The Car, Reinvented
mondaynote.com
mondaynote.com
I hate to be critical, but this really reminds me of a 1950s "visions of the future" piece more than a practical design choice.
For example, the skateboard is not a new concept [1], and there are reasons you don't see it. Modular design really only shines for low-volume production (or a large combinatorial space). When you are doing 10s of millions of units per year, as the auto industry does, modularity often gets thrown out in deference to volume/cost.
Also, distributed manufacturing is a huge pain, why would you ever want that? See Boeing's experience with the 737 supply chain. Just because you can doesn't mean you should.
It is a difference of just how independent the "skateboard" is from the top?
A true skateboard would be a static implementation of the frame and propulsion system, with defined interfaces for the "top hat" (industry term for the body of the vehicle).
Most modern cars use the monocoque/unibody type of body, where the body and frame ("skateboard") are combined into one, because of weight and safety (it can be made to bend in a more controlled manner, usually with different type of materials bonded/welded together). The frame+body is better because it's easier to repair (many unibody crashes are not economical to repair because a proper repair to the same strength as the initial car would need dismantling half the car).
Then we got vertically integrated mass production which cost less and built cars which handled a lot better as they'd been designed as a unit with bodywork and chassis having complementary characteristics. "I wish my sports car also had the ability to handle like a family car or a commercial vehicle if you slapped the right bodywork on," said no modern driver ever.
Calling it a skateboard (bah! so twentieth century) doesn't change the fact it's the nineteenth century approach to doing things....
The thing is companies don't really like using modular chassis like ladders anymore is because anything that can't use the roof for structural support is heavy and flexes like mad.
I don't understand why the author keeps banging on about CAN bus being an old standard. So what?
CAN bus is used in the automotive industry because of it reliability and low wire count. Why should it be changed?
Discussing the differences between a conservative and a liberal he used an analogy of a fence, where a liberal would want to take it down and a conservative would want to leave it up.
It got me thinking if this was true or not.
A little digging and I found this page:
https://en.wikipedia.org/wiki/Wikipedia:Chesterton%27s_fence
"Although the purpose of this fence is not obvious, there may be valid reasons for its presence."
(I'd like to think I'm in the mindful middle somewhere)
Imagine something more complicated where you need specialized knowledge to troubleshoot things.
It’s like they want to complicate things so the average person can’t fix things themselves. They look at John Deere not in askance but in admiration for making things irreparable by ordinary folks.
Navigation and playing music? Both are done by other systems as CAN already.
I mean, HTTP is quite old.
and CAN is a solid piece of tech. Ultra reliable, fast and low wire count. It's controllers are cheap and reliable.
I think the author is pointing at that concept, with the idea that CAN hasn't evolved over time in the same way as HTTP. Ideally there'd be a CAN2.0 for the cars that want it, that's backwards-compatible with CAN1.0 readers/etc.
If he works in this space, I would think he'd know that there are new things being worked on:
* https://en.wikipedia.org/wiki/CAN_FD
* https://en.wikipedia.org/wiki/FlexRay
Not that these seem to work differently at the OSI Layer 1/2 layer, so while older messages can be sent over the newer transport, it is not wire compatible.
As long as the new version of the protocol is still formatted on the higher OSI layers in terms of reusing the core messages from the old version, and then adding new messages that are optional to the understanding of what's going on from a CAN1.0 perspective, then a change in the PHY layer still isn't really throwing the baby out with the bath-water.
The cynic in me believes that Apple is so bought-in to its own hype that they have a need to change it for change's sake alone. Just like their pathological desire for thinner laptops with no practical benefit, they'll rail on about how this unnecessary churn is progress.
I think the Apple Car will be a buggy, unreliable, unrepairable mess that the buyers won't ever truly own.
Traditionalists like to point to its latency guarantees, but you can do just as well with a thoughtful Ethernet implementation. And there are single pair PHYs. Others argue in favor of the multidrop topology, but I see it as a common failure point. Plus it erodes the utility of the bandwidth.
For multiple drops I like CAN a lot better.
I don't know this space, but depending it may be there are more messages going to and fro, and so perhaps more bandwidth is desired? IEEE 1394 ("FireWire") seems to have a bunch of 'industrial' uses:
* https://en.wikipedia.org/wiki/IEEE_1394#Common_applications
The marine folks went from a CAN-like NMEA 0183, to a more bus-like NMEA 2000, and are not looking at an Ethernet-based system:
* https://en.wikipedia.org/wiki/NMEA_2000
* https://en.wikipedia.org/wiki/NMEA_OneNet
If all you're doing is getting sensor values at a few Hertz, then CANbus is probably fine, but if you need to distribute the images of a read-view camera, a bunch sonars, a bunch of radars, etc, then the 500 kbit/s in J1939/14 may not be enough.
