Sit in any new car, and it doesn’t take long to notice that a lot more electronic components exist in vehicles today than did five years ago. But the GPS, satellite radio, DVD/video screen, automatic climate control, and gas-mileage monitoring features, along with the AM/FM radio, CD player, and power doors, windows, and mirrors aren’t all of the electronics in a vehicle.
In fact, under the hood and in the body—in just about every system in which 10 years ago an actuator drove the mechanical or hydraulic system—there is now likely an electronic sensor and switch augmenting the mechanical and hydraulic systems or even replacing them. Many of theses systems, ECUs (electronic control units), are mission-critical, controlling the brakes, air bags, and steering. As such, they have to be 100% reliable, with tolerances nearing military and aerospace require-ments but at consumer electronics prices.
About 40% of the overall cost of today’s automobile involves electrical systems and software design, yet electronics is the least automated design discipline in the automotive sector.
“We used to joke that body and sheet-metal guys used an innovation called “CAD” to design every nonelectric piece of an automobile, and all they had left for the electronics department was an innovation called pen and paper,” says Jon Friedman, who before becoming director of the automotive-industry marketing group at The Mathworks worked in electronic design at an auto manufacturer.
Now, however, driven by consumer demand for new “electronic bling” and the need to make cars more reliable, automakers are in a tizzy trying to restaff and retool an industry that until recently mechanical engineers and design software had dominated. In this new age for the automotive industry, automakers need an increasingly electronics savvy staff and supplier-support structure that can deal with electronic-design issues that are unique to the automotive area.
In Reed Business’ 2005 Movers and Shakers electronics forecast, the head of General Motors research-and-development group, Alan Taub, details the business challenges GM faces as the company and the rest of the automotive industry move toward the “electronification of the vehicle” (Reference 1). Taub describes the electronics for enriching and making safer the driving experience, such as adaptive cruise control, OnStar telematics for monitoring the automobile and the well being of its occupants, and even advanced crash-avoidance systems that steer the car away from an accident.
Industry experts see electronics as the cheaper and safer alternative to slow mechanical systems that more quickly wear out over time in automobiles. Moving to more reliable ECUs instead of hydraulics and mechanical-based systems means a safer vehicle, fewer hits on a warranty, fewer recalls, and ultimately, automakers hope, greater profit for them and the rest of the design chain.
Electronics promises to play a bigger role as automakers move to more innovation in hybridization and explore alternative fuel sources, such as hydrogen and even carbon. But employing these advanced systems and networking them present numerous challenges all the way down the automotive eelectronics design chain.
Three-tier system Today’s auto manufacturers, or OEMs, don’t design electrical systems. Rather, they rely on “tierone” suppliers—companies such as Bosch, Delphi, Visteon, Bose, and 70 or so others that design the electronic subsystems that make up the automobile. Tier-one vendors, in turn, rely on tier-two vendors—essentially semiconductor and pcboard-design vendors—to supply and even custom design components for each ECU.
Automakers design automobiles four years in advance of their commercial release, so auto manufacturers today are working on 2009 cars, vans, SUVs, and trucks that they will release in mid-2008.
Traditionally, auto manufacturers devise a rough electronic specification for a vehicle that contains a list of ECUs, their behavioral specifications, and some data about the networks on which the ECUs will run, including low- and high-speed CANs (controller-area networks), LINs (local-interconnect networks), and even the emerging FlexRay. “Some manufacturers will specify performance and some will even [detail] vibration characteristics, but only in rare cases will they specify the electronic components that will be used to implement those specifications,” says Jack Morgan, senior director of the automotive group at Philips Semiconductors. “They do the ‘paper job’, and rely on suppliers for nitty-gritty details.”
After creating the rough specification, the auto manufacturers shop that specification around and take bids from tier-one suppliers. They sometimes select three or more tier-one suppliers for each ECU. The auto manufacturers typically demand that suppliers deliver a prototype—usually a breadboard— for a given ECU within six months to prove functionality.
