RF filters size reduction was identified as a key need a few years back when operating bands were expanding rapidly- from 2 (analog) to 4 (2G) to 9 (3G) to 58 (4G) and now 85 (5G). But the number of bands is not the only issue forcing smaller RF filters. With each new system generation(2G/3G/4G/5G..) the complexity of the protocol has increased. Relative to the RF, that complexity meant multiplying the number of filters required. Diversity doubled the number of receive filters required for each band and MIMO caused a similar multiplicity. It was conceivable from the onset that the actual PAs, LNAs, and switches could shrink or/and serve multiple bands from a single die. The filters on the other hand, that provide frequency isolation, could not serve more then a 1 or 2 bands each. The filters are also based on physical parameters so shrinkage is limited if performance is to be maintained. How are the filters keeping up when the needs are growing (over 20x since 3G)? In short, the sequence is- die shrinkage (continuous but limited improvement), wafer level packages (when embedded in front end modules(FEMS)), and the latest (within last 6 months)- multiple filters per die. With the latest developments, the filters occupy 7.5x less space then they did 10yrs ago. This is short of the 20x requirements caused by increased bands and protocols but a lot better then most expected 10yrs ago when the problem was in sight. In episode 2 we will see another 1.5x improvement.
Related Component: SKYWORKS SKY53735-11
Related Device: OPPO Reno 3 5G