The Car Computer – A radical proposal to solve the automotive e-waste conundrum

[draft version – still missing pictures, videos and a more engaging structure]

1. Setting the stage

This article is meant as a conversation starter: I know enough about technology to have ideas for how to improve it, but not enough about car manufacturing to know if the idea presented here is applicable the way I see it. Let’s figure that out together. Also, since a lot of current automotive tech is a closely guarded trade secret (allowing a company to charge 60,000 USD for their Tesla Model 3 reverse engineering results), I may be forgiven for quoting from obscure places.

Now, let’s state the problem of automotive e-waste in simple terms. In established e-waste recycling processes for cellphones, computers and home appliances, the first step is removing the electronic boards (PCBs), then shredding the rest and recovering metals and plastics with automated processes. PCBs are processed separately, which allows to recover the precious metals contained in them:

PCBs and elements such as plugs with gold-plated contacts are normally sent to copper foundries that specialize in the recovery of precious and special metals. […] While oftentimes the amount of such [precious or special] metals in each individual device is minute, the overall substance stream adds up to a considerable yield. For example, a ton of cell phones contains around 250 grams of gold, whereas a ton of gold ore contains only around five grams of gold. (source)

Now the situation is different for automobiles. Here, the PCBs are not extracted, but the whole car is shredded – have a look. Automated processes extract metals, plastics, glass and the like (see, see), but the best they can do in terms of recovering precious metals is separating the copper pieces from shredded up copper wires (see). The residue from this sorting process is called automotive shredder residue, and is simply landfilled. That residue contains the precious metals from PCBs, but now so diluted that in a study, they were not able to find them there, while on the other hand they found lots of gold and silver in PCB waste. economically unrecoverable. Our precious substances are now so diluted that they are economically unrecoverable. And all that just because they don’t extract the PCBs before shredding the car.

Now why would they not take out the PCBs and other electronic elements with valuable resources first? Well, because it’s not worth the cost of labour to extract them. First of all, the PCBs in cars don’t have a lot of valuables in them:

The price for printed circuit boards (PCBs), where most of the critical raw materialsand precious metals in car components are located varies widely depending on the PCBquality. The lower end of the range—2.80€/kg—represents PCBs with less metal contentwhereas the upper end of the range—ca. 26.00€/kg—is valid for, e.g., PCBs from mobilephones and notebook computers. The PCB components of ELVs [end-of-life vehicles] can be considered similarto low quality PCBs rather than the high-quality PCBs found in mobile phones.” (“Economic Viability of Extracting High Value Metals from Endof Life Vehicles”)

Now let’s assume a heavy PCB (by car standards) of 150 g and total labour costs of 32 EUR/h as in Germany. Such a worker must extract more than 11.4 kg PCBs per hour to be profitable:

image

So extracting a PCB of 150 g (very hefty for current cars but found in the engine control unit) must take 47 seconds or less to be profitable:

image

And that’s obviously hardly doable with current cars. It is even worse for the many other electronic control units (ECUs), which is a fancy name for the many small computer that a car has. A current top-of-line car with an internal combustion engine may have up to 70 of these ECUs, distributed throughout the car. This system architecture became so complex that a whole standard was developed to make all these little computers talk to each other, called the AUTOSAR Classic Platform.

2. A different architecture for car electronics

One of the first thoughts I head when looking at this problem was: automotive biomimetics. Let cars have a central nervous system rather than having their processing capabilities distributed throughout the whole structure. (The latter could be called biomimetics if you have the octopus in mind of course, where each arm can do its own thinking. But a central nervous system is a much more common thing in animals.) It seemed so much more logical and economical to me to concentrate and share processing resources, given that all these ECUs only need little processing power each.

Now to my amazement, I found centralizing the electronics in a car is already becoming the new standard paradigm. In February 2020, the Japanese magazine Nikkei published a widely cited article “Tesla teardown finds electronics 6 years ahead of Toyota and VW”. In it, they tore apart a Tesla Model 3 and found just a few separate electronic boards, with most of the electronics concentrated into one silver box:

Industry insiders expect such technology to take hold around 2025 at the earliest. […] There should be nothing stopping Toyota or VW from doing the same much earlier than 2025 […]. But technological hurdles are not the reason for the delay, according to the Japanese engineer who said “we cannot do it.” The real reason for holding off? […] Such systems will drastically cut the number of electronic control units, or ECUs, in cars. For suppliers that depend on these components, and their employees, this is a matter of life and death. So big automakers apparently feel obliged to continue using complicated webs of dozens of ECUs, while we only found a few in the Model 3.

