PSD Welch method - Self-implemented Routine

Hi all,

I need to process a signal I have and estimated its power spectral density. I have implemented myself a routine that basically does what the Welch method does (segmentation, average etc) and what I get is a PSD curve that corresponds to the one from the built-in function "pwelch" for each mode apart from the first mode that is always overshooting the real value. I am pretty sure the routine is implemented correctly and I have heard this is a common problem between persons who tried to implement their own version of the Welch method. From the documentation provided is not clear how the pwelch function deals with mean value or if there is a special treatment for the first mode (removed?).

I was wondering if you could clarify the implementation of the method and/or if anybody has a solution for this issue?

Thanks, Andrea.