Processor selection in automotive electronic systems (reprinted)

goodboy

Processor selection in automotive electronic systems (reprinted) [Copy link]

Automobiles are experiencing a digital revolution: gone are the days of pure mechanical systems and analog electronics. Today’s cars are digital cars, with dozens or even hundreds of embedded processors connected to each other through digital networks to control and optimize the operation of almost every system in the car. Future cars will integrate more processors because advanced applications and performance require more complex signal processing algorithms, including safety , engine and exhaust emission control, driver-car interface, and in-car information and entertainment systems.

The automotive market requires processor suppliers to make long-term commitments. For example, automakers sometimes require their suppliers to provide a 10-15 year supply commitment for a processor product. Below we will explore the various types of processors for automotive digital signal processing applications, as well as the advantages and disadvantages of each type. In addition, we will analyze the impact of the special requirements of automotive applications on processors for the automotive market.

Appendix: Processor types, representative suppliers and processor samples

Processor Selection in Automotive Applications

The choice of processor for automotive systems is influenced by many factors. The most important selection criteria generally include automotive certification, on-chip integration, performance, price and energy saving. The quality of software development tools and the availability of software components also affect the choice of processor. The processor supplier's commitment to its products and future development plans are also important considerations.

Because of the life-threatening impact, key automotive safety systems such as car engines, airbag control and brake systems have very strict reliability and durability requirements for processors. Therefore, automotive safety system applications are the most severe test for processor suppliers. These applications require processors to obtain automotive certification qualifications, and such processors require specialized design, manufacturing, packaging and testing methods.

There are many non-critical signal processing automotive systems that also require a large number of processors, such as in-car navigation and entertainment devices. Although automotive OEMs and automotive electronic system suppliers also require high-quality components for such applications, the requirements are not as high as those for critical safety applications. For example, processors used in in-car systems are generally not required to obtain automotive certification qualifications.

Today, the most demanding automotive signal processing applications are in-car navigation and entertainment systems. This may change in a few years as new safety systems begin to use video and radar processing, and engine and brake control systems will use complex model-based calculations, replacing the current popular lookup table reference methods with complex real-time calculations.

Integrating appropriate peripherals, memory, and I/O interfaces on the processor helps improve performance and stability, as well as reduce power consumption and system cost. The on-chip integration requirements for automotive applications are very different from those for other signal processing applications. Therefore, suppliers targeting the automotive application market must design their processors specifically for the special requirements of these applications. Multi-channel analog-to-digital converters are particularly useful for processors targeting automotive control systems. For example, an engine control system generally receives input signals from dozens of analog sensors.

On-chip flash memory is a key feature for processors targeting automotive control systems, because these systems use large lookup tables that sometimes need to be updated in the field. For example, the lookup table used in the engine control system contains tens of thousands of calibration points (or similar output values) from various control components (such as fuel pumps and ignition coils). The calibration point data is generally determined in the laboratory before the car leaves the factory, but some calibration points may need to be adjusted after the car has been used for a period of time. On-chip flash memory can be used to update calibration points or other parameters of the control algorithm in the field using data downloaded from the car dealer.

The biggest benefit of integrating flash memory on the processor is improved system performance and reduced cost compared to using a separate flash chip. Although integrated on-chip flash memory is valuable to system developers, it is not easy for processor suppliers to implement it. Automotive-certified processors require higher temperatures than mainstream flash technology can withstand. As you can imagine, processor suppliers competing in this market often need to invest a lot of resources to develop flash memory technology that can work stably on automotive systems.

Digital network transceivers facilitate communication between processors in distributed systems. There are various network protocols for different automotive systems. Processors for specific automotive applications generally integrate network transceivers for the relevant protocols. For example, the Controller Area Network (CAN) protocol is generally used for engine and transmission control networks. The Media Oriented Systems Transport (MOST) protocol is targeted at in-car infotainment applications such as audio, video, navigation, and communications.

