Improve and clarify light units #109404

New Issue

Brecht Van Lommel · 2023-06-27T11:57:03+02:00

Brecht Van Lommel commented

2023-06-27 11:57:03 +02:00

Fix wrong conversion from power to radiance of area lights #109153
Cycles: replace spot light disk sampling with sphere sampling #109329
Cycles/EEVEE: change point light to double-sided sphere light #108506
Cycles: Change sun lamp to have uniform intensity at high angles #108996
USD: improve light units conversion #109795

	Normalized	To PRMan	To Karma	Non-Normalized	To PRMan	To Karma
Point Light	`W`	N/A	`1/(4pi)`	`Wsr^{-1}`	N/A	`1/4` (?)
Sphere Light	`W`	`1/pi`	`1/pi`	`Wm^{-2}sr^{-1}`	`1/pi`	`1/pi`
Spot Light	`W`	`1/pi`	`1/pi`	`Wm^{-2}sr^{-1}`	`1/pi`	`1/pi`
Rectangular Area Light	`W`	`1/pi`	`1/pi`	`Wm^{-2}sr^{-1}`	`1/pi`	`1/pi`
Disk Area Light	`W`	`1/pi`	`1/pi`	`Wm^{-2}sr^{-1}`	`1/pi`	`1/pi`
Distant Light	`W/m^2`	`(4/pi) * 1/(2(1-cos(0.53180/pi/2)))`	`1/4` (?)	`Wm^{-2}sr^{-1}`	TBD	TBD

There is no point light in PRMan, only lights with some area. In Karma there is a discontinuity between normalized for a point light and a very small sphere light. Unclear how to handle this.
The normalized distant light in Renderman strangely uses the default sun angle of 0.53°. The Karma units are unclear to me. The Cycles non-normalized units are probably wrong.

Note this table assumes #109153 and #109795 to be included.

* Fix wrong conversion from power to radiance of area lights #109153 * Cycles: replace spot light disk sampling with sphere sampling #109329 * Cycles/EEVEE: change point light to double-sided sphere light #108506 * Cycles: Change sun lamp to have uniform intensity at high angles #108996 * USD: improve light units conversion #109795 | | Normalized | To PRMan | To Karma | Non-Normalized | To PRMan | To Karma | |---|---|---|---|---|--|--| | Point Light | \(W\) | N/A | \(1/(4pi)\) | \(Wsr^{-1}\) | N/A | \(1/4\) (?) | | Sphere Light | \(W\) | \(1/pi\) | \(1/pi\) | \(Wm^{-2}sr^{-1}\) | \(1/pi\) | \(1/pi\) | | Spot Light | \(W\) | \(1/pi\) | \(1/pi\) | \(Wm^{-2}sr^{-1}\) | \(1/pi\) | \(1/pi\) | | Rectangular Area Light | \(W\) | \(1/pi\) | \(1/pi\) | \(Wm^{-2}sr^{-1}\) | \(1/pi\) | \(1/pi\) | | Disk Area Light | \(W\) | \(1/pi\) | \(1/pi\) | \(Wm^{-2}sr^{-1}\) | \(1/pi\) | \(1/pi\) | | Distant Light | \(W/m^2\) | \((4/pi) * 1/(2*(1-cos(0.53*180/pi/2)))\) | \(1/4\) (?) | \(Wm^{-2}sr^{-1}\) | TBD | TBD | * There is no point light in PRMan, only lights with some area. In Karma there is a discontinuity between normalized for a point light and a very small sphere light. Unclear how to handle this. * The normalized distant light in Renderman strangely [uses the default sun angle of 0.53°](https://www.sidefx.com/forum/topic/73880/?page=1#post-312206). The Karma units are unclear to me. The Cycles non-normalized units are probably wrong. Note this table assumes #109153 and #109795 to be included.

Brecht Van Lommel added the

Type

To Do

label 2023-06-27 11:57:03 +02:00

Weizhen Huang commented

2023-06-27 12:07:30 +02:00

Point Light third column should be radiant intensity? It is divide by the solid angle of a sphere so (Wsr^{-1})

		intern/cycles/scene/light.cpp
		Line 1221 in 016d80ff6f
	
				      float eval_fac = (radius > 0) ? invarea * M_1_PI_F : 0.25f * M_1_PI_F;

/* Convert radiant flux to radiance or radiant intensity. */
float eval_fac = (radius > 0) ? invarea * M_1_PI_F : 0.25f * M_1_PI_F;

Point Light third column should be radiant intensity? It is divide by the solid angle of a sphere so \(Wsr^{-1}\) https://projects.blender.org/blender/blender/src/commit/016d80ff6f780b597e6ec5df65ce650aeb6a5a44/intern/cycles/scene/light.cpp#L1221 ``` /* Convert radiant flux to radiance or radiant intensity. */ float eval_fac = (radius > 0) ? invarea * M_1_PI_F : 0.25f * M_1_PI_F; ```

YimingWu referenced this issue

2023-06-29 11:47:28 +02:00

Regression: Lighting difference between 3.5 and 3.6 with Principled BSDF #109494

Brecht Van Lommel added the