Hacker News Reader

Comment by spectramax | Hacker News Reader

Btw, the new "fix" for the MCAS has just been announced and discussed by a 40-year experienced pilot [1]. Turns out that while we shouldn't be doing armchair analysis and assuming the worst of the engineering, management and executive team; the software fix released appears to be rather elementary and MCAS should have been designed with these safety checks in the first place. Why it wasn't is a huge concern and we wonder what other systems are at risk.

In summary, the software fix does the following:

1. Use inputs from both AOA sensors, if they disagree by 5.5 degrees, disable MCAS. Original MCAS system used only 1 AOA sensor and switched back and forth between the two after every flight.

2. Triple redundant filters, A) Average value reasonability filter B) Catastrophic failure low-to-high transition filter C) Left vs. Right AOA deviation filter

3. Limiting MCAS stab trim so that the elevator always can provide 1.2g of nose-up pitch authority for recovery. Furthermore, electric trim with the yoke switch will override MCAS.

Turns out that armchair analysis was sort of on point and goes along with the incompetency at Boeing affirmed in this reddit post - sometimes, I wonder how we, humans, collectively build extraordinary monuments while we individually rest on stilts.

[1] https://youtu.be/zGM0V7zEKEQ?t=370

View on HN

The software fixes as described sound sensible and it's hard to believe they were not in the original design. Unfortunately I don't see how these changes will allow them to keep a common type rating with the 737 NG. After the updated MCAS has made its one shot trim adjustment you will now be flying an aircraft with different handling characteristics if you keep increasing the angle of attack into the stall. Surely the FAA can't sign off on the common type rating and half hour of video training ??

Having a new type rating is a fair punishment, I think, for Boeing and the airlines who have been stringing along this obsolete airliner for way too long.

Note the 737 MAX was rushed out not just to fight the A320neo but also to counter this kind of jet:

https://en.wikipedia.org/wiki/Airbus_A220

which is cheaper to run than the 737 MAX but also has much better passenger comfort than the 737. I was in the similar

https://en.wikipedia.org/wiki/Embraer_E-Jet_family#E175

and even though it is a smaller plane, it feels more like riding in a widebody than in a 737.

Canada developed a fully modern small jet on it's own dime and taking risk, then Boeing nearly killed it by offering airlines a 70% discount on the 737 MAX. Airbus bought a controlling interest in it for free because that was the only way it could survive.

Common type rating is a big reason why airlines are sticking with the 1967 era 737 rather than upgrading to modern planes that will make flying much more pleasant, make less noise, use less fuel, less climate change impact, etc.

>then Boeing nearly killed it by offering airlines a 70% discount on the 737 MAX.

Actually, Boeing lobbied/sued to get punitive tariffs levied on the Bombardier CSeries. While they ultimately lost the case, while it was pending, Bombardier entered an agreement with Airbus to finish assembly of the planes for American customers as a US-based Airbus factory as a way to get around the tariffs. https://en.wikipedia.org/wiki/CSeries_dumping_petition_by_Bo...

You're right -- but Boeing provoked Bombardier into "dumping" its aircraft by offering that 70% discount!

Do you have any links? Because if true that was a clever tactic, dump your own products to force the competitor to price dump too and then sue them for price dumping.

See The Economist cartoon for this week: https://www.economist.com/the-world-this-week/2019/04/13/kal...

A slightly different event, but when Embraer accused Bombardier of being illegally subsidized by the Canadian government, Bombardier said "no u" and made the same accusations against Embraer. IIRC both companies ended up getting fined

Wasn't Embraer bought by Boeing?

Boeing has a marketing partnership w/ Embraer, but may have a hard time getting the Brazillian government to let go of it.

Not yet.

> then sue them for price dumping

you don't think Boeing has a few attorneys on staff providing legal review for its business decisions?

A new type rating wouldn't be punishment, but the bare minimum of common sense. It only seems like it would be unreasonably onerous for Boeing (et al) because Boeing set themselves in this position to call the FAA's bluff TBTF-style.

What should really happen, in the free market creative destruction paradigm, is that the entire design should be scrapped - designated as non-airworthy by the FAA. Since the design goals seem to have revolved around hacking FAA regulations, as opposed to good faith engineering, how can any of the results be trusted?

A reasonable pick-up-the-pieces approach would see the MAX grounded for a minimum of three years for an overall design review with no getting out of a new type cert, with production halted in case hardware changes are needed.

But what's reasonable can never happen politically, because Airbus, headquartered in that entirely non-allied and utterly foreign country France, would benefit. Mon Dieu!

Which is the general setup for Too Big Too Fail - ignore regulations/prudence, skim profit short term profits, and then make sure it is everyone's problem when the sham inevitably blows up. The Boeing executives that forewent actually designing a modern plane are the ones responsible for not having kept up with Airbus, and should frankly end up in prison for mismanaging a TBTF (ie quasi-government) entity.

The E175 is even better than a 787 in 9 seat configuration. It’s a superior experience in economy, and I hope every 737 and CRJ is replaced by it.

I was blown away when I experienced it. If you had an airline that could always put you on something better than a 737 and have a better cost structure... You'd put the other airlines out of business.

I'm surprised because I incorrectly assumed passenger comfort was mainly correlated with the mass of the aircraft. I.e. small planes get blown around more.

