Questions from an aircraft mechanic about cars.
Discussion
On all the modern aircraft I've worked on, there are multiple, redundant flight control computers.
Why is everything controlled by the same computer for a drive by wire car? For example, older generation cars had two separate computers to control both the engine throttle and the transmission - the Engine Control Unit, and the Transmission Control Unit.
Modern cars now implement a combination of those computers into a single computer - called the Powertrain Control Unit. This struck me as odd, especially after hearing about Toyota's having issues with Sudden Unintended Acceleration. In instances of SUA, shifting into Neutral apparently did not work. This makes sense if everything is controlled by the same computer and that computer freezes - nothing would work. If the same computer controlling the engine AND transmission has gone berserk, the transmission probably won't accept commands either, because it runs on the same computer. All the transmission stick/buttons do is send commands to the computer. There's no mechanical linkage between the controls/buttons and what you're trying to control. It's all just buttons/controls to send to the computer, which then does stuff. If they were separate computers, then in order for there to be SUA and inability to put the car into neutral, two separate computers (Engine Control Unit AND Transmission Control Unit) would need to fail at the same time.
You may be thinking, ''why not turn the car off?'' There is no key to turn. You push a button to turn the engine off. Electronic signal being sent. No mechanical contacts. Survivors have testified in the congressional hearings, for instance, that trying to turn off the engine, shift into neutral, take your foot off the accelerator, etc. have had no response from the car. Or at least no desired response.
On airplanes, there's anywhere from three to five flight control computers and also separate computers which control things like spoilers, autopilot, and the stabilizer trim. If one computer freezes or gives a bad result, like giving a result that would make the plane suddenly barrel roll, it's voted out by the computers that are working normally, and ignored. The reason the 737 MAXs crashed is because a system did not have redundancy. It took information from one AOA (Angle of Attack) sensor, and there was one computer. Since the incidents, Boeing has implemented redundancy to prevent such tragedies happening again.
So what's with the design philosophy of making everything computerized, and removing any redundancy (by making several important things be controlled on a single computer, producing a potential single-point failure mode which would always result in a high speed crash?) Airbus's design philosophy is that computerization makes things safer - if it's redundant. This is why an A380 that had an engine explode managed to land safely - if there was just one computer, the plane would have certainly crashed. It seems negatively redundant to computerize something and remove all redundancy!
If I were to make a comparison, this sounds similar to having a single computer which controls the elevator, ailerons, landing gear, flaps, and the engine throttle. If the elevator doesn't respond to input, you need to land ASAP. You need the flaps to slow down, the ailerons to steer, the engines to provide power, and the landing gear to land safely. You really need to be able to use the engine throttle to adjust pitch as well. But if all of those things are controlled by a single computer with no backup, and that computer freezes....you get my gist. It's a tragedy waiting to happen and a Single Point Failure mode..
My question is why is the design philosophy for car manufacturers so different from the design philosophy for aircraft manufacturers? Surely if car manufacturers design cars to have all kinds of wacky features, they could add in some redundancy as well? Or is safety not a major concern for car manufacturers? Honestly, negative redundancy should be illegal.
Why is everything controlled by the same computer for a drive by wire car? For example, older generation cars had two separate computers to control both the engine throttle and the transmission - the Engine Control Unit, and the Transmission Control Unit.
Modern cars now implement a combination of those computers into a single computer - called the Powertrain Control Unit. This struck me as odd, especially after hearing about Toyota's having issues with Sudden Unintended Acceleration. In instances of SUA, shifting into Neutral apparently did not work. This makes sense if everything is controlled by the same computer and that computer freezes - nothing would work. If the same computer controlling the engine AND transmission has gone berserk, the transmission probably won't accept commands either, because it runs on the same computer. All the transmission stick/buttons do is send commands to the computer. There's no mechanical linkage between the controls/buttons and what you're trying to control. It's all just buttons/controls to send to the computer, which then does stuff. If they were separate computers, then in order for there to be SUA and inability to put the car into neutral, two separate computers (Engine Control Unit AND Transmission Control Unit) would need to fail at the same time.
You may be thinking, ''why not turn the car off?'' There is no key to turn. You push a button to turn the engine off. Electronic signal being sent. No mechanical contacts. Survivors have testified in the congressional hearings, for instance, that trying to turn off the engine, shift into neutral, take your foot off the accelerator, etc. have had no response from the car. Or at least no desired response.
