Units

I am in the process of implementing Preetham's "A Practical Analytic Model for Daylight" into Radiance. I have some questions regarding Radiance units, which I understand are radiometric units. Here are some questions I came up with. My biggest confusion relates to luminous efficacy, and when to apply it.

1) If I have a spectral distribution of the sun in wavelength increments from 380 to 780, and apply cieresp.cal to obtain the XYZ CIE values, what units is Y in? When I convert to RGB using something like xyz_rgb.cal, should I divide by the luminous efficacy 179 or not? Reading Chapter 5 of Rendering with Radiance (pages 305-307), it seems I should just take the RGB values and use them.

2) If I have data in CIE XYZ, with Y in cd/m^2. Should I convert to RGB and divide by 179?

3) When using skygen, what units is the solar intensity in? What units does skyfunc return?

Thanks a lot for your time,
Alejandro

Hi Alejandro,

From: Alejandro Troccoli <[email protected]>
Date: July 15, 2004 4:56:17 PM PDT

I am in the process of implementing Preetham's "A Practical Analytic Model for Daylight" into Radiance. I have some questions regarding Radiance units, which I understand are radiometric units. Here are some questions I came up with. My biggest confusion relates to luminous efficacy, and when to apply it.

1) If I have a spectral distribution of the sun in wavelength increments from 380 to 780, and apply cieresp.cal to obtain the XYZ CIE values, what units is Y in?

It depends on what the spectral distribution represented. Oftentimes, these are only relative values, and you can't expect to get meaningful absolute units as a result. What you will have is a valid color, where CIE (x,y) chromaticity is computed by:

  x = X/(X + Y + Z)
  y = Y/(X + Y + Z)

Note that the absolute values drop out of the above formulae.

If on the other had, your units are watts/whatever, your result from cieresp.cal will be lumens/whatever. If you had radiance in watts/steradian/meter^2 on input, your Y output will be lumens/steradian/meter^2, or "nits."

When I convert to RGB using something like xyz_rgb.cal, should I divide by the luminous efficacy 179 or not? Reading Chapter 5 of Rendering with Radiance (pages 305-307), it seems I should just take the RGB values and use them.

Yes, you should divide by 179 to get back into RGB radiance color values. The pages 305-307 deal with reflectance, which is unitless and therefore has no conversion factor.

2) If I have data in CIE XYZ, with Y in cd/m^2. Should I convert to RGB and divide by 179?

Again, yes.

3) When using skygen, what units is the solar intensity in? What units does skyfunc return?

Do you mean "gensky?" If so, the solar radiance is in units of watts/sr/m^2 using the standard 179 value for luminous efficacy. Unless you know better, it is recommended to simply use this value, so it drops out of the calculation when you return to photometric units during the analysis phase at the end.

I hope this helps.
-Greg

A caution regarding gensky, luminous efficacy factors and spectral conversions.

The physical skydome is strongly blue-colored. This matters if you are attempting to represent the brightness of objects relative to sources of light with different spectra, and/or relative to room surfaces with different reflectance spectra.

Blue light is more efficacious, i.e., it both has more energy and appears brighter. The lumens per watt conversion factor of 179 is really only valid for pure white light (I think I've got that right?). If you look into the source code for gensky, you will see that a different lm/watt factor was used to derive the zenith brightness (203 ??). But this efficacy is not tied to the color of the skydome as modeled by the user in the input file in the skyfunc "glow" entity below:

skyfunc glow sky_glow

0 0

4 0.75 0.9 1.2 0

sky_glow source sky

0 0

4 0 0 1 180

I do not know your specific application, but if the appearance of the sky is not important for your simulations, you might consider changing the skydome glow color component to be equal R G B factors (1 1 1) so that you do not convolve different efficacy factors and spectral influences on your final result. Alternatively, you could re-write gensky so that it adjusts the luminance of the skydome (by adjusting the lm/watt factor) in response to the color of the skyglow primitives.

In a quick study of this effect several years ago, we found errors as much as 20%--which is small potatoes for daylight simulation purposes, but troubling nevertheless because it is (arguably) an internal consistency error, not a modeling error.

The real problem is that the international unit of luminous flux is defined in terms of a human perceptual factor rather than on a radiometric unit, imho. This results in skydome brightness measurements in candellas per meter squared (instead of watts/meter squared) whose conversion factor depends upon the color of the light source.

-Chas

3) When using skygen, what units is the solar intensity in? What units
does skyfunc return?

Do you mean "gensky?" If so, the solar radiance is in units of
watts/sr/m^2 using the standard 179 value for luminous efficacy.
Unless you know better, it is recommended to simply use this value, so
it drops out of the calculation when you return to photometric units
during the analysis phase at the end.

I hope this helps.
-Greg

···

[email protected] wrote:

Beware to those who attempt to read the Radiance source code -- it is not meant for human eyes! I have trouble interpreting it, myself. The best I can say for it is that it compiles well, most of the time....

Though it's true that gensky uses a different efficacy for the sky conversion, it only does this when the zenith brightness is being computed from turbidity, which is an iffy proposition at best. If you use either the -b or -B options to enter your own zenith radiance or irradiance (from TRY data or on-site measurements), it is YOUR responsibility to convert from whatever was measured to the right units for your simulation. If you took a luminance measurement on a gray sky, then you should divide by 179. If your sky has some color to it, you should take this into account as Chas recommends.

-Greg

From: Charles Ehrlich <[email protected]>
Date: July 16, 2004 9:56:39 AM PDT

A caution regarding gensky, luminous efficacy factors and spectral conversions.
  
The physical skydome is strongly blue-colored. This matters if you are attempting to represent the brightness of objects relative to sources of light with different spectra, and/or relative to room surfaces with different reflectance spectra.
  
Blue light is more efficacious, i.e., it both has more energy and appears brighter. The lumens per watt conversion factor of 179 is really only valid for pure white light (I think I've got that right?). If you look into the source code for gensky, you will see that a different lm/watt factor was used to derive the zenith brightness (203 ??). But this efficacy is not tied to the color of the skydome as modeled by the user in the input file in the skyfunc "glow" entity below:

skyfunc glow sky_glow
0 0
4 0.75 0.9 1.2 0

sky_glow source sky
0 0
4 0 0 1 180

I do not know your specific application, but if the appearance of the sky is not important for your simulations, you might consider changing the skydome glow color component to be equal R G B factors (1 1 1) so that you do not convolve different efficacy factors and spectral influences on your final result. Alternatively, you could re-write gensky so that it adjusts the luminance of the skydome (by adjusting the lm/watt factor) in response to the color of the skyglow primitives.
  
In a quick study of this effect several years ago, we found errors as much as 20%--which is small potatoes for daylight simulation purposes, but troubling nevertheless because it is (arguably) an internal consistency error, not a modeling error.

The real problem is that the international unit of luminous flux is defined in terms of a human perceptual factor rather than on a radiometric unit, imho. This results in skydome brightness measurements in candellas per meter squared (instead of watts/meter squared) whose conversion factor depends upon the color of the light source.

-Chas