The Internet of Things (IoT) is quickly becoming ubiquitous, whether in the home, on the factory floor, or say, at the doctor’s office (Fig. 1). For the system engineer, there are many approaches to architect an IoT system, but a basic design has three layers:
An end node gathers information about a process from a variety of sensors. Typical data includes temperature, pressure, speed, vibration, images, audio, physiological metrics, and much more. The node then sends the information via a wired or wireless link to a gateway layer that aggregates the information from multiple nodes and sends it upstream. Finally, the cloud or enterprise layer processes the data streams to control the operation of the various functions, as well as gathers data to identify longer-term opportunities for process improvement (data analytics).
End-Node Processing of Computationally Intensive Functions
There are, of course, alternatives to sending all of the data back to the cloud for processing and action. Pushing out processing functions to the gateway or the node—a strategy sometimes called “fog computing”—reduces the load on the network and gives a faster response. For the node, in particular, adding a microcontroller (MCU) can considerably complicate the design. If power-hungry digital-signal-processing (DSP) functions such as finite-impulse-response (FIR) filters and fast Fourier transform (FFT) frequency analysis are a requirement, the challenges proliferate, especially if it’s a battery-powered home or medical IoT application.
The MCU for a computationally-intensive IoT node must perform three main functions:
1. Acquire the data to be processed, which can arrive in either analog or digital format. A pressure sensor, for example, might require an analog front end to buffer and amplify the signal, feeding an analog-to-digital converter (ADC); or the data can arrive via a serial link such as SPI, I2C or UART. The design will be more compact and lower-cost if the MCU includes the analog or digital functions as internal blocks.