Generate the nine signatures from one lightness band on a 12-slot hue wheel

Context and Problem Statement

Three palettes were in circulation (the CSV, the colours inside the logo files, and the 0.2 draft’s optimiser output), picked one by one: lightness ranged from 0.40 to 0.75 in OKLCH, two signatures failed 3:1 on white, and the closest pair sat at ΔE 8.6. Demian set the direction on 3 Oct: coherence above all; AKRIBIS stays teal and AKRIMET orange with the tone free; SOLFARMA stays blue, more loosely; the other six are free; new units must be able to join. ADR-0010 adds a constraint: the ring is nine segments in a fixed order, so adjacent segments must be told apart.

Decision Drivers

Criteria 2 (one coherent, expandable family), 4 (accessibility), 6 (recognition where it exists), 9 (generated and tested) of ADR-0009.

Considered Options

• One lightness band for all nine on a 12-slot wheel; three levels (L 0.56, 0.60, 0.64), three chroma caps (0.16, 0.13, 0.10), two assignments (A: nearest slot; B: AKURALAB moved to green)
• Keep the delivered colours and solve legibility with a darker text step only
• Re-tune each colour independently (the 0.2 draft’s optimiser)

Decision Outcome

Chosen: wheel A; the signature at grid step 500 (L 0.635) with chroma cap 0.20; the text colour at step 600 (L 0.55).

Second amendment, 2026-10-04. On the page the L 0.55 set read numb. The cause was the decision itself: one colour asked to be both a vivid mark and 4.5:1 body text, which no hue can be. The two jobs are now one ramp step apart. The signature (marks, large graphics, UI components) sits at step 500, L 0.635, where every unit reads at least 3:1 on white (minimum 3.23, AKURALAB), the WCAG floor for graphics. Brand-coloured text and controls use step 600, L 0.55, the highest lightness at which every hue reaches 4.5:1, derived by rule. The chroma cap rises from 0.16 to 0.20; uncapped, red and magenta went neon. The table below carries the signature values; the text colours are one step down in each ramp (ADR-0015).

Correction, 2026-10-04. The band was first set at L 0.56 by eye from the study sheets. The ramps’ contrast report then showed AKURALAB at 4.41:1 and AKRIBIS at 4.48:1 on white, under the 4.5:1 text threshold the decision claims. The band is now defined by rule, “the highest lightness, in steps of 0.005, at which every signature reaches 4.5:1 on white”, which gives L 0.55; the colours move by about ΔE 1, below what the eye separates. The table below carries the corrected values.

• Wheel. Twelve slots every 30°, phased at 24° so AKRIMET’s orange (54°) sits on a slot. Every delivered hue lands within 12° of a slot: AKZOTEC 24°, AKRIMET 54°, AKLEIA 84°, AKURALAB 174°, AKRIBIS 204°, AKUTEMP 234°, SOLFARMA 264°, AKIoT 294°, SINGULAR4 354°. Free for future units: 114°, 144°, 324°.

• Two steps, one hue. The signature at step 500 (L 0.635, ≥ 3:1 on white) is the mark; step 600 (L 0.55, ≥ 4.5:1) is the same hue as text and controls; on a tinted surface text goes one further step down (700). The same step behaves the same on every site.

• Chroma. As high as sRGB allows at each hue, capped at 0.20. The teal family is gamut-limited (about 0.11 to 0.13); red, blue, violet and magenta reach the cap. Caps of 0.16, 0.13 and 0.10 were seen and declined as duller; uncapped as too loud.

• Assignment A keeps AKURALAB teal-green at 174°, 30° from AKRIBIS and its neighbour in the ring (ΔE 5.3, the closest pair). Assignment B (AKURALAB to the free green at 144°) separated the ring better but moved the hue 39° and left the 174° slot unusable for a future unit.

• AKRIBIS One has no hue: ink, and the nine signatures in the ring. The ring and the ring lockups are published in colour; the dot and letters lockups stay ink.

  Note, 2026-10-04. ADR-0016 proposes a colour for AKRIBIS One, the AKRIBIS hue at step 950 (the teal ink), and the ring in colour as one continuous spectrum in hue order; every lockup then has a colour version.

The resulting signatures, generated by brand/src/palette.ts from brand/units.yaml (hue) and brand/colour.yaml (band):

Consequences

• Good, because adding a unit is a hue slot in the register; its colour, its ring segment and later its ramp follow.
• Good, because the build asserts the band, the contrast floor, the hue distance to the delivered colour and the minimum separation.
• The warm colours keep most of their fire at step 500; SOLFARMA’s navy becomes a clear blue and AKLEIA’s gold an ochre. Accepted for coherence.
• The parent mark’s colour is decided separately (topic 2.1); the delivered AKRIBIS teal (#005e6e) is, within ΔE 1.5, the deep tone (L 0.44) of this wheel’s teal slot.