Pre-Safe Sound: Playing ‘pink noise’ in the split second before impact (2015) | Hacker News Reader

Pre-Safe Sound: Playing ‘pink noise’ in the split second before impact (2015)

mercedes-benz.com

345 pointsrosstex156 comments

This is a comment from Reddit on the same [0]; though I cannot vouch for its veracity, I found it interesting nonetheless:

Former Benz master tech here.

I've spent a good bit of time diving into the nitty gritty of these systems.

It uses all of the vehicle sensors to determine crash logic.

You want to fire the pretensioners before the airbags if at all possible. This locks the occupants in a much more secure and predictable psition.

Accelerometers can determine if something like a rollover is in progress.

Heavy braking + high speed + sudden activation of the parking sensors indicates a crash before physical contact happens.

It can use the radar on distronic equipped cars to predict a crash.

If the vehicle is in a slide, and a sudden yaw correction happens, this means a collision has occurred, even if the airbag sensors haven't been triggered. This is useful because the airbag sensors are only good at detecting force in one dimension.

It's even possible for the vehicle to determine to fire the pretensioners but no airbags based on occupant size and weight (determined through seat/steering wheel position and capacitive mats in the seats) and impact vectors.

It will even do things like full field the alternator up to 18+ volts to slam the windows shut super fast in the event of a rollover.

The actual detection logic fills a 4" thick book.

It's a big reason those cars are so damn expensive.

Mercedes doesn't fuck around with safety.

If I had to be in a wreck, it would be in an S class. Hands down.

0: http://www.reddit.com/r/todayilearned/comments/59qhaw/til_ne...

> If I had to be in a wreck, it would be in an S class. Hands down.

To be honest, I'd rather be in a Tesla, since no amount of smart sensors can replace a crumple zone. The S class in particular has a big engine (usually a V8), which won't play nice during a front end collision.

Crumple zones assume a nice, controlled, head-on crunch. Not so useful against getting whacked in the side by a truck. But then, probably not much is.

Tesla vehicles have the highest safety rating for resisting a side impact protrusion, one of the more difficult safety vehicle tests.

Also, they use the same materials engineering from SpaceX designed for collisions in space. One of the reason their safety rating was so high.

What does this even mean?

Spaaaaaaceee

https://electrek.co/2015/05/24/spacex-transferred-novel-weld...

Friction stir welding was developed in the 90’s, but Spacex perfected its own technique to adapt the technology to large sheets of metal like the ones use for the aluminium tank of their rockets. They developed their own technique and equipment and perfected the process. Here is Musk explaining the process to Wired back in 2012:

Instead of riveting the ribs and hoops, you use a special machine that softens the metal on both sides of the joint without penetrating it or melting it. Unlike traditional welding, which melts and potentially compromises some metals, this process works well with high-strength aluminum alloys. You wind up with a stiffer, lighter structure than was possible before. And your material loss is maybe 10 percent, just for trimming the edges.

Which materials?

A "collision in space" is generally a wildly catastrophic event.

I see this claim tossed around a lot, but is there really a documented significant effect on safety from not having an engine up front? I mean, with a car having that huge front, is there not so much space for crumple zones that engine or not is insignificant?

Now I know the Model S claims "best in class NHTSA rating", but no comparable luxury sedan has been crash tested by either US NHTSA or Euro NCAP.

That's right, neither the BMW 7-series, the Audi A8 nor the Mercedes S-class have been tested. So we just don't know.

Edit: Update: I've checked various other cars, and the only one I could find where both an EV and a non-EV version have been crash tested in the same year is the 2014 Kia Soul (a small MPV) Euro NCAP tests. For that car, the diesel one did worse than the EV on driver protection, specifically of the feet, which was caused by spot welds (presumably in the torpedo wall) breaking and sharp edges protruding near the driver's feet. But for child safety, which is more focused on just acceleration and thus a better proxy for crumple zones, the cars were equal. So this seems more indicative of a design fault with the Soul than a lack of crumple zones.

Euro NCAP have tested a E class Mercedess, which isn't that much smaller than an S class and it gets a higher rating than a Tesla Model S.

http://www.euroncap.com/en/results/mercedes-benz/e-class/253...

This is all fascinating stuff, thanks.

Of course, in the future when all cars are connected (even if they're not self-driving), combining all this sensor data and sending it to the vehicles behind you would save countless lives. The number of times you hear about pile-ups in fog and heavy rain - that could all be avoided by cars talking to each other.

But I'll bet there's still no open (or even closed) standard on inter-manufacturer comms yet?

and all the script kiddies will find it great fun to send out those "collision" signals on a busy roadway just to screw with people. the futures gonna be great!

That does raise interesting security problems though - my car receives a message saying to stop as quickly as possible, how does it decide whether this is a valid message or someone hacking cars for fun?

I don't have an answer for you, but maybe it stems from some sort of idea of identity? Maybe the car sends a message authenticated with your license ID. Cars are (hand waves) able to authenticate the validity of a message. If you send a message saying you were in a collision, that comes with the side effects of saying you were in a collision - emergency contacts and police may be notified, maybe something pings cars registered to you to see if it was actually in an accident based on sensor data. Your insurance rates may go up. If it turns out you were lying, maybe your car gets flagged for repairs next time you try to re-register your tags. Or maybe you are flagged for trolling and you get the "boy who cried wolf" effect.