We see that there are only legacy reasons and no good technical reasons for why a typical car today looks more like an Internet of Things than like a single computer with periphery for input and output.

Now Tesla did not introduce this innovative architecture with recycling in mind – though they arguably designed this system for easy over-the-air software upgrades (a recommendation derived from analysing the Tesla Model 3) and for after-sales electronics upgrades (such as the Model 3 upgrade to “Hardware 3.0” to make it full-self-driving).

But the same principle (that centralizing a resource makes it maintainable) also makes the car electronics recyclable as well. It allows to take actions on the electronics as a whole, and recycling is just another one of these. If you want to have a look: here’s a video of exchanging the MCU board in a Tesla Model 3. The MCU (multimedia control unit) board is one of the two main computers in a Tesla, the other being the ECU (engine control unit). Both are contained in the same silver box that you see in the video, called the “MCU & Autopilot ECU”.

Getting that box out might take 4-5 minutes if you don’t have to care to not damage the car because it’s end-of-life. With that, you’d hold maybe 80% (my guess) of a Tesla’s electronic resources in your hand. In the Model 3, Tesla went to quite some length to make the remaining peripheral electronic components as “dumb” as possible; for example, their triple front-camera module comes with no own microprocessor as is typical for competing products (see), so that all the processing power remains concentrated in their single silver box.

From a resource perspective the Model 3 design means that the peripheral PCBs are so small and light that we would not want to care for them as they contain only trace amounts of valuable raw materials. Note that we do not have to count its purely electrical components here (that is, those with no semiconductors) such as cables, small electrical motors and other actuators, as their only precious metal is copper and that is already reliably recovered with existing separation processes after shredding the whole car.

3. Further improvements

Taking out 80% of a car’s electronics in 5 minutes is quite good already, but can we do better? Certainly:

  • Quick remove box: Tesla’s main computer box could be mounted to the car with just a single bolt, reachable from within the glove compartment to provide theft protection. This simple change would allow to take out 80% of a car’s electronics in 1-2 minutes, especially when legislation would mandate a standard position and fastening bolt to use for these car computers.
  • When loosening a few more bolts, the box itself could be separated into plastic, metal and the actual PCBs. This de-manufacturing is the first step when recycling electronic devices (see) and the less time this manual process consumes, the better. The Tesla Model 3 computing box seems to be quite good from this perspective, as it just has two large PCBs and a metal frame and cover.
  • Of course, not generating e-waste in the first place would be even better. That is acknowledged, for example, by Germany’s federal office for the environment: “According to Germany’s statutory waste hierarchy, consumption avoidance (and thereby waste prevention) is the top priority, followed by recycling and disposal. But this hierarchy is for the most part disregarded in practice.” (source). Several ideas come to mind, including cars with 2-5 times the typical lifetime and car computer units that are so standardized that they can be re-used easily as spare parts in similar car models of the same manufacturer and even in cars of other manufacturers. Even better, these computers could establish a similar standard for mobile machinery as the IBM PC did for office equipment and home computing – any computer that complies with this standard could be plugged into any kind of computer-operated machinery, from cars to trucks to farming and construction machinery. All that however is only a realistic standardization project once the development of car electronics converges and slows down – which is certainly not now, at the beginning of the self-driving revolution, but 20 years from now.
  • There are some remaining ECUs such as for Xenon lighting that, it seems, were not integrated into the central computing unit even in the Tesla Model 3 because they are so standard and off-the-shelf that it was not economical to invest the own development effort to integrate them into the car’s central nervous system. But it’s certainly possible and technologically meaningful, and car makers would also do so if mandated by law, the same way they follow exhaust emission regulations. (Not counting the cheaters, of course. Looking at you, VW.)
  • Some ECUs cannot be integrated into the central car computer because they are needed in specific locations. This applies to the cameras and radar unit in the example of the Tesla Model 3. These units could be made so easy to extract without damaging them that it’s possible to take each out in 15 seconds, making economic recovery possible, especially when they can be used as spare parts for cars of the same or (ideally) also many other models.

4. Let’s not make recycling policy for legacy cars

The European Commission announced in June 2021 as part of their “Fit for 55” strategy that CO2 emissions of newly admitted vehicles must go to zero by 2035, which basically means the end of the internal combustion engine. It can be expected that the EU will be a hostile environment for ICE cars even much before that date, so that many or most of them will be exported to third countries as second hand vehicles rather than being eventually scrapped in the EU. So starting from around 2035, there will probably not be large numbers of ICE cars anymore that are being scrapped inside the EU. And the cars scrapped until then are either in use already or will be produced until 2025.

