Overall system complexity might be higher (due to more electronics), but this type of complexity tends to have a more favorable profile for economies of scale. One, generally, writes software once, but has to route hydraulics, reservoirs and so on for every car produced.

Separately, hydraulic systems are an inspection item with a serviceable fluid. Electronic systems work without servicing until they don't, at which point a lamp on your dashboard lights up. More convenient for the manufacturer/dealer, less convenient for you, but that's typical for modern anti-consumer (pro-producer/corporation) design.

Then servicing a hydraulic system is more manual labor intensive (replacing hoses, bleeding the system, replacing the fluid), while electronic systems come with components that are replaced wholesale, where the consumer pays the price for the new component. I guess it's obvious which ones the manufacturer prefers.

Driving experience suffers, but not in a way that measurably affects consumer behavior. Most people tend to care about other characteristics more than the particular issue of latency. As another example, see the issue of rev hang - a type of latency on its own.

My main concern is the same as with electronic gear levers/selectors, electric parking brakes, steer by wire, and throttle by wire:

What happens when the car has a failure that kills all the electrics?

That's a much more common failure scenario than people think. Batteries go flat. Wiring gets chewed and shorts with vibration. CAN buses need only a brief short to go out entirely and not reset. Even in a modest crash the car can lose its battery and alternator, or get a system short.

Tesla owners have learned this lesson repeatedly with getting locked out, locked in, stuck in park, unable to shift to neutral for a tow, etc. Other car owners have learned this with electric parking brakes getting stuck on.

If the electrics fail and the car is moving, I need steering, brakes, and the gear lever to all work to stop safely. If they're by-wire, I'm just a passenger.

Same thing that happens if a brake line pops, steering wheel locks up, or the throttle gets jammed open. There are ways of designing redundancies into those systems. Multiple communication channels, reserve sources of power. A lot of new model vehicles already have this stuff, drive by wire for example has pretty much been the standard for twenty years now. Typically, the entire system doesn't go down because they're designed for that not to happen.

But I'll take my mechanical overrides.

My one-off experience. Parked my car once for a long holiday (two months). Mice chew through the main wiring loom behind the dashboard and urinated over electronics (urine is slightly corrosive). 1) no mechanic has time to repair single wires 2) dashboard removal is more than day of labor 3) main can bus failures are hard to find and require pulling modules till canbus high/low get the right voltages I’d say a hard NO towards flybywire brakes.

Don't forget the possibility of adding digital locks so it can only be serviced with authorized parts by authorized dealers "for your security and safety".

Also, these systems are much more difficult to diagnose if they fail to function correctly. This means you are going to pay for 5 different things to be replaced, each after replacing the previous one didn't fix the issue. And you are going to pay big time for every replacement.

Not really. The issue with modern automobiles they don't really report much useful information about themselves beyond general upstream faults. One issue can flag errors in five modules, and it depends on the tech to know about the underlying technology in addition to diagnosing whatever mechanical issue. Techs also tend to be the type of people to throw up their hands and start throwing parts.

This is kinda where we need to shift more towards EVs because that would take out a lot of the complexity you get with modern automobile design.

> and it depends on the tech to know about the underlying technology in addition to diagnosing whatever mechanical issue. Techs also tend to be the type of people to throw up their hands and start throwing parts.

What do you think how well documented car internals are? What would you do if you open a control device and you see an unmarked ASIC?

It depends on the manufacturer. Some are better than others. Some document pretty well but are overshadowed by crap manufacturing.

Personally I really like how Tesla did their online service manual. It's a good balance of technical but not to technical.

https://service.tesla.com/docs/Model3/ServiceManual/en-us/in...

With regard to your edit. That's where there's a practical limit of what do you (a manufacturer) publish that the tech needs to know. So you need to cross it between technical know how, and how detailed you need to get. Me personally would pull out the oscilloscope and make some probes to get an idea of what to do. You set a carburetor on my desk and that would probably take me a bit longer over electronics. It's a skillset issue, basically, and we're transitioning off strictly mechanics to mechanics and electronics working together. Which naturally raises labor costs because people aren't use to that.

Not sure if my point came across. This is a knowledge problem. How can you test an ASIC for correct function if you have no knowledge of its intended function?

Right. That's what I added. You have to reverse engineer what's going on. As with anything if you don't know how it works. If your IC is producing crazy readings, or you look under a thermal camera, probably can blame the chip. Good luck getting a replacement.

