Edge Computing has become the hot new term, particularly, in the context of the Internet of Things (IoT) – a system of interrelated computing devices, mechanical and digital machines that have the capability to transfer data over a network without requiring human-to-human or human-to-computer interaction. “Things” live at the edge of the network, where data is collected from various sensors. In the old computing model (cloud computing), this data was uploaded to a server in the cloud, and then analyzed and processed remotely. However, scalability issues, excessive power consumption, connectivity, and latency are driving the demand for edge processing of IoT data (it is the new computing model). By handling the heavy compute processes at the edge rather than the cloud, one can reduce latency, and can analyze and act on time-sensitive data in real-time. The other often-cited benefit of edge computing is the bandwidth or cost savings of sending data to the cloud. There is plenty of bandwidth available to send data to the cloud but costs can rapidly accrue when the data traverses the WAN/Service Provider network en route to the cloud data center.

Fig 1. The Network Edge

According to Gartner, 91% of today’s data is processed in centralized data centers, but by 2022, almost 74% of all data will need analysis, processing, and action on the edge, where it originates.  New Industrial Internet of Things (IIoT) products from smart sensors to Human Machine Interface (HMI) systems, and more advanced industrial automation equipment, are being designed to address this trend. In this series of posts we will look at addressing the requirements for some of these products. This post focuses on smart sensors. 

What are Smart Sensors?

IoT applications use a large array of sensors to collect physical data to enable various functions such as predictive maintenance, flexible manufacturing, improved productivity, etc. Traditionally, sensors are simple devices that convert physical variables into electrical signals.  For IoT applications, they need to have additional properties:

  • Low Cost – to enable deployment in large numbers.
  • Small Size – to disappear unobtrusively.
  • Built-in compute resources in the form of a Microprocessor (MPU)/Microcontroller (MCU) – required to pass on sensed data to the remote system for processing/analysis.
  • Wireless – wired connections are usually unavailable.
  • Very Low Power – these sensors are typically battery-powered. The sensor should be able to operate for years without a battery change. Alternatively, the sensor should be able to operate using energy harvesting.
  • Robust Design – to minimize or even eliminate maintenance requirements.

A smart sensor incorporates predefined data-analysis and decision-making functions into an IoT enabled sensor. Incorporation of data-analysis and decision-making capabilities brings additional properties such as:

  • Self-Calibration
  • Data Pre-Processing
  • Self-Diagnostics/Self-Healing

The MPU/MCU can be used to calibrate the sensor so that it can be automatically set up for production changes. The MPU can also spot any production parameters that start to drift beyond acceptable norms and generate warnings for operators to take preventative action before a failure occurs. The sensor could work in a “report by exception” mode, where it only transmits data if the measured variable value changes significantly from previous sample values. This reduces both the load on the cloud compute resource and the smart sensor’s power requirements. If the smart sensor includes two elements, sensor self-diagnostics can be built-in. Any developing drift in one of the sensor element outputs can be detected immediately. Additionally, if one of the elements fails the process can continue with the second measuring element. Alternatively, the two sensing elements can work together for improved monitoring.

Figure 2 below shows the simplified block diagram of a smart sensor. This battery-powered device can measure relative humidity and temperature, process sensor data, and communicate the data over BLE. In this example, the smart sensor is built using the IDT TLSR8258 BLE controller chip and uses Avalanche Technology STT MRAM as boot, program, and data memory. Designers can quickly prototype such designs using any one of several available IoT development kits such as the Renesas/IDT BTG25.

Fig 2. A smart sensor designed with IDT TLSR8258 BLE and Avalanche Technology ULP MRAM

Fig 3. Front & Back of the Renesas BTG25 Development Kit Board


As sensors become smarter by including data-analysis and decision-making capabilities, they require memory with Persistence, High-Endurance, Fast-Write and Low-Power. Persistence is needed to retain data analysis programs and sensor data when power is removed. The memory needs to have high endurance because sensor data is continually written to it. It is analyzed in batches and deleted to make room for new data when power is available. This cyclical write-and-erase behavior requires high endurance. Very often, battery-powered systems are configured to go to sleep in between active states. Fast write helps conserve battery power because the processor and memory can be put into low power sleep states faster. Fast writes also help preserve data integrity in case there is unexpected power loss. Finally, low power helps maximize battery life. These requirements are met perfectly by Avalanche Technology’s STT-MRAM.

Avalanche’s STT-MRAMs are available in discrete memory, bare-die, and embedded memory – a memory macro for System-on-Chip (SoC) design. System designers can start prototyping with discrete memory devices, go to market with a Multi-Chip-Module (MCM), and finally use an SoC incorporating Avalanche’s Embedded MRAM in volume production. This approach provides the fastest time to market, maximizes system reliability, and supply chain availability, while also yielding the greatest power, cost, and board area savings. Avalanche’s Embedded MRAM is available through foundry partners such as United Microelectronics Corporation (UMC), to enable the next generation of low-power SoCs.

Fig 4. From Prototyping to Production with Avalanche MRAM

Avalanche Technology’s P-SRAM Gen 2 discrete STT-MRAM memory devices are available now in volume production.  Learn more about our discrete STT-MRAM devices HERE, and our Embedded MRAM HERE.


Smart sensors are the future of IoT.  Designers can easily implement these using a combination of sensor ICs, Bluetooth/Wi-Fi interface SoCs and MRAM. MRAM provides several benefits that ultimately enhance reliability and reduce total system cost in IoT applications.

Get our free Whitepaper on why MRAM has a clear advantage in Space applications

Download Whitepaper

Get our free Whitepaper on why MRAM is ideal for Industrial applications

Download Whitepaper