Your spectral landscape
Drag to zoom · hover to inspect · double-click to reset. Derivatives, when selected, use the right axis. Anchor selections change settings; process again to apply them.
Processing trail
Detected peaks
Widths are at half prominence. Local areas use a line between peak bases and may overlap. Click a peak row to populate a single-component fit starting center.
| Spectrum | Position | Signal | Prominence | Width (half prominence) | Local area | Plateau points |
|---|
Data quality & spectrum statistics
Noise is estimated as scaled MAD of adjacent signal differences / √2; real spectral slopes and peaks can inflate this estimate. Duplicate X coordinates are averaged and missing/nonnumeric rows are excluded explicitly.
Bounded multi-peak fitting
Positive widths, nonnegative peak magnitudes, bounded centers, and 0–1 mixing fractions. Dips are fitted after sign reversal. Narrow the preprocessing X range to focus on the feature of interest; maximum 3,000 points per fit.
| Component | Center ± 1σ | Center 95% interval | Amplitude | FWHM | Analytical area | Mixing η |
|---|
Sample-to-sample comparison
Concentration calibration
Enter known concentrations for standard spectra. Responses are interpolated at the chosen X position in the processed data. Blank subtraction and normalization affect the calibration; verify their physical meaning.
| Use as standard | Spectrum | Concentration | Response σ (optional) |
|---|
A reproducible spectroscopy workflow
- Import. Choose spectra, review the worksheet and columns, then add them to the workspace. Select the spectra to process. Text and workbook import settings support headers, skipped metadata rows and delimiters.
- Prepare. Select mode and axis units. Convert transmittance if needed. Subtract a blank covering the full sample range; select smoothing and a justified baseline.
- Inspect. Compare raw and processed signals, baseline and derivatives. Adjust prominence and spacing; inspect peak tables and quality warnings.
- Fit. Narrow the X range, process again, select 1–5 peak components, and enter informed starting centers if automatic detection is insufficient.
- Compare / calibrate. Compare only common spectral ranges. Enter known standard concentrations for a response-position calibration and check residuals before estimating unknowns.
- Export. Save data, peaks, fitted components, calibration, processing settings and reports. Workspace JSON preserves your raw spectra and settings for a later session.
Methods and limits
Local polynomial smoothing uses actual X coordinates and shortened/shifted edge windows; derivatives use the fitted polynomial coefficients. ALS baseline correction penalizes second differences in sample-index space and is sensitive to λ, asymmetry and sampling. Endpoint baselines assume peak-free endpoints. Anchor baselines need more distinct anchors than polynomial coefficients and should use peak-free regions.
Peak prominence follows local maxima/minima and higher-peak boundaries. Width is measured at half prominence by linear interpolation. Measured local areas and fitted analytical areas have different definitions; overlapping local areas are not additive.
Gaussian, Lorentzian and pseudo-Voigt components plus a constant offset are fitted in original processed signal units using bounded nonlinear least squares. Fitted FWHM and areas are derived from those model definitions. Local parameter errors and 95% intervals may be unreliable near bounds, with correlated components or unconverged fits. Fits can have local minima and do not identify compounds.
Calibration uses known concentration versus processed signal at one X position. Inverse estimates and intervals use local error propagation; unknown-response noise must be supplied to include it. Quadratic inverses must be unique in the calibration range. Residual-based LOD/LOQ values are provisional proxies, not validated detection limits. X and standard concentrations are treated as exact.
Formats and privacy
Excel XLSX/XLS/XLSM/XLSB, ODS/FODS, SpreadsheetML XML and HTML tables; CSV/TSV/TXT/DAT/ASC; JSON/JSONL/NDJSON; uncompressed numeric JCAMP-DX pairs and regular X++(Y..Y) tables. Compressed JCAMP ASDF/DIF/DUP, encrypted workbooks, binary instrument files, PDF/images and chemical databases are not supported. Excel formulas use saved cached values and are not recalculated. Convert unsupported spectra to two-column CSV.
Maximum 10 spectra, 20,000 distinct points per spectrum and 20 MB per file. Fits allow up to 3,000 points and 5 components. Invalid rows are excluded and duplicate X coordinates averaged. Data and processing remain in browser memory; export workspace JSON before reloading. No API, external CDN or paid key is needed. Browser downloads contain your data; keep them according to your own research practices.
Interpretation
Pearson/Spearman similarity is descriptive and is not chemical identification. P-values from densely sampled, autocorrelated spectra are not shown as evidence of independent replication. Preprocessing can distort shapes and alter areas; assess settings against standards and original spectra. This release does not provide automated band-gap estimation, database matching, instrument control or persistent cloud accounts.
03 Export & reproduce
Raw and processed spectra, peaks, fits, calibration and settings.