Issue link: https://resources.mouser.com/i/1442853
14 data and therefore choose different sensors," he says. "When I speak with design firms, I realize that the real dialogue we are having is not about the type of sensor they'll be using but about the type of data they need to make their product work." From Dupaquier's perspective, most smart solutions underuse the computing capabilities of their microcontroller units (MCUs). "In 98% of the use cases I see, the MCUs are underused. Most are used for simple power management and networking." He believes that if designers made better use of artificial intelligence and machine learning in their designs, their systems would deliver more intelligence, greater energy efficiency, and greater security at a lower cost. One reason they are not using MCU computing capacity more effectively is they do not understand the power of MCUs and smart sensors as software platforms. Dupaquier has demonstrated this power to many clients who believed that they needed to completely redesign a product to obtain greater functionality. In many cases, Dupaquier was able to add software that would run alongside existing system software to enhance its functionality. "We became specialists at hacking into microcontrollers. We would add our software next to the company software, and we always had room for it." As a broad design strategy, Dupaquier says that processing as much data as close to its origin as possible saves power and is more secure. "How much energy you use to run a cycle of knowledge is a well-known phenomenon," he explains. "You will increasingly see the data analysis happening directly in the sensor rather than in the MCU. Why? Because if you analyze at the sensor, the processor doing the first-level analysis uses less power than the MCU, which saves battery life. You also save on communication between the two components." There is an even bigger increase in power consumption and an accompanying loss of speed when uploading volumes of raw data to the cloud for processing. "If you analyze at the sensor, the processor doing the first-level analysis uses less power than the MCU, which saves battery life. You also save on communication between the two components."