Sure, the actual vehicle systems had better work and be properly authenticated, and that could be achieved in numerous ways... but none of this stops a hacker war-driving with a laptop, or sitting on an overhead bridge, or on the other side of the world.

There is no doubt that intelligent highways would be a massive target for terrorist and cyber attacks. However, one thing in our favour is the fact there is so little technology already in place, since they won't be burdened with decades' worth of insecure legacy systems: at least they'll be able to start out building infrastructure in full knowledge that it better be secure.

I'm tempted to qualify the second point as optimist.

US voting machines didn't have to deal with legacy technology and still turned pretty bad from an infosec POV.

Exactly, and it is not just this one case: serious, avoidable security cock-ups have happened every time a new class of device has been put on the internet, with the IoT DDOS debacle being perhaps the latest example.

not burdened by legacy, also means not burdened from years of lessons learned.

Sadly, the lessons learned in legacy systems do not necessarily address the problems we have now. Older systems were typically combating noise, signal degradation, power consumption, processor speeds, poor user interface technology, etc. Often they were extremely proprietary and if they included any kind of security it was commonly by obscurity, and even if not, the crypto used would be easily crackable by a modern script kiddie packing Aircrack or similar.

The lesson we do need to learn is future-proofing: whatever infrastructure we install on the millions of miles of public highway will need to be serviceable for decades.

In some ways we have methods to mitigate this already in place. They have, of course, been exposed as insecure when they were implemented badly[1][2]. That being said, there is prior art on this, in nature.

You have your own trusted sources of information, and you have your secondary sources of information, which you apply trust levels to. Your own trusted sources of information are your own senses, your sight, hearing, feeling (touch), smell, etc. You may have something communicated to you about your condition or environment through these senses, but it's silly to take that at face value without confirming with your own senses. If your friend next to you that you trust tells you that you're about to walk into a wall, you look ahead. If you can't see a wall, you might slow and exercise caution until you figure out why you were relayed that information, but unless you can completely stop without problem, you don't do that without cause.

I think it's a mistake to think of cards in the future communicating as a swarm. They need to be able to function independently, and also to take in extra information from the group when the group is available, and make decisions on that. That's less swarm behavior than social behavior, so we should consider groups of cars on the road a social groups, and the same information dynamics exhibited in those groups apply.

1: http://gizmodo.com/how-a-hacker-could-hijack-an-airplane-fro...

2: https://www.wired.com/2015/07/hackers-remotely-kill-jeep-hig...

So, a malicious actor could spoof the vehicle of somebody they dislike, and cause their insurance to go up, or pave the path for a later incident, where when police/emergency are contacted, the person is ignored?

I agree, that is probably going to be one of the most difficult problems to solve. However, as we dive headlong into self-driving tech, we are going to have to find an answer since the tech will be so much easier to implement if the vehicles communicate.

PKI, under the current proposals:

http://www.nhtsa.gov/staticfiles/rulemaking/pdf/V2V/Readines... (PDF link)

Couldn't this be solved better by some form of web of trust (WoT)?

Say your car A trusts other cars B, C ... Y to varying degree in a WoT. It meets a brand new car Z that is not yet trusted by any cars A-Y. Car A receives a message from Z, to which it assigns 0 credence points, meaning this data is not taken into account in any decisions. If car A then makes an observation confirming the message, car A now trusts Z by +1 point, and lets other cars know this. (I'm unsure how to make aggregation/distribution of trust work smoothly, but I believe the random mixing of traffic patterns will work in the system's favor.)

So gradually car Z becomes trusted, up to some maximum, e.g. 100. The only worry then is a malicious actor who mostly sends true messages, but that's hard to distinguish from a car with a sometimes faulty sensor. Those problems could be handled the same way, namely aggressively penalising a car for sending incorrect data (e.g. by -30), such that it's quickly distrusted, together with a prominent warning in the failing car when it goes below 0 trust that it needs maintenance (after which trust would be reset to 0). Possibly also increased insurance premium if your car remains with a trust below 0.

The same way you determine whether that stop sign is legitimate or not: you don't. You stop regardless of the possibility that the sign was placed there by potential carjackers. Until the hacking risk is demonstrated to be statistically more dangerous than ignoring potential warnings, all warnings will be trusted. Regardless of current proposals, any manufacture that ignores reasonable electronic warnings (ie STOP accident ahead) on technicalities (ie an expired cert) risks serious liability.

This touches on the much more likely scenario whereby the car's internal data doesn't mesh with the data from it's sensors. If the sign says "one way" it is a one way street. What google maps says doesn't matter. Current warning signs trump previously recorded data. "But google maps said I could do 70!" will be no defense.

That's a great way to rob people on an otherwise deserted highway.

As always Science Fiction swoops in - in the "Culture" series, AI have hardcoded (as in, on their physical hardware) automatic code to shut down external communication in the event external communication is attempting to takeover the AI. I mean your question is pretty much "what would that automatic code look like," and I have no idea. I'm just getting pretty excited about all this sweet stuff happening with AI, automatic cars, hijacking of the human nervous system. Hngg.

> AI have hardcoded (as in, on their physical hardware) automatic code to shut down external communication in the event external communication is attempting to takeover the AI

This sounds very much like you're describing an antivirus program, which we all know works perfectly well /s

> ... when all cars are connected ...

You're so funny. This is a world of proprietary technology.

What can I say? I'm an idealist :)

Exactly. At best we'll have 15 different standards, and Fords only talk to Fords, Toyotas only talk to Toyotas, and they all speak different incompatible protocols.