Advanced on-chip debug trace units are also useful for processors targeting critical applications. This trace capability provides system developers with detailed processor, software, and operating system status information, which is particularly useful for verification and debugging. The Nexus 5001 Forum standard for the global embedded processor debug interface defines the interface between software and on-chip debug hardware. The standard was first developed by the IEEE Industry Standards and Technology Organization (IEEE-ISTO) in 1999 and has been updated to IEEE-ISTO 5001-2003. The developers of the standard hope that it will encourage development tool vendors to add on-chip debug trace units or strengthen their support for them.

In-car information and entertainment systems are the signal processing systems with the highest computational performance requirements in current automotive applications, primarily because they involve applications such as video processing that require powerful signal processing capabilities. A high-end infotainment system may include a multi-channel audio system, a DVD player, a GPS navigation system, and a hands-free mobile phone, all integrated into one system. Processors for in-car infotainment systems include relatively high-performance DSPs, DSP-enhanced general-purpose processors (GPPs), and DSP/GPP hybrids. These processors typically operate at clock rates ranging from 200 to 750 MHz.

In contrast, processors for critical control systems such as engine and brake control are generally medium-performance processors. It is easier to meet the requirements of harsh operating environments such as high temperatures using larger chip manufacturing processes (such as 0.18 or 0.25 microns), and the processing speed requirements of control applications are generally not very high. Therefore, relatively low maximum processor clock speeds (40 to 150 MHz) and larger manufacturing processes are the best choices for such applications. However, the processing performance requirements of such applications are also increasing, and processor vendors must adjust their strategies to achieve higher performance while meeting high temperature requirements.

Automotive applications are particularly price sensitive. Processor vendors have been forced to develop highly integrated, specialized processors to reduce system costs. Although automotive applications are price sensitive, the automotive qualification process is expensive, and these costs increase the cost of the chip. As a result, automotive-qualified processors are generally more expensive than their non-qualified counterparts. In automotive signal processing systems, energy efficiency is generally not a major issue. Systems such as engine, chassis, and brake control are only in operation when the engine is running and the battery charging system is activated.

Nevertheless, energy efficiency is also important in some applications. Some systems are active when the engine is off, and their power consumption must be low so that the battery does not affect the engine starting. For example, in-car infotainment equipment is one such application. Other systems must be well sealed from the outside environment. In this case, the packaging of such systems may affect the heat dissipation, so the power consumption cannot be too large.

Signal Processors for Automotive Applications

In today's automotive systems, there are many types of chips used to complete signal processing tasks, from 8-bit MCUs to DSPs to FPGAs. In systems where signal processing plays an important role, 8-bit and 16-bit MCUs are no longer commonly used because of their limited processing performance. To reduce costs, system developers often choose processors that have just enough performance. But for some applications, it is wise to reserve some performance space, especially in-vehicle infotainment systems, which can benefit from the flexibility of this performance space, because some functional applications (such as voice recognition, navigation and audio control) are not yet fully developed when processors are selected.

32-bit embedded general-purpose processors (GPPs) are generally used for automotive signal processing control systems with medium performance requirements. Processors of this level generally use RISC architecture, with simple, common instructions and almost no parallel instructions. GPPs are particularly effective in algorithm processing that emphasizes decision-making and control flow changes, but in many cases their signal processing performance is also good. In addition, GPPs are also good compilation targets. Compared with some special DSP architectures that are difficult to compile, GPP compiled code is quite efficient. Popular 32-bit GPP architectures (such as MIPS, ARM, and PowerPC) have been widely used in automotive and non-automotive application systems.

The advantages brought by the wide market recognition include rich third-party software component supply and strong development tool support. Processors in this category include Texas Instruments' TMS470 series (based on the ARM7 core) and Freescale's MPC500 series (based on the PowerPC core). Both processors integrate automotive-specific peripherals on a 32-bit general-purpose processor core. Freescale's MPC500 series processors integrate peripherals, memory and dedicated I/O interfaces, mainly targeting engine and transmission control applications, with large-capacity flash memory, multiple CAN interfaces, a Nexus debug interface, multiple ADCs, and multiple advanced timing modules.

