Frequency spacing and oversampling ratio

The width of the interval between consecutive frequencies may be specified by the keyword freqspacing.

An alternative method is the automatic determination of the spacing by means of the oversampling ratio. In case of equidistantly sampled time series, the frequency spacing is defined by

$$\delta f := \frac{1}{T}\, ,$$ (6)

where $T$ denotes the width of the total time interval. For non-equidistant time series, it is advisable to use a denser frequency sampling,

$$\delta f := \frac{1}{\Omega T}\, ,$$ (7)

where $\Omega$ is the oversampling ratio. This quantity may be specified using the keyword osratio. The default value is 20, which is - in most cases - sufficient for practical use.

Example. The sample project limits illustrates the use of the keyword freqspacing, an example for the keyword osratio is provided in the sample project nyos. Both samples use the V photometry of IC4996#89 as input time series. The line

freqspacing 0.001

in the file limits.ini forces SIGSPEC to calculate Fourier amplitudes and sigs for every 0.001 cycles per day. In the file nyos.ini, there is a line

osratio 12

which overrides the default oversampling ratio of 20. Fig.9 compares the standard spacing from Example SigSpecNative, p.), with the spacings obtained applying the two above modifications.

Figure 9: Close-up for the significance spectra generated by the projects SigSpecNative, limits and nyos around the main peak for the V photometry of IC4996#89. Different settings for frequency spacing and oversampling ratio are applied.