Browser-based analysisChecking references…
SPECTRAL ANALYSIS
Color Science
Calculate spectral averages, compare color, and inspect measurement data.
REFERENCE FRAMED65 / CIE 1931 2°Relative colorimetry · fixed reference white
SPECTRAL WORKSPACE
Reflectance analysis
Add your spectral measurements
Drop files here, or browse. CSV, XLSX, and XLS · wavelengths in nm
Files are processed locally in this browser.
Up to 32 files / 50 MiB total; 25 MiB per file. One file selection is read at a time; wait before adding another. Analysis supports up to 1,000 spectra / 2 million normalized points, with a conservative 2 million integration-node budget. Oversized selections are rejected in full.
The demo contains fractional values: Neutral50 = 0.5. Choose Fraction for a 50% result. The unit setting applies equally to the demo and your uploads.
Loading verified reference data…
Input review
Your spectrum starts here.
Add measurement files or explore the neutral demo.
Coverage checks, color coordinates, and a complete run record follow.
Explicit unitsVerified referencesTraceable exports
Spectra included
/
Spectral profile
Fractional measurement across the selected band
No supported coverage for this plot.
Analysis results
Unavailable values are shown as a dash, never a measured zero.
Coverage, full sample IDs & diagnostics
Summary & process capability
Only valid, selected observations enter the summary.
KEEP THE CONTEXT
Your results, with their provenance.
The ZIP includes normalized inputs, settings, source hashes, cleanup, and summary. Preserve original acquisition files separately.
Inspect calculation settings & provenance
COLOR DIFFERENCE
Two colors. One clear comparison.
Enter CIELAB coordinates obtained under compatible observer, illuminant, reference-white, and measurement conditions.
CIE76 ΔE
—CIEDE2000 ΔE
—Acceptance limits depend on your application and viewing conditions. No universal pass/fail threshold is applied. Swatches are approximate, gamut-clipped D65 sRGB previews; they do not calibrate your display.
THE METHOD BEHIND THE MEASUREMENT
Know what your numbers mean.
A concise guide to the input contract, calculations, and practical limits of this suite.
Start with known units
Keep original acquisition files. Upload UTF-8 comma-separated CSV, XLSX, or XLS with wavelengths in nm. Select fractional values (0–1) or percentages (0–100) explicitly: 0.5 percent means 0.005 fraction.
Choose the header, wavelength, and measurement columns. Header row 0 means no header. Excel formulas require numeric cached values saved by the originating application; this suite does not recalculate formulas.
Missing rows are rejected by default. Optional drops, sorting, and identical duplicate collapse are recorded. Finite out-of-range values, nonpositive wavelengths, and conflicting duplicates are rejected.
Three ways to average a spectrum
Raw (%) is the wavelength-interval average of fractional reflectance or transmittance, multiplied by 100.
Photopic (%) is 100 × ∫q(λ)V(λ)dλ / ∫V(λ)dλ, with equal-energy incident spectral weighting.
D65 Y (%) is 100 × ∫D65(λ)q(λ)V(λ)dλ / ∫D65(λ)V(λ)dλ. V(λ) equals the CIE 1931 2° ȳ function in the active reference data. This is a relative luminous ratio, not absolute luminance or a complete brightness model.
Raw averages use the full selected band, including UV or infrared. Weighted averages use its positive-width overlap with 360–830 nm and are unavailable when there is no overlap.
A consistent reference for color
XYZ and CIELAB require complete 360–830 nm coverage, D65, and the CIE 1931 2° observer. The fixed reference white has Y=100. Changing the averaging band does not change the color calculation or white.
Integration retains measured wavelengths alongside 1 nm reference knots. Premultiplied D65 × color-matching weights are linearly interpolated. XYZ are color-matching coordinates, not direct cone responses.
Endpoints are interpolated only within measured coverage; no extrapolation occurs. The configurable 5 nm gap gate does not guarantee that the instrument resolved sharp features.
Read the curves correctly
The three curves show q(λ), q(λ)V(λ), and D65n(λ)q(λ)V(λ) on a fractional 0–1 scale. Percent material inputs are converted to fractions first. The photopic function V already has a unit peak at 555 nm.
D65 is relative spectral power, not percent. For plotting, D65n = D65 / 117.812, its fixed reference peak over 360–830 nm. This source-only scaling is independent of your sample and selected band. Individual curves are never divided by their own peak or area.
Only q is the material fraction; weighted curves are fractional relative responses. The white D65-weighted curve peaks near 0.879, because the illuminant and photopic peaks occur at different wavelengths. The reported averages still use their weighting integrals: the fixed D65 scale cancels between numerator and denominator, so averages and XYZ/Lab results are unchanged.
Unsupported results are unavailable, never fabricated zeros. A valid narrower band may have averages while full-range color remains unavailable.
Capability needs a process study
Statistics summarize valid selected spectra. Specification limits are optional; zero is a valid boundary. Cpk requires at least 25 valid observations, nonzero sample spread, at least one limit, and confirmation of a stable, representative process with suitable distribution assumptions.
The count gate is a software policy. Neither it nor the confirmation checkbox proves statistical stability, normality, or study adequacy. Sample standard deviation uses n−1.
Delta E compares compatible Lab measurements. CIE76 and CIEDE2000, including positive K_L/K_C/K_H factors, do not establish universal perception or acceptance bands.
Keep the evidence together
The analysis ZIP contains results, normalized spectra, summary, source hashes, parsing and cleanup records, selections, reference identities, code/build identities, and run notes. Rejected inputs remain in the review record.
Original uploaded bytes are not included. Preserve them separately. Files are parsed locally in the browser; the suite has no upload service or persistent measurement database.
Browser limits are explicit: 32 files and 50 MiB per workspace, 25 MiB per file, 200,000 rows / 2 million cells per file, and 1,000 spectra / 2 million normalized points per analysis. A conservative 2 million generated-node budget also bounds band and color integration. Oversized files or selections are rejected; data is never silently truncated.
This relative, nonfluorescent scalar model does not establish instrument calibration, uncertainty, angular response, standards certification, or production qualification. Only the verified 1931 2°/D65 references are active; historical modified 1964 data is excluded.