The microprocessors that help automakers add many new features and functions continue to evolve, following trends of industries that have less demanding environments. A growing number of automotive applications are adopting dual-core devices and floating-point processors, though cost justification remains a barrier for both technologies.
Texas Instruments recently unveiled a part that addressed both. Its TMS570F offers two ARM Cortex-R4F floating-point processors with a peak speed of 64 MHz. Following another trend, it is loaded with 2 Mbytes of flash memory and peripherals including FlexRay, CAN, and LIN modules as well as a timer and two 12-bit analog-to-digital converters.
TI’s processor is targeted at safety applications, which must have fail-safe designs. TI noted that the chip meets the International Electrotechnical Commission 61508 SIL3 or ISO26262 ASIL D safety standards for critical applications.
The controller is among a growing number of multicore devices designed for automotive applications. They are beginning to see use in powertrain, where processing capability is critical. Other multicores are aimed at infotainment, where the large volume of graphics for navigation and movies requires the high performance of multicores.
However, software is vastly different for multicores, so they are being adopted only when processing capabilities or lower power consumption requirements offset the expense of moving to new, more sophisticated software. Safety fits that requirement, since response times are important. Another limiting factor is that auto industry clock rates rarely go above 200 MHz, well below the multi-GHz speeds of PC processors.
TI’s device also offers floating-point functionality, which has been used in demanding areas such as powertrain control for years. That will help the part find a role in safety-critical applications.
One benefit of floating-point technology is it makes it easier for engineering teams to use autocoding, which reduces the initial cost of creating software while also facilitating reuse compared to handwritten code.
“With floating-point processing, it’s much easier to use automatic code generation,” said Charlie Murphy, Strategic Marketing Manager at Texas Instruments’ Automotive Microcontrollers group.
Floating-point math is becoming more widely used as powerful and inexpensive 32-bit chips gain usage. The cost penalty of floating point is lower on these powerful devices.
However, there are still areas where cost is king. While chipmakers offer more parts with floating-point capability, The MathWorks recently unveiled a tool that focused on cost-sensitive applications such as small motor control. Simulink Fixed Point 6 makes it simpler to design and simulate fixed-point systems and generate code for controllers that perform simple tasks such as opening windows and positioning seats.
“The classical development process for system engineers is to develop algorithms with floating-point math, simulate it until it’s right, then convert to fixed-point because floating-point processors are too expensive,” said Tom Erkkinen, Embedded Applications Manager at The MathWorks.