Wednesday, March 14, 2018
The automotive industry is in the midst of a sea change. Growing market needs for electrification, connectivity on the go, advanced driver assistance systems, and ultimately the goal of autonomous driving, are creating newer requirements and greater challenges. A chassis on four wheels is now fitted with cameras, radar and other sensors, which will be the eyes of the driverless car, as well as sophisticated ADAS/autonomous systems enabled by artificial intelligence, which will serve as the brain of the driver-assisted/driverless car for making informed decisions in real-time.
The amount of electronics is rapidly increasing with the level of automation. The cost of electronics in a car today — we are at Level 3 automation — is about 30% of the total cost of a car. It is on a growth trajectory to be 50% of the total cost by 2030 with Level 5 automation. Autonomous vehicles will be a game changer in the way we operate as a society, and they will dictate the next generation of mobility for mankind. This is all thanks to cutting-edge semiconductor technology, which has accelerated this vision for a driverless future.
As noted by Berthold Hellenthal of Audi (during Semicon Korea 2016), semiconductors will drive 80% of automotive innovation. The landscape is rapidly changing. Traditionally, semiconductor companies designed chips for automotive electronics using older, mature technology nodes. But with the rapid rate of innovation and consumer demands, automakers are leaning to semiconductor companies that specialize in consumer electronics to accelerate innovation in this industry.
Self-driving cars increasingly use sensors such as radar, LiDAR, ultrasound cameras and fusion sensors to enable 360° surveillance, object identification and classification. This is all aimed at preventing crashes and ensuring reliable operation in the worst weather conditions. The vast amount of data gathered by the sensors need to be processed in real-time, and decisions must be made dynamically. Imagine driving down a dark road on a rainy night with poor visibility and a racoon suddenly dashes across the road. The radar must detect that the object is indeed a raccoon and relay that information to the processor, which processes the information to perceive the context of the raccoon crossing and take corrective action to gently brake without hydroplaning. That’s a lot of key decisions to be made within a few micro seconds, any failure of which could result in catastrophe.
Whether it is Nvidia’s Drive Pegasus or Intel/Mobileye’s autonomous platform solutions, the goal is to create a supercomputer on wheels. That includes high-performance, multicore processors, which constitute the car’s brain, and various smart sensors, which serve as the eyes. These systems need to continuously sense, plan and act, and in the process they will have to learn to navigate scenarios that have never been encountered before.
The inferencing happens on the edge, while the deep neural network training happens in the data center. But these systems also need to constantly communicate with the cloud to get the latest and the greatest updates on new inferencing algorithms. To bring this all together, OEM’s are directly partnering with semiconductor companies to accelerate this range of innovation, and if all goes as planned we can ride in the first self driving car in just another four years.
As the saying goes, with great power comes great responsibility. As automakers race to get the next best self-driving car out in the market, they are faced with numerous challenges to meet the strict safety and reliability requirements for a safe driverless future. To accomplish this, they need to leverage simulation capabilities to design such complex, transformational products.
One of the goals of autonomous cars is to reduce the number of accidents. Nearly 94% accidents today are caused by human error. But at the same time, an autonomous vehicle will only be as good as the electronics inside of it. Automotive electronics, unlike consumer electronics, must operate in very harsh environments for extended periods of time. They need to be designed to be highly highly reliable, safe and have a zero-field failure rate over a lifespan of 10 to 15 years. Any failure can lead to a worst-case scenario involving a fatality, which is unacceptable. A better-case scenario involves a recall, which is costly to the carmaker. Hence, automotive OEMs, Tier 1 and semiconductor companies must work closely in partnership to define the semiconductor and functional feature requirements of such advanced systems.
EM, ESD and Thermal Reliability
The semiconductor chips that power these autonomous systems use advanced technology nodes to provide the required performance. According to Gartner, when compared to 16nm/14nm technology, 7nm offers 35% speed improvement, 65% less power, and 3.3X density improvement.
But reliability issues are very challenging at these advanced nodes. FinFET designs have high dynamic power density, and power directly impacts the thermal signature of the chip. Increased functionality and higher current densities cause localized self-heating of devices and joule heating of wires, leading to a large variation of temperature across the chip based on different modes of operation. Higher temperature, higher current and higher resistances are pushing the limit for electromigration (EM) and electrostatic discharge (ESD) failures on chip. For certain mission-critical applications such as air-bag deployment devices, ambient temperature of the chip also can affect the transient behavior.