Even the 'CAN folks' recognize this:
Every time I see someone say "all cars will be rentals in the future" I think if all the people I know who have stuff kicking around in the trunk that they use occasionally but rarely. The automotive version of "every-day carry". Some of it's basic car maintenance stuff. Some of it's not.
And then I also think about the people I know whose passenger or back seat is a giant pile of trash because there's no easy-to-empty integrated place for trash and it all just ends up on the floor...
Also to quote this incredibly relevant point from fartaspoobutt [dead]: [1]
The car I really want is a network of fast, affordable, widely available public transit.
One could say the same thing about clothes.
My car sits idle for 99.9% of the time (I make less than 2000 km per year, and vast majority of that in a few long trips each year). It would probably be cheaper for me to rent a car for vacation trip and go by taxi the other 10 times a year I need it for city driving. But we already have a car and it's been in the family since 1997 and it doesn't break so we keep it.
I would love to have a way to pay a reasonable yearly fee and have a good car available when I need it without paying for the parking space, insurance, maintanance, etc. But as it is renting a car is too much hassle.
Yes, conceptually the "skateboard" makes sense, but you're not going to have actual snap-in components like that. By Tesla's own admission, thinking of the Model X as just another body on top of the Model S skateboard set them back a long time. Whether you choose to go the traditional unibody design, or a more unconventional exoskeleton design like the cybertruck, each body is specifically made for each model type. While the Model Y is a "jacked up" model 3, the body is a completely new and different thing.
I had a truck with a hard to reproduce software bug where the electric steering assist would just turn off. Imagine going down a canyon road towing 5000 lbs and the steering suddenly gets extremely hard and the wheel rips from your hands. I had it happen. Luckily there is still a mechanical shaft going to the steering rack.
Indeed. I have zero confidence that the company who thought Catalina was ready to ship can deliver a safe drive by wire system. That's not the kind of system where you can just patch the bugs out later.
This is not only about engineering and sales. Current car culture is rooted in post-WW2 values: masculinity, personal freedom, power and control. Even Tesla plays that book with the SpaceX halo and the insistance on performance.
I wonder how Apple will approach this change of “soul” in car culture. It’s a shift not unlike what they brought to computing in the 80s, though they transformed the design then. Can Apple build a personal, shareable, fully automated electric pod? I can’t imagine them selling a vehicle with a wheel for instance.
All in all, doubtful we ever see the outcome of the Titan project.
Ever see a crazy classic car body dropped on top of a truck frame, i.e. a 4x4 60s mustang on a bronco frame?
> On top of that, the upper body will take the shape and size depending on the need: a sports or family car equivalent, a commercial vehicle, etc.
body-on-frame is not a great design for a family vehicle, as it tends to decrease the interior space and increase the weight for a vehicle of a certain overall size. it's particularly unsuited for a sports car due to the weight penalty and the higher center of gravity. AFAIK, it only really makes sense for large vehicles and/or vehicles designed to have a large towing capacity.
I have to wonder if they meant something more like the volkswagen mqb platform [0] but expressed it poorly.
[0] https://en.wikipedia.org/wiki/Volkswagen_Group_MQB_platform
But can be refilled in 5 minutes and driven for 300 miles, which I don't see happening with electric cars any time soon.
Also Toyota hybrids are not very responsive and fast. And they smell.
> And they smell.
Erm?
If you bike or drive behind a Toyota, you can notice the smoke and the smell. Especially during winter.
he compared cost, performance, regulations, but I didn't see him tackle the issue of range and recharge times not even in passing.
> It is the only way forward.
with claims like these, skipping the two major drawbacks seems quite intellectually dishonest.
Electrical is still "the only way forward" for automakers—politically, due to emissions regulations—whether consumers are willing to accept those drawbacks or not (i.e. whether it's profitable to build EV cars or not.)
Auto-makers just have to bite the bullet and accept lower profits as people refuse to compromise for a while, until those same consumers later realize that governments won't be letting them have things any other way.
Conveniently for Apple, their strategy has always been to leave enough of a profit margin for themselves (and market to those who can afford to pay it), that the business will pay back its capital outlay even from only niche uptake. Apple isn't Tesla, with a plan to eventually sell an EV car into every garage; they don't need general market adoption, just adoption by conspicuous consumers.
