In today’s world, wireless communication is playing an important role. Users of smartphones, tablets, laptops, and smartwatches use wireless communication for part of their social life, to follow the news, to virtually attend a meeting e.g. while commuting, to find their way, … and the expectation is that in the future more people will do that. This will give rise to even more data that is transported in a wireless fashion. Most of the data will consist of high-definition video and transport of data will happen with very low latencies. To realize all this, the fifth generation of wireless communication (5G) is about to come, with deployment starting in 2020. 5G will provide peak data rates of 10 gigabits per second (Gbps). Beyond the 5G era, the demand for higher data rates will not stop (see Figure 1): it is expected that 6G, the successor of 5G, will have to support wireless data rates up to 1 Terabit per second (TBPS). To support this tremendous data rate, the operating frequency will have to move to the so-called Terahertz band between 100 GHz and 10 THz, whereas 5G will use frequency bands below 100 GHz. In this THz band, lots of bandwidth is available for high data rate wireless communication, but the high operating frequency is challenging for the design of active circuits.
This Ph.D. research concerns with LO (Local Oscillator) generation and distribution for wireless communications beyond 100GHz. High spectral purity or low phase noise LO signals are desired for high spectral efficiencies which ultimately results in higher data rate for a given bandwidth. Phase noise in general increases with frequency hence synthesizing low phase noise oscillators is especially challenging at millimeter-wave. Also, both transmit and receive beamforming is implemented at higher frequencies to compensate for the electromagnetic path loss. Hence, LO signals must be distributed across multiple transmitters and receivers which are half a wavelength apart. Hence LO architecture must be designed in such a way that not only results in low phase noise but also minimizes LO distribution power. Also, direct conversion architectures are generally preferred at millimeter-wave due to simple system architecture. These architectures also require low phase error quadrature signals which adds to the challenge.
The usage of the millimeter-wave band beyond 100GHz will first start with the D band (110-170GHz). Hence during the first part of the Ph.D., LO generation architectures for D band communications will be studied and implemented in sub-nanometer CMOS technologies. After that, the frequencies around (275-325GHz) will be targeted.