I've been in the Bombardiers, and they are as you said a surprisingly comfortable plane, much more so than the 737s.

I'm not sure it's fair to put Embraer in the same category, it is a much less comfortable plane, both in size and handling in my opinion.

How is the 737 obsolete?

The 737 predates the 747 and the Concorde, and both of those are out of passenger service.

The circular cross-section, for one thing, makes it like the old American cars that were huge on the outside and small on the inside. The only way you'll really understand this is ride on an modern plane such as the E175 or A220 -- you'll get that you hate to fly not because you hate to fly but you hate to fly on a 737 and never get to fly on anything else.

The overall design of the aircraft is not compatible with high-bypass engines. Thus they bolted MCAS onto it. Even there the 737 Max does not support the geared turbofans that are available for modern airliners and the A320. These are so quiet that it isn't that you hate living near an airport, it is that you hate living at an airport where 737s land.

Modern airplanes use fly-by-wire which reduces weight (sometimes dramatically because the size of the control surfaces can be reduced) and enables flight envelope protection (like MCAS) in a way that is a lot more rational and thought-through than MCAS.

An all-electric plane like the 787 might also improve air quality on the flight. I don't think small airplane competitors have this yet.

Boeing has used it's market power to suppress alternatives to the 737 so you will pay more for tickets, suffer in transit, have noisier airports, and have more global warming. The MCAS system was just one part of this scheme.

The 737 Max is a case of the "undercompetition and underinvestment" connected with secular stagnation; it is like how your internet service sucks...

Why is this downvoted? It is 100% true.

The 737 has the _fundamental_ problem of having wings very low to the ground during ground operations. Taller landing gear is not an option, in fact the 737-max has slightly-extending landing gear to resolve this issue. It is part of the reason the plane is called "max" - they maxed out things like ground clearance, fuselage length, wing chord, and other parameters.

Back to the low wings: there is no room for a large (quieter, more efficient) engine. In the earlier 737s, the engine was just as high as it could go, you could even see the cowling squished at the bottom. In the max, the larger engine is placed in front of the wing, which means that it sits much further forward than the airframe was designed for. Thus, MCAS was needed to have the plane "behave in flight like the old 737".

These are _fundamental_ issues that cannot be fixed on the current airframe. Just as the 2019 Ford Mustang chassis is better (stiffer, lighter, safer) than the 1964 Ford Mustang chassis, newer airframes resolve many of the issues that were not designed into their 1960's counterparts.

> Taller landing gear is not an option

Why?

Because the plane wasn't designed that way in the 1960s, that's why. They can't change it without designing a different chassis, which means it's a different plane and a whole new type rating that pilots have to be trained for and certified on.

Nowhere to stow it.

747 is out of domestic passenger service in the U.S. but it's definitely still used a ton for international passenger service including to/from the U.S.

It's really old, and the main problem is that it's avionics are old, and don't have the same level of electromechanical flexibility that new planes.

Think of the difference between slapping IoT / real-time monitoring capabilities onto an old factory process versus building from the ground up with those capabilities kept in mind.

Obsoletion happens when there is a better alternative available, and there are plenty of better options to the aging 737 airframe.

> It's really old

The design has been iterated and improved many times. The field of aviation is very conservative. What's old is generally what works. New designs are prohibitively expensive. The 787 won't be profitable for some time, even a decade after it was launched.

> it's avionics are old

They're brand new. Where do you think they get them for brand new planes ?

>versus building from the ground up with those capabilities kept in mind.

They have done that, just the other way around: the avionics are built from the ground up with the capabilities of the well-understood (at least before -Max) aircraft in mind.

> > it's avionics are old

> They're brand new. Where do you think they get them for brand new planes ?

I have some industry experience- there are still new planes that get large, ugly beige Honeywell units in them that're slow and clunky while there are much better avionics packages available. The Honeywell packages are generally pretty newly built, but are still really old. The Raspberry Pi Foundation still sells the original Raspberry Pi, even though they've gone through 3 different generations and like 6 iterations since. The original Raspberry Pi is old, no matter when it was manufactured. The Raspberry Pi 3B+ is new, and has more features and power. Sure, they have similar names and functionality, but there is a world of difference between the two.

Imagine that you've designed a product around the original Raspberry Pi. The projects needs include GPIO, internet connectivity, and video output. The expectations have been trimmed appropriately to fit within the hardware available. Custom PCB's have been made for the Pi to socket into, and there's a standardized system in place for testing the Pi.

Later on, you decide that you want to replace the old Pi with a new one. Compared to the current Pi, the old one has less GPIO, less processing power, a weaker GPU, lamer internet, and less IO- and, although you can just drop in the new Pi with an adapter cable, you're still missing out on all the pins, and your code isn't exploiting all the available resources.

This "not everything gets updated" cycle has serious compounding effects the longer you go. After a few decades of Raspberry Pi releases, your product is a little bit antique, and needs some serious updating to be able to compete in the now-mature product category that you were in. Not starting over and working from the ground up has serious knock-on effects in the future, (for instance, the 737's wings were way too low, which is not future-proof for bigger engines).

This isn't just about avionics, either. It's about every single part of the airplane.