Some of these sound like upsides to me. I've owned a lot of older cars, and they're always a mystery box of "what will fail next?". So many moving parts, and they're often hard to diagnose.

Based on your comment, I expect newer cars to have lower maintenance and repair costs, fewer part failures, and less guessing about exactly which part is broken.

There is the crux:

> while electronic systems come with components that are replaced wholesale, where the consumer pays the price for the new component

Repair costs are skyrocketing with newer cars.

Off the cuff, I'd say at least half of my repair costs are the labor. Any chance you've seen figures on how much more expensive the parts are for newer cars?

It has to do with modularity and commoditization. Modern parts can be highly integrated and digitally customized to specific applications, in such a way that they become increasingly scarce and eventually impossible to find, whereas older cars are mostly made from commodity components where the same part is interchangeably used in tens or hundreds of vehicle models, and many become cheaper and easier to obtain as time goes on and tooling costs are amortized over decades.

Its not just the replacement parts cost. Modern cars are more tightly packed and not designed for individual parts being replaced afterwards. Sometimes its just things like a screw holding something behind the center console being accessible from the engine compartment in a position that forces a full engine/transmission removal. These are tiny things but can turn a 30 minute job into two person days.

Or a oil pipework that cannot be removed from the engine because there is a bracket in the way, which can only be removed by taking out the the master brake cylinder, which requires the brake pedal to be loosened. The brake pedal braket is mounted by 4 screws, one is only accessible by removing the cladding behind the steering wheel, the steering wheel blocks that, so you have to take that off, too.

.. fixing cars is a special kind of hell.

Newer cars do have lower maintenance and repair costs. This is why vehicles with over 100k miles still fetch a fair amount of money. Computers are very good at doing the same thing over again indefinitely, unlike fully mechanical implementations which tend to drift out of spec slowly over time. A side effect of that is they require more time to actually get in and fix things.

Unlike older cars, the way newer cars work is shielded within software and hardware black boxes and aren't particularly modular. You can't just pull a carburetor out of one car and put it in another. And you're fully dependent on the manufacturer to publish meaningful diagnostic information.

> this type of complexity tends to have a more favorable profile for economies of scale

Yeah, but what about economy of safety?

> The article mentions "saves weight and installation space" - it seems this savings come at the expense of massive complexity; how big can the savings be?

Take a step back and look at how complex brake systems have evolved at the present date:

On one hand, sure, it's just a foot-operated hydraulic pump and four hydraulic actuators.

On the other hand, the ABS module has valves to both cut off each brake caliper and to drain hydraulic pressure from each brake to prevent lockup. In an emergency braking scenario, the foot pump will quickly go the floor as pressure is drained, so the ABS module also has an electrical pump to put fluid back into the system.

All US-legal cars have some form of electronic stability control (and often have lane keep assist), which can use the same components to pressurize one or several brakes to make the car follow the driver's intent (as measured by a steering wheel angle sensor and a body yaw rate sensor. Europe-legal cars also usually some form of emergency brake assist which again uses the ABS module components to apply the brakes automatically.

To keep this all safe, you need pressure sensors to monitor valve and pump functions, sensors to monitor driver input, and sensors to monitor vehicle dynamics.

When you look at the system holistically, as it exists today, it is bandaids on top of bandaids and you realize that you no longer have just a simple hydraulic system.

I would make an analogy to the 737 Max vs. Airbus FbW here: We started with something very simple, and then incrementally increased complexity in (seemingly) appropriate ways. Then, many years later, we realized that our "simple" mechanical system is actually more complex and prone to more complex failure modes than a system which started out complex and rigorously managed that complexity.

Funnily enough, A380, A350, B787, and other new aircraft have their brakes connected by Ethernet (specifically AFDX, a modified variant of Ethernet) so braking actuator signals (and sensor data back) goes as UDP messages.

But it makes the system less complex and easier to handle, as there's one kind of cabling, and complex functions like ABS equivalent are handled from avionics bay

Aircraft don't need their brakes to be as responsive as cars do. There is virtually no chance of hitting a person or another plane while landing or maneuvering on the runway.

Aircraft brakes are life or death issues that require instant response times to changing conditions, which is why ABS originated in aircraft.

