LightDirection on shape with 3D effect is rendered different than specified

Regina Henschel 151

Insert a circle, set a large (72pt for 5.04cm circle width) rounded bevel and use a projection without rotation. That way you can see the direction of the lightning. When you rotate the camera, you should mark the axes with arrows. Use a gray fill color and no outline on the shape. Choose for example "Flat", "Soft" or "TwoPoint" lightRig. Compare the directions you see with the directions defined in section 2.1.1274 in [MS-OI29500]. It seems to me, that the X and Y coordinates are exchanged in column "Light Directions". "Soft" has only one light. In "Flat" the second light is blueish, so you can identify it. "TwoPoint" is suitable too, because its first light is more intensive than the second one. Of cause you can examine the other lightRig too, but the single lights might be difficult to identify.

Hung-Chun Yu 976 Reputation points Microsoft Employee

2024-02-21T15:27:59.51+00:00

@Regina Henschel Thank you for contacting Protocol Support, one of our engineers will be contacting you shortly. Hung-Chun Yu Microsoft Protocol Support
Mike Bowen 1,276 Reputation points Microsoft Employee

2024-02-22T00:34:29.5233333+00:00

Hi Regina Henschel,

I will look into this. When you say " the X and Y coordinates are exchanged in column "Light Directions"", do you mean the X and Y coordinates are reversed for all the Light Rigs, not just the ones you listed?

Thanks,

Michael Bowen

Senior Escalation Engineer - Microsoft Open Specifications
Regina Henschel 151 Reputation points

2024-02-22T09:05:20.5833333+00:00

I have not tested all of the 27 rigs because those with three or four light sources are difficult to examine. Besides the already mentioned rigs I see the exchanged coordinates too in the contrasting, legacy3 and legacy4 rigs. (legacy1 and legacy2 have same coordinates in x and y and thus are not affected.) It is much easier for you to see in the source code, whether the coordinates are used the same way for all lights.

In the meantime I have calculated the other way round:
I start with a light direction L(Lx, Ly, Lz) as specified. Then I determine the angles lat (vertical) and lon (horizontal) for the camera movement that are needed so that the vector from camera position to origin is the same as the light direction and thus the spot area is in the center of the sphere. I get lat=arcsin(-Lx) and lon=atan2(Ly/cos(lat); -Lz/cos(lat))

Thereby the light direction is assumed to be normalized. So when reproducing my findings you need to normalize the legacy light directions. The other light directions are already normalized, at least they look so to me. To reproduce my findings you only need to take a light direction from the specification, calculate the angles and enter them in the UI. The spot should then be in the center of the sphere. I have tested 'brightRoom', 'glow', 'morning', 'soft', 'sunrise' and 'threePt' so far.

When I now calculate the light direction in the 3D-scene coordinate system (x-axis right, y-axis down, z-axis out) from these angles lat and lon as in my mail, then I get for the light direction V the coordinates V(-Ly, -Lx, Lz).

So this seems basically to be the problem, that the used coordinate system is not specified.

Thus a solution might be to mention in the specification, that the coordinates of the light directions are not in the 3D-scene coordinate system, but they can be transformed to the 3D-scene coordinate system with the above rule. That would mean, that the normalized light direction of the specification has to be multiplied with the matrix

/ 0 -1 0 \
| -1 0 0 |
\ 0 0 1 /

to get the light direction in the 3D-scene coordinate system.
Regina Henschel 151 Reputation points

2024-02-29T14:59:37.83+00:00

[deleted because same as my edited comment above]
Mike Bowen 1,276 Reputation points Microsoft Employee

2024-02-29T18:03:12.34+00:00
Thanks for the update Regina, Regina Henschel. I am working to confirm your findings and I will let you know what I find soon. Best regards,

Michael

1 answer

Mike Bowen 1,276 Reputation points Microsoft Employee

2024-03-04T16:51:55.13+00:00

Hi Regina Henschel,

After researching this, you are correct that the description for how to render 3d effects is incorrect. An upcoming release of the MS-OI29500 2.1.1274 Part 1 Section 20.1.5.9, lightRig (Light Rig) will state that the light rig directions should be rotated by 90 degrees. Thanks for finding this, we really appreciate it!

Best,

Michael
Please sign in to rate this answer.
Regina Henschel 151 Reputation points

2024-03-04T23:17:21.2533333+00:00

Hello Mike Bowen, thank you for looking at the problem.

I think, "rotate by 90 degrees" is not precise enough for a 3D coordinate system.

Ultimately, it comes down to the question of which coordinate system was used for the directions in [MS-OI29500] - v20231113 and what coordinate system is used for extruded shapes. Unfortunately, there is no common opinion on whether a 3D coordinate system is right-handed or left-handed and opinions also differ on the direction of the z-axis.

Kind regards, Regina

Mike Bowen 1,276 Reputation points Microsoft Employee

2024-03-14T19:25:16.2233333+00:00

Hi Regina,

I wanted to check in and give you an update on this. There will be an update in an upcoming release that adds a note explaining how to rotate the vectors in order to transform the values into a 3D-scene coordinate system.

Thanks for your help!

Michael Bowen Senior Escalation Engineer - Microsoft Open Specifications
Sign in to comment