Without higher integration I don’t think 5G will be feasible, especially from an infrastructure side. On the handset side, I think process geometries are key to reduce the power levels so that battery powered devices can be enabled. Generally, smaller process geometries aid in reducing digital power. There are exceptions, but in general for digital processing, this is beneficial. On the infrastructure side, a typical basestation including small cell will require a very large number of radios. Our customers are looking at anywhere between 32 and 1024 per ‘sector’. This high radio count creates a lot of digital data that must be processed. Smaller process geometries will be required here. As for the radio, higher channel count will help some in terms of power but in general integration will not reduce power without careful attention to an optimized radio architecture. Additionally, for mmW 5G, synthesis and distribution of these frequencies, currently as high as 39 GHz, will be very challenging. Typical low cost PCB manufacturing doesn’t perform well at these frequencies and where entire mmW circuitry can be integrated into a single device, losses will be lower (and dissipation lower as well) and in the end systems should be less expensive and more manufactureable.