Just think of all the integration programming jobs that will be preserved! All of the pointless conference calls and GoToMeetings scheduled weeks in advance between developers and project managers and teams on the other side of the world because someone made a typo in an API specification that could have easily been cleared up in a two-second email.

Good times ahead! The hardware stacks will all likely be different too, so gravy train for all the HW and systems folks too. dozens of companies all re-inventing the same thing is why tech still has full employment.

Actually, the hardware guys already have it solved. We already know how to do bus arbitration, and cars are like little buses.

CANbus could work, and it's already used in the automotive world. It's nice because the higher-priority bus driver is not delayed. Unfortunately, the lower priority one gets run over. Might have to work on that.

You're being optimistic. The worst that can happen is that cars of different brands start lying to eachother.

Until regulators in a sufficiently large market decides that since the tech is available, either the manufacturers decide on a standard or they get one imposed on them, the same way EU sabre-rattling led to the standardization of mobile chargers.

You could mostly prevent pile-ups, but pile-ups generally don't kill people. Also, you don't need connected cars to avoid pile-ups, you just need to keep your distance and your eyes (sensors) on the car in front of you. Both of which works pretty well with current (unconnected/unilateral) technology.

None of the causes in listed in this article require interconnection to avoid: http://uk.businessinsider.com/the-cause-of-the-most-fatal-ca...

Most of the pile-ups that happen around where I grew up were due to thick fog. Where by the time you see the wreckage, it's too late.

Now, that can be solved by people slowing down due to conditions. But how slow is slow enough? Often it's slow enough that the person behind you is gonna wreck into you.

Maybe pile-ups aren't deadly, but wouldn't it be better if we had tech that helped prevent them? Or do maimings not matter?

Slow enough is slow enough that you can see an obstruction soon enough to stop before driving into it.

Either way, interconnections won't help you -- what if the car is crashed hard enough (or long enough for the battery to run down) that it's beacon is off/broadcasting wrong information? What if there's fallen tree or a pedestrian on the road? Generally, sensors (radar) can see through fog, and that's the most help you're going to get in a situation like this.

Yes, maimings do matter. But they don't count towards "save [ing] countless lives".

> what if the car is crashed hard enough (or long enough for the battery to run down) that it's beacon is off/broadcasting wrong information?

Doesn't that mean we're just back at the situation we're in today? The worse situation?

> What if there's fallen tree or a pedestrian on the road?

Also, not the situation at hand. Talking about the case where a car has crashed and is letting the cars behind it know that it has done so. It's like electronic flare deployment (or hazard lights.)

Don't you think that a lot of lives and livelihoods could be saved if we had such functionality?

I imagine typical car sensors can "see" through fog and rain just fine.

The pileups are 'due' to people going too fast for road and weather conditions. If you can't stop within your length of visibility, you're going too fast, end of discussion.

Without saying this is wrong, I will say it's not at all practical.

I grew up in the northeast. Abrupt and unpredicted snowstorms are common, and some of them are dense enough to drop visibility laughably low - 10-15 mph would offer visible stopping (and be a fairly acceptable speed for a crash regardless). People don't want to take six hours to get home, so they keep driving. If the roads are decent, they keep driving at ~30 mph or faster to make decent time.

So now you have a situation where all the other drivers are exceeding their visibility. If you drop down to 10 mph, you get to be the obstacle they hit at >30 mph when you loom up out of the snow!

The result is that everyone goes 20 mph or faster, because everyone else is. Yes, that's stupid and dangerous. No, there isn't any 'safe' behavior available if you get caught out on those highways.

Wow, I had no idea the situation was that bad in the northeast. That's honestly terrible for a supposedly first-world country. Here are the solutions I've enacted for myself in no particular order:

- Don't live in the northeast.

- Don't drive in bad weather conditions.

- Don't organize your life around automobiles such that you can't manage daily life without one.

Those are all pretty reasonable answers, honestly, and a lot of people who leave the northeast do it in general protest of the winters. Moving closer to the coast will also preserve you from much of this.

All I can really say is that I've definitely been on the road at moments where I realized that 100% of my options were dangerous accident risks. At that point you mostly try to minimize likely crash speed (if everyone goes 15-20 on a divided highway, probably no one will die) and look for an exit.

It is not that simple. "Too fast for road and weather" is a ex post facto determination. If there was a crash, you were going too fast.

Fog is tricky. Fog can give the driver a false sense of security. Modern overly-reflective road markers can, at night, make a driver think they are seeing further through fog than they are. They can see the lines, but can they see the black car parked ahead? Note too that large pile-ups in fog often involve trucks. Imho much of that is because truck trailers have poor running lights. We see the cars ahead. We follow their taillights through the fog. But that black truck flatbed trailer with two tiny red orbs is invisible. Or we think it is further away.

Fog also thickens unevenly. It blows around. Slamming on the brakes immediately upon entering a bank isn't going to make you any friends amongst the people behind you. That truck 4 seconds behind you (a reasonable distance on a highway at night) might not stop as quickly as your ferrari. Truck drivers are also much higher. They may have greater visibility than you and not realize that your sports car places your eyeballs in the low-level fog. The safest bet in traffic is generally not to act abruptly and unexpectedly.

(fyi, a sportbiker's helmeted head is also generally much higher than in cars, even SUVs. They too have better visibility in fog.)