The aircraft braking systems are very complex, including active thermal management and automated braking systems that take multiple data sources to manipulate breaking force, including understand how long the brakes survive when taken beyond melting point (RTO setting on autobrake)

I don't think ABS originated in aircraft because of the required braking response time. It originated in aircraft because aircraft must regularly operate in wet and icy conditions, and the performance of aircraft brakes is correlated to the length of runway required to land, which relates closely to the cost of everything in aviation. It was thus economical to use cutting-edge tech on planes long before it was economical to do the same for cars, which are mass-produced and must be affordable to individuals.

There is approximately zero chance of an aircraft pilot needing to stomp on the brakes because of a random pedestrian or animal, whereas that happens constantly for cars. Cars stop and go all the time and have to avoid each other. Aircraft hardly do that at all when they are on the runway. They use their brakes to not accidentally roll or be blown about, and in the process of landing. Those applications don't need epic response time.

Being slightly off in adjustment of brakes when landing an airliner results in dead passengers. It's not as visible for manual braking, but that's because the brakes have been "brake by wire" for some time. The scope of inputs in brake control loop is also much wider in aircraft than in typical car, at times including calculated mass, calculated ground speed, brake temperatures, tire temperatures, pressure on gear, etc.

Now consider that the same network used to control the brakes also is sufficient to control every other sensor and servo on the aircraft s well.

That is all pretty interesting but I think we've gotten far away from my original comment, which was that planes don't need brakes to be as responsive as cars do. The typical application you're describing is in landing, not in puttering around in the airport. Landing involves executing some braking program continuously until the plane is on the ground. What I'm really getting at here is that the pilot is not going to be taken by surprise and have to apply the brakes in the same way as a driver would. It stands to reason that despite the lower speeds that cars go, they need to brake spontaneously in all kinds of conditions (on worse surfaces than the typical runway, in many cases). I don't know why this point matters anymore but that's what I was trying to say.

I would like to add, brake by wire may be more reliable on planes than cars because planes are maintained very well by experts. Automobile design should strive for simpler and more failsafe options because owners are usually clueless people with limited budgets, who want to do as little maintenance as possible. The fact that hydraulic brakes work so well even in very old cars says a lot about how good they are. When they do get blamed for poor performance, it's usually due to something stupid like the owner not replacing worn out hoses after 20+ years of heavy use. I only found out about that one recently. Old hoses start breaking down and they gradually get congested.

A few things:

* Automation (emergency auto braking, full adaptive cruise, autonomy) is here and needs electrical signal input. AEB is very nearly standard everywhere now, so there's an electrical input braking component that has already been added to the system.

* If you look at modern brakes, it's hardly simple anymore. Brake boosters, ABS, master/slave systems. It's run on a corrosive fluid that is water sensitive and uses hard pipe plumbing.

Is DOT5.1 as corrosive now?

I agree that routing hydraulics everywhere is not nice, but this is the last component that I want to have a cliff failure mode profile, which is usually the caee with electronics - they work until they don’t. Hydraulic and pressure systems usually fail somewhat gracefully.

Imagine you wanted to install a system so you could pull a lever in your kitchen and have it open your front door. Would you string a wire to an electric motor controlled by your lever, or run hydraulic lines and install a pump?

Maybe it's just my bias talking as a senior software dev, but having spent a career watching software fail in inexplicable ways, and in the specific case of "if this door doesn't close, I will likely die" hand me the hydraulics please.

Also, my pinko commie ass would be remiss if I weren't to speculate about how if you fall behind on your car payments, these features among other drive-by-wire features would also allow a dealer to remotely lock out the vehicle until you pay up, and given the other shady things they already do to folks, I can't imagine this won't be an application for this tech too. Gotta fuck those poor folks till the line goes up, of course.

Edit: Or hell, doesn't even need to be if you miss payments. If you go too long for an oil change, your OEM could lock you out. If you aren't up to date on your Chevrolet+ subscription, they can make your steering wheel harder to turn. Or if you gasp try to service your own vehicle, ohhhh there could be all kinds of hell to pay.

Just, rampant opportunities to fuck with consumers just as hard as tech companies do. And isn't that what we all want in the end? Cars just as grating and irritating to use as our computers and cellphones are, in the thousand tiny ways per day they are?

The question was cost, complexity, weight, etc. Reliability is an different issue. But nothing says an electronic system can’t be extremely reliable too.

Your fear of being locked out makes no sense. That was feasible the moment you had a cellular modem and computer control of the ignition. Modern cars already have electronic control of the brakes for ABS, it’s just not in between the pedal and the calipers. Brake by wire changes nothing in terms of the ability to remotely disable a car.

