Embedded Systems Energy-Efficient Solutions for Wearables and IoT Edge Nodes
How to approach challenges in terms of performance- and energy- efficiency while designing IoT applications by utilizing synergies within a single product portfolio.
The Internet of things is mostly wireless – cables are hardly ever used in the IoT. The special aspects of the applications or remote installation sites often make it impossible to reach edge nodes with a cable. Therefore, the vast majority of things communicates by radio and is not connected to a centralized power grid. The same applies for wearables.
IoT applications are therefore largely dependent on batteries, rechargeable power packs and energy harvesting. However, all of these technologies have certain limitations. For instance, only relatively small energy can be generated by using energy harvesting. Battery and rechargeable power pack-operated devices, on the other hand, incur high maintenance costs since they must be regularly provided with new power storage units or be connected to the charging station. Energy is thus a scarce and valuable commodity with the Internet of Things. Therefore, in order to be successful in this market segment over the long term, the energy efficiency of your applications needs to be a main focus.
Advantages of near-/sub-threshold technologies
An analysis of the leakage currents in a CMOS IC shows that the optimal voltage range is between 0.4 and 0.5 volts for applications that are considered to be particularly energy-saving. This area is commonly referred to as near- or sub-threshold and plays an important role in the IoT. A basic representation of standard technologies is shown here in figure 1. This clearly shows that leakage currents dominate power consumption starting with a process-dependent threshold.
Fujitsu Electronics Europe (FEEU) is the distribution partner for the Deeply Depleted Channel Technology (DDC) from Mie Fujitsu Semiconductor (MIFS) and offers its customers the chance to realize applications in the low voltage range. Extremely energy-efficient near-/sub-threshold devices can be implemented by using a foundry model. The solutions that MIFS offers are intended for the 55nm and 40nm technologies because mask costs are comparatively low in this range.
The advantages low-voltage technologies have to offer for IoT are obvious. Whereas in most applications a constant increase in performance is the main goal, the IoT often only seeks to carry out certain measurements and then make this data available either on demand or at certain time intervals. The frequency of measurement data recording is often in the range of a few seconds. Therefore, speed is of secondary importance in most cases. This results in the breakdown of the building block shown in figure 2. The left side is permanently operated with the lowest power consumption. Processing and transmission of the data takes place only at certain intervals, which is why the right side can be operated dynamically.
The challenges for the low-voltage range
Working in the low-voltage range comes with a number of challenges. To start with, neither the typical transistor models nor the logic libraries are suited for use in near-/sub-threshold applications. Also, SRAMs no longer function reliably within this range, which makes temporarily storing the collected data practically impossible. The transistor itself, which has been optimized for 0.9 V operation, does not provide optimal results either. The transistor parameters are generally controlled only at rated voltage in subsequent series production.
MIFS addresses these challenges in three ways. First, they offer a transistor as well as the respective simulation models that are specifically designed for the ultra-low voltage range. Second, a logic library and memory compiler designed for low voltage were developed in cooperation with the Swiss research center CSEM. Third, MIFS monitors transistor parameters in production all the way into the sub-threshold region.