Spectra Lab
Join the waitlist
Free tools · nothing uploaded

Free NIR tools that run in your browser

Check a calibration, test whether a bias is real, and convert the units you meet every day. No sign-up, and nothing you type leaves this page.

Your data stays here. The calculations run in your browser. Nothing you paste is sent to us.
All tools

Pick a tool

Six tools, grouped by the job they do. Each one opens with a worked example so you can see what it returns.

Validate a modelValidation statisticsBias, SEP, RMSEP, slope, R² and RPD from reference and predicted values.Validate a modelIs the bias or slope real?t-tests that say whether a bias or a slope away from 1 is more than chance.Validate a modelLaboratory error from duplicatesSEL from blind duplicates, and the SEP your model can sensibly aim for.Convert unitsWavelength and wavenumbernm, µm and cm⁻¹ in both directions, with the spectral region.Convert unitsAbsorbance and transmittancelog(1/T) or log(1/R) and back, for one reading or a column.Convert unitsMoisture basisMove a constituent between as-is, dry matter and a fixed moisture basis.
Validate a model

Validation statistics

Paste the laboratory reference value and the model's prediction for each sample.

One pair per line, two numbers separated by a comma, space or tab. Nothing leaves your browser. The example loaded below is invented; paste your own pairs to replace it.

Validate a model

Is the bias or slope real?

A bias of 0.3 on ten samples can be chance. These two t-tests tell you whether to act on it.

The same pairs you would give the validation calculator. The example loaded below is invented.

Validate a model

Laboratory error from blind duplicates

Paste two readings of the same sample per line. Then compare your model's SEP with the target band.

One pair per line, two numbers separated by a comma, space or tab. Nothing leaves your browser. The example loaded below is invented; paste your own pairs to replace it.

Convert units

Wavelength and wavenumber

Instruments and papers report the axis either way. Convert between them here.

Type in any box and the others follow. 1450 nm, the first overtone of the O–H stretch that water absorbs, is filled in to start.

Convert units

Absorbance and transmittance

Choose what you have, then paste one value or a whole column.

Convert units

Moisture basis

Reference values are often reported on a different moisture basis from the one the model predicts. Put them on the same basis before you compare.

Use 0 for dry matter basis, the sample's own moisture for as-is, or your trade's fixed basis. The values filled in are an illustrative example.

In the SpectroScience course

More calculators for enrolled students

SpectroScience course students get a further set of calculators and Excel workbooks inside the course resources, including:

  • Beer–Lambert law
  • Calibration metrics
  • Model diagnostics
  • Feasibility checker
  • Calibration method selector
  • Concentration
  • Dilution
  • Absorbance
  • Wavelength converter
  • Moisture converter
  • Return on investment

Learn the method behind the numbers

32 lessons on building and validating NIR calibrations, with the calculators, handouts and lab sessions that go with them.

See the course calculators About the course

Definitions

The formulas

Every tool on this page uses these formulas. Check them against your own standard.

  • Bias is the mean of predicted minus reference. SEP is the standard deviation of those differences about the bias (n minus 1).
  • RMSEP is the root mean square of the differences. Slope and intercept come from regressing predicted on reference.
  • RPD is the standard deviation of the reference values divided by SEP. SEL is the square root of the sum of squared duplicate differences divided by twice the number of pairs.
  • Bias test: the bias is significant when its absolute value is larger than t × SEP / √n, with t from Student's t at the chosen confidence and n − 1 degrees of freedom.
  • Slope test: t = |slope − 1| × √Sxx / s, where Sxx is the sum of squared deviations of the reference values and s the residual standard deviation of the regression (n − 2 degrees of freedom).
  • Wavenumber in cm⁻¹ is 10,000,000 divided by the wavelength in nm. Absorbance is log₁₀(1/T), with T or R as a fraction.
  • Moisture basis: value at the new basis = value × (100 − new moisture) / (100 − old moisture).