I don't see how anything is wrong with "large, ugly beige" electronics. 737's fly everyday just fine. What benefits are they going to get from sleek, thin Jony Ive-inspired UX? Are they going to get to destinations faster? Are they going to have a smoother ride? What product requirements could possibly change that would necessitate upgrading the avionics?

All these high-tech upgrades have downsides. Fiber optic is very slow to be adopted in aviation because movement generally destroys it. those fancy HUD obscure the runway and lighting. AIRDRU's malfunction, sometimes spectacularly [1]. Sometimes they kill people when working exactly as designed [2]. I'm not surprised to see a fetish for the fastest, sleekest electronics on Hacker News, but I'm shocked to suddenly see such sharp criticism for the '37 on here (aside from the Max issue). It's one of the most reliable and safe aircraft ever, yet somehow it needs the latest glass cockpit technology.

[1] https://www.popularmechanics.com/flight/airlines/a26854898/p...

[2] https://en.wikipedia.org/wiki/Air_France_Flight_296

Your original claim wasn't "how anything is wrong with" the electronics, but whether they were old or not. You're moving the goalposts.

If you look at it, Boeing has put improved technology in all of the other planes it makes, just not the 737. For instance they have retrofitted the 747 with fly-by-wire.

Fly-by-wire offers huge avionics benefits, reduces weight, fuel consumption, etc.

Planes with fly-by-wire also have inertial measurement units that contribute to envelope protection so they would have another input to work from rather than just the AoA vanes.

the A220 is really meant to be an DC-9 Series 80/90/95 (MD-80, MD-90, MD-95/717) replacement, and doesnt match the 737max in either range or capacity.

Yes it is smaller, so airlines can run more frequent flights. The range is not so good, but there an awful lot of flights the 737 flies that don't go to the full range.

>The software fixes as described sound sensible and it's hard to believe they were not in the original design

Without the benefit of hind sight it's not obvious that it's a better decision. MCAS being turned off changes the flight characteristics of the plane. If you do that silently you might run into a situation where an aircraft stalls because a pilot is used to having MCAS correcting their input.

If both aggressive MCAS is unsafe because pilots can't control the plane and nonaggressive MCAS is unsafe because of stall risk that would seem to imply that 737 MAX should not be certified to fly.

An always-on system designed to always override the pilot should be at least as reliable as the control system used by the pilot. Because the 737MAX is supposedly flies like a 737, and because authorities signed off on this without mandating pilot training, this means this system needs to be as reliable as it can - to maintain the illusion.

And basically everyone knows that there are some subfields of software engineering where safety is critical. From nuclear reactor/warhead control to healthcare devices, and somewhere right in the middle there should be aviation, and other transportation industry related software (train signaling, braking systems, self-driving whatevers, and so on).

And we have Boeing, acting worse than Uber (when they put someone into a Tesla to watch it self-drive without hooking that someone up on intravenous Modafinil).

Do the proposed fixes do it silently?

> Original MCAS system used only 1 AOA sensor and switched back and forth between the two after every flight.

Who the fuck thought that was a good idea???

Let's make it 10x more difficult to detect that there is a problem.

If you're looking from an engineering PoV that's absurd.

If you're looking from a regulatory compliance PoV towards optimising sales - that's actually desired behaviour - if it's overly complex to diagnose it's less likely that somebody understands and point fingers back at us. You know, basically what happened after the Lion Air crash - Boeing being all over "it's not us, it's them"

Just to explain a bit more: It uses the main pilot side AOA sensor, and pilots switch seats every flight.

I can’t speak to foreign operators but in the US that most certainly does not happen. The captain flies in the left seat and the first officer flies in the right seat. Turns are taken on who is flying and the automation is selected to the side of whichever pilot is deemed the pilot flying.

> Turns are taken on who is flying

That's what I meant yes

Each side of the airplane has a set of sensors. The ones on the left (while sitting on the plate) are named Captain <sensor_name> and the ones on the right are named First-Officer (F/O) <sensor_name>. The MCAS software would switch from Captain AoA vane to F/O AoA vane between flights. The location where the pilots are sitting doesn’t affect how the airplane is flying itself. The person flying the plane can be sitting on either of the two seats, but the most senior pilot usually sits on the left.

It's not that the MCAS software would switch. It's that the sensors feed into two different flight control computers (Captain and F/O). And which FCC is primary would switch after each flight. Both FCCs run the same software, but with different inputs from the different sensors.

Does it switch automatically after each landing somehow, or because the pilots tell the plane which FCC is going to be primary for the flight?

So switching seats during the flight would have solved the problem?

That sounds so unlikely that most people would dismiss it as silly...

I'm not a pilot, but if the controls are linked -- why would the plane need to know where the pilot is sitting (or even have a switch for it)?

(Any 747 pilots reading? Can you please fill in the gaps for me?)

Not a 747 pilot (not sure why that matters here)... but a pilot. The flight positions are equal. The plane is designed to be flown the same from either flight position and this is routinely done. The guy saying otherwise is talking out their ass.

Thanks for the details :)

I said 747 because I was thinking Boeing at the time, and that was the first of their aircraft I could think of!

I was also thinking something smaller like a Cessna might not have the same setup because they're used for very different things (the air travel equivalent of a bus vs. a car).

Not a pilot either. Perhaps to show different information to the pilot flying vs. assisting? E.g. only the pilot assisting needs to see checklists?

