On the rationale and the workings of a filter-based approach to trend-cycle estimation