[1] note that it's misguided to claim you have to reach zero emission to compete with batteries, because batteries chemicals pollutes both during extraction, transformation and disposal.
"Let's completely reimagine from scratch. But we all know there's really only one option so let's ignore any other ideas."
If you write your own DIY electric power steering firmware, and it crashes or gets stuck it can just suddenly fail on the road and you're dead.
I think this is fair, but incomplete. (Disclaimer, I work for General Motors, these are my opinions alone).
It is hard to make things work. It is harder to make things work consistently. It is even harder to make things that will work consistently when you don't have control of the thing and users can use and abuse it in unforeseen ways.
Everyone who has spent more than a little time in a car has something they can complain about, regardless of the brand. Now think about how many parts are implicated in that defect. Now think about how many parts are in a car. A very very small number of defective or under-designed or mis-designed parts can lead to a bad experience.
Think about testing. Many people on HN are programmers or otherwise involved in software development. Think about writing tests for your code. How many end-to-end integration tests do you have? How many unit tests? What happens when you change platforms or update middleware? Tests in the real world are EXPENSIVE! When you test a new car, you're not testing 1 cheap part from a run of 1,000,000 parts; you're testing 1 expensive prototype part from a run of 10 parts.
The article touches on the dealership experience. For legacy automakers, there are laws in place that prevent them from ditching their dealers.
Legacy automakers also have agreements with labor unions. Adopting new work processes takes patience and understanding in the best of times. Re-arranging work can quickly become confrontational and not collaborative.
The other major partner for OEMs are suppliers. Suppliers do much of the design work, based on specifications from OEMs. Suppliers can sell the same or substantially similar parts, made from the same sub-components to many different OEMs; but the further down the sub-component tree you go, the smaller the profits. Also, if you want to do something crazy like ditch CANbus or upgrade your 12 volt architecture to 48 volts, you'll have to do a lot of work with your supplier base.
Cars aren't judged on the things that work well, they're judged by the things that don't. The car design and manufacturing process is all about limiting and eliminating those things. Uncontrolled changes can cause a huge spill of defects.
----
Edit: My main point here is that, Yes, the industry is ossified and slow moving, but you can't just speed up. You have to address the reasons why it is slow moving, and you have to do it WITHOUT RUNNING OUT OF MONEY!
If it was a racing series then you could supply the tyres and insist no parts came from existing cars with everything scratch built. That would make for an interesting race series.
But cars are about how they are made, so final assembly and the supply chain. Having the capital to reinvent that is way too capital intensive for where we are.
Might as well build a web browser in a computer where you have to build your own CPU working back from the HTML specs.
Tesla doesn't follow this approach. Their software is pretty unified and they can update everything including their controllers OTA.
According to quick googling cars got heavier: "The average new car weighed 3,221 pounds [1461 kg] in 1987 but 4,009 pounds [1818kg] in 2010." [1]. The weight of a trabant, which was probably a very badly engineered car, was about 600 kg [2] - 1/3 of a modern car.
I imagine we could make cars today even lighter and if weight of all cars would be similar than the safety standards would be much different.
[1] https://slate.com/business/2011/06/american-cars-are-getting... [2] https://en.wikipedia.org/wiki/Trabant
Also I am not a mechanical engineer but it looks like steel is really good at absorbing energy from impacts. Plastics tend to shatter when the impact force is too great. I actually experienced a minor accident in a fiberglass car. No one was hurt because the chassis held everything in place but the body was completely broken.
High end cars (and most/all race cars) use carbon fibre for both bodywork and structural components.
Having a modular car with a high degree of customization would be a huge hassle in keeping cars safe.
the real issue would be making the thing actually reliable. matings and fittings off the production line aren't the same you can get post assembly on a workshop, and that will affect component durability big time
This fantasy highlights some of the absurdity of modularity. A car with only 1/4 the batteries and no generator will handle completely differently. It would really be a different car. But "reconfigured" in less than an hour by a complete luddite.
And, "nuke"? Really?
Why not imagine the owner could just snap in and out between 2wd and 4wd drive trains? Choose to insert one electric motor or two (or four?) depending on the day?
What could possibly go wrong?
Still, it's fun to fantasize. And maybe some of it could make sense. Like unplugging the driver controls and just moving them over to the left or right seat and plugging them back in. Wait... what?
Let the batteries get cheaper, let there be more public transportation (classic variety and municipal bike sharing including cargo wagons, etc.), and hopefully the industry implodes on itself as described above.
And screw all this wall-e-esque infotainment stuff.