Welcome back everyone, I'm Jordan Geesege and this is The Limiting Factor. Last year at Investor Day, Tesla said the next generation vehicle would use a 48 volt architecture and an ethernet network that would eliminate cross-car wiring, which they later revealed was called Etherloop. The 48 volt architecture understandably attracted more attention and for good reason. The benefit of 48 volts is obvious, which is that quadrupling voltage from a conventional 12 volt architecture allows for 75% of the copper wiring to be removed from the vehicle. However, I'd argue that although the benefits of Etherloop are more abstract, it's an innovation that's more revolutionary than the switch to 48 volts, and therefore worth a deep dive.
So today, I'll walk you through my best guess at exactly what Etherloop is, why it's such a departure from CAN bus technology that's typically used in vehicles, and why I think it's the key innovation necessary to make Tesla's next generation vehicle a reality. Before we begin, a special thanks to my Patreon supporters, YouTube members and Twitter subscribers, as well as RebellionAir.com. They specialize in helping investors manage concentrated positions. RebellionAir can help with covered calls, risk management, and creating a money master plan from your financial first principles.
To kick things off, we first need to cover the primary parts of conventional automotive electrical systems. For the purposes of this video, there are three major parts. The first is of course the wiring, which is typically standard copper wire of different thicknesses depending on how much power is required, or whether it's just carrying data rather than power. The second is the endpoints that the wires ultimately feed electricity and data to and from, such as the brakes and the infotainment system. The third part of the electrical system is controllers, which are devices that manage collections of endpoints in the vehicle.
A typical vehicle has around 70 specialized controllers scattered throughout the vehicle to manage every switch, actuator, and gauge. Some of those controllers will have to manage endpoints spread across the vehicle. For example, a vehicle with traction control might coordinate traction through a central controller that's wired to sensors at each wheel. The fourth part of the electrical system is the communication network that ties together and is made up of the wires, endpoints, and controllers. For the last 40 years or so, vehicles have used a network architecture called CAN, which stands for Controller Area Network, which involves a hardware standard and a communication protocol.
On the hardware side, all of the controllers piggyback on the same wire that runs through the vehicle, called the CAN bus. And just to emphasize, every item on the network has to be connected to the same main wire that runs the length of the vehicle. For this video, it's not important to understand why that is. It's just a result of the way the CAN bus system was designed in the early 1980s. Besides the controllers being wired to the CAN bus, as I said a moment ago, they're also, of course, wired into all the endpoints they manage, which are sometimes on the opposite side of the vehicle. The result is a rat's nest of connections throughout the vehicle that uses several kilometers of wire.
With that understanding in place, let's move on to the history of Tesla's vehicle wiring, with the Model S, and then look at how that wiring system eventually evolved into the Etherloop architecture. For the Model S, released in 2012, Tesla used mostly off-the-shelf controllers, around 80%, and the remaining 20% were designed in-house. The total wire length in the vehicle was about 3 kilometers, and its total weight was about 36 kilograms. For the Model S, Tesla was just trying to get their first mass-market electric vehicle on the road, so it used a relatively conventional wiring system.
However, by 2017, when Tesla released the Model 3, 56% of the controllers were designed in-house. They achieved that by merging the controllers, so that one controller could look after more than one system or set of endpoints. A hypothetical example would be to use the same controller for both the front headlights and the windshield wipers. That would eliminate one of the controllers for either the headlights or the wipers, and would also mean one less set of wires to connect to the CAN bus. With dozens of controllers spread throughout the vehicle, that offered a target-rich environment. And compared to the Model S, Tesla was able to reduce the wire length in the Model 3, from 3 kilometers to 1.8 kilometers, which reduced the weight by 17 kilograms to 19 kilograms.
The next major improvement to the wiring system for Tesla is in the Cybertruck. We don't know exactly how long the wiring harness on the Cybertruck is and what it weighs, but Pete Bannon of Tesla said that they were able to reduce the total weight of the wiring by half as compared to the Model 3. That is, despite the Cybertruck being a much larger vehicle than the Model 3, the wiring in it probably weighs less than 10 kilograms and is less than a kilometer long. That's even more impressive considering the number of endpoints in the Cybertruck increased from 273 to 368, and more endpoints means more wires. As for what those endpoints are, I imagine a number of them are for new features like the rear wheel steering and steer by wire system. Regardless, how are they able to reduce the weight of the wiring package by half for a vehicle that's much larger and has 35% more endpoints?
