Summary:
Overview of ADAS architecture and MCU functions, two solutions for future ADAS architecture, and some thoughts on whether MCU will disappear from the ADAS domain
The installed capacity and penetration rate of ADAS will increase again, especially at the L2 and L2+ level; then, with the enhancement of the main control chip of ADAS domain controllers, will MCU still exist in the future?
Transferred from Zuosi Automobile Research
Contents of this article:
- Overview of ADAS architecture and MCU functions
- Two options for the future ADAS architecture
- Some thoughts on whether MCU disappears from the ADAS domain
0 1 ADAS architecture overview
As shown in the figure below, ADAS is generally composed of a functional safety MCU and a large computing power SoC. The SoC is responsible for sensor data fusion, running algorithm models, and performing calculations;
MCU is responsible for the safety of the system, monitors the working status of power supply/communication/other chips, and switches to a safe state when a failure of the automatic driving system is detected;
0 2Two options for future ADAS architecture
One is that the computing power SoC eats up the MCU, as shown in the figure below Renesas’ R-CAR-V3H, which integrates LockStep’s Cortex-R7@800MHz (2DMIPS), can complete the previous external functional safety MCU through the Real-Time CPU Function.
The other is an enhanced functional safety MCU, which can take over when there is a problem with the computing power SoC, to ensure the ADAS function of the minimum functional safety, such as Infineon's latest TC4X, which integrates a lightweight neural network acceleration core, PPU (Parallel Processing Unit), the PPU can perform post-data processing such as clustering and tracking on the imaging radar to ensure that the safety-related ADAS/AD function can also achieve Fail safe or even Fail operational with the help of ASIL D MCU in the case of SoC failure.
0 3 Thoughts on whether the MCU disappears from the ADS domain
1. Security
First of all, the most important feature of ADAS should be safety; so is the first architecture more secure or the second architecture?
Safety Xiao Er understands that security requires independent redundant backup, and it is necessary to ensure the absolute isolation and independence of the two units; Radiation interference or process defect effects or system interference are more likely to cause the same failure;
From this perspective, unless the security of AI SoC is certified by the industry, the second architecture should be more acceptable;
On the other hand, for the current functional safety MCU, the mainstream technology is still 40nm and 28/22nm, while the technology of AP is more advanced; the current automotive functional safety MCU can achieve a high operating temperature of 150°C; from this perspective, there is a External MCU seems to be more secure
2. Cost
Can computing power SoC integrated with MCU really bring about cost reduction? Here we do not take the pricing issue into consideration, but only from the design point of view, that is, to consider the area of DIE and the cost of packaging and testing; this part of Xiaoer has no account book, and it is understood that in order to ensure isolation and independence, the reduced DIE area will be limited; (wrong Welcome to correct me)
From the perspective of packaging and testing, it should be more advantageous to integrate MCU with computing power SoC;
However, there is a very important cost that is easily overlooked here, that is, software cost.
MCU needs the support of AUTOSAR, and AUTOSAR needs authorization fee; the customer system needs chips with different computing power, which may bring equal authorization fees; and choosing an MCU only needs a set of AUTOSAR authorization; Should be a cost issue that needs to be addressed;
(1) Simulation performance
According to some discussions on the Internet, the computing power SoC uses a higher technology, and its analog performance may be worse; Xiaoer thinks that this impact is relatively limited. Taking ADC as an example, in ADAS, it is mainly used for monitoring functions rather than precise measurement. Therefore, this part Simulation performance degradation should not be a major consideration
(2) Unspoken rules of the industry
Finally, there are unspoken rules of the industry: the current mainstream of the industry is the computing power SoC+MCU solution. To subvert the default rules of this industry, a leader is needed, and tangible benefits are needed;
One of the considerations behind the unspoken rules of the industry is substitutability. If you choose a computing power SoC to integrate MCU, in case of shortage or other impacts, is substitution more challenging than computing power SoC+MCU?
Source| Xinxiaoer's Afternoon Tea