Higher computing performance is rapidly becoming a necessity for automakers as they seek to optimize the air-fuel mixture and injection timing of today's engines. Using a scheme called direct injection, in which gas is injected into the combustion chamber via a common fuel rail (instead of into a cylinder port), automakers say they can boost their fuel efficiency by as much as 10 percent to 20 percent, in some cases.
The new MCU could help automakers do that because it enables them to boost the sampling rate from the engine's sensors and crunch the necessary numbers more quickly. Freescale said the sampling rate is a function of the 5746M's advanced sigma-delta analog-to-digital converters. The number-crunching capabilities are directly attributable to the device's three 200-MHz cores.
Freescale said its new device was also designed comply with the new ISO 26262 safety standard. ISO 26262 enables automakers to develop electronic systems that can prevent dangerous failures of airbags and steering, among other automotive systems.
The move to multicore is expected to gain momentum over the next two years as the auto industry migrates to more complex hybrid powertrains and higher-level safety and security systems. Other electronics suppliers are said to be targeting those areas with the development of their own multicore MCUs. In an email to Design News, a spokesman for Renesas Electronics said that automotive "multicore is on the roadmap" for the company.
Freescale said it expects the demand for the devices to increase quickly, especially in powertrain, largely because the technology gives automakers the opportunity to keep power consumption under control. "Low power has always been important in (automotive) body control," Veri said. "But it hasn't been as big a deal in powertrain until now."
I've seen two performance advantages from a multicore design. Since many high-end designs use an OS that time slices the various tasks performed by the application, a multicore can now devote an entire core to a specific task. The next boost is from designs that had multiple processors. Rather than having a communication link between the various processors, the processors now share the same resources and cohabitate together without having to communicate.
Two of the cores serve as basic microprocessors and one handles all of the I/O controls, which makes a lot of sense because I/O operations and handling various streams of serial data from sensors, microcontrollers, and wireless links could weigh heavily on the dual core portion of the chip. The 3-processor chip offers some redundancy as well as error-detection and error-correction technologies, mandatory for safety-critical equipment.
I would think that multicore might also help automakers do what they've been trying to do for years -- that is, cut the number of microcontrollers in vehicles. Some of the more complex high-end vehicles are now using 80 or 90 MCUs each.
Chuck, that is an interesting development. Lower clock speeds mean less power drawn, that is true. Automotive applications (especoally engine control) consist of a large number of calculations done repetitively in a short time. Couple the extra cores with virtualization software and you can get great performance with lots less power.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This radio show will show what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.