Final Activity Report Summary - TRANDSSAT (Total Realization of Analogue and Digital Systems on Silicon for Ambient Technology) The objective of the project was the implementation of the wireless link for 'Wireless sensor networks' (WSN). These networks consist of thousands of small nodes spread in the environment to collect and share data and being able to operate independently for time spans of the order of years. The single nodes are powered by small batteries or by energy scavenged from the environment (from light, vibrations, heat etc) and the consequently very low power budget imposes very stringent requirements for all nodes subsystems, including the wireless link block. Moreover, to be embedded in the environment without sensibly affecting it, the nodes must be very small and a natural way of achieving a small form factor is requiring the full integration on silicon of the electronic circuitry.Starting from those objectives, the first step of the project was to identify and purposefully design techniques enabling low power and fully integration of wireless transceivers. Since most of the energy in a communication system with very low packet rate is spent at the receiver waiting for packets and listening to the channel, energy reduction can be achieved by powering the radio only for a small fraction of the total time, i.e. with a very small duty-cycle, only when communication is needed. The duty-cycled receiver needs to be synchronised in time in order to be active in the same time periods when transmitters are sending data to it. Moreover, frequency synchronisation is required between each pair of nodes to be able to transmit and receive data in the same frequency band.However, if fully integration is required, usual communication techniques cannot be employed, as a very precise frequency and time reference, commonly derived from an external quartz crystal, is not available. A clock generator continuously run and turns on the wake-up radio when packets from other nodes are expected, according to an ad-hoc designed protocol. Since the clock generator is fully integrated is accuracy is not very high and the packet arrival time can only been estimated coarsely. The wake-up radio must then be a very low power transceiver whose function is only to detect the packet arrival and turn on a power-hungry main radio only for data demodulation.Using this approach, the main radio can be a standard receiver with high power consumption, since it is employed for a very small fraction of the total time, while ad-hoc fully integrated ultra-low-power circuits for the clock generator and the wake-up radio must be built. The main achievements of the project, apart from the study and proposal of this innovative duty-cycled wake-up radio scheme, are the design of the fully integrated frequency reference (or clock generator) and the fully integrated wakeup radio suited for this particular application.A frequency reference based on the electron mobility in a MOS transistor has been proposed. Mobility has very strong but well-defined temperature behaviour. If a reference frequency proportional to the mobility is generated, it must be compensated over the temperature range of interest. Consequently, a mobility-based oscillator and a temperature sensor for its compensation have been implemented. The temperature sensor alone represents the state-of-the-art for temperature sensors implemented in deep-submicron CMOS processes. It achieves a trimmed accuracy of 0.2 degrees Celsius (3sigma) over the temperature range form - 70 degrees Celsius to 125 degrees Celsius. The combination of the temperature sensor and the mobility based oscillator achieves a frequency inaccuracy of less than 2.6 % over the temperature range from -55 degrees Celsius to 125 degrees Celsius after a single-point trim and less than 0.5 % after a two-points trim with a power consumption of 51 JW.
