Internet of Things
The challenge of energy efficiency
In the Internet of Things, communication is predominantly wireless instead of wired. Most devices are not connected to the power grid, but draw their energy from rechargeable batteries or batteries. What does this mean for manufacturers of new products?
The Internet of Things is constantly growing. According to an estimate by analyst firm Gartner, 6.4 billion networked devices were in use in 2016. This figure is set to rise to 20.8 billion by 2020. For 2016 alone, analysts expect growth of 5.5 million devices per day. However, only a few of these are wired: the special features of the applications or remote installation locations often make this impossible; sometimes, however, it is simply not economically viable to connect the devices to the network via kilometers of cables. Instead, communication often takes place via radio. GSM modules are very popular in the Internet of Things - despite the capacity limits of this technology. In future, applications that require short latency times and high data rates will be based on the standards of the fifth generation of mobile communications (5G), for example NarrowBand IoT (NB-IoT).
Like all other components, the radio modules must operate with as little power consumption as possible in order to qualify for use in IoT devices. By dispensing with cables, the applications are completely reliant on batteries and rechargeable batteries - or on energy harvesting, i.e. the generation of electrical energy from sources such as air currents, ambient temperature or vibrations. All of these technologies have severe limitations: Energy harvesting can only generate relatively small amounts of power. With batteries and rechargeable batteries, maintenance costs increase the more often they have to be replaced or connected to the charging station. Energy is therefore a valuable commodity in the Internet of Things - and the limiting factor in many new applications.
What must applications look like if they are to be particularly energy-efficient? Analyzing the current consumption in a CMOS IC results in a voltage range between 0.4 and 0.5 V to ensure maximum energy efficiency. We are talking here about the near or sub-threshold range.
The depths of the low-voltage range
Near-/sub-threshold technologies play a crucial role in the Internet of Things. Their advantages are obvious: while increasing performance is the key measure of success in most areas of technology, the IoT is usually about comparatively simple applications, such as carrying out certain measurements and transmitting the measurement data to a central server. The focus is not on the performance of the individual devices, but on the potential of the entire network. This finding applies down to the smallest detail: measurement data is usually not made available as a continuous stream, but either on demand or at fixed time intervals. In addition, the frequency of measurement data acquisition is often in the range of a few seconds or more. How quickly a circuit works is therefore of secondary importance.
The low voltage range: If the voltage falls below a process-dependent threshold, the leakage currents dominate the energy consumption.
© FujitsuThe functionality of IoT devices can be divided into two areas: on the one hand, the area of data acquisition and the temporary storage of data, which can be operated permanently with minimal power consumption, and on the other hand, the area of data processing and communication. These are energy-intensive tasks that only take place at certain intervals. Back-bias technology can therefore be used here, which dynamically adjusts the bias voltage of a circuit to the current requirements.
The low-voltage range poses a number of challenges: Since neither typical logic libraries nor transistor models are suitable for near/sub-threshold applications, special components are required. In addition, SRAMs (static volatile memories) no longer function reliably in this range. This makes the intermediate storage of recorded data more difficult. The transistor itself also causes problems. Optimized for operation at 0.9 V, it often delivers unpredictable or inadequate results. However, potential errors associated with this often remain undetected, as the transistor parameters in series production are usually only checked at nominal voltage.
Data is recorded and temporarily stored in near-treeshold mode. The more energy-intensive data processing and communication takes place in dynamic back-bias mode.
© FujitsuFujitsu Electronics Europe (FEEU) acts as a distribution partner for Mie Fujitsu Semiconductor's (MIFS) Deeply Depleted Channel (DDC) technology. The company understands the challenges of the low voltage range and addresses them as follows: First, it provides a transistor and associated simulation models specifically designed for the low-voltage range. Second, in collaboration with the Swiss Research and Development Center Centre Suisse d'Electronique et de Microtechnique (CSEM), it has developed a logic library and memory compiler designed for near/sub-threshold applications. Thirdly, MIFS controls transistor parameters in production down to the sub-threshold range, allowing problems to be reliably identified.
DDC technology has two further advantages: Firstly, the variance of the transistor parameters is significantly lower compared to standard technologies, the SRAM memory content is therefore stabilized in the targeted voltage range. Secondly, the back-bias sensitivity should be mentioned: The characteristics of the logic can be influenced or modulated by applying a substrate voltage - with standard technologies, such an influence is only possible to a very limited extent. DDC technology allows such modulation to be carried out either statically or dynamically: A static implementation works like trimming discrete circuits. In the dynamic case, circuit parts can be modulated according to the current performance requirements of the application. This approach has a positive effect on leakage currents and therefore on the energy balance of the application.
Synergies between components
It is not only the energy efficiency of individual components that is important, but also their interaction on a PCB. The correct combination requires experience in the low-voltage range, and suitable components often have to be collected from all over the world, as the number of suppliers is limited. FEEU, for example, offers MCUs and RTCs from the American manufacturer AmbiqMicro and is therefore aimed in particular at customers for whom in-house chip development is not an option. Both products operate in the near-threshold range and are therefore designed for energy efficiency. They significantly undercut the power consumption of typical MCUs. The processor core of the MCUs is an ARM Cortex M4F, whose power consumption is even lower than that of the Cortex M0+. In addition to a 13-channel ADC, the standard interfaces of the Apollo MCUs include configurable GPIOs. These are suitable for implementing I2C and SPI interfaces, among other things. A maximum of 512 KByte Flash and 64 KByte RAM are available as memory. The Apollo MCUs are available in two different packages, either in a 41-pin CSP with 27 GPIOs or in a 64-pin BGA with a total of 50 GPIOs. In addition, the RTC integrated in the MCU is also available as a single component. The RTC variant with integrated power management unit is particularly suitable for users who do not want to or cannot change their existing MCU. Efficient RTCs are particularly indispensable when data is only recorded at relatively short intervals. During the long inactive phases, power consumption must be reduced as much as possible.
Other components are also important for IoT applications. It has already been mentioned that data storage in the low-voltage range is a challenge - and non-volatile memories in particular have a comparatively high power consumption during write accesses. FRAMs, which are highly energy-efficient, are an alternative. In addition, standard components, such as crystals and chip resistors, must be compact and low-power in order to meet the requirements of the IoT.
Author:
Dr. Klaus-Peter Dyck is Senior Manager Marketing & Application at Fujitsu Electronics Europe in Langen.