Three ways. First, they continued to merge controllers in the vehicle, just like they did from the Model S to the Model 3. Second, they switched to a 48 volt wiring system for the Cybertruck. As Tesla said at Investor Day, 48 volt wiring allows for four times less current than a typical 12 volt wiring system, which means much thinner wires that use about 75% less copper. Yes, as we've seen with the carasoft teardown of the Cybertruck, the vehicle isn't using 48 volts everywhere, but where 48 volts was used, as Monroze teardown showed, it significantly reduced the weight of the wiring. Third, for the Cybertruck, Tesla started using what Monro refers to as an ethernet ring, and Tesla refers to as Etherloop, which replaces the CAN bus used in the Model 3. By using that Etherloop, not only were they able to reduce wire weight, they were also able to reduce the number of cross-car connections from 490 in the Model 3 to 155 in the Cybertruck. As you'll recall from the beginning of the video, Tesla said the Etherloop would eliminate cross-car wiring.
So although the Etherloop and the Cybertruck reduced cross-car wiring by 68%, it didn't completely eliminate it. Why is that? In short, it would have been nearly impossible to jump from a 12 volt CAN bus electrical system to a 48 volt Etherloop system in one step. So the Cybertruck was a necessary intermediary step between the wiring architecture in earlier vehicles and the wiring architecture that'll be used in the next-gen vehicle, which must use a full Etherloop network that's at least mostly 48 volts for their unboxed manufacturing process to work. But we're getting ahead of ourselves. Let's back up for a moment. Why would it be nearly impossible to jump from 12 volt CAN bus to 48 volt Etherloop in one step? Let's start with the controllers.
Most auto manufacturers have, for the most part, outsourced the hardware and software for their controllers. That's as opposed to Tesla, which, in effect, created a startup within the company to design in-house controller hardware and build the supply chain for it, as well as write the software. So, over time, as they gathered engineering resources and built their supply chain, the controllers startup within Tesla swallowed more of the controller designed for each new vehicle. The Cybertruck is the penultimate version, and it not only increased the in-house controller content to 85 percent, it also adopted a predominantly 48 volt architecture, the first mass produced vehicle by any manufacturer to do so. That meant the controllers had to be a native 48 volt design as well, which meant a full redesign of most of the controllers that went into the Cybertruck.
Why? Because 48 volts enable smaller controllers that use smaller wires, traces, effuses, and thermal management, such as heat sinks. That is, the Cybertruck is the biggest leap in wiring architecture that Tesla has deployed so far. But that's not all. The vehicle also uses Etherloop, which is an entirely new wiring architecture and communication protocol that replaces the Canbus network. So what exactly is Etherloop and how does it differ from Canbus? Bear in mind that much of what I'm about to cover here is speculation. As usual, it's speculation that emerged after I spent some time researching the topic. If you're an expert in vehicle networks, let me know what you think in the comments below.
Kicking things off, as I said earlier, with each new vehicle, Tesla merged more of the controllers. The implication of that is that the controllers went from being system-specific to increasingly general purpose. That means the controller hardware was increasingly defined by its software. This is what's known as software-defined hardware. As I pointed out earlier, one immediate benefit of more generalized controllers that looked after more endpoints was that it reduced the amount of wiring to the Canbus. However, for more distributed systems like brakes, cross-car wiring was still needed so that the controller could still send and receive communications to each brake. These are of course the cross-car wires that Tesla talked about eliminating on Investor Day. How could the cross-car wires be completely eliminated? In my view, by jumping in with both feet on generalized software-defined controllers and ditching the Canbus network standard. Let's take a closer look.
What if, rather than running dozens of meters of wire from a specialized brake controller to brakes on each of the four corners of the vehicle, there was a general purpose controller at each corner. That would mean the wire from the controller to the brake could be less than a meter long. That controller would also be responsible for all the other endpoints near the wheel, for things like tire pressure, suspension, and traction control, and maybe even the lights and cameras. But if each brake had its own controller, how would braking be coordinated for the vehicle as a whole? The general purpose controllers would be wired to each other with Ethernet, or network cable, to form the network. But in that network, whether it's on the controllers or central vehicle computer, there would be firmware and software of varying complexity serving each system in the vehicle, such as the braking system. That would allow the brakes to be controlled as a group with any other endpoints involved in braking, such as the brake pedal or the vehicle's central computer.
