An arsenal of refined codes for the frequency analysis of time series resulting from ground-based observations, which typically consist of hundreds or thousands of data and contain tens of different periodicities, is currently available for asteroseismology. Nevertheless, satellites like COROT, MONS or Eddington are expected in the next future to produce light curves containing millions of measurements which, due also to their very low noise levels, will allow us to detect hundreds or thousands of excited modes: a jump which entails a great necessity of new tools. With a view to helping with this requirement, I am setting up a new method for the deconvolution of extensive high resolution frequency spectra from their spectral windows. The new approach can be considered a development of the well-known CLEAN algorithm with two significant improvements: I) in order to increase the accuracy in frequency determinations, an iterative non-linear adjustement is performed at each step; II) in order to recover part of the former computational speed, at each step a fraction of the spectral signal due to a set of independent components is subtracted from the spectrum instead of a fraction of the largest component. Finally, no re-convolution with the main peak of the spectral window is operated on the resulting spectrogram. A first test performed using a synthetic light curve produced by H. Kjeldsen in the framework of the MONS H&H exercises yielded encouraging outcomes: in spite of a time base of one month only, a mean square difference between output and input frequencies of about 0.2 Hz could be determined. Moreover, rotational splittings of the order of 0.3 Hz (assigned, in the simulated data, to Cen A and B) could be detected.

 
Print this abstract