Peak Finder — Advanced

What this page adds. The basic calculator finds peaks by local maximum and fits Gaussians in your browser. This page adds the derivative and wavelet finders, more filters, valleys, Lorentzian and Voigt shapes, and a peak editor to set each peak's start by hand. To do that it sends your table to the server, computes there, and sends the numbers back. Nothing is stored, but the data does leave your machine. If that matters for your data, use the basic page. The peak finding article explains each method.

Noise makes a small local maximum every few samples; the prominence filter decides which bumps count. Two peaks closer than their width merge into one bump: raise the number of peaks, place each one in the editor, and fit. A drifting baseline or heavy noise is best removed first, with detrending or smoothing.

1. Paste your data into the Input grid — straight from a spreadsheet, or as CSV/TSV text — or import a file. Column x, then y1 and y2. Up to 20,000 rows (5,000 for the wavelet finder). Rows are sorted by x; one y per x.
2. Peaks are found at once. As soon as the input is applied, the server finds the peaks of each y, they are marked on the chart, and the Output grid lists each one's position, height and half-prominence width (HPW): the width at half its prominence, read straight off the data. FWHM and area wait for the fit.
3. Find picks the method.
   • Local maximum: every sample higher than its neighbours, then the filters.
   • Derivative: where the smoothed slope crosses zero from plus to minus.
   • Wavelet: ridges of the continuous wavelet transform, peaks that hold across widths.
4. Filters. Prominence is how far a peak stands above its surroundings, as a fraction of the column's range (0.1 = 10 %). Under More filters: Min height, Min width (samples), Plateau (the smallest flat top, samples) and Threshold (the smallest step to both neighbours) are off when blank. Distance drops a peak that sits closer than that many samples to a taller one. valleys finds the dips instead. Any change finds again.
5. Peaks to fit sets how many peak functions the fit uses, 1 to 10. It starts at the number found; raise it when a bump hides a second peak in its shoulder - the widest one is split in two.
6. Edit one peak opens the editor, which sets each peak's start: pick it with Peak #, then its Type (Gaussian, Lorentzian or Voigt; shapes can be mixed), its Position, its FWHM and its Amplitude, the height above the offset (0 or above; 0 or below with valleys). Each change redraws the peak functions at once, with no fit. The amplitude starts from the data, and the fit frees it.
7. Edit-on-chart. Click a peak function (a dashed line; before any preview, a found peak's dot) to choose it - its line turns bold and the editor opens on it. While the pointer stays on the chart:
   • Shift + wheel or ← →: the position.
   • Ctrl + wheel (Alt + wheel on a Mac) or ↑ ↓: the amplitude.
   • Mouse wheel or Shift + ← →: the FWHM.
   Every change shows in the editor boxes and redraws like a ▲▼ step. Click empty chart space to let the peak go.
8. Real time range — on prominence, min height, distance, position, FWHM and amplitude.
   • ▲▼ buttons or the mouse wheel over the box: the value moves and the page follows - one request a moment after the last step.
   • Digit wheel on min height, position, FWHM and amplitude: the wheel steps the digit just left of the mouse pointer.
   • Typing takes effect when you leave the box.
9. Press fit. The server fits every peak and one constant offset together; the time it took sits beside the button. The chart shows the input as a light black line, each fitted peak as a green dashed line and their sum as a red line. The editor then shows the fitted values, ready for another round.
10. Read the result. The Output grid lists each peak's shape, position, amplitude (its height above the offset), FWHM and area, each with its standard error. Each fitted amplitude is held at 0 or above (at 0 or below with valleys): a peak stands up from the offset, so the fit never answers with a negative peak. Under the chart: R², the residual RMS and the solver's evaluation count of the shown y. Show picks the y; each keeps its own peaks and editor. Export saves the Output grid as a CSV/TSV file.