For my undergraduate thesis I worked with the RMS calculation by means of the signal expressed either in time domain or frequency domain. Since I worked also, considering a non-equally spaced frequency span I had to use the Parseval's theorem in a slightly different way...
I attached an extract of my thesis (sadly it is only in Spanish) where you can check this little adjustment. The only comment is that I used digital amplitudes (not calibrated SPL), but I think should work well for calibrated levels, since the most important consideration in the Parseval's theorem is related to the frequency and time axis...
My complete thesis it is in my google site...http://sites.google.com/site/alejandroossesven/publications
I could give you more details if you want...
--- El lun, 6/20/11, Merri J. Rosen <mrosen@xxxxxxxxxxx> escribió:
parseval s theorem.pdf