In addition, advanced 2.5/3D and wafer-level packaging technologies are bringing the die and wafer together, while creating more thermal hot spots that will impact both the chip and system level EM and ESD. These approaches also can increase the chance of thermal-induced stress, which can lead to warping and contact separation, and causing long-term reliability issues that ultimately will render the product useless.
Automotive electronics systems require detailed thermal modeling for reliability analysis as typical operating temperatures can range from -40°C to 50°C while certain ADAS and power management systems under the hood can be subject to device junction temperatures as high as 135°C to 150°C. Thermal issues are very serious, for advanced FinFET technology nodes, self heating and joule heating can cause local temperatures to rise greater than ten degree centigrade as the surrounding materials of low K dielectric cannot dissipate the heat well into the silicon substrate, hence heat gets trapped in the wires and devices. This high delta T may lead to EM failure on the chip interconnects/device layers and extra efforts are required to quantify the temperature rises on devices and wires. Traditionally, automakers have assumed a constant power dissipation across the chip while performing system level simulations, but this simplistic model is not realistic anymore, the temperature variation across the chip can be very significant and should not be ignored.
The problem we are trying to solve is one of multiphysics, and that requires comprehensive reliability analysis and signoff solution that spans across the entire design chain — IP, chip, package, system — which addresses growing interdependency of various multiphysics attributes. Among them are power and thermal integrity, and reliability in sub-16nm designs, which are required for accelerating design closure.
EOS and Aging
Electrical overstress is another big concern, especially in automotive applications where the electronics must last for a very long period. EM is a time-based failure, where large magnitudes of currents flowing through thin wires over a long period of time can cause the metal atoms to be displaced, resulting in opens or shorts. ESD, meanwhile, is an event-based failure, where an externally applied high voltage event, typically of very short duration, can breakdown the device. EOS, in contrast, is a time-based failure that can be triggered by some operation of the device outside its normal parameters. EOS impacts the normal operation of transistor devices when a particular stimulus may result in a large unexpected voltage on a transistor gate, causing dielectric breakdown. This degradation of the device can happen over a period of time and it can cause irreversible total failure of the transistor.
Aging refers to the degradation of the device with time. Electric fields across transistor gates slowly degrade the dielectric of the device. Main physical effects are bias temperature instability (BTI) and hot carrier injection (HCI). Electrically, this manifests itself as a shift in threshold voltage, which results in reduced drive currents over time. That, in turn, causes increased delays, which eventually causes timing failures. Aging effects typically occur over a period of years (2, 5 or 10). Typically these show up as a slow degradation of individual transistor performance. Some transistors may significantly degrade while similar transistors might not be impacted. The aging phenomena is extremely sensitive to usage patterns. Common design techniques, such as clock gating, can exacerbate problem.
Electromagnetic compliance is a key requirement for automotive electronics. EMC constitutes both EMI and EMS. EMI pertains to the impact of electromagnetic emissions from an electronics system on other systems. EMS is the opposite, and pertains to the susceptibility of the electronic device to external electromagnetic fields. EMI compliance, starting from chip level, is cost-effective.
Simulation-based design changes are a necessity, especially with the use of complex multicore designs, which have to operate in a variety of modes to cater to various scenarios. Compact and accurate chip power and ESD models that can be used in system-level EMC simulations will become essential to meeting EMC requirements.
Functional Safety analysis is another very important requirement for automotive electronics. Safety standards like ISO 26262 require performing multiple analysis methods in a consistent and thorough manner. Efficient application of safety and reliability engineering methods at concept, system, software and hardware level can reduce up to 55% of effort and time-to-market for safety and reliability assurance. Various qualitative methods (e.g., FMEA, qualitative FTA) and quantitative analysis (e.g., failure-rate prediction, diagnostic coverage, residual and latent faults analysis, FMEDA, FTA), as well as dependent failure analysis are required. These methods need to be carried out at various levels ranging from IP design and parts and sub-parts of integrated components up to sub-systems and systems.
Certainly, designing automotive electronics systems is more challenging than designing semiconductors for a traditional mobile market. Working together, ANSYS and TSMC have defined workflows that enable electromigration, thermal and ESD verification and signoff across the design chain (IP to SoC to package to system). Within the comprehensive workflows, multiphysics simulations capture the various failure mechanisms and provide signoff confidence, not only to guarantee first-time product success, but also to ensure regulatory compliance.
Copyright © 2018 CST, Inc. All Rights Reserved