Cabling and its weight are helping to drive integration and a possible shift towards wireless communication within cars, Lars Reger, CTO of NXP Semiconductors’ automotive unit, explained during his keynote at CDNLive EMEA this week (April 28).
Reger said the company is looking at the possibility to move some of the body electronics to use wireless communications to report their status and relay commands rather than wired cabling to save weight. A key stumbling block lies in the reliability of wireless and its vulnerability. He said researchers are determining “where are the security constraints low enough that we can rely on wireless?”
Wiring is one of drivers behind the development of a specialized form of Ethernet that car companies now want to distribute through their vehicles, and is driving integration decisions at NXP that will influence the technologies needed for self-driving cars such as vehicle-vehicle (V2V) communications and radar.
“Regular Ethernet has four wires, which means we are doubling the amount of wires and weight [compared to existing automotive networks]. That’s nasty,” Reger said. “Can’t we put ethernet over regular twisted-pair cabling. Broadcom had developed this with BroadR-Reach. We have licensed the patents for it and developed our own silicon.
“With BroadR-Reach, you tend to see a lot of passive components around the silicon PHY. We have integrated this using SOI to bring many of these passive components inside the silicon.”
Ethernet meets RF
The shift to Ethernet is changing decisions over integration in subsystems such as radio receivers, Reger said. “Since 2008 we have tried to up-integrate tuners and basebands into RF CMOS, building complete system-on-chip receivers.”
The result of this integration was the Dolphin SoC, which uses Tensilica’s processor cores for the software-defined processing it performs. “It’s a one-chip, receive-everything solution that it will end up in the RF antenna on top of the car and will be connected by Ethernet. We can take out €15 to €20 of copper cabling using these integrated receivers.
“With this, we have shown we can do this in sub-gigahertz RF,” said Reger. Now the aim is to take integrated RF into the millimetre-band spectrum and integrate other wireless subsystems so they can use a single network port, starting off with 6GHz for automotive WiFi.
Then, two years ago, work started on a CMOS RF design that could handle radar at in the 77GHz band. “On Christmas Eve they called and said we had taped it out to GlobalFoundries,” Reger said, adding that at the previous Consumer Electronics Show, “people said ‘you are a lunatic it won’t work’. Now we are out with a prototype front-end RF CMOS device.”
Reger added: “We will have little radar stamps, to replace the parking distance controllers we have today and able to use the same wiring harness. But they will be able to provide complete radar cocoon.
“Radar is one of the key technologies for self-driving cars. These sensors can also measure big bags of water: if I’m standing in front of traffic light, it can detect if I’m crossing. If children are crossing, the car can use V2V communications to send message to cars nearby to tell them ‘don’t approach this junction at 60km/h’.”
The need for multi-gigahertz RF integration into commodity CMOS and SOI processes needs better support from EDA, Reger said. “With analog and mixed-signal parts we would like to get to digital design standards. We need the best-qualified, least-surprising flows, especially on the V2V part. We can’t wait for silicon to come back from the factory. We are missing three months of learning. I need to know what I taped out immediately. That works well in digital. But in mixed signal today, that’s more difficult.”