When you think about the IoT, what comes to mind? Do you think about large industrial systems: automated factories, turning out locomotives and jet engines? Do you think about the electronic components needed to monitor and control the system, the reams of data required to assemble predictive algorithms or the tempos of a modern city?
And how much attention do you pay to the sensors, the signal conditioning devices, and the microcontrollers that inevitably populate any version of the IoT you can think of? Whether you’re thinking of Industry 4.0 (Siemens’ term for digitally-controlled factories), or your remotely-controlled garage door openers — of which Amazon’s Echo is among the latest and most sophisticated incarnation — you’re tapping into arsenals of versatile microcontroller and processor hardware, and libraries of widely-sourced software.
Analysts tell us that more than $20 billion worth of IoT devices will be deployed within the next few years. Whether they serve as replacements for previous-generation programmable logic controllers (PLCs) or hosts for new-generation smart homes, it’s the low-cost semiconductor building blocks (think: 50-cent processors) which are enabling the IoT enabled revolution.
The precision required for IoT applications
The building blocks for an IoT sensor node invariably include carefully chosen sensors, signal-conditioning electronics (amplifiers and data converters) to elevate the low-level sensor signals, a microcontroller which uses local memory to evaluate the data pulled in by sensor inputs, and a communications module (increasingly wireless) which publishes the microcontrollers’ reaction. The building block functions are largely the same whether you’re monitoring an automated factory or a baby asleep in its crib.
What is different from application to application is the speed, precision and cost parameters within which the components must operate. The speed with which sensor data is collected and evaluated is a function of the clock rate of the controller, as well as the conversion speed of the sensor and data converter front end. The resolution is typically a function of the data word fed to the controller by the data converter.
Invariably, the speed and resolution of the sensor node electronics are greater than the requirements of many applications. Current microcontroller technology, for example, allows low-power microcontrollers to process 32- or 64-bit data at billions of clock ticks per second. Current data converter technology, for instance, allows 12- and 16-bit converters to sample data at high MHz rates (150- to 250-Msamples per second in broadcast receivers). Yet, for machine tool applications — automated factories — the 12-bit converter with an 8-bit processor (and a folded data word) has long been something of the standard. The electronics of, say, a manufacturing robot arm could pick out the rotational pattern of the arm much faster than the mechanical arm could travel there.
And this process of grabbing sensor readings, converting them to digital words, and comparing those words with predicted patterns remains the same for practically all other monitoring and control application, from robot manufacturing arms to automotive lane change warning systems and hospital patient monitoring equipment.
Pay attention to the electronics components, sensors, data converters and low-cost/low-power microcontrollers as these provide the clues as to the gadgetry that will show up in the very near future. The processor word-width (8-, 16-, 32- or 64-bit) and clock rate will point to the applications enabled by processors and data converters. A sampling of IoT processors suggests that the ARM Cortex architecture is among the most popular microcontroller cores.
A sampling of IoT processors and development kits
Intel, whose processors dominate the computing space (both PCs and data center servers), hopes to extend its footprint in automotive realms with its Atom-based E3900 processor series. There is, be aware, an ongoing debate as to where to place microcontroller intelligence on a network of sensors and controllers. Analyzing large data sets would encourage placement near the computational hub. A fast response to sensor query would suggest placement near the edges of the network.
Intel suggests that the low-power consumption and fast-response time of its IoT processor enable the processor to be physically placed close to the input sensors. This is especially valuable in automotive cockpit displays, where drivers must remain aware of the smart car’s operation at all times. Additionally, Intel’s specifications point to the graphics capability (4096 x 2160 via DisplayPort) of the processor for dashboard readings.
In contrast, Asus’ Tinker Board targets multi-media applications. It’s powered by an ARM-based graphics processor (GPU), though, like other processors, Tinker Board promotes its versatility. Targeted applications include high-resolution media playback, gaming, computer vision, and gesture recognition. Specialised video processing algorithms in the development kit include H.264 and H.265 playback for HD and UHD video. The Tinker Board is equipped with a high-definition data converter for 24-bit/192kHz digital audio.
Arduino remains among the most popular development boards available. Arduino promotes the versatility of its prototyping kits, calling out “artists,” “beginners,” and “hobbyists” as potential users. The company’s multi-function Lucky Shield, for example, reads a variety of environmental conditions like barometric pressure, altitude, luminosity, temperature, motion and proximity with its onboard sensors. Many current IoT systems have undoubtedly begun life as an Arduino prototype.
The Arduino Otto architecture is targeted towards home security. STMicroelectronics’ implementation includes Wi-Fi and an on-board stereo microphone (ideal for voice-actuated applications). STMicroelectronics is also active in 9-axis inertial sensors — MEMS sensors — which are used by athletes and fitness buffs for performance analysis.