In other words, any automotive recycling policy targeting cars with internal combustion engine will only affect car manufacturers in the next five years. It makes little sense to create policy for legacy cars.

Instead, recycling policy should focus on the (very different) electronics and architecture of electric cars. It can be expected that Tesla’s centralized car computer architecture will become the standard due to its implications for cost and the necessity for it based on the latest hype – autonomous driving requirements.

In many ways, the transition to electric cars makes automotive electronics recycling easier: electric cars do not have many of the components where precious metal recovery is borderline feasible right now, for example the oxygen sensor that contains platinum and palladium (source). They do also not need the massive amount of electronic sensors and actuators that are needed for controlling internal combustion engines (see). And several key components of electrical vehicles are so valuable that their recovery is always economical and will be done for that simple reason. This includes the drive motors (containing massive amounts of copper and often neodymium for the magnets) and the battery (often used for a decade in grid balancing batteries after they are end-of-life for automotive use).

Policy should focus (at first) on the few remaining key issues with electronics recycling in electric cars. My proposal is specifically to obligate car manufacturers to do three simple things: (1) integrate the electronics into a central car computer so that 90% of all valuable raw materials in the electronics reside there and (2) to place the car computer into the same standardized location and fasten it with the same standardized type of bolt to facilitate disassembly and finally (3) similar standardization and marking would be applied to other significant electronics components that are necessarily peripheral, such as cameras and radar, making it easy and fast to recover these before a car is shredded.

These three steps should solve 95-98% of the car electronics recycling issue, and the central change involved here – the switch to a central car computer – is even in the economical self-interest of car manufacturers. Enforcing this by policy just means speeding up a change that seems bound to happen anyway, and that speedup will be welcomed by manufacturers.

Three simple steps that make car electronics a nearly circular system sounds great, and a bit too good to be true. So it’s your turn now – you’re welcome to try and shoot holes into this, and to provide your own alternative and better ideas. Let’s discuss!

4 Likes

This is great, Matthias.

One stumbling block is Conway’s Law – that the structure of a product replicates the structure of the company that makes it.

So “centralize all the electronics” means “persuade the multiple teams working on electronical components to collaborate within a different hierarchy

Fortunately for us, it’s totally fine for the components to be logically separate, as long as they are physically co-located. i.e. you can still have 6 separate teams working on 6 separate components, just so long as they get jammed into a single box. Whether or not they eventually join forces on one computer is a matter for the manufacturer.

1 Like

Centralization is also tricky when different systems have different requirements for reliability. You don’t want the brakes affected by a bug in the satnav.

I also wonder what constitutes electronics.
e.g. sensors basically need to be located next to whatever they are sensing. wiring is probably a lot of ‘electronics’ by volume, but can just be extracted as raw metals.

Thanks :slight_smile: Hadn’t heard about Conway’s law before. I like the concept. In the case of cars, it’s imaging not just the structure of the company, but the structure of their extensive network of suppliers. 70 electronic control units in a modern car means 70 components obtained from different suppliers, each coming with its own little PCB.

You’re completely right – for the purpose of simple recycling, tens of little boards located in one single box would also do. Working on a single common computer is just another way of logical separation, the result of moving functionality from hardware into software. Like, “apps” running on that common computer, adhering to a well-defined software interface. In that case, a supplier that now provides a sensor plus electronics board combined into one hardware unit (“ECU”) would then supply the sensor plus a piece of software running on that car computer.

Absolutely not. And it seems that a good system architecture for that is already around. In the Tesla Model S, they have two separate computers (user interface and actual car control) only connected by an API and another piece of hardware called the gateway. There’s this segment from a DEF CON 23 presentation where they talk about this highly. (I highly recommend that car hacking presentation. Very neat demonstration of how one would go about probing car infosec.)

So in actuality there would be two computers that can run suppliers’ contributed “apps”. And just like you observed for other separate pieces of electronics, there is nothing against co-locating these two computers into a single box. That’s not done in the Tesla Model S, but it is done in the Model 3, where a single silver box behind the glove compartment contains both of these boards.

I also wondered about that. In the recycling context, I settled on this: if it can be extracted automatically during the crushing and sorting process, there’s no need to extract it first; that applies to cabling, solenoids, small electrical motors and other pieces of copper larger than 3 mm. If however it has raw materials that are not captured by the automated crushing and sorting, then that’s electronics to care about. This refers mostly to PCBs. Camera PCBs are a good example of electronics that are necessarily peripheral.

2 Likes