By Ann Steffora Mutschler and Ed Sperling
As automotive OEMs come up to speed on electrification of vehicles, each at their own pace, they are starting to embrace novel packaging approaches as a way to differentiate themselves in an increasingly competitive market.
Wirebond used to dominate this market, where most of the chips were relatively unsophisticated and product cycles were slow—sometimes as long as five to seven years. But as automakers begin planning for an increasingly fast-moving driver-assisted or fully autonomous future, they have begun considering advanced packaging as a way to keep their technology current and to get products to market more quickly.
“We are seeing an increase in new and different packaging choices and considerations for automotive applications,” observed Edward Fontanilla, deputy director of product technology marketing at STATS ChipPAC. “One example is high-frequency radar devices for 77GHz Advanced Driving Assistance Systems (ADAS). These high-frequency radar devices require a much tighter RF signal isolation and aggressive performance targets. Fan-out wafer level packaging (FOWLP), particularly embedded wafer-level ball-grid array (eWLB), is becoming a prevailing packaging choice.”
Fig. 1: NXP’s FOWLP radar module. Source: NXP
Fontanilla pointed to a number of advantages to using a fan-out approach in automotive electronics. Among them:
• It eliminates the need for a laminate substrate, replacing it instead with a copper redistribution layer (RDL). That RDL has a shorter connection distance, which in turn significantly reduces impedance.
• As with many advanced packaging approaches, it also uses a smaller form-factor with lower parastics in the interconnection, which is critical for high-frequency applications.
• A lower tolerance from wafer-level processing enables better yield, which is cost-effective for high-frequency applications;
• And like all fan-outs, it provides more design flexibility because there is less routing interference and sufficient isolation for RF channels.
Others report a similar increase in attention by automotive manufacturers in advanced packaging.
“There is a direct interest from car manufacturers themselves, and more specifically from Tier 1s,” said Jean-Marc Yannou, senior technical director for ASE Europe. “OEMs such as Audi recognize that differentiation in cars more and more comes from electronics because car manufacturers have solved most of the problems with internal combustion engines. So they go electric or hybrid. That’s one differentiator. Another differentiator is that cars go higher-end all the time. Consumers wants a lot of gadgets in the car such as to be able to electronically adjust the seat, heat the seat, open the trunk or sunroof just by pressing a button.”
He noted that automotive OEMs are well aware that vehicle differentiation is possible because of the electronic components. They also understand they can’t differentiate by using a different CMOS technology, so it wouldn’t make a lot of difference if they stay at an older node or move to 10/7nm. For automotive applications, the volumes are not high enough to see a significant price drop with device scaling.
“The investment/CapEx levels are so high that very, very high volumes are needed for automotive makers to justify this move,” Yannou said. “They will follow, but they follow well after the wireless industry. They are still about five years late on new CMOS technology node adoption.”
When it comes to packaging, though, it’s a totally different story.
“In the past it was the same as for silicon nodes,” he said. “We were only following suit for the wireless industry. Now we do more and more things that are actually specific for automotive. Audi said all the differentiation of semiconductors, which in itself is the differentiating factor for electronics and for cars in general, will be packaging. Packaging will be used to bring the performance to a higher level with better thermal dissipation and lower electrical losses by shrinking the whole size.”
While the size of the electronics package can be a factor in design, given that there is plenty of space in a car, the real focus is on the ability to hide that circuitry. “Electronics have to be invisible to the eye of the user,” Yannou said. “Miniaturization is one important factor.”
What kind of package?
This isn’t a simple swap out of one technology for another, however. There are a number of factors that need to be considered, such as how these devices will be used, the environment in which they will be used, and which type of packaging is best. Current bets are on fan-outs, but that also could shift to include other packaging types.
“The environment it will be used in is critical,” said Tom Salmon, vice president of collaborative technology platforms at SEMI. “You need to think about whether it will be under the hood, in the infotainment system, and how that will all look in 2025. Then, what are the materials that you will need for packaging? This may be very different because you no longer have Tier 1 and Tier 2 players defining things for the next 20 years like they did in the past.”
Packaging provides some flexibility because not all of the components need to be changed out for every new application. Working groups for the Heterogeneous Integration Road Map—now being developed as a successor to the now-defunct International Technology Roadmap For Semiconductors—working on a set of reference platforms for various markets such as automotive. Along with those markets different working groups are developing guidelines for 2.5D, 3D and fan-outs.
“The goal is to create optimal platforms to develop the right components and still have the flexibility to use wafer-level packaging or chip-scale packages or whatever is optimized for a particular environment,” Salmon said.
Fan-outs have been getting much of the attention lately. Jan Vardaman, president of TechSearch International, described fan-out wafer level packaging as a disruptive technology because there is no substrate and no traditional underfill, and all packaging can take place at the foundry or at the OSAT. But she noted that it does require chip-package co-design.
“Fan-out wafer-level packaging is used for radar modules because of performance,” said Vardaman. “This is RF, so lower parasitics are very important. There also is a lot of wirebond today and a lot of QFNs (quad-flat no-leads). The higher-pin-count microcontrollers are moving from wirebond to flip chip, and there are some system-in-package modules in various parts.”
