Skip to content

resample_linear

Resample a spectrum onto a uniform wavelength grid by linear interpolation.
Category Resampling
Backend numpy - implemented here on top of numpy primitives
Version 2.1.0
Reads ctx.spectrum (a Spectrum1D)
Writes spectrum

Linear interpolation is fast and adequate when the target sampling is similar to the source. For large changes in resolution prefer a flux-conserving method. v2.0.0 — uncertainty propagation : σ_out = sqrt(interp(σ²)), the linear interpolation of the VARIANCE onto the new grid (v1.x interpolated σ itself, which is not a propagation rule). For an output sample at fraction t between two inputs the exact variance of the interpolant is (1−t)² σ_a² + t² σ_b² ≤ (1−t) σ_a² + t σ_b², so the per-sample value is conservative ; the covariance between neighbouring output samples (they share input samples, and oversampling duplicates information) is NOT tracked, which under-states the errors of anything fitted to the resampled spectrum. Without an input uncertainty the brick is unchanged. v2.1.0 — oversampling: the grid holds round((npix − 1) · oversampling) + 1 samples over the same bounds, i.e. oversampling output samples per input sample (2 doubles the sampling, 0.5 halves it) ; mutually exclusive with step and n_points. The step, n_points and default paths are byte-identical to 2.0.0.

Parameters

Parameter Default Required Description
wavelength_min None - Lower bound of the new grid (Å); null keeps the input minimum.
wavelength_max None - Upper bound of the new grid (Å); null keeps the input maximum.
step None - Constant sampling step (Å). Mutually exclusive with n_points.
n_points None - Number of samples in the new grid. Mutually exclusive with step.
oversampling None - Output samples per input sample: the grid holds round((npix - 1) * oversampling) + 1 samples over the same bounds. Mutually exclusive with step and n_points.

Use it

from spectro_kernel import run_algorithm

output = run_algorithm("resample_linear", ctx, {
    "wavelength_min": None,
    "wavelength_max": None,
    "step": None,
    "n_points": None,
    "oversampling": None
})

spectro run resample_linear --input spectrum.fits \
  --param wavelength_min=none \
  --param wavelength_max=none \
  --param step=none \
  --param n_points=none \
  --param oversampling=none

{
  "tool": "resample_linear",
  "arguments": {
    "session_id": "<session_id>",
    "params": {
      "wavelength_min": null,
      "wavelength_max": null,
      "step": null,
      "n_points": null,
      "oversampling": null
    }
  }
}

Every algorithm is an MCP tool of the same name; describe_algorithm returns this page's metadata as JSON.

References

  • numpy.interp — piecewise-linear interpolation
  • For large resolution changes, prefer flux-conserving resampling (specutils.manipulation.FluxConservingResampler)
  • Bevington & Robinson 2003, Data Reduction and Error Analysis for the Physical Sciences, 3rd ed., McGraw-Hill — §3.2, variance of a linear combination.