A busted AOA sensor is going to set off a whole bunch of alarms since it's used in air speed and stall calculations. It's not something that is going to be missed with or without MCAS.

Unreliable hardware fails unreliably. It's entirely possible to have all your checks in place, and still not be able to detect that the sensor is faulty.

Independent comparable sensor input is invaluable and irreplaceable.

The plane might not be able to identify that the AoA sensor is wrong, but the pilots will surely know something is wrong. They will be getting stall warnings and airspeed warnings. Again, maybe they can't say it's the AoA sensor but they will definitely know something is wrong.

You're assuming all failures are detectable by a computer. That's not necessarily the case.

I was wondering the same. What is the logic behind this setup? That basically means that it is no better than having one sensor. Because if any one of them breaks, you have broken sensor active on next or following flight. Also I recall there was something said about Lion Air crash that the crew on the day before experienced some flight control problems which was not communicated to the crew involved in crash? Did they use other sensor on the day before and that was faulty too, or isn't that is how it actually worked?

I've heard it suggested [1, 2] the original design for MCAS could only make a small adjustment a single time, which the pilots could easily overcome with the stick, classifying it a less safety-critical component as it couldn't possibly crash the plane.

Later, the MCAS ended up adjusting the trim multiple times, by an amount the pilots couldn't override with the stick. Thus it should have been safety level DAL-A (loss of plane) [3] - but the safety analysis and MCAS design weren't updated to reflect that.

And the FAA, in what with hindsight seems like a ridiculously naive move, allowed Boeing to self-certify their own faulty safety analysis.

[1] https://www.seattletimes.com/business/boeing-aerospace/faile... [2] https://news.ycombinator.com/item?id=19419454 [3] https://en.wikipedia.org/wiki/DO-178B#Software_level

It's worse than having one sensor, the chance that one of two breaks is larger than that just a single one breaks. And by alternating you ensure that the broken sensor will be used.

Was talking to my flight instructor about twin engine planes. He was adamant that they were LESS safe than single engines because 1) you are twice as likely to have an engine failure and 2) the statistics indicate (non-intuitively) that an engine-out in a twin is more likely to lead to fatalities. He said the insurance companies used to think twins were safer but they have adjusted their rates after looking at actual data. I was fully on-board with point #1, but #2 was a surprise to me. I suppose that's what Burt Rutan tried to address with the Boomerang design.

For my commercial MEL training, most of the time was spent on assymetric thrust handling due to a single powerplant failure. When both are working, or neither are working, the plane handles about the same as a single engine.

When one engine fails, while you've lost 50% horsepower, you've lost more like 90% of excess horsepower which is what you need to accelerate and to climb. And assymmetric thrust is dangerous at slow airspeed. You must react exactly correctly and quickly. Failure on takeoff roll means an immediate abort. Failure at takeoff with runway remaining means cutting power and landing on what remains. Failure once airborne and you must quickly identify the failed engine, apply full opposite rudder (toward the good engine), if you don't do this while in a full power low airspeed climb the airplane will absolutely roll over. Many accidents are the result of continuing when they should have aborted, and being anemic or slow with corrective rudder.

Turbine powered airplanes are way different, they have a lot more excess thrust available so powerplant failure is not dire. They have so much more power that they will still easily accelerate to a safe single engine takeoff speed and climb away for a circle to land, which is safer than aborting after certain speeds that smaller twins won't get to on takeoff. That's all part of the computation for "balanced field length".

Yeah, there's a pilot proverb about how having a second engine "helps you get to the crash site faster".

Or "the plane is certified such that the remaining engine has enough power to take you to the crash site". Indeed.

As outlined above, you double the probability that a mechanical problem occurs. (For p small, 1-(1-p)^2 is around 2p).

Now, if the pilots are very much on top of things (well trained and current), then they can control things and safely land at the nearest suitable airport. However, if they make a mistake (and that's surprisingly easy [1]), it could easily be fatal.

A single-engine plane with an engine failure, in contrast, becomes a bad glider, with a decent chance of surviving the ensuing forced landing.

[1] For example, the TransAsia ATR 72 twin turbo prop clipping a bridge in Taiwan after the crew erroneously shut down the "good" engine.

https://en.wikipedia.org/wiki/TransAsia_Airways_Flight_235

https://www.youtube.com/watch?v=RSoZOsIC25M

Just want to point out that "shutting down the 'good' engine" isn't some isolated incident - that was a contributing factor to the British Midland Flight 92 crash [0] as well.

0: https://en.wikipedia.org/wiki/Kegworth_air_disaster

Would it be possible to make single-engine planes usable for commercial passenger flights with similar economics?

To your second point, my assumption as a non-pilot is that a symmetrically balanced glide is more controllable than lopsided thrust that could put you into a spin?

Loss of one engine at low speed will leave a twin uncontrollable. See Vmc (https://en.wikipedia.org/wiki/V_speeds?wprov=sfla1).

If you promptly shut off the other engine, does it become equivalent to single engine failure?

Not really. Twin engine aircraft don't have to meet the same crash worthiness standards and usually have higher stall speeds (in this case minimum crash speeds).

To be fair the statistics are swayed slightly by a number of factors, training accidents and different risk profiles of the flights.

In general, a twin will weigh considerably more than a single as well as having greater drag, making for shorter glides.

