Anatomy of a smart sensor: a non-technical perspective
Recent interests in technology advancement to improve our lives have sparked the rise of the Internet of Things (IoT), automation, Industry 4.0 and Artificial Intelligence. Such phenomena are signatures of a new era in which all facets of our daily lives are improved by technology. Embedded in the core of all these new toys is a smart sensor. Smart sensors are intricate and compact devices that translate physical information into a digital form. This digital form is a critical data source that enables the development of most modern technology. In this article, Jimmy Tran — Electronics Engineer at the NSW Smart Sensing Network (NSSN) — takes us on a discovery journey inside smart sensors and elaborates on their design and functionality.
Most sensors have a similar structure. Their body consists of (A) a microprocessor (a tiny computer brain), (B) an energy management system including a power source and, possibly, a power harvester, (C) a communication system and most importantly, (D) a sensing system – a product of science, maths and engineering — Figure 1.
Figure 1: An abstract structure of a smart sensor (enclosed purple box) with respect to sensing physical data and sending information to the network.
In a smart sensor, the sensing system (D) will utilise its module to collect physical information such as temperature, moisture, air quality, light, chemical reaction and so on. The sensing module usually translates such information to a voltage signal. For instance, a temperature value becomes a voltage value and its range is mapped into a range of voltages. This translation can be modelled, experimented and must be calibrated rigorously to accurately reflect the physical data. This is where science (how to collect data), maths (how to model the data) and engineering (how to implement the data collection and calibration) come together.
Once this translation is completed, the sensing system transforms the voltage signal into a binary value — analogue to digital conversion (ADC) — as computers only understand and represent data in such form. This part is as crucial as the sensing part since erroneous conversion corrupts any operation that relies on the data. However, in saying that, this conversion is never perfect and always introduces a small error.
Once the digital data is read by the microprocessor (A), it is then processed (filtered, interpolated, extrapolated or converted). For example, in sensing temperature, the data can be converted back to the actual temperature value in digital form by a formula before being sent. The processor can also control other components for optimal performance such as improving the power efficiency and connecting and executing commands from the server via the communication system. This is the smart component in a smart sensor.
The communication system (C) is there to find an optimal way to pack and send the data off, either via a wired or a wireless system depending on the geographic requirements of the project. The choice and the design of this system rely heavily on the amount of data (data bandwidth) to be sent at a given time, power system design, and the considerations of signal degradation and losses over data transmission (link budget), which links back to the geographic requirements of the project. A battery-powered system will imply a low bandwidth design while a wired system might guarantee constant streaming of data.
The power system design should always be considered before any sensor design or sensing module. This is because the power system design determines the total amount of available resources for the whole system (power budget), analogous to how much budget you have for purchasing household items. An energy harvester is needed for a battery-powered and wireless application of a long field service while a wired system consumes the power delivered by the connected wire. Power efficiency is also required and mainly managed by the hardware design and the pattern of sleep and operational modes of the microprocessor. It will help with the choice of battery that can reduce the footprint of the whole sensor.
“Well, that is it” says Jimmy — “a crash course on smart sensor architecture design.” Once a smart sensor is running and sending data to the server, the information will then be displayed for a monitoring project or can be further processed for an AI project. But these front-end applications will be a topic for later articles.