Start with the quantity represented by the spectrum
A spectrum assigns an optical quantity to each wavelength. Its vertical scale determines what can be calculated. A total luminaire spectrum in W/nm integrates to radiant watts. A receiving-plane spectrum in W·m⁻²·nm⁻¹ integrates to irradiance in W/m². Their photon-weighted results are PPF and PPFD respectively; an entire luminaire’s output does not specify the amount reaching one point.
An uncalibrated relative curve preserves spectral shape but has no absolute optical scale. It can produce relative integrals and CIE chromaticity coordinates. It cannot supply absolute lumens, lux, PPF, PPFD or photon efficacy merely because its peak is labelled 1 or 100. If data has been calibrated externally, enter the resulting absolute density with its units and calibration basis.
Unit-area normalisation divides every density by the integrated area over the entered range. The resulting relative density has units nm⁻¹ and integrates to 1 under the same representation. For the artificial 1 W/nm band on 659–661 nm, its 2 W integral gives a normalised density of 0.5 nm⁻¹ across that band. This changes the representation of the shape, not the original absolute results; a zero-area spectrum cannot be normalised.
Make the wavelength representation and coverage explicit
Two supported representations describe different functions:
- Linear nodes: each row gives wavelength and density. Density varies linearly between adjacent nodes, and wavelengths must increase strictly.
- Constant bands: each row gives a start wavelength, end wavelength and density that remains constant over that band. Bands must be adjacent and ordered. A known zero gap is an explicit zero-density band, rather than an omitted interval.
These choices matter near sharp edges. A band from 659 to 661 nm at 1 W/nm is a rectangle with area 2 W. Linear nodes at 659 nm and 661 nm, both at 1 W/nm, also describe a constant value between those nodes. Adding zero nodes outside them instead creates linear ramps and changes the integrated area. Neither representation can recover narrow features that were not measured.
The 400–700 nm photon result requires that whole band to be covered, or a justified declaration that output is zero outside the entered range. CIE xy and the photopic integral use the observer’s 360–830 nm range and require the corresponding coverage. An instrument that measured only part of a band has left the remainder unknown. It has not established zero output there. The integrated area over the entered range remains useful and is labelled on that basis.
Photon integration weights radiant energy by wavelength
For monochromatic light, each photon has energy E = hc/λ. Longer-wavelength photons carry less energy, so the same radiant wattage produces more photons at a longer wavelength. Using wavelength in nm and spectral radiant density Pλ in W/nm:
PPF = [10⁻³/(hcNₐ)] × ∫₄₀₀⁷⁰⁰ λPλ dλ
The factor converts nm to metres and mol to µmol. The calculation uses the exact NIST constants: h = 6.62607015 × 10⁻³⁴ J·s, c = 299,792,458 m/s and Nₐ = 6.02214076 × 10²³ mol⁻¹. The DLC horticultural requirements describe the conventional 400–700 nm PPF band and PPE in µmol/J; their North American qualification thresholds are not used here.
For an artificial exact rectangular spectrum of 1 W/nm from 659 to 661 nm, with known zero output elsewhere, the radiant integral is 2 W. The wavelength-weighted integral is (661² − 659²)/2 = 1,320 W·nm. Its PPF is 11.0343383424 µmol/s, displayed as 11.0343 µmol/s. The wavelength integral is taken before display rounding.
With irradiance-density inputs, the same operation gives PPFD in µmol·m⁻²·s⁻¹, on the stated receiving geometry. With total-flux inputs and independently supplied matching whole-system input power, PPE = PPF/input W. Electrical watts are a separate input; radiant watts do not establish the system’s electricity use.
A single-wavelength model has a narrower meaning
The monochromatic mode applies photon flux = radiant W × λnm × 10⁻³/(hcNₐ), or its inverse. At 660 nm, 2 radiant W gives the same 11.0343 µmol/s as the symmetric artificial rectangle above. A monochromatic wavelength outside 400–700 nm still has a photon flux, but it is outside conventional PPF. A broad white LED cannot be represented by one wavelength chosen from its colour temperature.
The same relationship gives energy per photon = hc × 10⁹/λnm, in joules, using vacuum wavelength. Its inverse is λnm = hc × 10⁹/energy per photon. At 660 nm, one photon carries approximately 3.00976645 × 10⁻¹⁹ J. An independently specified energy of 3 × 10⁻¹⁹ J per photon corresponds to approximately 662.1486 nm. The energy input uses multiples of 10⁻¹⁹ J, so the latter example is entered as 3; results identify the full joule value.
Energy per photon is distinct from radiant power. Knowing a photon’s energy establishes its wavelength, but total radiant watts also require a photon arrival or emission rate. A result for one photon therefore does not infer a lamp’s output, receiving-plane illumination or electrical consumption.