On airplanes, there's anywhere from three to five flight control computers and also separate computers which control things like spoilers, autopilot, and the stabilizer trim. If one computer freezes or gives a bad result, like giving a result that would make the plane suddenly barrel roll, it's voted out by the computers that are working normally, and ignored. The reason the 737 MAXs crashed is because a system did not have redundancy. It took information from one AOA (Angle of Attack) sensor, and there was one computer. Since the incidents, Boeing has implemented redundancy to prevent such tragedies happening again.
So what's with the design philosophy of making everything computerized, and removing any redundancy (by making several important things be controlled on a single computer, producing a potential single-point failure mode which would always result in a high speed crash?) Airbus's design philosophy is that computerization makes things safer - if it's redundant. This is why an A380 that had an engine explode managed to land safely - if there was just one computer, the plane would have certainly crashed. It seems negatively redundant to computerize something and remove all redundancy!
If I were to make a comparison, this sounds similar to having a single computer which controls the elevator, ailerons, landing gear, flaps, and the engine throttle. If the elevator doesn't respond to input, you need to land ASAP. You need the flaps to slow down, the ailerons to steer, the engines to provide power, and the landing gear to land safely. You really need to be able to use the engine throttle to adjust pitch as well. But if all of those things are controlled by a single computer with no backup, and that computer freezes....you get my gist. It's a tragedy waiting to happen and a Single Point Failure mode..
My question is why is the design philosophy for car manufacturers so different from the design philosophy for aircraft manufacturers? Surely if car manufacturers design cars to have all kinds of wacky features, they could add in some redundancy as well? Or is safety not a major concern for car manufacturers? Honestly, negative redundancy should be illegal.
Edited by Charlesitis on Saturday 26th February 01:44
Just my guess but the answer is probably because it is cheaper, simple as that!
Car manufacturers will do anything to save a penny, and if that idea saves weight then even better as cars are expected to become ever more economical whilst carrying more components for safety or marketing reasons.
Car manufacturers will do anything to save a penny, and if that idea saves weight then even better as cars are expected to become ever more economical whilst carrying more components for safety or marketing reasons.
I am no engineer and I saw stats about how many ECU's are in various cars getting thrown around on the EV threads here.
Someone was trying to make the point that Tesla was better because they are cheaper to build and use fewer ECU's and hence arent chip supply constrained. VW and BMW cars had 2-3 times the number of ECU's.... so yeah, some OEM's can and will cut corners. Others use tried and trusted systems with multiple ECUs for different purposes. Does it help in these situations? Good question - thats where my knowledge runs out.
However, it was widely reported that Tesla has stopped fitting the secondary steering angle sensor on Model Y's due to lack of supply. You legally require a second to provide a supplementary signal for where the wheel is turned and where the tires are pointing - its a legal requirement for the use of any auto-steer technology to get corroborated data. Clearly Tesla will retro-fit them, but for me, it shows that there is a degree of redundancy built in from a legal standpoint. Someone thought about this and put it in the standard - so thats good. Not sure it fully addresses the real requirement though.
Someone was trying to make the point that Tesla was better because they are cheaper to build and use fewer ECU's and hence arent chip supply constrained. VW and BMW cars had 2-3 times the number of ECU's.... so yeah, some OEM's can and will cut corners. Others use tried and trusted systems with multiple ECUs for different purposes. Does it help in these situations? Good question - thats where my knowledge runs out.
However, it was widely reported that Tesla has stopped fitting the secondary steering angle sensor on Model Y's due to lack of supply. You legally require a second to provide a supplementary signal for where the wheel is turned and where the tires are pointing - its a legal requirement for the use of any auto-steer technology to get corroborated data. Clearly Tesla will retro-fit them, but for me, it shows that there is a degree of redundancy built in from a legal standpoint. Someone thought about this and put it in the standard - so thats good. Not sure it fully addresses the real requirement though.
It's a reasonable question but your analysis is a bit simplistic and yet missing some big contextual stuff.
First you should probably ask 'if' rather than 'why'. There are multiple controllers and sensors for redundancy purposes in things like brake-by-wire. Putting that aside...
The most obvious is that the consequences are hugely different. If you experience significant control failure on an aircraft then it endangers the lives of everyone onboard and potentially a lot of people not onboard. Even in the worst case scenario the same doesn't hold true for cars.
Safe aircraft operation is almost wholly reliant on its systems - think about IFR. This is supported by sensor data and good data requires multiple sources. The same simply does not apply to cars where hopefully you can see where you're going. This vastly limits the scope.