And if you forget to turn on your car's taillights, all is lost. There is a reason that motorcycles have them hardwired on.

> If you can't stop within your length of visibility, you're going too fast, end of discussion.

In the real world pileups typically happen when you suddenly go from full visibility to dense fog (or snow) with no warning.

So maybe instead of blaming people who could not do anything about it, see if there is a way to help them.

Laying the blame isn't a useful way to solve a problem.

Where I grew up, it was dense snow, which behaves pretty similarly.

I don't really see the value of saying "human error, so who cares about solving it with tech!" It's human error that happens all the time, and causes plenty of injuries and expense. Falling off a cliff is human error, but we still put up railings where it's likely to happen.

And as you pointed out, it's not human error that a single driver can solve - if you go slow enough to be safe, you're likely to get hit from behind by someone who didn't. Since the reality is that drivers are faced with having no safe strategy, it's a problem I'd love to see solved.

Assuming the occupant didn't see it coming, the car will have made all its decisions and initiated all active safety measures before a human being inside the car would even notice a collision is taking place.

The car is effectively 'done' progressing the event before the human even starts.

It actually seems a lot slower than I would have expected, to be honest. The seatbelt tensioners take 150ms to activate, during which time I would have travelled 2m at 30mph (1). That effectively makes them useless, no?

1) https://www.wolframalpha.com/input/?i=distance+traveled+in+1...

No, the pretensioner just has to beat the deceleration of the car. The goal is to retain (or even pull in) the passenger in the seat to reduce the impact on the airbags.

Typically the pretensioner will pull in on the order of 10-12cm of webbing, this is done via a pyrotechnic charge. To be fully effective, the pretensioner would then need to deploy about 110-120 milliseconds before the airbags (which will deploy in 40 ms or so.)

So then it falls on the car's sensors and computer to detect with sufficient advance notice the impending crash. With modern radars (like the ones in newer mercedes cars...) they should be able to gain most of that time at least for frontal crashes.

Depends on what kind of crash it is. My dumb guess is that most crashes can be predicted by sensors in advance of the moment of impact, perhaps by significant multiples of 150ms.

For example consider a crash where your car slips on ice and drives right into a tree. Before you hit the tree there's gonna be all kinds of crazy steering and braking attempts that sensors could detect, and the accelerometers are gonna show weird lateral ice-slipping motions.

> most crashes can be predicted by sensors in advance of the moment of impact,

Provided you dont have any false alarms. For instance, someone making an unsafe pass on a 2-lane road, lets say a fast motorcycle. radar sensor sees incoming object with relative velocity of 150mph, see's driver slam brakes, yet motorcycle dives back into his lane at the last second. In this case, deploying the airbag without accelerometer input would actively harm the driver.

Once when I was on a walk, a car activated all airbags just before it passed me while being driven in the opposite direction. I was more or less looking at it all the time, so there was no external factors - just a very loud thud.

The speed was perhaps 25 km/h which was good, because the young lady in the car just kept driving straight ahead - no braking or steering or anything - until the car slowly hit the kerb and then a sign post 75 meters down the road and came to a stop.

There was no external damage to the car - or perhaps the fender was slightly dinged.

I walked there, while figuring out what had really happened and noticed that the driver was just sitting there in the car while it started to be filled with white smoke.

When I got there the car was more or less full of smoke, and the young lady still in the drivers seat looking straight ahead and appeared to be very confused.

Someone else got there a few seconds before me and opened the doors to the now totally smoke filled car, and the lady sort of woke up - and she got out just as I got there.

I think she was physically fine - no arms broken although that could happen when an airbag deploys - but she was really confused although she was starting to get a grip after a minute or so, and started to worry about the blown up interior in her car...

I guess it's one of those things you really don't know how you will react - she totally froze.

If that would have happened at any sort of speed, I'm pretty sure it could been a fatal accident.

Why did the car activate the airbags? What car is this? I've never heard of airbags deploying improperly before (except when we have a busted set that we use to launch a tire a million feet into the air).

I don't remember what car it was, it was somewhere between 7-10 years ago. Possibly a small Toyota, or another asian car.

I remember that I had to look it up at the time - and apparently a spontaneous deployment is not unheard of, but rather uncommon. Old news article: http://www.autonews.com/article/19980824/ANA/808240708/airba...

> I guess it's one of those things you really don't know how you will react

In my case of the hypothetical close-call, I think if I had a car with one of these "impending doom" tones, just hearing that tone in conjunction with a near-miss on the road would probably give me a heart attack.

I read "It takes just 150 milliseconds to deploy the reversible seat belt tensioners" as 150ms being the time to fully tension them (and then, possibly, for the case where they have to be tensioned most, for example when a passenger is leaning forwards), so some effect will be visible before that.

And useless? There certainly are situations where this will help, for instance when the driver is braking hard. Also, if you hit a wall head-on, a Mercedes will have over a meter of crumple zone. Pulling the passenger away from the car interior by the time that crumple zone has done its job will surely help, even if it only a few centimeters.

Seat belts also should not crush the passenger to death. That puts limits on the speed at which they can be tensioned (it wouldn't even surprise me if the weight of the passenger would be used to adjust the pretensioning force)

Yeah, I wonder if that's a mistake on the webpage. Some other site I found[0] says the decision to deploy tensioners happen at 0.015s after impact, and they "go into action" at 0.020s. Someone might have accidentally added a zero when converting units?