When it comes to brakes, the first three priorities are reliability, reliability, and reliability. If indeed the brake-by-wire is as reliable as it's cousin, fine and dandy. But I've worked on brake systems and I, to be blunt, like the idea that if you slam on the brake pedal, oil travels through lines and closes calipers upon brake discs. You can do this with a car turned off, with it's engine disabled, with it's electrics completely dead, because the linkage between your brake pedal and the brakes is a mechanical one.

I'm not saying there aren't ways to make this equally safe with software, I can think of many ways. Do I necessarily trust the built-to-cost auto industry to do them? Ehhhhh....

Agreed. In a perfect world, wire-based systems should be cheaper and more reliable. But as a software engineer, I can tell you that current software development practices do not produce software that is reliable enough to entrust it with your life.

And when corners get cut, the software development lifecycle tends to be an easy target.

There are plenty of places where we entrust software with our lives. We don't yet have brake by wire, but we certainly have software-controlled brake actuation on any car with ABS, and that could easily kill you if it activates at the wrong time in the wrong way. Uncommanded acceleration is another possibility that might have actually killed people (or maybe it was just floor mats). Outside of cars, your microwave could burn your house down in the middle of the night with the right sort of software bug. So could any device with a lithium-ion battery. And of course numerous airliners are fly by wire and they remain the safest form of travel out there.

Now, I certainly wouldn't trust anything written with a typical app development process with my life, but it's not the only way to do things. I'm totally on board with being pessimistic about most software reliability, but it's not realistic to say that no software can be reliable enough to trust with your life.

> We don't yet have brake by wire, but we certainly have software-controlled brake actuation on any car with ABS

The software is limited to being able to flutter pressure going to the actuators within the calipers. Basically it de-pressurizes/re-pressurizes the actuator very quickly. ABS cannot engage or disengage the brakes in totality, that is exclusively done via the pedal. (In the systems I've worked on anyway)

> Uncommanded acceleration is another possibility that might have actually killed people (or maybe it was just floor mats).

Rather famously, early Prius models had issues where the drive-by-wire throttle would become stuck on due to a software malfunction.

> Outside of cars, your microwave could burn your house down in the middle of the night with the right sort of software bug.

Um... how? I could see this if something suddenly let go quite violently in the power supply maybe, but that's far likelier to happen when the unit is in use, not when it's sitting idle. And more to the point even if the computer inside had a bug that caused the magnetron to engage at full whack with nothing in the microwave... I mean it probably wouldn't be good for the microwave, but I don't see how that equals a fire.

> So could any device with a lithium-ion battery.

That one's true though. But good electronics are monitoring all the voltages across all the lithium cells involved, along with temperature, and many have perma-kill switches if anything looks too off for this exact reason.

> And of course numerous airliners are fly by wire and they remain the safest form of travel out there.

They're also absolutely drenched in aeronautics gear that is both managing the aircraft and verifying that the redundant systems are ready to take over, reporting the same data, and that all is well. And aircraft regularly get grounded if anything even seems like it might be off with it.

What is checking "if anything looks too off"? Pretty sure there's software involved there.

And yes, I know airliners have all sorts of stringent standards. But the fact remains that many of them do rely on software to keep the occupants alive.

Making software reliable enough to bet your life on it is hard, but not so hard that we should consider it to be an unattainable goal.

> What is checking "if anything looks too off"? Pretty sure there's software involved there.

It's not just software: is software running on redundant systems that can check one another for consistency and performance. You're never going to see that level of engineering in an automobile, cars would be far too expensive if they were as complex, electrically, as an aircraft. Aircraft are only getting away with it because, as you correctly observed, if the computers malfunctioned, they're going to fall out of the sky and kill all your passengers. If your car malfunctions, it just won't move/will stop moving.

That “it anything looks too off” was battery management systems. I don’t think they have redundant systems.

In any case, the sentences I was replying to was:

> But as a software engineer, I can tell you that current software development practices do not produce software that is reliable enough to entrust it with your life.

All I’m saying is that this is wrong and we do routinely trust software with our lives. If you want to explain how it’s done and why it’s done that way, go for it, but it sounds like you think you’re arguing against it.

I might use hydraulic lines if a wiring fault could kill me lol...

