Why chip design gets relevant for carmakers
There are many commonplace facts regarding the electronics content in today’s and tomorrow’s cars. For instance, that in every car generation the value of chips and control units in cars is rising, and that some 80 percent of the innovation in vehicles is associated to electronics and software. Without few exceptions — for instance, Audi’s semiconductor cooperation with a number of semiconductor manufacturers – the automotive industry contented itself with the role of the customer.
Current trends in automotive electronics as well as in chip manufacturing however could cause these two distant ends of the value chain to get closer together, says Robert Schweiger, Director technology Solutions Automotive at Cadence EMEA. So what does car design has to do with chip design? “This is what we have been asked frequently by automotive engineers”, Schweiger said. “They say: Since we do not make chips, we see no need to get in touch with companies like Cadence”. Perhaps they should think twice, because the software from CAE vendors like Cadence is the key to the IP contained in the SoC which run the software the carmaker or tier one is developing. “For this reason, it is not entirely unreasonable to have an active influence on this IP.“
Things like the Connected Car, autonomous driving and integration of consumer gadgets into vehicles are driving the trend towards higher integration. In future ECUs and Advanced Driver Assistance Systems (ADAS), SoC with complex building blocks and IPs will displace today’s microprocessors. “The ECUs for the next generation of cars will be higher integrated and implemented as SoCs or SiPs (System in Package)”, Schweiger said. Higher integration at chip level offers multiple benefits, from IP protection to saving space and weight. “Space is increasingly an issue in modern vehicles. There is almost no space available to place the ever-growing number of driver assistance systems.”
Another big factor, according to the Cadence expert, is the computing power required to run the complex algorithms. From 2012 to 2015, the number crunching capacity in dashboards and under the motor hoods has tripled, and it looks very much like future car systems will require a multiple thereof. “This leads to the realization that conventional microcontrollers will no longer be sufficient to run demanding applications like ADAS which have to process radar and video signals in real-time. In the infotainment domain, SoCs are already established, now they start to enter the ADAS markets.
This brings us back to the question why an automotive OEM should busy himself with chip design. The reason is that SoC architectures, semiconductor functional blocks determine the functionalities the carmaker wants to implement. “In such applications, you don’t see standard SoCs but increasingly custom-made designs”, Schweiger said.
The rising demand for computing power and the likewise rising market volume shifts automotive applications in the focus of leading-edge chip process technology. Many technologies recently developed for consumer markets are now qualified for automotive use with AEC Q-100 certificate. Examples are 28nm FDSOI and 40nm Low Power from GlobalFoundries or TSMC’s 28nm process. (At the Cadence event, GloFo showcased “automotive-ready” versions of its latest processes) “This gives chip customers in the automotive value chain – including OEMs – the chance to make use of available IP and drive integration higher,” Schweiger said.
Another motivation for automotive electronics designers to get granular on chip design is speed. Car design cycles continue to lag way behind the chip design cycles. In times when electronics and connectivity in cars is a differentiating factor, carmakers feel the pressure to close the gap and speed up their design cycles. Utilising complex SoCs however confronts carmakers with a problem: They only can start building prototypes when first silicon specimen is available. Or they can start right away if they use simulation and virtualization. “Chipmakers are used to do it this way,” Schweiger said. “But automotive OEMs and tier ones will have to realign their development processes.” Simulation and virtualization would not only cut their time-to-market but would also enable them to master the complexity inherent to their designs and applications. With such techniques, they can start developing their software and test it on a virtual prototype and make better use of the time the chipmaker needs to implement the SoC designed according to the carmaker’s specs. “Both processes no longer run sequentially but in parallel”, Schweiger says “Users can save a lot of time”.
And it is not only about saving time. The virtual test system also enables automotive electronics designers to test the functionality of their complex ADAS algorithms and examine if the hardware they have chosen is performing good enough for the software, or if the interplay between hardware and software works properly. “It is like a hardware-in-the-loop (HiL) test – just with a virtual hardware in it”, Schweiger said.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