[0] http://www.takata.com/en/around/seatbelt01.html

No, they lock your seatbelt before you move. Modern ones do this during heavy braking: all you notice is that when you step on the brake your belt is already locked.

That's not the pretensioner. The pretensioner is a one-shot pyrotechnic device (i.e. explosive) that pulls in the seatbelt. It's only activated if the car senses a crash is happening.

It's an ambiguous term. You're talking about a pyrotechnic system that can only be used once, but there are also electronic systems that are a part of daily driving. My point is that the 150ms happens before the person moves at all, so 30mph is not part of the equation.

I defeat this by driving with my left shoulder far forward.

This is because it nearly killed me several times. When in a dangerous situation, I may need to brake and/or turn my body to look behind me. It's the "and" which is trouble. If I brake and then need to look, the shoulder belt might stop me, leading to a crash.

The fundamental problem is that restraints are designed for 100% passive crash dummies. Restraints are not designed for people who must actively control the vehicle.

>It will even do things like full field the alternator up to 18+ volts to slam the windows shut super fast in the event of a rollover.

I imagine this breaks the window mechanisms. However, you've been in a rollover, probably doesn't matter anymore.

How is this of benefit? All I can see it doing is chopping off my arm if I've got it hanging outside the window.

If it prevents a child from being ejected, it will likely save their life.

How does a secured child get ejected from a vehicle? Sorry, that just doesn't compute..

E.g. my six year old is in one of these https://www.britax.com.au/car-seats/britax-safe-n-sound-enco...

childen go through glass. Its about keeping arms restrained.

As someone who frequently drives with my arm out the window, that would certainly be concerning. I can see some considerable safety benefits to containing unrestrained people in the vehicle (rollovers tend to result in ejections, which have a high mortality rate), but at least from a management perspective, I'd much rather let a negligent person be ejected 50 times than keep 49 of them in the vehicle and have 1 sue us because we cut their arm off.

As long as their sensors correctly identify a rollover vs. another kind of accident, I think it's an acceptable risk. If you're in a rollover situation you're likely to have both hands on the wheel already. If your arm is out the window during the rollover it's not likely to fare well anyway.

Which hand signal (brake, left, right) do you most often give when having your arm out the window? Do you mean it, and follow through, or is this the old-fashioned equivalent of leaving your turn signal on?

Note: hand signals are still legal, still tested on driver exams, and (as I saw a few weeks ago) still actually used for real.

Why isn't the middle finger an option?

Your concern should be for the safety of your customers, not for whether a measure that drastically increases safety might lead to a suit in rare cases.

If your arm gets flung out the window when that side of the car is facing up then it could get crushed when that side rolls over.

If you're in a rollover you won't have great control over your arms in the first place, so if your windows are open there's a good chance they'll be going out the window at some point. If you watch professional drivers in accidents, they attempt to cross their arms across their chest to keep them from flopping around.

If your arm is hanging outside the window and the vehicle is actively rolling, you're likely going to lose that arm.

Otherwise, based on all the other logic described, the decision to roll the windows up probably takes into account if there's an obstruction (just like auto-up windows do).

I'm not sure of the official reason, but I'd imagine it prevents ejections (or worse, partial ejections) from the vehicle.

I was the first person on the scene of a rollover where a woman was pinned underneath the car (fortunately only suffering broken bones) because she was partially ejected through the open window.

After extensive testing it was found that a clean amputation provides a quicker recovery than a degloved, torn up limb. There were more amputations that would be ideal, but net morbidity was reduced.

Source or sarcasm tag. Needs one or the other.

Come on, it's completely obvious that it's the second.

It will help keep your head from bobbing outside the vehicle as it rolls, preventing your death from a crushed skull.

Edited to add:

"partial ejection" are the magic google words. Here's an example:

"It was found that, in standard impact tests on high containment barriers, partial ejection of the head through the side windows occurs systematically"

from: https://trid.trb.org/view.aspx?id=750825

"Keep your arms and legs inside the vehicle at all times."

Motor & gear mechanisms usually have a lifetime that decreases at something like voltage^3. So you can double the voltage and get 1/8 the life. Since they have to be rated for thousands of cycles normally, you can overpower them a lot for a single cycle.

It sounds like it also may cause damage to other electronic components. They essentially switch the alternator into different mode that doesn't limit the output voltage in order to generate as much energy as possible.

I can't imagine them running a separate wire in the harness only for this one feature. It makes more sense to simply increase the voltage of the 'mains' 12V power inside the car, dump as much energy into it as possible, and design all critical components to handle the event gracefully. Sacrificing the less important electronics sounds like a viable option during a rollover event.

Automotive components are already designed to be very robust to input voltage transients.

Here's an interesting paper on the conditions automotive electronics are exposed to, including very high voltages during a load dump: http://www.ti.com/lit/an/snva681a/snva681a.pdf

Anybody that's accidentally dumped a capacitor into the car's electrical system when trying to install a subwoofer can attest to this. I was genuinely shocked anything worked in my car when I did it.