> Is it really as simple as "cheaper to produce"? Is it cheaper to replace? > it seems this savings come at the expense of massive complexity

Less parts, less things to break, less things to install, less things to maintain. Etc. All good things when it comes to cost. You'd be surprised how that stuff adds up quickly.

This will not be less parts. It will be integrated modules Instead of tubing, with each module being about as many parts as the original entire abs system. Cheap to manufacture because cars can be engineered more easily, but each module will be 1000 dollars and unobtainable in 20 years, because it will be very specialized digitally for that specific vehicle models, sub year, etc. non-maintainability is a feature, not a bug.

> As an aside, IMHO every time a component changes to "by wire" the driving experience tends to suffer due to latency.

_window-by-wire_ & _mirror-by-wire_ seems fine.

I've seen a non-zero number of electric windows and mirrors that have failed in my time; what does that look like for a braking system?

I was replying to a comment saying that, replacing mechanical actuation by electrical actuation, in a car, is always a downgrade of the user experience.

The discussion was not about failure.

My electric drivers window is dead and my driving experience is definitely worse. Electrical stuff is so hard to DIY and work on as well and parts are ridiculously expensive.

Apologies, you're correct. I had my wires crossed (pun intended).

No problem!

Cheaper.

Cost.

Space and weight. It saves on both while also being more flexible and presumably much simpler to install. Install times will save money at the factory and the space savings can be used elsewhere. In theory it can also be more reliable. It may also be able to provide smarter brake pressure to come to a stop quicker and more safely. The only real downside will probably be brake feel. You aren’t going to get a lot of feedback, but for day to day driving that isn’t much of a concern.

You can imagine the robot bolts the integrated motor breaking assembly to the body. And then a tech snaps the high voltage and control connectors together. That's it.

> Why is there a seemingly unending shift in the automobile industry toward digital/electric controls and away from anything mechanical/physical?

Because it is cheaper.

Hydraulic systems are themselves complex and require many components all of which may fail. Detecting failure in a hydraulic system requires additional systems and is generally only indicative of overall system state rather than specific failure in a specific sub-assembly, on account of the shared nature of the medium. Assembly and testing are expensive.

Electronic systems are much simpler by part count, lighter, and smaller.

When communication fails, you know it's failed, and you know where it's failed.

Adding sensors for real time feedback is extremely cost-efficient. If your current/torque/temperature are not within an expected profile, a braking system failure may be inferred. This is extremely cheap, reliable and fast on low-end hardware.

In this case, they do not need left and right forms, they can just produce one part. Right away that's a huge saving in mold cutting time, mold storage, part production, distribution, etc.

In a mechanical system, if something fails, you have to disassemble the subsystem, identify the failure, search the supply chain for the part in question (possibly not labeled), obtain or manufacture the part, re-install it, and then hope for the best.

In a digital system, it can probably tell you what fails. It's probably cheap. It probably has the part number printed on it.

About latency, the key point is that humans suck. There's no way 1ms latency is going to alter driving outcomes. If you care about latency, you're autonomous driving. And that's something that electronics excel at versus humans. Let's say signals begin to come in suggestive of brake failure on one of four brakes while rounding a corner at high speed. ABS systems can compensate for the failure in real time and intervene to guarantee safety. Mechanical systems will leave the driver to 'discover' the 'new handling characteristics' at the point when it's already too late.

So the summary is: cheaper to produce, cheaper to assemble, cheaper to debug, cheaper to maintain, cheaper to replace, less unique parts, more freely positioned within a vehicle design, more readily reusable across vehicle designs, safer in some conditions, more suited to forward-looking applications such as autonomous driving.

(Disclaimer: I am not a mechanic, however I do have a fairly solid understanding of multi-disciplinary mechatronic design problems.)

Electric brakes were avoided up to now for safety reasons, despite potential savings. You don't save money by destroying your reputation with such absurd compromises.

Cost and the potential for vendor lock-in drive this. How long does anyone seriously think electric brakes will last on cars that bang around in mud and potholes full of water? What is the failure mode? Can you even tell when the brakes are failing? There are hydraulic brake systems many decades old, still in service. There's nothing new about this electric brake stuff that has only become feasible or cheaper today... It was deliberately avoided by engineers of the past because they did not trust the technology. They didn't trust hydraulics either. That's why emergency brakes are cable-driven. Of course, the cable can also work when there is no hydraulic pressure, such as when the car is parked and turned off. It's a smart design that serves multiple purposes.