*4.3. Sensitivity to the Number of Samples*

In Figure 6 the standard deviations of the errors versus *N* are reported, for a given signal with two tones of the same amplitude and for two different noise levels. As expected, the standard deviations decrease when the number of processed samples increases. For both noise levels, the effect of the tone distance is less significant for high *N*. In the case of the lowest SNR value (5 dB), the trend is quite the same for all the methods, since the variability due to noise is comparable to the systematic effect of IFFT; meanwhile, for the highest SNR value (40 dB), the parametric algorithms and IFFTc show better performances.

**Figure 6.** MSE of *δ* versus the number of processed samples for a two-tone signal with *A*<sup>1</sup> = *A*<sup>2</sup> = 1, *d*<sup>12</sup> = 3, *N* = 256, *f*1/Δ*f* = 40.2, random phases, SNR 10 dB (on the left), and 50 dB (on the right).