A lot of your question is predicated on the idea of the computer becoming unresponsive like a desktop PC. Fortunately that's not really how embedded, safety-critical systems behave. Nor is it really clear that real world failures can be attributed to anything like this, rather than either physical problems like jammed accelerator pedal or pedal sensor failure.
There is a significant usability problem in that if this scenario does occur, there are various barriers to safe intervention. For example, holding down the stop/start button on a car for say three seconds probably shuts it down. But who is going to know that, much less try this when confronted with SUA?
First you should probably ask 'if' rather than 'why'. There are multiple controllers and sensors for redundancy purposes in things like brake-by-wire. Putting that aside...
The most obvious is that the consequences are hugely different. If you experience significant control failure on an aircraft then it endangers the lives of everyone onboard and potentially a lot of people not onboard. Even in the worst case scenario the same doesn't hold true for cars.
Safe aircraft operation is almost wholly reliant on its systems - think about IFR. This is supported by sensor data and good data requires multiple sources. The same simply does not apply to cars where hopefully you can see where you're going. This vastly limits the scope.
A lot of your question is predicated on the idea of the computer becoming unresponsive like a desktop PC. Fortunately that's not really how embedded, safety-critical systems behave. Nor is it really clear that real world failures can be attributed to anything like this, rather than either physical problems like jammed accelerator pedal or pedal sensor failure.
There is a significant usability problem in that if this scenario does occur, there are various barriers to safe intervention. For example, holding down the stop/start button on a car for say three seconds probably shuts it down. But who is going to know that, much less try this when confronted with SUA?
In any event of an uncontrolled accelerating vehicle, the brakes can overpower the engine and bring it to a stop.
There are many 'computers' in a modern car but they will be called modules, ie. Body control module, airbag control module, suspension etc. Most automatics on sale today will have separate transmission module to the ECU, the ECU will control engine functions only and everything will run through the CPU via multiplex system. Their memory will only be 'look up' tables and do very little actual computing so delays freezing and crashes are impossible.
There are many 'computers' in a modern car but they will be called modules, ie. Body control module, airbag control module, suspension etc. Most automatics on sale today will have separate transmission module to the ECU, the ECU will control engine functions only and everything will run through the CPU via multiplex system. Their memory will only be 'look up' tables and do very little actual computing so delays freezing and crashes are impossible.
rigga said:
Car won't fall out of the sky.
Simplistic answer, but basically that's it.
Exactly. Stopping for a car and stopping for a plane are not quite the same. Still very impressed by the fact that aircraft, statistically, just never really go wrong, the failure rate would be considered zero under "ordinary" metrics. Both mechanically and electronically. Simplistic answer, but basically that's it.
There’s redundancy in automotive systems where deemed necessary too, but sometimes it’s not so obvious. Things like split braking systems so you only lose half in case of failure, to multiple redundant calculation within electronics modules. Have a read up on ISO26262, ASIL and SOTIF if you’re interested in finding out more.
Charlesitis said:
On all the modern aircraft I've worked on, there are multiple, redundant flight control computers.
Why is everything controlled by the same computer for a drive by wire car? For example, older generation cars had two separate computers to control both the engine throttle and the transmission - the Engine Control Unit, and the Transmission Control Unit.
The new stuff I work on, has more than one module for powertrain. The gear selector in your hand is the transmission switch, but you could argue that’s “computing”, as it sends it’s information/signals along the CAN.Why is everything controlled by the same computer for a drive by wire car? For example, older generation cars had two separate computers to control both the engine throttle and the transmission - the Engine Control Unit, and the Transmission Control Unit.
Engine ECU, obviously.
And the gearbox has it’s own transmission module, also working on CAN.
And then nearly every Switchpack/module/ECU/CAN is connected to the central body unit.
What manufacturer are you referring to, that does that lot on One module ?
That said, you could stick a hundred modules in if you wanted to, just to do a gear change. But they’d still have to talk to each other. So if your comms network goes down or a rogue module brings it down, you’d still lose all hundred.
Edited by RUSSELLM on Sunday 27th February 09:47
Sump said:
Is this just a thread for you talk about being an aircraft mechanic?
When a car ecu fails the car doesn’t move anywhere. It isn’t gliding in the sky, it’s sat dead on someone’s drive…
When a car ecu fails the car doesn’t move anywhere. It isn’t gliding in the sky, it’s sat dead on someone’s drive…
Surely equally likely you'll actually be driving somewhere at the time...
Sump said:
You don’t need as much redundancy failure points. Also MCAS with Boeing?
Forums | General Gassing | Top of Page | What's New | My Stuff