Beyond the braking system, this would of course mean that every endpoint in the vehicle would be available on the network. Whether that's video from the cameras, accelerometers in the suspension, or temperature sensors in the battery pack, they would all be able to communicate with each other to provide vehicle level surfaces, such as self-driving, ride control, and thermal management. In a sense, everything that was once running on dedicated hardware would be virtualized on the network and running on general purpose hardware. That hardware and software technology stack is what I think Tesla is referring to as the Etherloop. Let's look at CANBUS versus Etherloop as diagrams to provide another conceptual angle. Feel free to jump to the next time stamp if you feel you already grasped the Etherloop concept.
The first thing to note for the image on screen is that for CANBUS and Etherloop, there would be both power and data wires running between the controllers and endpoints. However, the way that's designed will vary by the system. That means the lines we see here represent where there could be data and or power wires, and I'm not being specific as to which will be needed. That's because the important part is the broader system architecture. The second note is that when Tesla implements the Etherloop network in the unboxed process, it might make sense to use controllers for each sub-assembly. That's as opposed to just the four controllers I've shown here for simplicity. With that in mind, starting with the CANBUS network, each controller is responsible for a specific system, like brakes, suspension, or lights.
To control the brakes, for example, there needs to be a direct wiring connection from the controller to each set of brakes, which can result in cross-car wiring. Any communication that needs to occur to and from the brakes, to, for example, the brake pedal, computer, other systems, or gauges can be passed back and forth along the CANBUS. With an Etherloop network, each general purpose controller is assigned a specific zone in the vehicle and would control all of the endpoints in its local area, also known as a zonal architecture. If one of those endpoints was part of a broader vehicle system, again, like brakes, the controller would communicate through the Etherloop with all the endpoints responsible for braking. The way that communication would occur is through a series of point-to-point connections between the controllers on the ethernet cables, which form a network in the vehicle.
That network would share all the data in the vehicle for all the endpoints, making every endpoint a shared resource. The end result and key takeaway here is that compared to a CANBUS network, the Etherloop network would use much less wire. That's because the controllers would be placed closer to the endpoints, and because the CANBUS would be replaced with a series of point-to-point ethernet connections, rather than using a separate wire for the data from each controller to a main CANBUS wire. So with all this potential to reduce wiring, why did no one implement an Etherloop before? First, as I said earlier, auto manufacturers have outsourced the design, manufacturing, and software of the electrical systems in their vehicles to third-party suppliers. And to make matters worse, one component will often contain parts or software from more than one supplier. So in order to make even a minor change to one part of the electrical system, it takes months to get all the parties together to start the redesign process.
Now multiply that by up to 70 controllers throughout the vehicle, and not just one change to one controller, but re-architecting all the controllers in the wiring system from the ground up. The second reason no one did Etherloop before is because on the software end, moving to an Etherloop network isn't as simple as just using the typical communication protocols used in a home network or on the internet. Let's take a closer look. The reason CANBUS has been so successful is that it's highly reliable for critical functions like brakes and has low latency. That's because with a CANBUS network, messages are prioritized in a way that, if two messages are sent at the same time, the message with the higher prioritization arrives first at its destination and is executed.
That's as opposed to a typical ethernet protocol, where if two messages are sent at the same time, they collide and cancel each other out. When that happens, each of the sender's weights a random amount of time, and then resends the message. That creates greater latency. Latency is the time it takes for a network to transmit a message. Long latency times aren't acceptable in vehicles, because a fraction of a second delay could mean life or death. To solve the latency problem, like they do with most things software-related, Tesla's likely developed their own in-house ethernet communication protocol that gives priority to critical systems or messages. With all those difficulties in mind, it should explain why, said earlier in the video, that it would have been nearly impossible to jump directly from a 12-volt CANBUS system to a 48-volt Etherloop system, and why the Cybertruck was a necessary intermediary step between earlier vehicles and the next generation vehicle.
It was just too much to do in one step, and we can see evidence of that in the low voltage system, where some of the wiring is still 12 volts rather than 48 volts, and in the cross-car wiring, which should have been completely eliminated if Tesla had fully moved to generalized controllers on Etherloop. It should also explain why Etherloop is just as critical, if not more critical than moving to 48 volts. For the next generation vehicle, if they continue to use 12 volts for some things, that's fine because it would have little effect on how the vehicle is manufactured. However, moving to a full Etherloop system for the unboxed process for the next generation vehicle will be a necessity. That's because the vehicle won't move down the production line as a single cage or box. It'll be built from several pieces that are merged at the end of the line. That means the vehicle can't use a typical CAN bus wiring system that has to be installed in one piece and touches every part of the vehicle.