FWIW: 5 years ago or so I was on a plane that lost[0] an engine in rough weather shortly after take off.

The only reason I know is because the captain informed us about it and added that since it was technically an emergency landing we would be greeted by the fire department.

He then turned the plane around and landed as if nothing had happened.

The bloke across the corridor, an old pilot, complained about how there never happened anything remotely exiting on any of his flights ;-)

[0]: to be as accurate as possible I think what he said was they'd lost oil pressure on one engine and decided to shut it down voluntarily.

A broken AofA sensor has other consequences, such as the differing pressure altitude and airspeed on the left and right panels, so I would guess that it is expected that the problem would be found and fixed after the flight it occurred on (though this did not happen in the Lion Air case.)

They replaced "something", but probably haven't confirmed that, and haven't coordinated with the incoming crew to watch out for that, etc.

This came up in an earlier discussion. Someone suggested that, in attempting to fix the problem reported from the previous flight, the relevant systems were shut down, and this reset the AofA sensor selection mechanism.

Well if you have two sensors you never know which one is right.

You either trust both when they are nearly in agreement or disengage any autopilot type system like the mcas. The engineering solution is three sensors voting on agreement and disregarding the one that is out of agreement with the other two.

I'd favor one sensor or three... though there's something to be said for one simple structure being more reliable than 2+ complex ones.

coming from an engineering and climbing background... i favor redundancy.

> Let's make it 10x more difficult to detect that there is a problem.

A problem no one can detect is a problem you don't need to fix ! /s

> I wonder how we, humans, collectively build extraordinary monuments while we individually rest on stilts

My take is that we used to have brilliant engineers in the time where computational ressources were scarce (60s, 70 , 80s...), and they thought out their design thoroughly starting with a fucking paper and pencil. Aerospace indutries still rely on the designs from these times.

But now engineering seems to be : let's ask my computing power what is the good design, with little further questioning.

Brilliant engineers are still there. We haven’t gotten stupider as a species in 30 years. It’s the economics that changed.

As people say “They don’t build stuff like they used too anymore”.

First we started going cheap on the materials. Now we are going cheap on the engineers too.

Why pay a senior “brilliant” engineer $$$$ when a fresh out of college “average” engineer can write the “same piece of code” for $$

> “They don’t build stuff like they used too anymore”.

In fairness part (but far from all) of the reason we don't build stuff like we used too, is because we have much better understanding of material limits so we don't need to over-design structual elements like we used to.

How that understanding of material limits feeds into design obsolescence is where I do get annoyed though.

With everything we know now we could (obviously) build things better (for whatever criteria you chose) than they did 50 or 100 years ago, if that was the goal but it rarely is.

My favourite piece of civil engineering in the world (because it's nearby and I've ridden over it many times) was built in 1980 though (it's the exact same age as me) and its still awe-inspiring to look at.

https://upload.wikimedia.org/wikipedia/commons/b/b1/Humber_B...

The problem is that if a company gives customers (both b2b and b2c) the choice between a cheaper product with minimal guarantees vs. a product that is indeed better and can last longer, not enough customers will choose the better (more expensive) product to justify development cost (there are exceptions obviously).

Partly, companies are even forced if the lowest bidder for pre-defined specification gets the contract.

I believe there are more than enough engineers and companies that would like to build better and more durable products but demand just isn't there.

The problem is you can't trust anything/anyone any more... Brands are generally trusted until you get burned by one. And then they relabel or buy out another brand to burn.

I often make choices that are anywhere from 50% more cost to 200% if I'm relatively certain I'm actually getting a better product. Unfortunately that trust is usually limited and even then, you often don't have a good trust case.

There are also times where I look at industrial options for comparison, as that's often usually skewed towards more solid/maintainable production. Though again often means more expensive.

I tend to find the best value (my own experience) about 2/3 up the pricing scale for whatever products I'm looking at.

If I could find appliances in general with a 15+ year warranty (that didn't have weird loopholes), I'd be more inclined to buy. Unfortunately, they often don't cover interface boards, and even then they redesign in often incompatible ways each year and stop making the previous version entirely.

As someone who has spent most of their working life so far trying to make it easier to store data on materials, I'm not sure things have progressed all that much.

I do work with people from Boeing (and Airbus) but they are not yet using all the features of stuff I wrote 30 years ago.

I think it may be that brilliant engineers still exist somewhere, but I suspect they're less likely to go into aerospace than in the 50s, 60s or 70s. Heck or even the 40s.

Immediately after the war aerospace was red hot - there were countless countries and companies putting out new jet designs - both commercial and military. The competition and pace was intense and there was quite some cachet and public interest in aviation.

Since roughly, I guess, the 90s aerospace is rather settled and a very mature market. There's only a few major players left, and very very few new products; many of those are variations on the previous. Not nearly as much place or appeal for the brightest stars.

No, there's more to it than that.

Firstly, from 1903-1968 aerospace was in a period of exponentially improving performance (range, speed, altitude) comparable to Moore's Law. But from the mid-60s onwards the rapid improvements stopped, as developments hit hard limits imposed by material science and thermodynamics: it's hard to get a material much stronger than carbon fibre or titanium alloy, it's ridiculously hard to go faster in air (drag increases faster than the square of speed: if you throw more energy at the problem via your engines it comes out in the form of friction-induced heating).