Mix matching spectra before calculating their metrics
For two channels, the combined density is Pmix(λ) = aPₐ(λ) + bPᵦ(λ). Both channels must use the same wavelength representation, density units, calibration basis and measurement geometry. The calculation applies these multipliers to every wavelength before integrating.
A multiplier represents a declared optical contribution. It does not predict how an LED’s spectrum, input watts or temperature changes with an electrical control setting. For relative curves, independently normalising both peaks to 1 can remove the relative scale needed for a meaningful physical mixture. Establish that scale before assigning channel weights.
CIE xy describes chromaticity rather than colour rendering
The workbench integrates the declared spectrum against the CIE 1931 2° colour-matching functions to form proportional X, Y and Z values. It then calculates x = X/(X + Y + Z) and y = Y/(X + Y + Z). Multiplying the entire spectrum by one positive factor changes X, Y and Z together, leaving xy unchanged. An integrated zero signal has no defined chromaticity.
The data is from the CIE 1931 standard colorimetric observer dataset, DOI 10.25039/CIE.DS.xvudnb9b, published by the International Commission on Illumination in 2019. Its 1 nm table covers 360–830 nm. The dataset is adapted to TypeScript with unchanged numerical values under CC BY-SA 4.0; the source format has changed. Colour-matching functions are linearly interpolated. Integration splits at their 1 nm knots and integrates the product with the chosen spectral representation; this numerical rule does not remove measurement or interpolation uncertainty.
For an independently reproducible artificial 1 W/nm band from 555 to 556 nm, known zero elsewhere, averaging the two CIE endpoint rows gives proportional X = 0.520173, Y = 0.99992835, Z = 0.0055267995. Thus x = 0.3410 and y = 0.6554 after rounding to four decimal places. The full-precision values remain in the calculation.
The photopic result weights the same absolute spectrum by y-bar and 683 lm/W, following the CIE luminous-flux relationship. Total spectral radiant flux produces lumens; spectral irradiance produces lux. These weighted integrals depend on spectral shape, so there is no universal lux-to-PPFD factor. Chromaticity also does not establish CRI, TM-30, CCT, Duv, a crop response or a health outcome.
Convert existing xy, uv or u′v′ coordinates
A product report or colour meter may already supply chromaticity coordinates. Those two numbers and their coordinate system are sufficient for conversion; a spectrum is not required. Keep the source, standard observer, operating state and measurement conditions with them. Prime marks matter: CIE 1960 v and CIE 1976 v′ are different quantities. Assigning the same numbers to another coordinate system changes the point they describe, so the input-system selector reinterprets entries rather than replacing them with converted values.
The CIE 1976 UCS definition, from CIE S 017:2020, gives the transformations below. With D = −2x + 12y + 3:
- u = u′ = 4x/D
- v = 6y/D
- v′ = 9y/D = 1.5v
The inverse transformations follow algebraically. For CIE 1960 uv, set q = 2u − 8v + 4, then x = 3u/q and y = 2v/q. For CIE 1976 u′v′, set q = 6u′ − 16v′ + 12, then x = 9u′/q and y = 4v′/q. All three result pairs describe the same supplied chromaticity. Results from spectral mode also include both UCS pairs, calculated from the unrounded xy values.
For the artificial input x = 0.3127, y = 0.3290, D = 6.3226. Thus u = 1.2508/6.3226 = 0.197830, v = 1.974/6.3226 = 0.312213, and v′ = 2.961/6.3226 = 0.468320, to six decimal places. These are given coordinates for checking arithmetic, rather than a claim about a particular lamp. As a reverse example, u = 0.2, v = 0.3 gives q = 2 and x = y = 0.3, with u′ = 0.2, v′ = 0.45.
Conversion retains the standard observer of the supplied coordinates. A result derived from 10° data remains on that basis; the transformation cannot turn it into a 2° measurement. The spectrum calculation above uses the CIE 1931 2° observer because it integrates that specific dataset. Retain the observer and measurement conditions when comparing reports.
For nonnegative normalised XYZ, the coordinates must satisfy x + y ≤ 1, u + 10v ≤ 4, or 3u′ + 20v′ ≤ 12, as applicable. The tool checks these boundaries using the entered decimals before rounding. These are mathematical constraints, not a check against the visible spectral locus. Passing them does not establish that a real light source can produce the entered point. In particular, the mathematical xy origin represents the Z axis, while a zero optical signal has undefined chromaticity.
Coordinates alone do not recover a unique spectrum or establish brightness, colour rendering or a product match. The tool does not calculate a distance between two colours, a MacAdam-step count, or signed Duv from a black-body locus. Keep those tasks distinct when reading a specification.