The automotive IC industry is currently focused on the challenge of building advanced driver assistance system (ADAS) SoCs that deliver very high performance image processing from limited power budgets with high reliability. What’s less of a focus at the moment is the process of reliably turning real-world analog signals from the many sensors involved in automotive systems into a digital data stream for onward processing.
Manuel Mota, product marketing manager, analog IP, Synopsys, addresses exactly this issue in a recent posting on the DesignWare Technical Bulletin.
His piece considers various aspects of designing semiconductor IP that will be suitable for use in advanced SoCs for automotive applications, looking at design for reliability issues, the challenge of meeting functional safety requirements, and the specific functional safety requirements of data converters.
One of the challenges of achieving the necessary level of reliability for automotive IP and SoCs is meeting the requirements of the US Automotive Electronics Council’s AEC-Q100 standard. This defines four maximum ambient temperature grades (0 thru 3) at which components should still work properly.
Mota describes how designers have to translate these ambient operating temperatures into junction temperatures on the silicon, by taking into account the average activity in the SoC and the packaging’s thermal resistance and applying a temperature profile that describes how long the device is expected to operate at any given temperature range during its lifecycle.
The article also describes the factors that go into defining the maximum failure rate of a device, which has to be less than one defect part per million throughout an automotive product’s lifecycle of 15 years. Mota describes the impact of factors such as design variability, transistor aging, and electromigration. He also explains the impact of automotive devices being required to work instantly at the expected maximum functional temperature, even if the temperature profile says that the SoC or IP may only work at that temperature for a fraction of its operating lifetime.
The second major focus of Mota’s article is the challenge of meeting the automotive industry’s preferred ISO 26262 qualification for functional safety, especially in safety-critical applications. Key functional safety considerations for automotive IP include whether it can detect, report and self-correct a fault; detect and report a fault quickly enough to enable the rest of the system to execute safety measures in time; and whether external measures can protect the system at a level above the IP.
Meeting these requirements in data converters is a challenge. Traditional protocol-level fault detection and correction mechanisms are not applicable. The data converters don’t have the processing capabilities or knowledge of signal dynamics to determine if a signal is corrupted. Self testing may be impossible, since it would interrupt the normal operation of the converter. And so the automotive safety requirements of these converters must be addressed in other ways. These can include better characterization and production testing, automated testing strategies, and in-operation system level tests made possible by adding test features to the converter.
Mota goes on to argue that these features alone may not be enough to meet automotive functional safety requirements, and should be supported with external functional safety measures that can identify and address a converter’s safety risk at the system level without impacting the overall system safety. One simple way to do this is through functional redundancy, as shown in Figure 1.
Figure 1 Functional redundancy for identifying operational fails in a data converter (Source: Synopsys)
For the full story, read the Data Converter IP for Automotive SoCs white paper.