Without data, the IoT would be useless. The electromechanical and electronic sensors used to collect information are just as important to the IoT’s digital brains as sight, smell, hearing, and touch are to human organic ones.
However, unlike the limited range of human senses, engineers can use the IoT’s sensors to detect almost anything: from low frequency vibration to high humidity, or low light intensity to high carbon monoxide concentration.
The IoT’s sensors form the critical connection between the real world and its cloud-based digital twin. And the trueness of the digital twin and how it evolves is largely dictated by the frequency, accuracy and precision of the data gathered.
As humans live in an analog world, light, heat and sound vary constantly, and our biological sensors allow us to detect even the tiniest differences in color, temperature, or audio frequency. However, computers need discrete values for their calculations. To achieve this, engineers use analog-to-digital converters (ADCs) to transform the continuous signals of the inputs into digital representations.
An ADC works by sampling an analog waveform at regular intervals and assigning a digital value to represent the magnitude at each point. Because only a finite number of digital values represent a continuous waveform, the digital representation will vary in discrete steps. This process is called quantization because it produces a quantized (countable) version of the original waveform. The number of digital values determines the resolution, typically given in the number of bits supported by the ADC. A 10-bit ADC, for example, quantizes the analog signal into 1024 (210) discrete values.
Sensors can either directly incorporate ADCs—measuring an analog signal and outputting a digital one—or they can output an analog signal proportional to the analog input so that an ADC inside a computing device can later convert it.
The wide range of commercial sensors allows engineers to choose exactly which they need for their IoT application. But it turns out that while dozens of sensors are available, IoT applications rely on relatively few mainstream types. For instance, 31 percent of the sensors used by the Industrial IoT (IIoT)—which has embraced sensors primarily to keep an eye on machines such that maintenance is performed only when necessary—monitor vibrations. 18 percent detect light and/or color, 15 percent pressure, and 12 percent temperature, according to Automatika, an open access journal for control, measurement, electronics, computing and communications. That’s just four types making up 76 percent of all IIoT sensors.
And it’s not just the industrial sector that makes use of these common types. Smart buildings, for example, use occupancy sensors to determine if office lights need to be switched on because there are people in the room. And color sensors are a staple of wearable IoT devices, for example, to check the oxygen level of the wearer’s blood.
Further, the accelerometers that are the key to monitoring vibration and movement are widely used in wireless devices, including smartphones, wearables, and asset trackers. One novel application is to keep track of the speed and acceleration of unmanned aerial vehicles (UAVs) to make sure they are flown safely.
Even though many applications use just a few sensors, they are available with different measurement ranges, resolutions, and sensitivities and come with different methods for reading out the data. The integration of the sensor into a hardware application can be overwhelming for the software engineer.
Prototyping kits like the Nordic Thingy:53 help short-circuit sensor integration challenges by providing platforms that include already integrated measurement devices for typical IoT applications, making it the path of least resistance to IoT application development. Thingy:53 boasts a suite of sensors, including a built-in IMU, a low-power accelerometer, temperature, humidity, air quality and -pressure sensors, a digital MEMS microphone, and color and light sensors.
For the Thingy:53, Nordic has collaborated with Edge Impulse, a U.S.-based machine learning (ML) specialist, to include ML firmware with the prototyping platform. The firmware enables developers to rapidly collect sensor data and test embedded ML models on the Thingy:53. The prototype kit points to a future where sensors provide a continuous stream of data to feed ML algorithms running directly on the embedded devices they are a part of. This enables IoT systems to learn from collected data, identify patterns and make informed decisions with minimal intervention from humans.
Building advanced applications with today’s and tomorrow’s sensors will be vital if we are to maximize the potential of the IoT, optimize its functionality, and make the world a better place.