.g. 1-3 m). Now what you are doing is compensating for this infinite projection by moving the projection points closer. For the 2D image it should not be a problem as there you only match the image at the same distance, but for the 3D images in a normal application this would probably mess up the “angular verity” of the scene.

But, if you feel comfortable with the new setting then I guess I cannot object.

On the other hand, I have checked this pattern on my 5k+ and even if I adjust the IPD offset, I have the green line in the middle i.e. when I count the boxes in one or the other eye from the green line to the inner (nose) edge, the counts are same for both eyes. So your observation is indeed strange.

7 Likes
5

Hard to say. We’ve asked Pimax for the “focal distance”, but only got the distance to the lenses.

To me, this sounds like it might be a bug, which could be why some people have so much trouble getting a good experience. I agree with the OG, the squares should be centered. Perhaps the offset is only being set for 1 eye.

@Sean.Huang, would you please have someone verify that software IPD is working as expected? Thanks!

1 Like
6

It seems to be working on my 5k+ as expected, i.e. symmetrically reducing (or enlarging) the eye offset of each eye.

3 Likes
7

I’ll try it on my 8K later tonight.

What version of PiTool are you running? (It might matter.) I’m using 180.

1 Like
8

I am running PiTool 1.0.1.132.

1 Like
9

I’m using 180 too
With this new settings everything looks right, the scale of the world and the objects seems more accurated.
I don’t know if it’s better for the eyestrain though, because I have the eyes tired of all the tests. I’ll try again tomorrow

3 Likes
10

Today I was working with my IPD and I understand how IPD offset works…

When you move IPD offset you change the “virtual distance” between the nearest and the farthest object.

Example. I was using Red Matter, which has an incredible graphics and very realistic in measures. And in one room I see that between the nearest and farthest object in the room there are 6 meters for me eyes. But it isn’t real. It must be 20 meters.

Then I change my IPD Offset to 2.0 and then changed again the IPD to my near IPD again (66.9). And now the objects has a real distance between them.

Now, I close my eyes. I relax them… and open again. And I see everything without eyestrain because my mind try to find the real distance and logical measure in the virtual room.

That the reason because we have eyestrain. Is Not the lens. Is about a bad Vr calibration of the virtual distances.

Happily we have a lot of tools in Pitool to correct this. But it need a lot of work. And seriously, the best way is trying with a realistic VR game and trying to see it like the real world. Quickly you will see that the objects aren’t in the correct place or distance between them. Sometimes you will see the background objects are too close, or far. Then you need to adjust the IPD offset to calibrate it, and compensate the distance.

2 Likes
11

One more thing. My thoughts is that we can’t calibrate it with an image. Because the measure changes in some games, and between Oculus and Steam VR games too.

Sometimes they use different “world scale”, that is the name of the option in Unreal Engine and Unity for this problem. That the reason why some games has a bad scale for the objects.

1 Like
12

When you change an IPD offset you basically change the IPD for what the rendering concerns, which means changing the distance between the two virtual cameras shooting the virtual world. In other words you may perceive it as a change of the distance, but this is not correct, because what it actually does is changing the angles at which you observe the objects. (https://community.openmr.ai/t/pitool-ipd-offset-how-does-it-work/22344).

13

Yes, that is the reason why the “depth” changes and the distance between object changes.

Is the same that when you changes 3D in the 3DS. Is like you adjust the “depth” of the 3D vision.

In my case it was wrong by default in Pimax. I needed to change it to be correct.

1 Like
14

I discovered the same thing. I set the hardware ipd to my normal ipd(also 67) and then adjust the ipd offset for the correct depht scaling, also ended up with +2. I find skyrim the best to set the scaling wich is natural for me. A huge step up for immersion!

3 Likes
15

Me too. The software IPD created a realistic 3D effect I hadn’t experienced before and it made me yell out in excitement, it was that much of a game changer for me. So the conclusion was that Pimax got IPD all wrong and maybe right for a select number of people only

3 Likes
16

Are we expecting the eye tracking module to, at minimum, automatically adjust the software IPD once HW IPD is dialed in correctly?

In all honesty, this very issue is what is keeping me from using my 5K+ over my Index and I don’t have the Patience to dial it in manually.

1 Like
17

I have +2 too!! It will be good if we create a Poll hehe

1 Like
18

You mean this one:

https://community.openmr.ai/t/ipd-settings-poll

19

Well… is not the same, but it’s possible it can be related the difference a bit. But I think that it’s not linear. It’s exponential. For instance, with IPD offset 0 I used 65.5, and with offset 2 I use 66.9 (my real near IPD).

1 Like
20

Strictly speaking, if you dialed the hardware IPD correctly, there should not be any need for an additional software adjustment. That said, I see two potential sources of problems:

  1. Pimax does not calibrate the view geometry according to the optics, so if the panels are misplaced or simply have some matrix misplacement themselves, or if the lenses are not coherent, there is no way to fix it. The IPD adjustment can only mask it somehow, but one would need full access to the view geometry configuration to do it properly.

  2. Pimax does not say which model they used to define the lens corrections and the geometry, so they may have used particular eye depth which is far off for some and since their canted panel design is sensitive to eye depth as well it may cause problems too.

I am afraid that all those “success stories” about how an IPD offset solved the issue are just an anecdotal examples of having an arbitrary side effect (of an IPD offset) countering the real problem in the headset and generalizing them won’t really help much (and may even be counter productive).

9 Likes