It'll have to use a point-to-point Etherloop network where you can connect any controller to any other controller to gain access to the network, which in turn will allow the vehicle to snap together like Legos. Lastly, it's worth covering a few other benefits of Etherloop. First, the typical data rate for a CAN bus system is 1 to 10 megabits per second. That's as opposed to Tesla's Etherloop system, which is capable of at least 1 gigabit per second, which is 100 to 1000 times higher. That means Tesla can pipe more than just basic functions like the brakes through the Etherloop. They can put high data rate services through it like audio and video. Put another way, if you had the right connector, you might be able to plug a monitor into any of the controllers on the vehicle, access the entertainment system, and watch movies. Tesla isn't likely to allow that, but it gives you an idea of what's possible. In the Cybertruck, they're already piping the audio system through the Etherloop. In the past, the audio system would have had its own separate wiring system that piggybacked onto the CAN bus system.
Now, the audio system is fully integrated into the broader vehicle network where it has access to shared power and data. A second benefit of Etherloop is that as the controllers become increasingly general purpose, Tesla may get to a point where they only have a few standardized controller designs or even a universal controller design for their vehicles. That would simplify their supply chain, vehicle design, and production, and create economies of scale that would drive down vehicle production costs. In summary, at Investor Day, Tesla said that for the electrical system in the Cybertruck, they would be moving the controllers closer to the endpoints and using ethernet to connect the controllers, which would in turn reduce cross-car wiring. I was excited about that possibility at the time and what it meant for Tesla's upcoming vehicles, but I didn't realize what a fundamental improvement it was until I started looking into it.
If my speculation is correct, Tesla is completely redesigning the electrical architecture of the vehicle and bringing it into the 21st century. The controllers will become a high-throughput network of communication hubs that allow access to any endpoint on the vehicle by any other endpoint and make the vehicle maximally software defined rather than limited by its hardware. That may allow Tesla to finally achieve less than 100 meters of wire in the next generation vehicle, which is something they were aspiring to do with the Model Y over four years ago, once again making the impossible merely late. Tesla's accomplished a lot in the past to dramatically simplify vehicle manufacturing, with innovations like gigacastings and the structural battery pack. But in my view, the Etherloop may end up being one of the most impressive and could take the longest for any other auto company to duplicate.
That's because it's something that Tesla's been working on for about 15 years, component by component, since they designed the Model S. Other automakers that haven't developed in-house controller design expertise are probably seven to ten years behind Tesla. Some CEOs like Jim Farley appear to have recognized this and are making big moves to insource their electronics, so I'm eager to see how fast they can implement Etherloop. Lastly and importantly, Etherloop is the final and possibly the most daunting piece of making the unboxed process work for the next generation vehicle. Without it, the vehicle can't be manufactured as separate subassemblies, because using a prior art canbus wiring harness would require everything to be on the same canbus wire. With Etherloop, those subassemblies can snap together and come to life as an integrated network. More on that in the upcoming unboxed video. If you enjoyed this video, please consider supporting the channel by using the links in the description. Also consider following me on X. I often use X as a test bed for sharing ideas, and X subscribers like my Patreon supporters generally get access to my videos a week early. On that note, a special thanks to Asriel the Cat, my YouTube members, X subscribers, and all the other patrons listed in the credits. I appreciate all of your support, and thanks for tuning in.
这主要是因为自从特斯拉设计Model S以来,他们已经在逐一改进这一技术,已经有大约15年的时间了。而其他没有开发自有控制器设计专业技术的汽车制造商,可能比特斯拉落后七到十年。一些CEO,例如Jim Farley,似乎已经认识到了这一点,并且正在大动作地内部开发他们的电子设备。因此,我非常期待看到他们如何快速实施Etherloop。最后也是最重要的,Etherloop是使下一代汽车的无盒制造过程工作中的最后、也是最具挑战性的一环。没有它,汽车就无法作为独立的子组件进行制造,因为使用传统的canbus线束会要求所有东西在同一条canbus线上。而有了Etherloop,这些子组件可以组合在一起,并作为一个集成网络运作。更多内容将在即将发布的无盒视频中介绍。如果你喜欢这个视频,请考虑通过描述中的链接来支持这个频道。另外还可以考虑在X平台上关注我。我经常使用X作为分享想法的试验场地,X的订阅者和我的Patreon支持者通常能提前一周看到我的视频。特别感谢Asriel the Cat,我的YouTube会员,X订阅者和所有在片尾致谢名单中的赞助者。我非常感谢你们所有的支持,谢谢收看。