We are launching today two new inertial modules, the LSM6DSRX and ISM330DHCX, with the former serving high-end consumer products, such as virtual and augmented reality headwear, while the latter will fit in industrial applications. They are the newest additions to our smartest family of iNEMO microelectromechanical systems (MEMS), which is recognizable by the letter X at the end of their nomenclature and the presence of a machine learning core. The new models offer greater precision with an angular rate capable of reaching 4,000 degrees per second (dps) or twice as much as the previous generation. Additionally, the angular rate typical zero-rate-level change vs. temperature, which measures the variation of bias as the environment gets hotter or colder, drops from ±0.010 dps/ºC to ±0.005 dps/ºC, meaning that the new components are more stable and can thus offer more accurate results.
LSM6DSRX and ISM330DHCX: A New Front End to Open the Door to More Applications
Readers of the blog will remember the LSM6DSOX that￼ made a public appearance in 2018 and officially launched in 2019. The new LSM6DSRX is not its replacement but a different approach since it offers greater precision and performance but increases the typical power consumption. In high-performance mode, the gyroscope needs 0.90 mA, while using the gyroscope and the accelerometer requires 1.2 mA. Comparatively, using the two modules draws 0.55 mA in the LSM6DSOX. The numbers of the LSM6DSRX are still low, and a manufacturer designing a drone, a VR headset, or a vacuum robot will undoubtedly take the precision over the slight increase in consumption. However, we’ll also continue to offer the LSM6DSOX for the engineers that don’t need the extra accuracy but work with small batteries.
The difference in specification between the LSM6DSOX and the LSM6DSRX comes from the fact that they both use a different front end, which is responsible for reading the moving mass. However, both generations of MEMS share the same architecture and digital engine, which takes the analog signal and transforms it into a digital one. They also have the same First-In-First-Out (FIFO) buffer and compression algorithm, thus testifying to the robustness and performance of the underlying system of both of these components. The front end reflects our desire to produce a module with greater performance and accuracy to satisfy a broader range of use cases. Additionally, the LSM6DSRX inaugurates the use of the S4S interface in the X family of MEMS to facilitate sensor synchronization with a Qualcomm host.
LSM6DSRX and ISM330DHCX: A New Industrial Module and a More Accessible Machine Learning Core
If the LSM6DSRX is not the first inertial sensor with a machine learning core, the ISM330DHCX is the first version of this architecture for industrial applications. As asset tracking becomes increasingly more prevalent and predictive maintenance is ever more accessible thanks to cost-effective development platforms and cloud solutions, an industrial-grade sensor with machine learning capabilities becomes even more critical. Compared to the 5 degrees per hour of the LSM6DSRX, its industrial brother has a bias stability of only 3 degrees per hour, thus signifying the even greater stability of its measurements over the same period. The ISM330DHCX also has a broader temperature range of -40ºC to +105 ºC, whereas the more general-purpose device goes up to +85ºC, and it includes a temperature compensation system to better maintain its accuracy, even under the harshest conditions.
Both devices also feature a similar machine learning core as the previous inertial sensor, but whereas the LSM6DSOX had 256 nodes, the new models have 512. It uses eight configurable decision trees to process the information it captures and make certain deductions, such as throwing an interrupt or classifying certain activities. The ability to undertake these processes within the sensor instead of a microcontroller dramatically reduces the power consumption. We are also announcing that we’ll soon release a new version of our Unico GUI that will help build the decision tree out of Python, MATLAB, Weka, or RapidMiner. Developers will thus be able to streamline their workflow, and even enthusiasts or startup will be able to use our tools to create the configuration files that they will then load onto the LSM6DSRX or ISM330DHCX. Unico can also take advantage of the 16 Finite State Machine present in both devices.
Where to Begin: Development Boards
The best way to start testing the new inertial sensors is to get their respective development boards. The LSM6DSRX is on the STEVAL-MKI195V1, and the ISM330DHCX is on the STEVAL-MKI207V1. Both are compatible with the STEVAL-MKI109V3 motherboard, which embarks an STM32F401VE microcontroller. It’s the quickest way to start experimenting with the Unico GUI and begin working on a prototype. We even offer Finite State Machine and Machine Learning examples, so developers can experiment with our scripts and data to grasp the components’ capabilities. Hence, even an enthusiast with no knowledge of data science can follow the steps outlined in the README files of each example package and run the demo applications. We also have a community dedicated to MEMS machine learning and artificial intelligence that can offer quick support and where anyone can ask questions.