A new benchmark suite for automotive microcontrollers responds to a change in chip architectures while also focusing on power consumption. The test, developed with support from Volkswagen and a number of semiconductor suppliers, focuses on the peripherals that surround microprocessor cores while also giving engineers a way to compare the power consumption of different microcontrollers.
The Embedded Microprocessor Benchmark Consortium (EEMBC), which has provided an automotive benchmark suite since the late 1990s, unveiled its new test suite in May. It was developed by a working group chaired by Volkswagen with support from 11 top-tier semiconductor vendors including Freescale, Fujitsu, Infineon, Microchip, NXP, Renesas, STMicroelectronics, and Texas Instruments.
The new benchmark suite addresses a change that has occurred in recent years as automakers have taken more of a systems approach based in part on the high integration levels of microcontrollers. The circuitry around the microprocessor cores is at least as important as the core itself. That made it necessary to develop a more flexible follow-on to EEMBC’s existing AutoBench test suite, which focused only on cores.
“We still need to benchmark the cores, but intelligent peripherals are where a lot of other intellectual property is invested these days,” said EEMBC President Markus Levy. “With peripherals, you can’t just write generic code, the CAN techniques for one company are different from those at another company. We’re doing more at the system level.”
Currently, EEMBC is using a National Instruments PXI Chassis with a CAN (controller area network) module to develop LAN (local area network) packets. Once the CAN version is completed, EEMBC plans to expand the peripherals list by addressing pulse width modulation (PWM), analog to digital converters (ADCs), FlexRay, and universal asynchronous receiver/transmitters (UARTs).
A key aspect of the new test suite is its versatility. It will be able to test a range of peripherals, helping engineers choose microcontrollers using identical parameters.
“The requirements for a door lock controller and an instrument cluster controller will be quite different, with different peripherals,” Levy said. “We’re building in flexibility so Tier 1s and OEMs can specify how they want the benchmark to run. Today, everyone does proprietary testing to compare parts; there’s no apples-to-apples comparison. We provide a standard methodology.”
The new test also focuses on power consumption. Individual tests of the microcontroller measure the power consumption of the CPU (central processing unit) and peripherals under various loads, the amount of time that it spends in low-power modes under various CPU/peripheral loads, and the time required to wake the MCU (microcontroller unit) from its various low-power states to resume processing. The working group plans to align the benchmark suite with AUTOSAR.
“If a trunk latch controller continuously polls the network instead of waiting for a wake-up command, it will take a lot more power,” Levy said. “If you’ve got 20-30 or more controller each doing small functions, this can add up to put a fair draw on the alternator. If companies can optimize this power consumption easily, they might as well do it.”
Levy noted that more automakers and semiconductor suppliers are likely to join the effort now that it’s gone public. Not surprisingly, Volkswagen plans to implement the benchmark it helped develop.
“Following completion of this new benchmark suite, we will demand the Tier 1 suppliers and semiconductor vendors to provide results for the microcontrollers that will be integrated into the next generation of electronic modules,” said Volkswagen’s head of electric and electronic development, Dr. Volkmar Tanneberger.