Integration and performance
Whether it be in consumer electronics or industrial applications, the trend towards increased integration is one which is continuing.
Having a single box responsible for a whole raft of functions is not just cheaper, but also reduces overall system complexity, and makes maintenance simpler.
According to congatec’s Christian Eder, the traditional configuration which required programmable logic controllers (PLCs) another system with the human machine interface (HMI) is giving way to industrial PCs.
“The functionality can be melded into one single box quite easily,” Eder explained. “With multi-core platforms, if you can separate the cores.”
Utilising virtualisation solutions such as the Real Time Hypervisor from Real Time Systems, one core can take on the function of the PLC, running a real-time operating system, while another core can run an embedded operating system and serve as the user interface. These systems can run the same software as traditional PLCs and HMIs.
These two completely separate systems exist in the one industrial PC, and are connected to each other by a virtual Ethernet controller.
Integration is also apparent on the chip level: notably, AMD’s merger with ATI in 2006 allowed the semiconductor company to merge the central processing unit (CPU) and a graphics processing unit (GPU) into one chip, which the firm dubbed an accelerated processing unit (APU). The GPU is also integrated into AMD’s SoC offerings.
As a result, Eder says even some of the low power AMD processors which have computing performance akin to Intel’s Atom processors come equipped with very high performance graphics engines, which provides additional advantages for 3D graphics applications.
According to Cameron Swen from AMD, the integrated GPU can also be used to accelerate applications which support multi-threaded processing.
“The GPU in the heterogeneous architecture of the AMD APUs and SoCs can be programmed through OpenCL for a variety of compute intensive functions to deliver excellent performance per watt,” Swen explained. “For applications like PC-based machine vision cameras, a fan-less design can be created that delivers excellent image processing performance.”
While chipmakers continue to bake features into their processors that present compelling capabilities for industrial applications, the role of the engineers who build the industrial PCs and boards on which the processors reside remains key.
According to Eder, it is the engineers at companies like congatec who ensure that the advantages provided by the processors are translated into useable features in the final solution.
“Processors themselves are quite general purpose,” Eder said. “Our systems are tailored for industrial functions through the features we provide on the modules.”
These include BIOS/UEFI features, microcontrollers which handle battery management or watchdog timers (critical for detecting and recovering from faults), and implementing support within the firmware for remote management capabilities such as Intel’s Active Management Technology.
When working with ARM-based processors, such as the Freescale i.MX 6 SoC, which do not have BIOS or UEFI for the initialisation functions, the onus is on congatec and other system manufacturers to provide the software support systems, such as bootloaders in addition to the drivers which support the embedded features provided by the processor.
Security is also a focus for congatec, and keeping up with the changes in technology is an important part of the job.
“Today we equip our modules with an external security chip,” said Eder, “But upcoming processor generations will have enhanced security features like Trusted Platform Module (TPM) built in.”
But perhaps the most complicated issue that engineers have to deal with is power management. The demand for reduced power consumption and thermal output means many industrial processors run in low-power states most of the time, which necessitates careful power management.
“Parts of the processors are switched off when not needed,” Eder explained. “This brings the most power use improvements, but this also means a high change of load on the power supplies.”
For a high-performance CPU, the electrical load can drop from 50A down to almost zero within microseconds, depending on the application. Software with repeated cycles of high processing demands and idle times can aggravate the situation, and special knowhow on the part of the board engineers is needed to avoid destabilising voltage swings caused by this high change in loads.