You're provided with two different ways of computing zeroth (i.e. smoothing), first or second order derivative of the data.

- Fourier transform. The data are convoluted with the nth order derivative of a windowing function (a Gaussian or a box). This is done in the Fourier (frequency) domain and, therefore, is fast. The commands are f, d, and D to smooth, take the derivative once or twice respectively.
- Cubic spline interpolation. The data are approximated by a few carefully chosen points (markers) and interpolated in between by a third order polynomial. (see b)

Christophe Leger 2009-02-24