FFT or 1/3 Octave analysis?

When expressing SPL of a continuous broad spectrum noise source is it best to express this as RMS generated from FFT, or RMS generated from 1/3 octave band analysis?

08 October 2013 808 1 View