Secondly, there have been huge incremental improvements since 1968. Consider the 737 series. The 737-100 had a maximum range of 1540 nautical miles on 17.86 metric tonnes of fuel; today's 737MAX-10 can do 3850nm on 25.94 tonnes of fuel: it's 72% more fuel-efficient. It's also gone from 118 to 204 seats, so about the same upgrade is passenger capacity. The cargo capacity has almost doubled, too.

This being HN, the point of comparison should be an Intel-cpu laptop circa 2009 with an equivalent laptop today. The headline clock speed is probably similar (stuck around 2.4GHz), as is the number of hardware cores. The modern laptop probably has more RAM and more storage, but only by a factor of 2x or 4x. But it has double or triple the battery life, 4x or 8x the pixels on screen, a screamingly fast SSD instead of a rotating-platter drive, and a GPU that the OS can offload a chunk of work onto other than simply accelerated graphics. It looks disappointingly similar, but if you could give your 2009-self your 2019 laptop they wouldn't be in a hurry to give it back again.

There have also been huge improvements in overall safety, the 737MAX MCAS problem aside; compare accident rates today with accident rates in the 1980s, let alone the 1950s if you want an eye-opener. (Of 1010 Boeing 707s delivered, there have been 173 hull losses in accidents.)

Hmm, aren't we agreeing here? :)

As a brilliant star with a newly minted engineering degree I might go into aerospace pre-1968 to play with blank sheet developments, X planes and XB-70, Concorde, Harrier, SR-71, 707 etc etc. Much the same with at engine makers - Merlin to Olympus and Pegasus. An absurdly long list, often with remarkably small teams. I might not want to take my 2019 vintage degree into a world of incremental improvements at one of the few global giants. Even if those improvements are of enormous significance over the years. It doesn't feel earth shaking any more, Farnborough isn't showing astounding free PR every year, just steady incremental progress (mostly). No more flying bedsteads.

Same goes with the laptop analogy actually - the fun was probably in the earlier years than now with new model every year with an extra few percent on the numbers, or slimmer. Green fields are usually more fun, and when they succeed, more satisfying; even with the odd Osbourne effect or total failure.

Were I that newly minted genius, I'm not sure where I would park myself today. Space X for some, perhaps. Probably not Boeing or BAe. I'm sure there's plenty of good (enough) engineers at all of them still. Working as one of thousands. That's not so alluring.

> 1010 Boeing 707s delivered, there have been 173 hull losses in accidents

Never knew that. That is frankly amazing.

I forgot the point I was trying to make while giving the background!

It's not that the engineers of yesteryear were brilliant while today's engineers are second-rate: it's that what they were doing back then was in some respects easier. (And we remember the good planes: as often as not they invented utter turkeys: look at the history of the Supermarine Scimitar for example — https://en.wikipedia.org/wiki/Supermarine_Scimitar — TLDR: yes, it's military, but to lose 51% of your aircraft in 9 years of peacetime service takes some doing. And then compare with the F-35, two crashes out of 308 airframes in service over a 5 year period, even though it's a vastly more complicated and higher performance aircraft.)

Yeah. People tend to forget that as each problem is solved, the next problem is typically more complex.

I like to say we never solve problems, we try to replace bigger ones with smaller ones.

And more equally distributed ones: there are not many easy solutions anymore that are pareto improvements, ie do not have downsides elsewhere.

For example, the door to the cockpit is locked now to keep terrorists out - which allows a suicidal co-pilot to fly the machine into a mountain undisturbed. (Having said that, a suicidal pilot can take down the machine one way or the other, so maybe locking the doors was a pareto improvement.)

Best minds of our generation are trying to get more ads in front of people and trying to get more people to click on them.

They’re at Elon Musk’s companies being underpaid and overworked so they can do something legitimately cool.

> Why pay a senior “brilliant” engineer $$$$

there are very few brilliant engineer of yesterday whose job role today is to train the next generation of brilliant engineers so that their vast experience doesn't get lost in time; the brilliant engineers of today have to relearn their experience walking the same mistakes, so that even companies that pays those premiums for senior engineers will have to face the same generational problem as everyone else, because they don't want to waste highly expensive professional to something as mundane as training on the job.

>We haven’t gotten stupider as a species in 30 years

I'm honestly not sure that's true. On top of anecdotes (in my grandfather's generation, the norm was to fluently speak 3-4 languages), go ask any STEM prof. with 20+ mileage about the general math, science and logic literacy of students nowadays.

> On top of anecdotes (in my grandfather's generation, the norm was to fluently speak 3-4 languages)

I'm not aware of anyone in my grandfather's generation that can speak more than one language fluently. The fact that my grandparents grew up on farms on the edge of the Dust Bowl may have something to do with that, but I rather suspect you're attempting to compare a tiny subset of the population in the past against the majority of the public today.

I agree and that’s why it was framed as an anecdote.

I'm a math professor at a US state university. I once asked one of my older colleagues whether our students had gotten better or worse over the years.

About the same, he said.

I think that professors judge their own generation by their peer group, and the current generation by their classes.

That's almost certainly "rosy past" bias. Even Aristoteles is said to have complained about the terrible youth.

There's also the prevailing management theory that "all staff can do the same job" -- I've seen this in various perverted and butchered variants of Lean and Agile.

