Automotive design has captured the public’s imagination, and their expectations are creating countless technical challenges for engineers working in this space. For SI/PI/EMC/EMI engineers, the challenges are unique inside the car due to harsh environments as well as crowded systems. In this special report, we talk with five experts in automotive applications who are working in simulation/test & measurement (Rohde & Schwarz, Keysight Technologies), connectors (TE Connectivity, Samtec), and automotive chipset design (Valens) to address the critical needs of the latest automotive applications.
It Is Not Like A Server
It is well known in automotive electronics that reliability and longevity are critical factors for selection in an automotive design, so components must be specifically designed for this application. Harsh environments, including large temperature variations (−40°C to +125°C) and vibrations require stringent specifications. Automotive EMC requirements may be less well known. “The electrical modules are packaged in an extremely tight environment with multiple electronic control units next to each other. This requires applications to be designed around tight EMC, SI, and loss tolerance budget,” notes Huzefa Bharmal, senior manager, product management, TE Connectivity.
Engineers need to rethink their process design and milestone plans if they are accustomed to designing for standard consumer products. For instance, “required process controls lead to additional documentation, testing, and engineering reviews that add up over time,” observes Jim Koch, global sales manager, automotive/transportation at Samtec.
There is a power story here as well, as all of the power for automotive electronics has to be generated and stored locally. A standard internal combustion engine (ICE) uses a single electrical bus associated to a 12 V battery, while a hybrid electric vehicle uses a 48 V battery and bus, according to Fred Weiller, automotive & energy solutions lead, Keysight Technologies. Weiller notes that, with an ICE/alternator architecture, heavy electrical loads require the systems to tolerate many electrical transients and variations in electrical loading that can affect the other components in the vehicle, especially when the air conditioning unit is turned on.
Bus Speeds On the Rise
It stands to reason that more electronic devices (radar, sensors, infotainment, safety, etc.) in the car will increase the need for bandwidth and speed within the car. “In-vehicle connectivity must deliver the necessary speed, while guaranteeing low latency, high-performance, and maximum resilience against electromagnetic interference (EMI/EMC),” says Dana Zelitzki, VP, marketing, Valens. Zelitzki notes that in terms of speed, 12 Gbps is the more immediate goal of the market, to cater first to cameras and sensors. Then, as cars evolve in terms of autonomy and numbers of devices, the bandwidth will subsequently increase, as suggested by the recently announced MIPI Alliance®’s A-PHY specification, currently in development and expected to support speeds of 16 to 24 Gbps, with a roadmap of 48+ Gbps and beyond. Valens’ automotive technology is the underlying basis of the A-PHY connectivity standard.
Zelitzki also notes there are issues with the need to support uncompressed data transmission, particularly in advanced driver-assistance systems (ADAS) applications that cannot tolerate the latency associated with compression. She notes that compression also adds complexity in the overall architecture, which increases cost.
Dr. Nik Dimitrakopoulos, manager automotive, Rohde & Schwarz, points out that automotive Ethernet, a specification from the IEEE 802.3 and 802.1 groups, is a scalable and future proof standard. The IEEE is working to ratify a specification for 2.5/5/10 Gbps with groups looking at data rates up to 50 Gbps. Dimitrakopoulos reports that the OPEN Alliance, an industry group that encourages wide scale adoption of Ethernet-based networks as the standard in automotive networking applications, is working on achieving speeds of 10 Gbps with possible growth up to 40 Gbps. With this work, he sees the potential for sensor fusion and high-speed cable backbones.
In addition, Dimitrakopoulos notes that if 10BASE-T1S can achieve standardization, it could potentially replace the controller area network (CAN) bus, making it possible for future vehicles to run exclusively on automotive Ethernet. Weiller cautions, however, that “data rates this high in a vehicle, may present new challenges. For example, channel availability combined with the extremely volatile and noisy EMC and EMI car environment, may lead to considering optical links as an alternative to wires.”
Despite the approach taken, all seem to agree that there will be need for high-speed busses in the car. Bharmal is reporting current customer demand for solutions that go up to 25 Gbps, noting that OEMs are looking to use high-speed buses to support the need. For reference, a 4K camera sensor running at 60 FPS generates about 12 Gbps of uncompressed data, and a long-range radar for detection generates approximately 10 Gbps of uncompressed data. Koch agrees and notes that he is already fielding automotive OEM customer requests for information on 112 Gbps PAM4 solution in anticipation of next-gen architectures.