> If I had to be in a wreck, it would be in an S class. Hands down

I do wonder why rally and track cars have roll cages with mandatory helmets, windows made of plastic and prominent switches for cutting the electrical systems and fuel. This is in the lower racing classes that have otherwise standard road going cars. Rally cars go on dangerous tracks, track cars go fast on safety standards approved tracks. Road cars combine the ability to achieve dangerous speeds and they go on roads as dangerous as the rally cars. At 155 miles an hour you are far from guaranteed to be safe in an S class. I would go for the roll cage, 4 point harness, Hans device, helmet and fireproof suit if I was to thi

> If I had to be in a wreck, it would be in an S class. Hands down

That way of thinking aside, Mercedes are the true pioneers of safety even if the liked of Volvo claim more credit. Fantastic as these innovations are, we need to restrict vehicles to their safety ratings. This then means that the safer cars are quicker, the S Class becomes quickest on the road, Tesla aside...

I can't find the article but I read that you're more likely to experience a head injury as a pedestrian in a major city than a cyclist, and yet I always wear my helmet on my bike and never when I'm walking around.

I guess it's about finding that balance between protection and, what, fashion? I don't know.

Vehicles with rollover cage are forbidden from driving on public roads in most of EU.

> If I had to be in a wreck, it would be in an S class. Hands down.

Most "car guys" you talk to would say the exact same thing. The luxury flagships are super-interesting from a technology standpoint. You can look at them and see what new features will be coming to Camry's and Taurus's in the next decade or so.

All that effort to protect the people inside, and they still won’t try to do anything to prevent them from running people over.

I wish car safety was also based on what happens to people outside the car.

People inside a Mercedes are more likely to be Mercedes customers than those outside. They're even more likely to become repeat customers if they just totaled their car and walked away from it without a scratch.

We need to make car manufacturers pay for externalities if we want them to care about people outside the car. They won't do it on their own.

The interesting bit:

If an impending collision is detected that would be expected to produce a loud crash, the vehicle’s sound system plays a short interference signal. This causes the stapedius muscle in the ears to contract, which for a split second changes the link between the eardrum and the inner ear and so better protects it against high acoustic pressures. Most importantly, the reflex reduces the damage to hearing.

Here's the sound: https://www.youtube.com/watch?v=bCOw5lisrIU&t=25s

I might be imagining this, but I could swear I felt my ears becoming "full". Sort of like tired of the noise and numbing down? It's hard to describe but it felt similar to being on an airplane for a long time where sound becomes kind of dull because of all the engine noise.

You are not imagining it. (I guess it depends on the definition of "imagining", but that's philosophy.) It's definitely a real effect.

You aren't if you use headphones and play it at close to 80db the feeling is like when you put on the active bose noise canceling heaphones and turn them on.

You feel a slight pressure on your ears which is what this intended to do.

You might enjoy this site then, which has several different noise "colors", each has a different kind of effect or feeling.

http://onlinetonegenerator.com/noise.html

It had no discernable effect on me, in a fairly noisy domestic setting including a washing machine in the background. I was expecting it to be much more penetrating.

Perhaps I've ruined my hearing after years of listening to shortwave white-noise.

I felt that as well.

I wonder if it as cool hypothesis that they decided to implement as a cheap bit of marketing, or whether the protective factor is actually evidence-based.

That's a bit cynical of you, isn't it?

Also, would you say the same thing if it was another car company that did this?

> Also, would you say the same thing if it was another car company that did this?

Why would I have something against Mercedes?

Did you actually read the article, or are you just here to snark on HN?

Here, let me do your research for you. First link on a search for "stapedius reflex" is https://en.wikipedia.org/wiki/Acoustic_reflex

A quick search on Google Scholar for "acoustic reflex noise" turns up this paper from 1962 which states in the abstract that the acoustic reflex is more easily triggered by wide bandwidth noise than a single sinusoidal tone. http://scitation.aip.org/content/asa/journal/jasa/34/9B/10.1...

I think we can safely say that this work is pretty evidence based.

Well, I read the marketing piece, yes - but thank you for pointing me at the evidence. The level of 'snark' may have been influenced by me previously reading the Soylent story. Marketing masquerading as science is certainly a thing, though not in this case.

A friendly reminder: "a quick search on Google Scholar" is not research.

That paper you mention is just a functional description of the "accoustic reflex". I don't think the parent comment was putting the existence of the "accoustic reflex noise" in doubt; rather, and I agree with him, it's its effectiveness to protect the driver in a car crash what is in doubt.

The page makes a reference to a human trial in 2011. It would be great if they included a footnote with the reference to the actual paper.

It’s not "research research" but it is "does this phenomenon actually exist in the literature" research.

Insisting on testing the phenomenon in an actual car crash seems superfluous to me. A car crash produces a loud noise that can potentially damage hearing, which can be alleviated to some extent by triggering the acoustic reflex with a quieter noise immediately before impact. So long as the generated noise is well below the threshold of damage then I’m really failing to see why more specific research is required in order to believe that this would be an effective intervention.

The bit about the noise from a car crash being loud enough to damage hearing after a one-time short exposure is what I doubt. Sure it's loud, but does a split second exposure to it really do permanent damage? I don't know what the answer is, but I am skeptical about the usefulness of this feature.

Well, consider that an air bag is basically a bomb going off in your face. And I'm pretty sure explosions from up close cause at least temporary hearing damage.

> Deployment of a driver’s side front airbag will generate mean peak sound pressure levels of approximately 160 dB (decibels).

> Studies have shown that the pain threshold from noise is about 140 dB and that a single exposure to sound pressure of this level can cause permanent, severe hearing loss.

> The deployment of a side airbag generates a mean peak sound air pressure of 178 dB.

So yeah ... a car crash will definitely mess with your hearing permanently.

