Above the pixel grid’s Nyquist limit. At this viewing distance and pixel pitch the display cannot carry the frequency you asked for, so what is drawn is an alias. Move closer, or ask for less.
The contrast you asked for does not fit in sRGB. A grating swings either side of a mean — a grey of the same cone luminance for the chromatic carrier, the base colour’s own luminance for the luminance one — and here one of those two phases lands outside the gamut. The excursion is capped at the largest one both phases survive, and the figure above says what was achieved — clamping instead would flatten one phase and quietly change the contrast you thought you had set.
There is no detectability reading here. The page printed a “Detection” percentage from a peak-normalised bandpass value, which is a sensitivity and not a probability — and its temporal term was a Gaussian centred on 8 Hz, so a static grating could never read above 41.1% however visible it was, and the default stimulus is static. A detectability figure needs a threshold criterion and a calibrated model, and this page holds neither. The weighting curve it did have is kept under its own name, and the two readings above say where it sits.
Neutral off-axis is the check the opponent axes have to pass: a grey has to sit at the origin of both chromatic channels. It reads 2.2×10−4 rather than zero because the published cone matrix is given to four decimals, and it read 0.0877 before the chain passed through XYZ — four hundred times larger, and visible as a grey ramp climbing steadily up a channel that should not move at all.
| Input | L | M | S | L − M | S − (L+M)/2 | L + M |
|---|