The automotive sector could become one of the key markets for 3D integration according to the head of Audi’s progressive semiconductor program, but it depends on greater transparency between chipmakers, tier-one OEMs and the vehicle manufacturers themselves.
Berthold Hellenthal explained in a keynote at CDNLive EMEA on Tuesday (20 May 2014) how important understanding the progress of semiconductor technology is to Audi even though it does not design its own chips. Despite not being directly qualified for use in cars, he said companies such as Audi have found ways to incorporate leading-edge devices into their designs and are looking to use more of them. A possible method for that is likely to involve 3DICs, he said, and could open doors for more fabless suppliers to compete in the sector.
“There are no devices in advanced nodes that are made for automotive specifically. But we have to use the parts that are available,” Hellenthal said, noting that chipmakers for the car market cannot expect to stay on older nodes with products that replicate functions from the consumer space simply because they are designed for the full automotive environmental specifications.
“If the customer wants LTE in the car, we will not tell the customer it’s not possible. We have to find ways to use the devices that do support LTE in cars. If you are not involved in 28nm or don’t plan to go to 16nm, how can you still be in the market for the next generation of cars and take advantage of the automotive market?” Hellenthal asked.
New market players
The move into 3D modules presents opportunities for chipmakers, Hellenthal said. To some extent it has already started. “Today, nVidia is not just selling semiconductors but providing modules. We are driving semiconductor manufacturers to provide us more than the single piece of silicon. They are starting to provide solutions.”
Hellenthal said potential opportunities for using module-level integration in cars had been missed in the past because the tier-one OEMs did not see the same cost-performance tradeoffs that a carmaker like Audi would. This realisation is pushing Audi to look more closely at what goes into ECUs and similar modules although the tier ones will continue to do the actual implementation work.
“We have started to work on semiconductor specifications. We influence them so we can benefit from the next generation the most. We will see a shift in the value chain and a shift in competencies.”
TSVs for automotive
As an example of the kind of integration Audi wants, Hellenthal pointed to an experimental sensor-node system-in-package fabbed by Fraunhofer ISM. This combines a sensor, micro controller, passives and power converter in a multilayer block. “It is a complete, intelligent solution to a problem that we have. “Integration offers new business cases. For example, it makes more sense to put more money into semiconductors [for communications] so we can use less expensive, lighter cables. From a system perspective, the benefits are great. This is very beneficial to us but it does change the value chain.”
“When it comes to thru-silicon vias, people are debating whether it is suitable for automotive or not. What we have seen in the past is that there is no obstacle to being suitable for automotive. This is why we get involved with thru-silicon vias today. To find how we can make these things feasible.”
Hellenthal referred to the use of 3DIC technologies as the “intelligent integration puzzle”, treating the individual elements as pieces of a jigsaw puzzle.
“We can have intelligence right at the sensor where it’s needed. It gets the digital traffic [between modules] down to where it’s bearable. And it is perfect for adding new things. You may see some new digital driver for a MOSFET. You can change the driver but not change the MOSFET. You can change technologies without having to design everything new. We can also change things to suit different application needs,” Hellenthal explained. “It’s very interesting on the volume and investment side. For situations where it doesn’t make sense to make new silicon, it makes sense to work on the module side.”
Hellenthal said integration would provide smaller, lighter ECUs. “Why not integrate all of it within a system-in-package. Put a connector on it and it’s an ECU. It is not something we have done yet, but in the future we are looking for these things.”
“I’m also talking to a lot of fabless people. Fabless people have one big advantage. They don’t have a big problem taking a part they don’t make themselves and putting it in a module. They can take a PMIC and sell it in their module under their name. The larger IDM does have a problem picking products from other vendors.”
Hellenthal added: “System-in-package helps in other aspects. For example, the latest generation of NAND flash doesn’t come in 0.8mm pitch BGAs. As the footprint of these devices becomes smaller, we can use package-in-package for them. So we can use parts that are not [designed for] automotive very quickly.”