Hi All,
I’m back again evaluating Omniverse Create/Composer as a tool for lighting design, and have noted some issues / quirks with accuracy in the false-colour illuminance view in RTX Accurate (iRay) and the interpretation of IES files for shaping.
And before I get to the quirks, I have to praise the OV and iRay teams for their amazing work!
I’ll preface this by noting that I am a qualified lighting designer / engineer, and have been working with lighting design / calculation / rendering software since the late 1990’s, so when I refer to ‘expected behaviour’ or ‘correct interpretation’ it’s not just based on whatever software I’ve used last. Until very recently I’ve been using Bloom Unit (essentially iRay for Sketchup) as well as iRay for Rhino - but neither of those is supported any more. I am checking, editing, and comparing photometric files in Photometric Toolbox (the industry standard tool).
The expected behaviour then, is that with the intensity set to match the IES file’s lumens, and the light emitter sized to match the photometered fitting, for direct illumination only, directly below the luminaire, at a distance of not less than 10X the luminaire’s largest dimension, the value should correlate with the IES file’s 0/0deg (nadir) intensity as per E=I/d². This is ground truth.
For example, if the IES file has a downward intensity of 5,000cd and we place the light source 10m above the calculation plane, we should see an illuminance value of 50 lux (5000 / 10^2).
I am using the following settings for the lights:
- Rotate: 0 / 0 / 0 (downward emission)
- Scale: 1 / 1 / 1 (no rescale)
- Colour Temperature: Enabled, 3000K (colour temps are energy-preserving)
- Intensity: To match the total flux of the luminaire as reported by Photometric Toolbox (generally also very close to the lamp lumens value)
- Exposure: 0 (no adjustment)
- Normalize Power: On
- Radius: As needed, typically 0.025m (50mm diameter) for test purposes.
- Diffuse/Specular multiplers: 1 / 1 (no adjustment)
- Visible in Primary Ray: Yes (but doesn’t affect results)
- Disable Fog: No (but not relevant here)
- Enable Caustics: No (but not relevant here)
- Shaping File: Selected IES file
- Shaping Cone Angle: 180 (this is how the DiscLight is created, though not the default value?)
- Shaping Normalize: On, but doesn’t seem to do anything when an IES file is used
- Shaping all other parameters: Default values (no adjustment)
THE GOOD NEWS:
-
Visible In Primary Ray: Works as expected, including correct(ish) luminance values when viewing IES-shaped sources from various angles.
-
Colour Temperature: Works as expected - energy-preserving
-
Light Source Colour: Works as expected - non-energy-preserving (acts like a coloured filter/gel over the light). An option for energy-preserving source colour would be helpful also (like for RGB LEDs with lumens per colour).
-
Emissive Materials (OmniSurface.mdl): Works as expected, emitting in cosine pattern. Both emission_nt (cd/m²) and emission_lx (lumens/m²) are emitting as expected.
-
IES Candela Multipliers: Works as expected, OV is using the correctly multiplied candela values, not just the raw values.
-
Guided sampling and/or caustic sampling produces illuminance / luminance values that are about 3% higher than standard sampling. This is tolerable for the benefits they provide when needed.
THE BAD NEWS:
-
In general (with correctly configured lights with IES shaping) the RTXA Illuminance view is reporting illuminance values about 11% higher than ground truth.
-
For some maybe 15% of the IES files I’ve tested, RTXA will give erroneous results, usually in the range of 15% to 30% above the correct value. I can’t for the life of me figure out why this is occurring, there is nothing in the IES files that should be driving incorrect results. But it I have checked, and OV is correctly applying the candela multiplier. These are files that I’ve used in other software without issue. I’ve included in the attached, some example ‘problem’ files from Faze, Gamma, and Hunza Lighting - plus a fully synthetic IES file out of PLM that shows this issue.
-
When shaping DiskLights or RectangleLights with IES files having any emission above the horizon (90deg vertical), behaviour is ‘odd’. It would appear that OV is somehow re-assigning any upward flux back to downward, so that you get much more downward illumination than you should. Using the ‘Ledil Zorya’ IES in the attached with a disc light gives about 3X the correct value below the light.
-
Similarly, when shaping CylinderLights with IES files having all-around illumination, some flux seems to be re-assigned out of the dead ends of the cylinder, resulting in higher than expected downward values.
-
Using a SphereLight with a downward-only IES distribution (e.g. a downlight) results in erroneous low illuminance values (about 27% of ground truth in the examples I tested).
-
The Normalize Power setting I think is not working as intended, and is out by a factor of Pi. That is, a non-normalized surface of 0.319m radius (1/pi m²) produces the same emission as when normalised. The expected behaviour would be that non-normalized intensity is in lumens/m² i.e. a 1m² surface (e.g. circle of 0.5642m radius) would produce the same emission as the normalized emitter.
-
Shaping Normalize doesn’t seem to do anything (at least with IES files)?
-
I can’t for the life of me figure out how Shaping Angle Scale is meant to work, if it all?
-
Also can’t figure out how Shaping Cone Angle is meant to work? It does something.