Fig. 2: Fan-in vs. fan-out. Source: TechSearch International/STATS ChipPAC
Reliability and the unknown
Underlying this shift of how to put chips together is a concern for reliability, which always has been a concern in automotive and other safety-critical markets. This used to be a relatively straightforward discussion, but as more electronics are added to vehicles it has taken on a whole new level of complexity. Packaging is just one more facet of this discussion.
“There needs to be a systematic approach to functional safety, but it’s difficult to determine how we’ll achieve that,” said Robert Bates, chief safety officer for the Embedded Systems Division of Mentor, a Siemens Business. “You need to plan for the possibility of random failures. They do happen. And you need to do this with the understanding that both the hardware and the software are getting more complex. For machine learning and neural networking, the software is not directly implementing what it’s doing. And for safety-critical hardware, typically this is a few generations old and you’re adding unproven and different conditions into autonomous systems.”
While advanced packaging allows the ability to swap components in and out, the process is more predictable if those components were developed to work within those packages. The problem is that automakers are in the middle of a mad scramble to stay current with this technology, so they increasingly are making use of various commercially available components that were never developed for advanced packages and which have not been tested under extreme conditions for extended periods of time.
“This puts Tier 1 and Tier 2 suppliers in a difficult spot,” said Bates. “They need to provide more data.”
While that data can be simulated, there is far less data available from real-world testing, which in turn puts the onus on companies to do more verification and testing of the components.
“There will be lots of testing as we ‘follow the chips,’” said Anil Bhalla, senior strategic marketing manager at Astronics. “It will be necessary to verify that designs and manufacturing operations produce solutions that are defect-free and safe according to regulatory standards. Automotive will require the use of much more proven technology than consumer devices such as smart phones. And this testing will need to happen both at the component and system-level.”
That’s just the starting point, too. “Safety-critical applications will require additional testing as we better understand the defect mechanisms,” said Bhalla. “The industry is assuming this will work based on early trials. More trials on a broader scale will support the gradual roll-out of this new technology. The economics of autonomous driving is motivating the entire semiconductor ecosystem to evolve during this transition.”
Supply chain shuffle
The push toward autonomous vehicles and advanced electronics can be traced at least partly back to Tesla, which until several years ago was largely ignored by most big automakers as a niche player. Tesla’s rollout of autonomous driving technology changed all of that, disrupting the supply chain and calling into question relationships that had been in place for decades. Carmakers suddenly had to scramble to stay relevant, ratcheting up their efforts in driver-assisted and fully autonomous engineering.
But rather than develop everything from existing components, they began searching for the best technology from wherever it was developed. This isn’t just one or two components, though. It requires an assortment of electronic content, including RF sub-systems, power management units, motor controllers, a collection of sensors, and advanced logic that uses CPUs, GPUs, and a variety of hardware accelerators.
Moreover, many of these components are being developed by companies that until recently had little or no interaction with each other. ASE’s Yannou recalled one recent meeting set up by an OEM between ASE and the Tier 1, Valeo. “The Valeo people were a little bit upset in the beginning because they were asking what I was doing in the room with one of my sub sub sub contractors. The OEM told Valeo, ‘You want to work with ASE, and together you will do this integrated module. And if you can bring everything on one side of it, then you can build the Wi-Fi module on the other side by freeing space on the second side of the PCB.’” This project is running today.
The OEMs also realize that today, building a car is like assembling modules together. “When you look at the chassis of a Tesla vehicle it’s overly simple. It’s just tubes of aluminum welded together, a lot of batteries, an electrical motor, some electronics and that’s it. Other car OEMs—especially the European ones, but also the Toyotas and GMs and Fords of the world—all used to be internal combustion engine specialists. But since the industry is moving away from those engines, things are changing. When you think about it now, a battery expert or an electronics expert can build a car as well as an engine expert like Ford, GM, Toyota, Renault, Peugeot, Volkswagen, Audi, BMW. All of these traditional car OEMs actually fear for their future business,” Yannou said.
That has a direct effect on the packaging technologies being developed for automobiles. Wirebond is still used in a majority of automotive infotainment applications, such as the GPS, audio controller and USB devices, as well as in the MCUs used in electric and hybrid vehicles, the instrumentation, the power management systems, and body systems, which includes Ethernet, transceiver and interior lighting components. But for new systems, such as ADAS, as well as more current infotainment, instrumentation and body systems, fan-outs are beginning to gain traction, according to STATS’ Fontanilla.
In addition, those packages are beginning to include the same kinds of chips that have been developed for the mobilility market. “Today the dominant pitch is, ‘If it works in smart phones, why wouldn’t it work in cars,’” said Yannou. “And as a matter of fact, very often there’s very little — if nothing — to be changed so that it works. It’s just a matter of qualifying it long enough and hard enough. The technologies are quite robust actually.”
Exactly how this will play out in the packaging world isn’t obvious, though.
“Large IDMs like Infineon or large Tier 1s like Continental tend to use mature technologies, which are very well proven as being robust for automotive like QFP (quad flat package) or even through-hole type of packaging,” Yannou said. “They would refrain from using organic substrates BGAs, LGAs. They would refrain from using QFN because you don’t see the leads, so you can’t check on the solder connections to the board. And these are valid reasons. Smaller players may dare to use new technology, so that kind of shakes the whole supply chain forces large players into rethinking whether to move forward with some new technologies, as well.”