Challenges in Automotive ASIC Design Based on True Stories published at EET ASIA
Inomize is a professional research and development company specializing in ASIC design and manufacturing of hardware and software solutions for OEMs and ODMs.
In recent years, the automotive industry is undergoing a tremendous revolution, with more and more electronics systems being introduced into cars by all major manufacturers.
Many of these systems are built on ASIC infrastructure, where both ASIC players and car manufacturers must collaborate to provide the required solutions for the market.
In such area, one cannot rely on existing solutions. In many cases, there’s a need to develop the technology from scratch. Also, in the case where one is using some off-the-shelf IP’s or modules, they must have a deep knowledge on how to integrate these and make sure all IPs are in sync and conform with the requirements. The challenges and risks that such development faces are huge, and one would like to have an experienced team which have already been there and done that.
With rich experience in various ASIC projects in the electric vehicle domain (EV), Inomize is in a good position to assist EV manufacturers, module makers, and ASIC companies (supplying to the car industry) and address the needs of this growing market.
Inomize’s unique advantage is in understanding the overall technical system requirements, the target market, and limitations, paving the way for the best-in-class ASIC solution.
Inomize CEO Udi Shaked says, “On every moment, the company is working on more than five projects, where each of these could have been a basis for a startup by itself. The company and its engineers are dealing with the forefront of technologies in every possible domain.”
Inomize engineering team is intimately familiar with the various standards, protocols, and concepts associated with designing an SoC for the automotive market, such as:
Advanced driver assistance systems (ADAS)
ISO 26262 – Road vehicles – Functional Safety standard
Automotive Safety Integrity Level (ASIL)
In-vehicle infotainment (IVI)
Light imaging, detection, and ranging (LIDAR)
Connected car
Autonomous driving
EV battery control and power management
Inomize: Expertise in Automotive Market
EE Times Asia met with Inomize to get a glimpse of the company’s previous experience in projects involving automotive functional safety and what the key takeaways are from these projects.
Automotive Sensor Fusion – This is a classic case where a huge idea started in a coffee shop meeting. Back in 2015, Inomize got a request from a Tier-1 company to develop a sensor fusion chip for autonomous driver chipset on 16nm process node.
The chip interfaced multiple image sensors simultaneously and supported image processing, starting with a standard ISP, distortion correction, and AI-based features enhancement by multiple dedicated CNN (convolutional neural network) hardware processors.
Image streaming, formatting, and extraction is done with an optimal shared memory subsystem that was based on multiple LPDDR4 and multi-way Last Level Cache (LLC) of several MB.
Quad-core CPU managed all data flows and traffic. It also communicated with the host companion chip via various high-speed interfaces.
A network-on-chip (NoC) containing almost 100 masters and slaves with bus width up to 512 bits provided the required infrastructure for the chip.
The challenges Inomize faced were:
Providing floor planning-oriented architecture by splitting the NoC and the LLC with the multiple LPDDR4 into four corners of the chip and reach fully abated solution with no wiring at the top.
Implementing a mesh structure to reduce data interfaces between physically remote hardware pieces.
Using several ISPs and distortion correction units as a shared common hardware for all sensors and reach high utilization of these machines for several type of sensors.
Applying functional safety (ASIL B level) by adding hardware agent near each NoC master and slaves that add ECC for all data transfer over the NoC for both data and address, as well as monitoring all transactions based on unique ID, timeout, and firewalls.
Supporting logical BIST per each CNN machine, memory BIST on demand, parity for all configuration registers, and ECC on each memory with address parity as well, all managed by dedicated ASIL D safety CPU.
Performance optimization of the fabric to provide continuous and optimal data flow.
Supporting secure boot mechanism by serial flash, cryptographic engine based on firewalls, ASIL D CPU and LLC to bring secured software image to the quad Core CPU, all other data flows were blocked by the hardware during the boot process.
Supporting wires tunnel as hardware macro without real clock tree synthesis due to dimension of the tunnels (narrow and long).
Supporting two-boot configuration for full and half performance (half memory, half CNNs).
In-Vehicle Networking ASIC – A different customer asked Inomize to architect, design, and tape-out an ASIL B chip using the TSMC 28nm process node (Automotive Grade 2). To reduce time to market, Inomize incorporated a previously designed ARM Cortex subsystem into this ASIC.
Inomize team had held FMEDA sessions and assisted the customer to maintain FMEDA tables, identifying failure modes and defining failures detection mechanisms. As part of this process, Inomize utilized its knowledge of previous automotive projects.
Some of the IP’s that were integrated in the chip had built-in safety support, which needed to be studied and integrated into the design. For other IPs, Inomize have built safety solutions of its own.
An example of a safety solution tailored around an IP is a complex AHB matrix with a total of 20 masters and 20 slaves with parity checks over AHB matrix transactions. The parity bits are transferred as user bits of the control (Address), Write data, and Read data channels.
Other safety elements were:
The Main Safety block manages all safety activity of the chip. For example, it has Error counters for each indication and Error injection mechanisms to verify safety indication logic is functional.
All SRAMs support ECC (SEC/DEC) with address parity and with error injection.
LBIST of the main IP’s. The LBIST is inserted over the netlist together with DFT insertion.
Clock monitors to make sure the clocks (PLLs and critical clock dividers of the chip) are in their correct frequency.
Safe clock generation. The need for an external crystal was eliminated by implementing a proprietary internal oscillator.
Watchdog timers allowing reset of the chip in various modes. Once started, the watchdog timers cannot be accidently disabled by the software.
Temperature and voltage sensors to detect temperature and voltage conditions.
E-FUSE with multi bit protection.
A dedicated safety pin to report faults to the system’s safety manager which was defined by Inomize.
A functional safety assurance tool was used to check the efficiency of our fault detection mechanisms.
All synchronizers in the design were verified not to create an MTBF issue by selecting synchronizers FFs with special parameters for synchronizers reliability awareness during placement.
DFT at automotive grade: ATPG of 99% Stuck-At faults and 85% transitions faults, full MBIST, etc.
Analog LiDAR ASIC – A pioneer company in analog LiDAR development asked Inomize to design an ASIC incorporating their laser-based object sensing solution for ADAS and autonomous driving.
Inomize analog team, together with the customer’s engineers, have implemented the company’s next generation FMCW LiDAR technology in an advanced ASIC solution that will enable superior vision capabilities for future vehicles.
The combined engineering teams developed a single mixed-signal silicon chip that incorporates state of the art low-noise trans-impedance amplifiers array with all the supporting circuits needed to shape and drive large amount of data generated by the customers’ sensor.
Summary
This article discussed several projects and examples from Inomize experts in the automotive market. Apart from this market, Inomize’s expertise spans across many other parameters not discussed in this short article. To learn more about how Inomize (https://www.inomize.com) can help you reach your design goal, don’t hesitate to contact us at +972-72-277-5400 or e-mail info@inomize.com. See article at EET ASIA