next up previous contents
Next: Normalization and Fitting Up: Spectral Analysis Previous: Spectral Analysis

Rebinning and Interpolation

Raw spectra are usually not sampled at constant wavelength or frequency steps, and sometimes even with gaps in between the bins. At some stage the independent variable will have to be converted into linear or non-linear functions of wavelength or frequency units and gaps will have to be filled with interpolated values. Frequent cases are: wavelength calibration, redshift correction, $ \log(F_\lambda)$ versus $\log(\lambda)$ presentation, and the comparison of narrow-band filter spectrophotometry with scanner data. Related commands are REBIN/LONG, already described, REBIN/LINEAR for linear rebinning, i.e. scale and offset change, REBIN/II (IT,TI,TT) to do nonlinear rebin conserving flux (see below) and CONVERT/TABLE to interpolate table data into image data.

Note that in our implementation we make a conceptual difference between straightforward interpolation and rebinning. REBIN/II (IT, TI, TT) redistributes intensity from one sampling domain into another. There is no interpolation across undefined gaps and no extrapolation at the extremities of the input data. If you need these, you will have to manipulate the input data first (generating non-existent information !). REBIN/II (IT, TI, TT) conserves flux locally and globally.

Petra Nass