Spectral Estimation Analysis with Different Segment Lengths

Investigation of three different L values: L=256 (3 data segments), L=128 (7 data segments), and L=64 (15 data segments). The corresponding spectral estimation plots demonstrate that windowing effectively reduces false spectral peaks but further smoothes actual peaks. For L=64, the spectral line at ω=0.8π is clearly identifiable, but adjacent peaks remain difficult to distinguish. L=128 provides optimal balance between resolution and detection capability, while L=256 yields superior results with clear identification of three spectral lines and their amplitude relationships. Beyond Welch's method, alternative approaches like Bartlett's method can be implemented using periodogram averaging for power spectrum estimation.