The four windows and their settings
What the four windows show
One trace, read four ways: we follow light back from your eye, through the stone, to the part of the dome it came from. In ISO, COS and SC2, white is a facet returning light from the brightest part of that dome and black is a facet returning none.
Light Return
Where the light comes from:
- Direct, 45° to 75° up
- Indirect, below 45°
- Contrast, above 75°, where your head is
- Leakage
ISO
A dome lit evenly from every direction. The most forgiving test: it asks only how much light comes back, not from where.
COS
Bright overhead, fading to black at the horizon. Hardest on shallow cuts, which lean on low-angle light.
SC2
A ring of light peaking at 45°, dark overhead and at the horizon. Closest to how a stone is lit on a hand.
None of the four depends on scene lighting: they model the cut’s geometry, not any real stone. ISO, COS and SC2 follow the definitions in the GemRay user guide by Robert W. Strickland; the implementation is our own.
Face-up brightness
Per cent of the dome’s available light returned into the viewing cone, averaged over the stone’s face-up image. Whole stone is the full outline; table only is the area inside the table facet, the figure that moves most between a good cut and a poor one. Leakage sits below the rule because it is not a light model: it counts the light paths that came back with nothing at all, so it does not move when you change the head shadow.
| Light model | Whole stone | Table only | vs T57 |
|---|---|---|---|
| ISO | |||
| COS | |||
| SC2 | |||
| Leakagelight that never came back |
Brightness against tilt
The stone tilts under a fixed dome, which is what happens on a hand. Solid lines are the whole stone, dashed lines the table only. A cut that holds its height across the sweep keeps working when it is not being looked at straight on.
The chart as a table
What we count
Brightness here is energy, not just direction. At the crown a Fresnel share of the light reflects straight off without entering the stone, and that light still reaches the eye: face-up, the table’s own reflection points at the zenith and lands in the head shadow, which is the dip you see near zero tilt. The rest goes through and pays Fresnel again on the way out. Internal bounces are left alone, because they are overwhelmingly total internal reflection, which loses nothing.
That choice is not cosmetic. Measured across these 21 cuts, leaving the energy out and counting direction alone reorders 8 of them on ISO, 7 on COS and 10 on SC2, while adding the exit face to the entry face reorders 0, 4 and 5. The model settles; counting direction alone is the outlier.
What the model leaves out. We do not count the energy lost at partial internal reflections. That is why these figures sit above those of a full physical simulation, and why the dip near zero tilt is shallower here than in GemRay.