The move toward autonomous or driver assistance technologies is dramatically changing the requirements for data processing inside the car, and it is shifting some computation to the cloud. According to Dimitrakopoulos, autonomous cars are expected to be processing 4 to 5 TBytes of data every day. Most of the computation is expected to be done within the vehicle with some data being offloaded and processed by nearby networks (edge computing). How much computing that can be offloaded will likely depend on the capabilities of 5G networks.
Bharmal and Zelitzki note that due to latency and safety reasons, automotive OEMs prefer to keep the critical computing within the car until technology advances and more computations can be safely offloaded to the cloud. This is challenging given these considerations: Level 5 (L5) autonomy will require greater than 25 Gbps to support +4 Terabyte/hr of data generation. For reference, Tesla’s ADAS chip for L3/L4 autonomy is designed for 72 trillion operations/s. ZF’s ProAI RoboThink can process 150 trillion operations/s. NVIDIA Pegasus (in development for L4/L5) can process up to 320 trillion operations/s.
Zelitzki sees an interest in smart-vehicle architectures, explaining that each application/device has its own ECU/CPU, which adds complexity and costs to the overall vehicle architecture. A smart-vehicle architecture would permit remote resource sharing within the car, through high-speed, low-latency connectivity among different units. In addition, it can reduce the number of computing units because different applications and devices can share the same unit through long-distance transmission of data. Her company specializes in converging different interfaces (video, data, USB, PCIe, Ethernet) over a single wire, so they are seeing a great deal of interest in this architecture.
In addition to these considerations, Koch notes, “We are seeing some hive mindset systems—what one car learns, all cars are taught. This amount of data will require fast, low-latency connectivity which 5G can provide.” Weiller adds that new standards and low latency communication initiatives such as 5G C-V2X will enable offloading onboard systems to edge computing without negatively impacting response time for critical ADAS applications.
EMI and Power Integrity
EMI is a major issue in all connected cars, and even more so in electric vehicles (EV), given the power and number of cables involved. Bharmal observes that the largest source of EMI in an EV is the power converter. For this reason, components near the high-voltage stack or even radiating antennas are designed to operate at different frequencies and impedance levels.
Weiller notes that transients in an EV may also happen as the result of quick acceleration or deceleration when the energy flows to/from the inverter/generator and from/to the battery. The DC:DC converter takes the high-voltage from the battery and converts it down to the low voltage (usually 12 V) used by other components, protecting the rest of the circuitry in the vehicle.
Automotive EMI engineers need to focus on low frequency magnetic disturbances and high-frequency digital emissions, according to Koch. He adds that low frequency EMI design typically leans toward a single point ground system, whereas high-speed digital is multi-point as much as possible.
Zelitzki explains that reducing the number of cables, connectors, and devices improves resilience against EMI. During EV operation, hundreds of amps of current can flow from front to back of the car. This large current generates voltage drops between different ECUs in the car. Specifically, she notes that these voltage drops can result in currents of up to 6 A through the shield of data cables, which can disrupt the integrity of the transmitting channel. Valens approach is to use unshielded twisted pair (UTP) cable, which they say prevents such issues from occurring.
Future Safety Systems
The consensus is that cars of the future will use both radar and cameras as they move toward L5 autonomy, with tradeoffs of cost versus effectiveness (safety).
Dimitrakopoulos sums it up this way: Radar and cameras are complementary sensing technologies. While radar is robust against rain and snow and can instantaneously measure the speed of moving objects, cameras have a much higher resolution with benefits for object detection. Both technologies will be used in parallel for 360 degrees surround sensing.
Weiller points out that the availability of 77 to 81 GHz radar has dramatically improved its ability to differentiate smaller objects from one another. The main advantage of radar technology is that it is immune to weather and light limitations, so it provides a level of consistency and availability not found in the cameras’ realm.
“Ultimately it is not a matter of either-or, but a rich combination of multiple sensors from multiple technologies providing a set of overlapping data points that benefit the vehicle’s artificial intelligence (AI) in the same way that humans benefit from information coming from complementary senses of vision, hearing, and touch, to evaluate their environment and make the best decision possible,” says Weiller.
Test & Measurement Wish List
Design and technology shifts typically require new ways of testing and measuring, and the cars of the future are no exception. Of course, standards bodies are assisting with this, but we asked our non-T&M panelists what they wished for from their vantage points.
Bharmal: EMC chambers that can perform vehicle level testing as well as a multiport VNA that can operate at high- frequency would be very helpful. In addition, environmental chambers that can perform live testing at high temperature and vibration would be important.
Zelitzki: A major issue we see in tests today has less to do with the type of instruments, but rather with the test standards themselves. We believe that, given the increased bandwidth expected, more applications and devices, we need stricter test guidelines to guarantee safety and performance.
