We propose a new method for the small-signal stability analysis of power systems based on the spectral decomposition of a square H2 norm of the transfer function. Compared with the dynamics of H2 and H-infinity norms of the transfer functions, the analysis of the behavior of individual eigen-components allows the earlier identication of the pre-fault condition occurrence. Since each eigen-component is associated with a particular eigenvector, the potential sources of instability can easily be localized and tracked in real time. An important class of systems operating under the pre-fault conditions near the boundary of stability is considered. We demonstrate that in such cases several ill-stable modes can increase the system energy up to a critical level much earlier due to their synergetic effect. In particular, an ill-stable lowfrequency mode can act as a catalyst increasing the energy in the system. An illustrative test for the stability analysis of a real small power grid at Russky Island is provided.