Computes the power spectrum of a 1D data set using an FFT for speed. Data are assumed to be regularly spaced, and no weighting of data is performed (use SINFIT instead if weights are very important). The data set can be of any length, but the programme will run most quickly if there are 2**n points. If the number contains a large prime factor then it can be very slow, the TRUNCATE option allows truncation (with loss of no more than 10% of the data) to a less awkward number. All data are assumed to be of good quality.
Normalization of power spectrum is:-
power = (wave amplitude / 2)**2
The data can be tapered to eliminate spurious power arising from
a jump between the first and last data values (since the data
are effectively assumed to be periodic). If this option is
selected (answer Y) then a cosine bell is applied to 10% of the
data at each end, after removing the data mean.
The data mean contributes only to the zero frequency power, but
it may be desirable to remove it to avoid it dominating an output
plot.
A data object containing the power at each frequency is created.