Koch: Automotive Ethernet requires screening attenuation and coupling attenuation testing per IEC62153-4-7. The test gear required includes a triaxial fixture, which is not something most SI labs have. These tests require “craftsman” level attention to detail to ensure shield termination is repeatable and representative of typical installations. Test equipment and systems which simplify these tests would be welcome for development and production level tests.
There are some specific test and measurement concerns about the emerging 10G Automotive Ethernet and SDC11 requirement of −65 dB, specifically, how will we be able to make the measurement? Koch and Bharmal help frame the problem. “From Samtec’s perspective, those are extremely tight mode conversion requirements. Maintaining this level of balance in the field over time will be challenging for the industry,” says Koch. Bharmal adds, “As we move towards safety applications based on 10G Automotive Ethernet, proper measurements and testing will be critical. Special end of line testers and fixtures will be needed to measure and qualify products. VNA calibration at high-frequency will be a big challenge, but I am sure we will find ways to automate the process to ensure repeated and reliable performance every time.”
His confidence is matched by our T&M panelists who reply, “To date 10G Automotive Ethernet is still under discussion; the standard is expected sometime in the middle of 2020. For measuring 10GBASE-T1, a higher bandwidth oscilloscope and VNA will be required. R&S provides the right tools to assist engineers with the new challenges by offering high-end and high-precision equipment,” says Dimitrakopoulos.
Weiller notes that Keysight brings decades of measurement and calibration expertise within the VNA/component test world, providing both electronic and mechanical standard open-short-load (OSL) calibration kits, “enabling customers to ensure the measurement plane is where it should be, at the end of any test fixtures instead of at the analyzer port itself.” The company also supports more advanced calibration techniques, such as fixture removal/on wafer/probe calibration and multiple mixer calibration techniques to improve measurement reliability and repeatability.
What Is Next
At the end of the day, automotive designs have always had significantly different component reliability requirements (10 years versus a few years) than regular consumer applications. Achieving the necessary reliability to avoid catastrophic failures is challenging, and it will get even harder with autonomous vehicles.
Bharmal believes that perhaps one of the greatest challenges will be to achieve high-reliability in mass production. He notes that these products are designed as per Automotive Electronics Council (AEC) reliability requirements and are manufactured per IATF 16949 guidance. As we move towards full autonomy, cars will run non-stop, 24/7 and the reliability requirement will be extremely stringent to simulate +400,000 operational miles under harsh automotive environment.
Weiller’s concern is that without industry-wide standards bodies and policies, it will be difficult to ensure that all testing is performed to the same high standard. His team is actively involved in multiple standards organizations such as ETSI for automotive radar and Open Alliance for automotive Ethernet to drive standardized testing at the industry level. This is important because in the worst-case scenario, a critical system failure could cause an accident, or a fire in electric vehicles with large batteries. In industrial applications, a failure may impact costs and timing, but not lives. Thus, reliability and quality are paramount in automotive applications. Properly segmented electrical systems, with redundancies in place for safety, are a design requirement. If all systems run off the same electrical bus, one failure would affect all systems with adverse consequences for an AV. For power modules used in EVs to drive the powertrain, transistors need to be qualified and known to be reliable in conditions that could otherwise lead to a catastrophic failure.
The need for robust standards is a global issue. Dimitrakopoulos reports that the automotive industry, especially in Germany, has much higher safety requirements than consumer industries. He notes that the EU initiated a program called “Vision Zero,” aiming for zero serious injuries in road traffic by 2050. “Without strong efforts in functional safety (the automotive standard is known as ISO 26262), security efforts and advanced driver-assistance systems, Vision Zero cannot be achieved.”
Engineers working on designs for current and next-generation automotive applications have a distinct challenge, but they seem aware of the probable obstacles along the way.
1. OPEN Alliance, http://opensig.org/about/about-open/.
3. Automotive Ethernet: An Overview, https://support.ixiacom.com/sites/default/files/resources/whitepaper/ixia-automotive-ethernet-primer-whitepaper_1.pdf.
4. SAE International: SAE International is a global association of more than 128,000 engineers and related technical experts in the aerospace, automotive and commercial-vehicle industries. Its core competencies are life-long learning and voluntary consensus standards development, www.sae.org/.
5. “SAE International Levels of Autonomous Driving Chart,” www.sae.org/news/press-room/2018/12/sae-international-releases-updated-visual-chart-for-its-“levels-of-driving-automation”-standard-for-self-driving-vehicles.