After the prototype passes the functionality test at an OEM, the tier-one suppliers begin designing the actual scale-model ECUs with some help from tier-two suppliers. It is common for a tier-one vendor to supply ICs for two tier-one suppliers manufacturing the same ECU for a given auto manufacturer.
The suppliers submit the scale model ECUs to the OEM, who designs the wire harness with essentially giant-scale routing software that helps the auto manufacturers weave the miles of wires and hundreds of points of wires that will connect the various ECUs. During this time, the specifications for ECUs can often change. The fact that some automobiles can have multiple derivatives that car dealers call “options” further complicates the process. This scenario means that vendors must supply multiple variants of a given ECU to ensure not only that it functions properly on its own but also that it does not interfere with the performance and reliability of other ECUs and in multiple system configurations.
“Traditionally, the automotive companies only start testing when the last piece of hardware becomes available; that [approach] doesn’t work anymore,” says Friedman. “The automotive industry has a build-then-test mentality: You bend sheet metal, forge iron, and see how [it] works. When you had one engine and one vehicle, you [could] build it and test it. But in vehicles I worked on, we had 2000 to 3000 buildable electrical combinations; you can’t build all those.”
Indeed, every automotive company faces the tough task of quickly balancing long automotive design cycles and the fickle demands that consumers place on electronic gadgetry. “There is especially a lot of the pressure in the infotainment area, where consumers want a seamless experience between what they do inside their car and what they do outside their car,” says David Stone, NEC’s technical-marketing manager. “The industry is pushed to figure out how to bridge this history of enormous reliability and super long planning cycles with the idea that people are changing out their consumer products every 16 months or even every year.”
OEMs are trying to meet many of these demands by incorporating standard interfaces such as wired and wireless USB, Bluetooth, and others into car designs and by increasing the use of platform design or even programmable logic. All of these technologies show promise for allowing OEMs to move more easily with consumer demands and possibly update the electronics on an automobile that could be on the road for more than 10 years—the equivalent of five or six consumer-product generations.
In the last 10 to 15 years, the number of ECUs in most automobiles has climbed from three to 15, and to as many as 70 in high-end automobiles, such as the Mercedes S-class and 7-Series BMWs (Figure 1). Although you can attribute much of this increase to new electronic gadgetry for entertainment, once-simple functions, such as interior lighting, have become much more sophisticated.
Stephan Lehmann, strategic marketing manager for Freescale’s global automotive business, says the Volkswagen Beetle is a prime example of how the number of ECUs is increasing in a car. Lehmann notes that the old Beetle had a simple electric switch, so that when you opened a door the overhead light would come on. That light would shut off when you closed the door.
That simple switch has evolved into a complex ECU that networks into the overall body controller. That same light switch links to the doors’ opening and closing and helps dim, then eventually shut off, the interior lights after a driver starts the car. It also networks to the air-bag controller to turn on when the air bag deploys, so car occupants can better evaluate the situation following a collision.
Even analog-power devices, which here to fore transferred battery power to ECUs, have become more sophisticated. Today, power-control units have replaced traditional relay-control units.
“The power technology in lamps and windshield wipers today are really quite sophisticated,” says Morgan. “It is under the control of the microcontroller, which tells it how much power to allow. Rather than just turn on, it says “I know this load on this lamp to be 4A,” and if the load exceeds that value, the lamp will turn off immediately. A traditional relay control would just turn it on, despite the short, and would burn up. Then you would have to buy all kinds of expensive pieces instead of replacing a bulb or a small wire.”
Networks and buses ECUs coordinate and connect by means of an in-car network. Each automobile typically employs a minimum of three networks running simultaneously. A standard bus configuration would use high speed CAN for power-train and safety controls, low-speed CAN for body functions, and even a subnetwork, such as LIN, for functions controlling power windows.
