Colour Measurement Uncertainty & Tolerance Guard Band

Quadrature addition assumes the three contributions are independent, which is a fair description of drift, presentation repeatability and inter-instrument disagreement but understates the total where a single cause feeds two of them - a failing lamp, for example, degrades repeatability and drift together, and the combined figure will then be optimistic. The shares are variance shares rather than shares of the DE figure, which is why they sum to a hundred while the DE contributions do not; that is the correct way to read an uncertainty budget and it is also why reducing the largest share is worth so much more than reducing the smallest. A capability ratio below about one means the measurement system cannot reliably distinguish a pass from a fail at this tolerance, and the honest responses are to agree a wider tolerance, to move both parties onto instruments with better agreement, or to exchange physical standards rather than numbers. Averaging readings helps only the random term, and it helps as the square root, so going from three readings to twelve halves a contribution that is usually the smallest of the three to begin with. Nothing here covers sample presentation - aperture size, backing, folding and directionality on a napped or corduroy fabric routinely dwarf every instrument term in this budget.