One of the keynotes at the recent DesignCon was by Robert Heath of UT Austin titled 5G for Vehicle-to-X Communications. He started off with an overview of 5G, pointing out that there is a move from mobile handsets to other markets: e-Health, energy, factor of the future...and of course automotive. As he put it:

You are a carrier, everyone has a phone, who do you sell to? Can we sell phones to cars?

One issue with autonomous driving and even ADAS is that it is limited by line-of-sight. We can't see around a truck, and neither can radar or lidar. But one vision for how that could change is given by the example of overtaking a truck. The truck could say pass or don't pass. An oncoming vehicle could say it. Or the truck could just broadcast a video of what it can see at the front. Or a nearby cell-tower, which is a lot higher than a vehicle and so has a better view, could say something.

Robert pointed us to a document 5G PPP White Paper on Automotive Vertical Sectors that he took a quote from:

Full automation requires Gbps data rates and ms latencies.

I remain a skeptic, to be honest. One is that the communication industry always says this (because they hope it is true, at a minimum), and the automotive industry never does. There is also the existence proof: we drive our cars without needing access to basestations at all. True, these days we use GPS in our phones and use the connectivity of our phones to get maps and directions. But even our phones don't have Gbps data rates nor ms latencies.

Connectivity Options

There are a lot of connectivity options, 5G is not coming out of a vacuum.

At the low-end there is DSRC (Dedicated Short Range Communications). It seems that everyone in the automotive industry is waiting for everyone else in the ecosystem to implement this (other vehicles, infrastructure like traffic lights). It's unclear whether this will ever be a factor. The FCC allocated a band for this...in 1999, twenty years ago. ETSI in Europe did too, in 2008, ten years ago. So far it has only been used for electronic toll collection. It has millisecond latency but low bandwidth. Now the government wants the spectrum back, since it is in a band with great characteristics and so too valuable to let lie fallow.

Then there is cellular, with LTE (the dreadfully named Long Term Evolution that everyone simply calls 4G). This has 10-100ms latency and 1-10Mbps bandwidth. Up with LTE Advanced Pro, latency comes down to the millisecond range and bandwidth up to 10-100Mbps. True 5G should have under 1ms latency, and higher bandwidths still, up to 1Gbps.

V2X sounds like it might have a lot of choices, but X stands for either V or I: vehicle-to-vehicle or vehicle-to-infrastructure. There are clearly some possibilities for traffic efficiency with things like platooning, reducing braking shockwaves due to congestion, more efficient use of intersections. For this the latency needs to be low but low bandwidth connectivity is adequate.

Infotainment requires higher bandwidth with more passengers doing things like gaming or having a mobile WiFi in the car. As cars get more automated, that even includes the driver. Again, since everyone can already do this on their existing phones, this barely seems like part of the automotive market to me.

Here are the requirements that the 5G industry partnerships came up with. But, as Robert said:

I'm not sure these numbers didn't come out of someone's backside.

The status of 5G V2X development is that the first release of the standard doesn't have special support for vehicles. The main thing that seems to be going on with the standard is that people want it to evaluate, so that they can decide who gets what intellectual property when. Nothing has been developed yet related to applications (like automotive in particular).

mmWave in Automotive

mmWave spectrum is really valuable for vehicles, and is not just a spectrum multiplier. But there are implementation challenges. There are lots of tiny antennas. It scatters more than it reflects, resulting in fewer paths from transmitter to receiver. However, we can't increase the transmit power to compensate since "the FCC doesn't want us baking people." As a result, the same power has to be spread over a larger bandwidth and the signal-to-noise-ratio (SNR) is lower.

5G mmWave is beam based. When you turn it on, it has to find one of the beams being broadcast, and then decide how to adjust its own beam. This is challenging for high mobility. There are some optimizations, such as using the sub-6GHz bands to identify the direction of the basestation and to have it provide other information to get the connection set up correctly, and so reduce what is called the training length—the time to get the transmissions between basestation and car synchronized.

There are further opportunities for multi-antenna use to multiple basestations, so that the connection is not lost when something (a truck, say) obstructs one of the beams.

Some other blue sky ideas are to use the communication signals for radar, since the car being pinged and the basestation can listen, too. They are similar frequencies.

Conclusion

My big takeaways were, despite the title, that 5G is not going to be a critical part of automated driving for some time. In fact, it's unclear how effectively smartphone usage will work from inside a moving car, keeping all the beams aligned, switching between large numbers of basestations. I don't think mmWave goes through glass, so some sort of repeater would be required. Of course, the sub-6GHz bands should work pretty much the same way as with 4G.

In the UK, many phones are not sold by the carriers but by specialized stores. One of the largest is called Carphone Warehouse, a name that seemed a bit archaic once our phones were in our pockets and not our cars. It would be ironic if the name turned out to be prescient not archaic, and we end up needing carphones again, even if they are communicating with our smartphones rather than requiring a bulky in-car handset.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.