Once again this year, the embedded world proved its reputation as the leading trade fair for the worldwide embedded community. With the latest developments around AI/KI, the fair achieved its second best visitor result in its history - congratulations and thanks for three eventful days full of fascinating insights into our digital future!
The event chose "Embedded Intelligence" as its main topic for 2019; an absolutely logical choice not only because of the connection to high-performance computers in the cloud, but also the performance of processors and SoCs (System-on-Chip) in the embedded area. Furthermore, the availability of suitable algorithms have increased so much, or have become so cheap, new technologies such as artificial intelligence (AI) with machine learning and deep learning are being used in more and more applications. This enables completely new systems which perceive the environment independently, draw conclusions from it, and make partially or fully autonomous decisions.
Edge Computing: The key to digital transformation
Kontron presented its range of hardware, software and services under the motto: "Digital Future - From Edge to Cloud - Standard and Custom IoT Solutions" together with Kontron Electronics (formerly exceet electronics) from Germany, Austria and Switzerland as well as its sister company S&T Technologies within the S&T Group.
Our motto does not come from out of the blue sky as the key to digital transformation lies in so-called edge computers. They are becoming increasingly powerful with more "embedded intelligence" and taking on things that were previously only possible in the cloud including analytics, predictive maintenance or even AI tasks such as machine learning/deep learning or object recognition. The 8th generation Intel® Core™ i or Xeon® server class processors offered by Kontron on various modules/boards and systems as well as the AMD Ryzen V1000 with extremely high GPU performance are suitable for this purpose.
Lack of qualified software engineers slows down developments
AI was THE topic of the embedded world and at many booths one could marvel at prefabricated demos with face recognition or applications from the industry. Most of the demos at the embedded/edge computer exhibitors were standard demos of the hardware manufacturers of Neural Network Processors/VPUs, and if you asked for efficient tools and software support for individual customization, you often didn't get the details. But this support is extremely important. The reason for this is that many companies lack qualified software engineers so that many projects can either not be realized at all or only with delays, which of course severely impairs competitiveness on the market. With the help of its sister company S&T Technologies, Kontron can offer competent support in consulting and the implementation of software projects from a single source.
The following example of AI/machine learning illustrates this:
- Easy start; prepared training environment from an embedded server/industrial server together with pre-installed software focusing on visual inspection/machine learning - saves approx. six weeks of preparation and training time.
- Special hardware accelerators and interfaces in this environment so that .Net and Java developers can also handle it. Instead of approx. 2000 lines in C or C# e.g. on the middleware of Intel® OpenVINO™, only approx. five commands in .NET or JAVA are required for programming the inference part.
- Complete process from a single source for the training and inference phase of AI, so that the project can be implemented efficiently.
Based on the AI plug-in for the SUSiEtec IoT software framework from S&T Technologies, a live demo on the Kontron booth demonstrated machine learning with object recognition of a module or system and automatic link to the respective product.
New open standards and compact solutions for the 'last mile' from cloud to sensor/actuator
With the future requirements of AI and 5G for higher processor performance, higher bandwidth and larger memories, the need for a new open standard for Computer-on-Modules which supplements the previous widely used standards SMARC, Qseven and COM Express®, is also increasing. Since 2018, several companies, including Kontron, have been working under the umbrella of the PICMG® Association on the specification of this new standard with two form factors: Server Type and Embedded Type. The Server Type supports CPUs up to 125W TDP, higher bandwidth with PCI Express® Gen5, LAN interfaces up to 100GB, USB 3.2 and up to eight memory bars. The Embedded Type supports CPUs up to 65W TDP, PCI Express® Gen5 and USB 3.2, as well as high resolution graphics channels, ethernet with TSN support and up to four memory bars.
At the same time the implementation of the "last mile" to actuators/sensors from IT to OT within the scope of industry 4.0 projects requires increasingly compact and cost-effective edge computers and IoT gateways. New Arm® based SoCs (System-on-Chip) with integrated industrial I/O interfaces such as digital and analog I/O, PWM, serial interfaces and field buses such as CAN up to the future TSN standard for deterministic ethernet allow the implementation of cost-effective IoT/IIoT edge computers and gateways. These transfer the data at the field or controller level via Ethernet-based networks and standards such as OPC UA and optionally TSN, if the transfer is to be deterministic, to the IT and/or an on-premise cloud. At embedded world Kontron presented the KBox A-230-LS, a compact BoxPC with multi-core architecture and up to six Gigabit Ethernet ports, whereby up to five ports are TSN-capable, four of them via an integrated switch in the SoC. There are no additional costs for an external switch, so compact and cost-effective IoT systems/gateways with direct TSN connection can be implemented.
Raspberry Pi goes Industry
Originally, the Raspberry Pi was primarily designed to give beginners easy access to hardware programming. Since then it has become an integral part of the embedded market and is increasingly used for applications in industry and medical technology, for example, in addition to the existing maker and home constructor scene. What’s more, the new CM3+ Compute Module has a long-term availability until 2026. Due to the increasing breadth of applications, there is also a steadily growing support of operating systems - there are more than 40 different ones - which are more or less tailored to specific requirements. With the release of Codesys V3.5 SP14 the functional scope of 'Codesys Control' for SoftSPS applications has also been extended. The runtime system is now available in a 64-bit version for all Arm®/Cortex-based platforms, and optionally with multicore support for Raspberry Pi, among others.
Kontron already had a SBC with the CM3+ module in 4.3" (diagonal) format and a 4.3" PanelPC in its range via Kontron Electronics. At embedded world, Kontron presented the 'KBox A-330-RPI' based on the long-term available Raspberry Pi Compute Module CM3+. This allows Kontron to leverage the huge software pool of the Raspberry Pi community. The system with the four Cortex-A53 cores is compatible with the Raspberry Pi 3B+ and has been enhanced with industrial features. These include a second ethernet interface and a serial port that can also be configured as a CANbus. For the 'KBox A-330-RPI' Kontron guarantees an availability of at least eight years.
Compact solder-on SOMs open up new savings potential
At embedded world ST Microelectronics presented a new SoC with the ST32MP157 with three processor cores, including the widely used Arm® Cortex-M4 core which has been used for many years in numerous microcontroller applications. The scalable, Linux- and Ethernet-capable hardware with comprehensive industrial I/O interfaces is predestined for cost-effective and compact industrial IIoT edge computer and edge gateway applications, from sensor and actuator to cloud with only one chip. With this latest processor architecture Kontron introduced a 2.54cm x 2.54cm solder-on SOM at the show, along with a 4.3" diagonal SBC carrier that fits behind a 4.3" display or in a compact DIN rail enclosure.
The SOMs for soldering are not only more compact than the Computer-on-Modules (COM), but also more cost-effective due to the elimination of connectors between the module and the carrier. Like other components, the SOMs can be machine soldered, eliminating the need for manual connector assembly.
So far, there are SOMs from different manufacturers in different versions, proprietary and adapted to the different SOCs and processors. There is an increasing call for a standard for solder-on modules - in addition to the previous COM standards downwards - so that users can choose between different providers for second source.
It remains to be seen when this will happen. What is certain, however, is that the world of IoT/embedded computers will remain exciting. And that we will report on the upcoming new module standards and other news here in the course of the year! In the meantime we are working on our digital future - be ready! ????