No abstract

Sensors with long lifetimes create new applications in medical, infrastructure and environmental monitoring. Due to volume constraints, sensor systems are often capable of storing only small amounts of energy. Several systems have increased lifetime through V DD scaling [1][2] [3]. This necessitates voltage conversion from highervoltage storage elements, such as batteries and fuel cells. Power is reduced by introducing ultra-low-power sleep modes during idle periods. Sensor lifetime can be further extended by harvesting from solar, vibrational and thermal energy. Since the availability of harvested energy is sporadic, it must be detected and stored. Harvesting sources often do not provide suitable voltage levels, so DC-DC up-conversion is required.An 8.75mm3 sensor platform capable of nearly-perpetual operation is proposed. The system includes a 73kHz near-threshold ARM Cortex-M3 core that is powered by two series-connected 1mm 2 solar cells and a Cymbet thin-film solid-state Li battery through an integrated power management unit (PMU) (Fig. 15.8.1). It is suitable for volumeconstrained long-term wireless sensing applications such as intraocular pressure monitoring to detect and track the progression of glaucoma. In the 7.7µW active state, the system collects data from on-chip temperature and capacitance sensors, performs data processing using a 16kb non-retentive SRAM (NR-SRAM) for temporary storage and writes the results to a 24kb retentive SRAM (R-SRAM) (Fig. 15.8.2). Between sensor measurements the system enters a 550pW sleep state by disabling SRAM accesses, power gating the Cortex-M3 and NR-SRAM and switching the PMU to sleep mode. While asleep, the R-SRAM, wakeup controller and sleep timer are powered by a 50Hz switched capacitor network (SCN) that converts energy from the solar cells and battery. If sufficient light is available, solar energy is used to recharge the battery. When the next sensor measurement is scheduled, the wakeup controller switches the PMU to active mode by enabling a 1.2MHz clock for the SCN and a linear regulator (LR). Then power gating is disabled, allowing data collection and processing to begin.The Cortex-M3 achieves 73kHz operation at 400mV and 1MHz at 500mV while running a 64-point DFT program ( Fig. 15.8.3). The energy-optimal point for active mode operation is 2.1µW at 400mV, because further voltage scaling increases total energy consumption due to excessive leakage [4]. During sleep mode, the processor and NR-SRAM are power gated. When the system enters active mode mode, the Cortex-M3 begins program operation with pointers retained through sleep mode that denote the program location and allocated R-SRAM for sensor measurements. The sleep power is 100pW at 400mV and 460pW at 500mV, including R-SRAM, wakeup controller and balloon latch leakage plus sleep timer switching power. The idle processor lifetime is 49 years based on the 12μAh 2.9mm 3 Cymbet battery, which included in the system volume of this work.A custom SRAM was developed to minimize leakage power during sleep mode...

No abstract

The Internet of Things (IoT) is a rapidly emerging application space, poised to become the largest electronics market for the semiconductor industry. IoT devices are focused on sensing and actuating of our physical environment and have a nearly unlimited breadth of uses. In this paper, we explore the IoT application space and then identify two common challenges that exist across this space: ultra-low power operation and system design using modular, composable components. We survey recent low power techniques and discuss a low power bus that enables modular design. Finally, we conclude with three example ultra-low power, millimeter-scale IoT systems. IoT Application Space and ChallengesIoT devices are unique in that they are focused on physical interfaces, allowing them to sense and actuate the world around us. This is in contrast to previous computing platforms, such as desktop and handheld devices, which have primarily focused on human interfaces. These human interfaces require relatively large input and output devices and require the devices to be co-located with their users. In contrast, IoT devices receive input through sensors that are often highly miniaturized through MEMS technology and send out information through wired or wireless interfaces to mobile and cloud computers. This enables the placement of IoT devices in myrad new locations and applications where computing was previously absent. As a result, IoT presents the semiconductor industry with a market opportunity that may exceed that of all previous computing classes.The diversity of our physical environment results in a similarly diverse IoT application space that ranges from tiny implanted heart-rhythm monitors, to large, extremely long lifetime HVAC sensors, to sensors for oil reservoir diagnosis [1]. To categorize this wide application space we recognize four ontological properties that are critical in determining IoT device requirements: sensing modality, form factor (size), energy source / sensor lifetime, and connectivity. Table 1 lists example values for these properties. For instance, physical form factor can range from tiny millimeter size for implantable sensors to 10s of centimeters for infrastructure monitoring. Sensing modalities can include pressure, chemical, strain, and temperature, etc. In addition, the sensing modality can be characterized by measurement frequency and possible triggers , such as a timer, a detected event in the monitored value, or an external event such as a radio message.Similarly, communication connectivity is characterized by communication interval, distance, data rate/size, etc. Clearly the listed items are not exhaustive. Each property can take a very large number of possible values resulting in a highly varied application space. Some examples include 1) tiny sensors that measure temperature every 10 minutes using a secondary battery and harvesting with infrequent communication for studying Pika's in the Rocky mountains [2]; 2) continuous monitoring of brain activity in seizure patients with a contin...

No abstract