Design of Frequency divider with voltage vontrolled oscillator for 60 GHz low power phase-locked loops in 65 nm RF CMOS

Resumen

Increasing memory capacity in mobile devices, is driving the need of high-data rates equipment. The 7 GHz band around 60 GHz provides the opportunity for multi-gigabit/sec wireless communication. It is a real opportunity for developing next generation of High-Definition (HD) devices. In the last two decades there was a great proliferation of Voltage Controlled Oscillator (VCO) and Frequency Divider (FD) topologies in RF ICs on silicon, but reaching high performance VCOs and FDs operating at 60 GHz is in today's technology a great challenge. A key reason is the inaccuracy of CMOS active and passive device models at mm-W. Three critical issues still constitute research objectives at 60 GHz in CMOS: generation of the Local Oscillator (LO) signal (1), division of the LO signal for the Phase-Locked Loop (PLL) closed loop (2) and distribution of the LO signal (3). In this Thesis, all those three critical issues are addressed and experimentally faced-up: a divide-by-2 FD for a PLL of a direct-conversion transceiver operating at mm-W frequencies in 65 nm RF CMOS technology has been designed. Critical issues such as Process, Voltage and Temperature (PVT) variations, Electromagnetic (EM) simulations and power consumption are addressed to select and design a FD with high frequency dividing range. A 60 GHz VCO is co-designed and integrated in the same die, in order to provide the FD with mm-W input signal. VCOs and FDs play critical roles in the PLL. Both of them constitute the PLL core components and they would need co-design, having a big impact in the overall performance especially because they work at the highest frequency in the PLL. Injection Locking FD (ILFD) has been chosen as the optimum FD topology to be inserted in the control loop of mm-W PLL for direct-conversion transceiver, due to the high speed requirements and the power consumption constraint. The drawback of such topology is the limited bandwidth, resulting in narrow Locking Range (LR) for WirelessHDTM applications considering the impact of PVT variations. A simulation methodology is presented in order to analyze the ILFD locking state, proposing a first divide-by-2 ILFD design with continuous tuning. In order to design a wide LR, low power consumption ILFD, the impacts of various alternatives of low/high Q tank and injection scheme are deeply analysed, since the ILFD locking range depends on the Q of the tank and injection efficiency. The proposed 3-bit dual-mixing 60 GHz divide-by-2 LC-ILFD is designed with an accumulation of switching varactors binary scaled to compensate PVT variations. It is integrated in the same die with a 4-bit 60 GHz LC-VCO. The overall circuit is designed to allow measurements of the singles blocks stand-alone and working together. The co-layout is carried on with the EM modelling process of passives devices, parasitics and transmission lines extracted from the layout. The inductors models provided by the foundry are qualified up to 40 GHz, therefore the EM analysis is a must for post-layout simulation. The PVT variations have been simulated before manufacturing and, based on the results achieved, a PLL scheme PVT robust, considering frequency calibration, has been patented. The test chip has been measured in the CEA-Leti (Grenoble) during a stay of one week. The operation principle and the optimization trade-offs among power consumption, and locking ranges of the final selected ILFD topology have been demonstrated. Even if the experimental results are not completely in agreement with the simulations, due to modelling error and inaccuracy, the proposed technique has been validated with post-measurement simulations. As demonstrated, the locking range of a low-power, discrete tuned divide-by-2 ILFD can be enhanced by increasing the injection efficiency, without the drawbacks of higher power consumption and chip area. A 4-bits wide tuning range LC-VCO for mm-W applications has been co-designed using the selected 65 nm CMOS process.