[1]http://hearinghealthmatters.org/hearinginternational/2012/th...

Stand by some one at the range without ear pro and ask them to shoot one shot.

The sound is only a fraction of a second but that ringing you'll hear is the death sound of the frequency you'll likely never to be able to hear again.

You also need to understand that sound and loudness isn't what causes hearing loss it's the pressure shockwave.

If a blast "noise" is mostly outside of your hearing range the pressure will still do permanent damage.

When a bomb goes off most of the energy can very well be outside of the normal frequency range of the human ear, it will still however rip your ears apart.

that ringing you'll hear is the death sound of the frequency you'll likely never to be able to hear again

or you'll hear it all the time

Any noise over a given decibel threshold damages hearing.

I think the problem in car crashes is probably the airbags though: airbag inflation is very loud. One article I found ( http://www.sae.org/standardsdev/tsb/cooperative/airbag.htm ) suggests 170dB, vs 140dB for the car crash itself.

170dB is well within the 'can cause permanent hearing loss' range. 140dB is 'this hurts' (and probably causes hearing loss).

Yes, sorry - to be clear, I wasn't questioning the existance of the actual reflex. More the protective effect - however I understand that my post probably wasn't worded very well.

FORMER LIBRARIAN:

> "a quick search on Google Scholar" is not research.

Yes it is the start of the process. Stop being snarky please.

Here is another Librarian approved good start: https://en.wikipedia.org/wiki/Acoustic_reflex

There are plenty of sources listed on the bottom.

From https://en.wikipedia.org/wiki/Acoustic_reflex#Protection_rol... :

  According to the article Significance of the stapedius
  reflex for the understanding of speech, the latency of
  contraction is only about 10ms, but maximum tension may
  not be reached for 100 ms or more.[6] According to the
  article Le traumatisme acoustique, the latency of
  contraction is 150 ms with noise stimulus which SPL is
  at the threshold (ATR), and 25-35 ms at high sound
  pressure levels. Indeed, the amplitude of the
  contraction grows with the sound pressure level stimulus. 10]


  Because of this latency, the acoustic reflex cannot
  protect against sudden intense noises.[10][6] However,
  when several sudden intense noises are presented at a
  pace higher than 2–3 seconds of interval, the acoustic
  reflex is able to play a role against auditory fatigue.[10]

I did not bother to check more than the abstracts of the references, but it does seem like evidence exists.

Thanks. Enlightening.

I have read somewhere of a click being played through tanks' crews headphones when the main gun is fired, for this reason. This was wrt 70s-era Soviet tanks.

Forwarded this onto my dad who is an ENT specialist. He responded:

"The Stapedius reflex has been use for artillery gunners for decades but the pink noise use is certainly novel."

I was going to ask if this would work at the gun range

I worked with an audio engineer on real-time video encoding hardware and he said that in audio compression, you can throw away audio data immediately after a loud noise because the ear can't pick it up. This is probably an oversimplification, but sounds like a similar technique in this safety feature. I like it.

Interesting! Can you completely discard it (as in, complete silence immediately after) or just compress the immediate bits after a lot more than normal?

I'm guessing that the data couldn't completely be silenced because then any heavy dynamic compression at playback would bring out the optimization and break the sound.

Sadly I don't know the answer. I was just the GUI engineer fascinated by the cleverness of how to compress audio.

This is actually a big part of how MP3 compression works. Rather than using straight up data compression, it uses a psychoacoustic model to simplify different components of the signal that are affected by masking.

There's a nice high-level (human readable) technical overview here if you'd like to learn more: http://arstechnica.com/features/2007/10/the-audiofile-unders...

I should add, as another side note / rabbit hole to those new to the art of 'fucking with brains through audio' — masking is also frequently put to use in mixing, particularly EDM and more percussive genres.

To help squeeze every last drop of perceived dynamic range out of a track you can use a technique known as sidechain compression (http://samplesfrommars.com/blogs/tips-tricks/18999227-how-to...) to dynamically compress other components of the mix in response to (usually) a kick drum.

Is this what causes the volume to dip a bit after strong bass/drum hits?

Yes, exactly that; the compressor turns down the rest of the mix as it is triggered by the kick drum (which isn't going through the compressor). Easily heard at the beginning of "Call on me" by Eric Prydz [1] if you concentrate on the synth string sound. Lots of people think this is the only use of it (it's not), but it's probably the most popular.

[1] - https://www.youtube.com/watch?v=L_fCqg92qks

Thanks for that informative link.. I would say the advice to simply re-write a musical composition so that bass and kick don't happen at the same time seems to put the cart of mix aesthetics before the horse of musical intent! :)

"Sideband compression" sounds like an interesting dsp project! But I think the term is "sidechain compression", right?

Yep, nice catch. Edited.

Re-arranging is definitely one 'solution' and the only real approach when you're talking music played live on acoustic instruments. It's one of the elements of good orchestral music I find most fascinating - that many sources of noise, each playing their part and either perfectly blending or completely juxtaposing one another.

I'd argue that using all the various forms of DSP at your disposal when working with sound in the recorded / amplified sense (and especially electronic music) is an equally valid aesthetic choice. Different limitations, different capabilities and difference aesthetic possibilities for different mediums.

In electronic music, mix aesthetics is quite often exactly the musical intent - you're trying to create a track that sounds a certain way first, and think about melodies and harmonies (if you even have them) second. You would often pick specific key and bassline because of how low frequencies (55-110 Hz octave and below) sound in the mix on your intended equipment. You would even distinguish between open-air and club tracks sometimes because of how open space affects sonic qualities of your mix.

