Comparison between G 185-32 (18 May 2005, upper panels) and synthetic data. Each synthetic dataset was constructed by adding to the VLT data a sequence of 20 sinusoidal waves between 1.265 and 2.5 Hz, equally spaced in frequency (Δf = 0.065 Hz). All the 20 sinusoids have same relative amplitude of 2 × 10^-4, 4 × 10^-4 and 6 × 10^-4 respectively. The various columns represent respectively: light curve with synthetic sinusoids superimposed (1), amplitude spectrum (2), amplitude spectrum of the amplitude spectrum or FT² (3), autocorrelation function of the amplitude spectrum (4). The ordinate is given in relative amplitude  ×  1000 for Cols. 1 and 2, and in σ units for Cols. 3 and 4. The dashed lines show the 4σ detection threshold, which correspond to 4 times the average noise for Cols. 2 and 3 and to 4 times the standard deviation for Col. 4. In Col. 2 the mean noise was calculated using a (smooth) cubic spline interpolation in order to consider also the noise increase at low frequency due to transparency variations.