nx_skinspectra_lib.nx

nx_skinspectra_lib.nx -- MEASURED ABSORPTION SPECTRA, PARSED NOT TRANSCRIBED (2026-09-03, rung ST6) WHY PARSING AND NOT A TABLE OF CONSTANTS. Every coefficient on /compare/skintwin was a PROBE VALUE, and pinning the published spectra was supposed to fix that. It nearly fixed it the wrong way: the obvious move is to read the mirrored table and copy numbers into a conf. That is transcription, and transcription is exactly the failure this rung exists to remove -- MEASURED THE SAME DAY, the probe melanin coefficients carried a blue-to-red ratio of 5.0 against the published law's 2.72, so the shape was right and the slope was 1.8x too steep. A number typed by hand is a number that can be wrong. A number parsed out of a sha-pinned mirror is the mirror's responsibility. SO THE ONLY VALUES TYPED HERE ARE THE ONES THE PUBLISHED LAW ITSELF STATES -- 1.70, the decade count 12, the exponent 3.48, and the source's own conversion constants 2.303 and 64500 -- each written as it appears in the source rather than pre-multiplied. The haemoglobin numbers are never typed at all; they are read from the mirror at run time. SOURCES, BOTH MIRRORED AND PINNED (skintwin.refs): MELANIN [@omlc_mel] Jacques and McAuliffe 1987; Jacques, Glickman and Schwartz 1996. Melanosome absorption measured by the threshold for explosive vaporization under pulsed lasers. Fitted spectrum for skin: mua = 1.70e12 * lambda^-3.48 [cm-1], lambda in nm. HAEMOGLOBIN [@omlc_hb] Prahl, compiled from Gratzer (MRC Labs, London) and Kollias (Wellman Labs, Harvard Medical School). Tabulated MOLAR EXTINCTION e in [cm-1/(moles/litre)], converted by the source's own formula mua = 2.303 * e * x / 64500, x in grams of haemoglobin per litre. THE PREFACTOR IS A DISTRIBUTION, NOT A CONSTANT, AND THE SOURCE SAYS SO IN ITS OWN WORDS: "The concentration of melanin within melanosomes is quite variable. Ten-fold variation is to be expected." So ss_melanin_mua returns the LAW at unit concentration and takes concentration as a caller argument. A lib that folded a tenfold-variable quantity into a fixed coefficient would publish a constant wearing a measurement's clothes. NUMERICAL NOTE THAT MATTERS: 1.70e12 is far outside Q30 and must never exist as a value. mua is computed as ONE exponential of (ln 1.70 + 12 ln 10 - 3.48 ln lambda), so the huge prefactor is never materialised and the result lands in range at every visible wavelength. MEASURED: 993.73 / 494.30 / 365.16 per cm at 450 / 550 / 600 nm, and the prefactor-free ratio reproduces (600/450)^3.48 = 2.721365 EXACTLY. license_tier: ORIGINAL No hw writes (Rule 26). LIB (no main).