Sidechain compression doesn't do anything more for overall dynamic range than regular compression. Rather it provides a channel for some audio to completely bypass the compressor while the rest gives way to it. The point of it really is purely stylistic, the fact that it reduces the overall dynamic range is incidental.

Keyword there was perceived dynamic range. Completely agree that it's a stylistic choice, the goal of which is usually a 'huge kick/bass' etc. It's simply a technique to create that illusion without needing to increase actual dynamic range.

A great example of this technique being used stylistically is Titanium by David Guetta (ft. Sia) - listen for the kick drum masking out all of the other instruments in the mix at the end of the chorus. In this case, it's used more as an obvious effect than a transparent mixing technique (as compressors usually are).

Doesn't all lossy compression use models of human sensory processing to determine which information is important?

Yup! This is called temporal masking [1].

There's also simultaneous or frequency masking. If you've ever wondered why loud noises prevent you from hearing something that would otherwise be perfectly audible (e.g. a loud motorcycle driving by your apartment can temporarily drown out a noisy air conditioner), that's why.

Both of these psychoacoustic concepts are used in perceptual audio encoders like MP3 - the idea being that certain bits can be discarded because the listener wouldn't be able to hear them anyway.

https://en.m.wikipedia.org/wiki/Auditory_masking

This is why I only listen to music in PCM 24-bit WAV's instead of compressed audio. The difference is huge due to the missing frequencies in MP3 and similar formats.

Dear everyone,

Mercedes is a good marketeer and with best intentions they included protection of hearing in their campaign. I wouldn't blame them for not knowing and after having spent lot's of money on this it could be dissapointing to hear that in fact, the acoustic reflex arc, otherwise known as the stapedius reflex does not protect hearing against loud sounds. It is simply a myth that even among many professionals is believed.

The facts are that the ear's transparency for sounds is traditionally tested with only a single frequency, around 226 Hz, and around that frequency the sound levels get reduced by the acoustic reflex. Thus it seems as if the ear protects itself for loud sounds and the myth is born although it has only been measured at a single frequency.

Technology to measure the same prinicples for a wider frequency range is currently only available for a selective group of researchers and for the industry developing medical instruments in this direction (for which I work). We know that above 1000 Hz the stapedius reflex is doing the opposite of protection. It will increase the sound levels.

I could explain why the reflex is doing this, why did it evolve like this during the last many thousands of years. It is certainly easy to understand that a couple of hundred years ago (before industrialization) no protection for loud sounds was actually needed and thus hearing protection is not a logical candidate for explaining why we have such acoustic reflex...

Though this is not the point. The point is that Mercedes' feature is not protecting your hearing during a collision impact. They made it with best intentions and for that they deserve a compliment and credits.

A little more info about the acoustic reflex that I consider to be correct can be found here: https://youtu.be/3a3Eeuhkh-c?t=1164

Feel free to try to get in touch with me as well if you want to find out more about this topic.

Cheers Jos

Mildly interesting: when I played the sample pink noise in the article, my dog went mental.

Of course he did. Poor thing, he realized you were going to be involved in a car crash in the next 100ms.

(sorry couldn't resist :-))

Mine didnt - she continued to gaze at me wondering if I had any food.

I wonder if that noise is going to be a major ptsd trigger for people involved in an accident afterwards. (Of course probably any loud noise would be). It will be interesting to see any data they can collect out of this on the benefit in practice.

Fascinating stuff. Kudos to Mercedes-Benz for pushing the envelope and making driving safer.

The pink noise in the video has some strange dropouts and is compressed. If you want to listen to real pink noise I suggest this: http://stenzel.waldorfmusic.de/post/pink/ Source Code: https://github.com/Stenzel/newshadeofpink disclaimer: I made this.

I often listen to pink noise while I work to suppress room noise and mask tinnitus. Ironically, I find it is also good for relieving headaches! :) I have a CD with hour-long tracks of just white noise (sounds like static), pink noise (sounds like waves), and brown noise (sounds like airplane hum).

http://whitenoisemeditation.bandcamp.com/

You can also use SoX!

`[audey@feather ~]$ play -c2 -n synth whitenoise lowpass 400 highpass 40`

Sounds more like white noise to me. What makes it pink??

I don't know really know what this means tbh, but here's an explanation: http://www.acousticfields.com/white-noise-definition-vs-pink...

It's white noise with a different power distribution; low frequencies are louder:

http://i.stack.imgur.com/MMr0a.png

Redshift

The sound is distinctly crisper on BigQuery, though. Of course, cables can make a difference but not always -- use CAT-8.2 if possible.

Because humans hear frequency in log scale, pink noise has more "even" distribution of amplitude to our ears. Pink noise -> 100-200hz has the same power as 10k-20k ie each octave has equal power AFAIK.

White noise is defined as having equal energy per frequency; that is, a 10hz wide slice of spectrum centered on 100hz will have as much energy in it as a 10hz wide slice of spectrum centered on 1khz.

Pink noise has equal energy per octave; the slice of spectrum from 40hz-80hz has as much energy as the slice from 80hz-160hz and the slice from 1khz to 2khz.

Pink noise can be thought of as white noise with a very low-q (3dB per-octave, specifically) low-pass filter applied.

Reply on news.ycombinator.com