The theory is sound -- at least two engineers on a project should know how a given functional unit works. It's a practice for lessening the impact of "bus factor", where you have one engineer everyone relies on because they're "the one person in the whole company who understands it".

The issue is when this kind of thinking reaches management, and you see someone being pushed into a role when their skill set is radically different to what's needed.

And then there's surprise and finger-pointing when things (inevitably) go wrong...

Here's the fun bit. A lot of the brilliant engineers that worked on incredibly important historical projects were not getting paid $$$$.

Only in silicon valley ad-tech do 'brilliant' engineers make 40 times more then the janitor.

> “They don’t build stuff like they used too anymore”.

Okay maybe, but we have been saying this literally forever.

Indeed, and I'd add that we are going cheap because we can, thanks to computing power.

As a matter of fact : computing tools make a lot of things easier and cheaper (which is good). But the setback is a tendency to "laziness" in the whole process...

I'm glad we don't build stuff like they used to: https://www.youtube.com/watch?v=joMK1WZjP7g

> But now engineering seems to be : let's ask my computing power what is the good design, with little further questioning.

I see this quite often at the undergrad level. Even for the simplest problem that can be solved with a pencil and a couple of calculations, many students launch their finite element software and blindly trust the results, even if it is telling them something obviously wrong, like orders of magnitude larger or smaller than expected, or with the opposite sign.

I acknowledge the problem, but I do not know where it comes from, or how to stop it.

I don't find that particularly surprising. Engineering students want to learn how to build machines to find solutions to everything, so it's intuitive that they would place a high confidence in a solution provided by a machine.

Good observation.

There is a nice analogy to programming here. The same that one of the main tasks of the programmer is to handle errors, one of the main tasks of the engineer is to know the limitations of the tools he is using.

With no back checking?

That's why I said "engineering students", not "engineers". Not blindly trusting your outputs is (or, at least, should be) part of the curriculum.

Is this for real?

That's startling to me. I don't really feel like I "know" something well enough to bother with the overhead of developing computer code until I can at least get the essence of it down on paper.

Computers are for enumerating and increasing confidence in proofs. I find pen and paper to be far more creativity friendly.

SpaceX is landing rocket stages on a drone ship after having them drop from 7400 km/h at 100 km altitude. JPL has a city car sized rover craned from a hovering lander on Mars and DeepMind has a computer win against the acting Go champion. Pretty sure that rules out the "today's engineers are bad" hypotheses.

Indeed we still make great technological achievements, and thanks God there are still brilliant engineers out there. But they may not be equally distributed in the industry.

And many of them (in my humble experience, at least) rely too much on computing tools and algorithms they do not completely master, expecting a "magical" answer out of them.

There absolutely are brilliant engineers out there.

The problem is, they're probably not working at Boeing.

There were a lot of failures and accidents back then.

But in general, designers did get more experience because there were more aircraft programs and they were way faster. So if you designed, say, landing gear, you might have done quite many during your career. It was more "agile" in that sense.

Nowadays companies can not go so quickly to production. The whole business has turned into oligopolies both in the main integrators and the suppliers.

But we do. Look at SpaceX. Landing Rockets, crew dragon going to be half the cost of starliner.

It’s just that LargeCorp eventually has the profit motive as the overriding goal and stops being innovative. It’s schumpters creative destruction in reality

I think that's a bit unfair; designs have gotten so much more advanced and efficient in part thanks to computational design. The efficiency, range and maximum weight that airplanes can carry has pretty much doubled since that time.

However, the original comment and the video it is a reply to explains a bit of the countering force; economy and competition. It's management that pressures the engineers to cut corners, basically.

I don't doubt the Boeing engineers are some of the best people in the industry. I'm very much skeptical of management (and whipping them are shareholders) that optimize for time and by extension profits.

We still have many brilliant engineers. The problem is that the brilliant ones from that age have all retired or died, and the brilliant ones these days didn't go into aerospace engineering, they went into other fields. Even when I was in college 25 years ago, we were making fun of the Aerospace Engineering majors as being in a dead-end field where they could look forward to few jobs and lay-offs. It's only gotten worse since then, I'm sure.

The pay and career potential and job security is much better in computer science and engineering, so that's where all the brilliant engineers have been going for quite some time now.

Why do you say we shouldn't do armchair analysis right before you acknowledge that said analysis was on point? I think it's good our populace is eager to analyze. Ultimately we aren't making the decision, just forming a branch of decentralized checks. Long as witch hunts like the Boston bomber don't happen I say people should analyze till their hearts are content.

I agree with you but the general view is to avoid doing armchair analysis until official word is out.

As far as armchair analysis doesn’t cause any harm, I think it’s fine!

> 1. Use inputs from both AOA sensors, if they disagree by 5.5 degrees, disable MCAS.

Great, and now you have the other problem the MCAS was supposed to mitigate (the tendency of the MAX to do backflips).

Shrugs, if the MCAS disables with a clear notification to the pilots I think at this point most 737-Max pilots would know the procedure to deal with that.

Better than the plane flying itself into the ground.

> Shrugs, if the MCAS disables with a clear notification to the pilots I think at this point most 737-Max pilots would know the procedure to deal with that.

The plane is only certified to be 'identical to the 737' when MCAS is on, and operating correctly.