The most common network bus in automobiles is the CAN. By many standards, the CAN is showing its age, with critics claiming it is too slow and not deterministic enough for you to accurately set or analyze timing of signals across the network. New networks, such as FlexRay, are much faster and are more deterministic; however, there has been slow but sure adoption and support for the standard. Designers are also kicking around multiple networks for the infotainment portion of an automobile, such as MOST (Media Oriented Systems Transport) and IDB (Intelligent Transport System Data Bus).
Unique design challenges Automotive electronics is unique in that devices have to be 100% reliable across stringent running conditions but must at the same time be low cost. “The automotive environment is indeed a unique environment,” says Morgan. “What is commonplace in automotive, no consumer product can survive, so all our semiconductors must work over a broad range of temperature—typically from -40 to +125°C—and typically the devices are lower power and have stringent ESD/EMI requirements. If they get a failure in a car, it is nearly a catastrophe. Everything has to work in automotive. That puts a lot of constraints on an electrical system.
They are not sloppy constraints either, because as you connect all the electronic components together, you get a lot of tolerance stack issues that don’t allow you to be very sloppy.”
Semiconductor and tier-one companies have to conform to OEM PPAT (production-part-approval process) requirements and numerous standards, most notably the AECQ100 qualification flow, ISO9001, and ISO/TS 6949 (2002), which are quality guidelines for tier two suppliers.
Vendors such as Freescale, Philips, Infineon, International Rectifier, and NEC, which also offer analog-power devices, have to be extremely careful in building and testing to these tolerances as analog devices are more susceptible to malfunction. Infineon, for example, does a lot of work on package and die reliability.
“We really look at how die and package work as a system,” says Ray Notarantonio, Infineon Technologies’ automotive-power applications manager in North America. “We qualify our devices to 175°C, and that is not a trivial task. You also have to get the packages qualified to that level, too.”
But although playing in the market means IC vendors must do extensive testing, they say it is well worth it, because although no electronics segment is experiencing booming growth and the automotive market and volumes aren’t extremely high, automotive semiconductors provide a relatively sizable, stable revenue stream (Table 1). In fact, just about every IC manufacturer offers something for the automotive space.
The recent trend is for OEMs and tier-one suppliers to use more standard products and ASSPs (application-specific standard products) than ASICs. “Our suppliers have traditionally used mostly custom ASICs, and to some degree, custom processors. This [situation] is changing,” says Fred Huntzicker, engineering group manager at GM. “We are driving to more commodity solutions, and standard solutions are being made available that meet many of our needs.”
Still, ASICs in the segment grew 6.2%, according to a report from Strategic Analytics, and 9.2%, according to Gartner (Figure 2). However, because of reliability and cost constraints and volume, tierone vendors typically implement ICs in older and more proven process geometries. OEMs would like to see tier-one vendors provide the electronics in more bleeding-edge silicon processes, but EDA software and methodologies aren’t reliable and cost-effective enough for most IC vendors to go that route. Today, engineers design most digital ICs for 2008/2009 vehicles using 0.18-micron CMOS. Freescale’s PowerPC and Renesas’ SH-series microprocessor units use 0.13- and 0.12-micron silicon processes for power-train ECUs. Engineers probably implemented the most advanced digital chip in the car you are driving today in 0.25-micron CMOS.
In the face of a slow- to no growth EDA industry, some vendors have made the leap of faith that a jump in automotive-electronics content will lead to greater demand for EDA tools. But the automotive vendors and suppliers say that contention is only partly true, and there won’t be a huge demand for more standard IC-design tools. Most chip companies and tier-one suppliers already own the tools to design pc boards and 130nm ICs. What the industry needs are tools targeting automotive-design problems.
SOC means “system on car” In particular, the industry needs streamlined communications across the three design tiers and for real system-level-system-on-cardesign and verification. There is a need for modeling from chip, to ECU, to network levels to allow the OEMs to proactively design a network and the required ECUs with model-based design and software and pass that system-level specification down the design chain.