DSPs, DSP/GPP hybrids, and DSP-enhanced GPPs are typically used in in-vehicle infotainment systems and control systems that require signal processing. These processors have special features, including multi-accumulation hardware, large memory bandwidth, and instructions for running multiple algorithms. These features combined can greatly accelerate digital signal processing algorithms, much faster than GPPs with the same clock rate.

DSP/GPP hybrid devices and DSP-enhanced GPPs are designed to integrate the best features of DSP and GPP: the signal processing capabilities of DSP and the high efficiency of GPP in decision-intensive algorithms and compiled code. This combination of functions is particularly important for systems that require both signal processing and decision processing. Such processors include Texas Instruments' TMS320C2000 series, Freescale's MC56F83xx series, Renesas' SH7760, and Analog Devices' ADSP-BF53x (Blackfin series).

FPGAs may seem like an unlikely fit for automotive processing applications because they are known for being expensive. However, in recent years, FPGA vendors have introduced a range of low-cost, high-efficiency devices, making FPGAs an option for automotive systems. Unlike traditional fixed-architecture processors (such as DSPs and GPPs), FPGAs are not restricted by a pre-defined instruction set. Instead, FPGAs offer system designers tremendous design flexibility to develop a processing architecture that is tailored to a specific application.

Because FPGAs have powerful parallel processing capabilities, their signal processing speed is faster than the fastest fixed-structure processors. But high performance comes at a price: the development cost of FPGA-based signal processing systems is much higher than the cost of fixed-structure software development. Although the role of FPGAs in automotive systems will gradually expand, they are currently mainly used for interfaces in in-vehicle infotainment systems. Of course, once FPGAs enter automotive systems, they will have more other uses and may replace the functions of other system components.

For example, with the advent of "soft" processor cores implemented in FPGAs, such as Altera's Nios II and Xilinx's MicroBlaze (both 32-bit RISC processor cores), microprocessors may increasingly be implemented in FPGAs rather than separate chips. This can save costs because soft cores can be customized (designers can include and exclude certain features and make trade-offs between functionality and resource consumption), and they can also be easily interfaced with specialized hardware (such as specific algorithm accelerators) that uses the FPGA structure.

Digital signal processors are everywhere in the car

As automotive applications become more electric and electronically controlled, digital signal processing will be everywhere in the car. Applications that already use digital signal processing will increase the computational load, prompting the development of a new generation of high-performance automotive processors. For example, Freescale's new MCP5554 processor runs twice as fast as its predecessor, the MPC566, and the addition of SIMD instruction execution can further improve its signal processing performance.

New applications for digital signal processing in the automotive field include both computationally intensive applications that require high signal processing performance (such as lane tracking systems) and applications that only require general processing performance (such as tire pressure monitoring systems - TPMS). Processors for automotive signal processing applications have a wide performance range and will become more diverse in the future. High-end applications such as video-based safety and infotainment systems will require higher signal processing performance, while low-end applications such as TPMS will require energy-efficient and efficient processing performance.

More processors, wider performance range, when will this trend end? Perhaps it will take until embedded processors penetrate every corner of the car system. Imagine this scene: in addition to the air pressure monitor integrated into each tire (new cars will be mandatory), each tire also has a built-in processor to collect and forward information about its status and performance. For example, the tire may automatically issue a warning: "This is the right front tire. I noticed that the road is wet, but my tread depth is not enough to cope with this condition."

You may think this is a bit too early, but the trend of using more processors in automotive systems is irreversible. Given the continuous reduction in the cost of semiconductor products and the potential benefits of smart automotive devices, it is foreseeable that one day our cars will be equipped with smart tires.

By Bjorn Hori

BDTI Analyst