If MCAS has to be disabled in response to a common mechanical problem, then the plane must be re-certified.

Which raises a rather obvious question ...

Although at least now you're trying to solve a more logical problem. Increasing angle of attack => counter with elevator, if not sufficient reduce thrust

Reducing thrust will increase angle of attack. The angle of attack is the angle of the wing relative to the airflow, not the angle of the plane relative to the horizon.

Yes conceptually, no because of signage error.

If you were in a position MCAS would have to kick in, you were at positive AoA, reduced thrust would impart a negative AoA moment.

> you were at positive AoA, reduced thrust would impart a negative AoA moment

I don't think that is correct. I think you're still thinking of the angle of the plane with respect to the horizon, not the AoA.

If reducing thrust reduced AoA, then you could recover from a stall by reducing thrust. In reality, the opposite is the case. The only way to recover from a stall by manipulating engine thrust is to increase it (though this won't always be sufficient by itself).

AFAIAK the procedure is to not increase thrust, quite the opposite (disable the autothrust thingie), get the nose below the "horizon" (which is of course wind dependent), gain airspeed, and then start to increase thrust (which will automatically start to up the nose a bit), then climb back to whatever altitude you should be at.

I don't mean to say that increasing thrust is the standard stall recovery procedure in most aircraft, or that all stalls could be recovered from merely by increasing thrust. But if one could only manipulate thrust, one would increase thrust to recover from a stall, not decrease it.

Ah, yes, of course, theoretically if you can just increase thrust arbitrarily, that always solves the issue. (Again, of course, this might mean the whole plane becomes a rocket because the wings already fell off, so you might need some thrust vectoring too.)

Okay, I get what your saying. I just wasn't reading right, and while what I've said may be theoretically correct, under most circumstances, the control surfaces are the primary contributors to the aircraft's orientation.

Fiddling with engines in an MCAS activation situation would just offset the flight envelope into a more dangerous zone, that could be even more difficult to recover from.

Took me a bit, but I got there.

We're not talking about the aircraft's orientation, we're talking about the Angle of Attack. The AoA can increase even if the nose of the aircraft is dropping relative to the horizon.

Exactly, I'm only talking about in the MCAS-scenario.

> 1. Use inputs from both AOA sensors, if they disagree by 5.5 degrees, disable MCAS

This can only mean that MCAS is disabled even more frequently. This is a terrible idea. This is like RAID0.

The failure mode will increase greatly in complexity. MCAS might cut in and out, the plane might behave erratically.

What the need to do is add a damn switch for 'disable MCAS' and tell people how and when to use the damn thing.

No, they can't do that, because then they'd have to retrain the pilots, or maybe even have a different type rating, which would make them uncompetitive against the Airbus and Bombardier planes.

Remember, profits are more important than passengers' lives.

Just booked a flight. Price between two major carriers was about the same. One was a Boeing, the other Airbus. Guess who got my business?

> because then they'd have to retrain the pilots,

I think that ship has sailed already.

I don't get it, Boeing seemed to have a positive slope, with their plane outselling the competition. Fragile system design is hard to understand in this context. But maybe MCAS were conceived at a time where Boeing was more willing to cut corners.

I thought that the flight immediately before this one had MCAS trouble too. Were both sensors busted, or was that actually an even number of cycles previous to the crash, and only the previous day?

The plane went through a maintenance cycle in-between the flights; it was suggested that the maintenance caused the chosen AoA sensor to flip twice.

I'm not sure of the exact mechanics of this, but it's conceivable the maintenance crew performed a full engine or board computer boot cycle. Therefore, when the pilots started on their flight, it was the second warm-up since the previous flight, hence the same sensor was selected.

There was an extra pilot in the cockpit uninvolved with flying the plane.

His participation in Crew Resource Management allowed the previous flight to correctly identify and remediate the issue before things became unrecoverable.

I remember that, but if the issue was that one AOA instrument was broken, then only half the flights should be affected.

I know very little about aircraft, but wouldn't making the MCAS less powerful increase the risk of a stall? How do we know that there would be a net decrease in crashes?

We know nothing about how many times MCAS got engaged, maybe Boeing does. Who knows.*

Furthermore, the problem seems to be that there is no explicit feedback about this from the pilots, because some of them did not even know about this. Consequently there was no widespread beta test in simulators, meaning there was not much data to base the fine tuning on.

* actually, since MCAS activates only when the plane is over the critical angle (according to the AoA sensor), and there are other warnings, probably those can be counted.

I wonder if the new iteration will also get an intermediate disable state, where you keep powered trim but shut out all automatic inputs. Defense in depth.

>Furthermore, electric trim with the yoke switch will override MCAS.

Does that mean it didn't override MCAS before? Really?

The point was, that they "just" put that shadow system there which will perfectly encapsulate the 737MAX to match the old 737 "interface".

Now, they completely fucked that up. They did a so mindblowingly bad job, that anyone involved at the FAA who should have known about this should be in jail. Either for incompetence (negligence) or worse.

They designed a system that relied on one input, operated periodically, told no one, documented nowhere, made it too powerful, always on, and un-turn-off-able, and spent the rest of the day corrupting/pleasing the FAA, then made the AoA input mismatch light an upsell feature, and finally called it a day.

Yep, this being one of the reasons the planes went splat.

Reply on news.ycombinator.com