In the last decade, the filter community has innovated both design techniques and the technology used for practical implementation. In design, the philosophy has become “if you can't avoid it, use it”, a very practical engineering approach. Modes previously deemed spurious are intentionally used to create in-line networks incorporating real or imaginary transmission zeros and also reduce the number of components and thus further miniaturize; spurious responses are re-routed to increase the passband width or stopband width, frequency variation in couplings is used to create complex transfer functions, with all of these developments using what was previously avoided. Clever implementations of baluns into passive and active networks is resulting in a new generation of noise-immune filters for 5G and beyond.
Finally, the use of a diakoptic approach to synthesis has appeared an evolving approach in which small blocks (“singlets”, “doublets”, etc.) are cascaded to implement larger networks, (reducing the need for very complex synthesis), with this new approach promising a large impact on the implementation of practical structures. Filter technology has migrated towards “observe it and then adapt it”, pragmatically repurposing tools not specifically originally intended for the applications. Combinations of surface wave and bulk wave resonators with L-C networks are improving the loss characteristics of filters in the region below 2 GHz.
Lightweight alloys and other materials designed for spacecraft are being used in filters intended for space, to provide temperature stability without the use of heavy alloys such as Invar. Fully-enclosed waveguide is being replaced in some cases by planar and quasiplanar structures propagating quasi-waveguide modes. This is generically referred to as SIW (Substrate Integrated Waveguide). Active filters trade noise figure for insertion loss but perhaps will offer advantage in terms of size and chip-level implementation. Finally, the era of reconfiguration might be approaching, as the basic networks are evolving, perhaps lacking only the appearance of lower-loss, higher-IP solid-state tuning elements.