To date, traditional EDA companies, such as Mentor Graphics and Synopsys, have dabbled in automotive tools. Mentor has for many years offered its VeSys wire-harness tools for routing cables throughout car, airplane, and household-appliance chassis, and Synopsys has for years offered a mechanical-, electrical-, and thermal-simulation tool with Saber. Over the last year, Mentor has also added to its portfolio a multidiscipline simulation tool that allows users to describe signals with standard VHDL-AMS rather than a proprietary language, which Saber uses.
Mentor has seemingly gone one step further, jumping into network design by purchasing Volcano Communications Technologies in May 2005. Mentor’s not breaking into new territory with that acquisition nor is Volcano by any means the only game in that area; companies such as Vector-CanTech specialize in automotive-network design and simulation. And companies such as The Mathworks and Vast Systems Technology offer advanced ECU models that take strides in offering OEMs tools to model an entire electrical system and create more precise constraints and derivatives more rapidly.
Volcano and Vector tools attack automotive-network design from opposing angles. Simply put, Volcano requires users to know all their ECU constraints and then allows them to fit networks to the ECUs. Meanwhile, Vector-CanTech offers bus-network design and simulation into which users plug their ECUs. Both companies support multiple bus standards, including CAN, LIN, and the emerging high-speed-bus protocol, FlexRay.
Vast focuses on offering ECUhardware modeling, and The Mathworks offers ECU-hardware and -software modeling. Companies such as ETAS Group and Enea focus on software-embedded modeling and design.
Real-time design and testing are imperative in the automotive industry, says Jeff Kessen, director of strategic marketing and communications for ETAS. “An eight-cylinder engine needs fueldelivery computations nearly 500 times per second at high rpm, while an ABS [antilock-braking system] can adjust braking force more than 10 times per second,” he notes.
But it isn’t just hardware working with software that requires simulation—it’s working with extreme heat conditions and getting bounced around in automobiles over a long period of time. “Some OEMs are now integrating a transmission ECU with the hydraulic control valves, and the ECU is immersed in transmission fluid,” says Kessen. “With aggressive driving on a hot day, the fluid can reach 150°C (300°F). The motherboard of a PC could never survive those conditions.” That scenario means that a need also exists for EDA vendors to expand or partner with other disciplines—thermal and mechanical and perhaps even fluid-dynamic tool suppliers.
One of the reasons that few electronic tools for automotive exist is that many auto suppliers are secretive (or embarrassed) about their electrical-system design flows and don’t wish to participate in joint standards. One of the most promising efforts in standardization is AutoSAR (Automotive Open System Architecture), which promises to help define ECUs and subfunctions in hardware and software that are common to all automobiles. The group has the support of several European auto manufacturers and suppliers and hopes to standardize hardware and software of common functions. Doing so, the group hopes to streamline communications between design tiers and speed automotive-electronic-design cycles. Mentor, through its acquisition of Volcano, has a director’s role in AutoSAR, which gives the company more power in influencing the direction of the standard.
AutoSAR, which gives the company more power in influencing the direction of the standard. Mentor and Synopsys are not the only big EDA companies drawn to the potential opportunity in the automotive industry. Cadence Design Systems in mid-2005 said it is working on a design kit targeting automotive design. It remains to be seen whether the kit will include specialized tools or simply bundle existing IC, pc-board, and package tools at a price attractive to automotive groups within tiertwo, tier-one, and automotive companies.
The need for a true system-oncar-level tool is not a new concept for Cadence, EDA’s largest vendor. The company had offered a tool, VCC (Virtual Component Co-Design), which turned out in retrospect to be ahead of its time and certainly ahead of the auto industry’s need for system-level modeling.
For more information... For more information on products such as those discussed in this article, contact any of the following manufacturers directly, and please let them know you read about their products in EDN Asia.
Keithley Instruments
With more than 60 years of measurement expertise, Keithley Instruments has become a world leader in advanced electrical test instruments and systems from DC to RF (radio frequency). Our products solve emerging measurement needs in production testing, process monitoring, product development, and research...
