The example aircraft is a modernized Boeing 737. First flight in 1967 and still being manufactured, 47 years later. That's a long production run.
The description of the controls is long, but there are two main divisions - flying the aircraft, and systems management. One of the big problems in cockpit design is keeping the pilot from spending too much "head down time" on systems management.
The 737 was the first jet transport designed for a flight crew of two. Previous jets had a flight engineer as well. Further back in aviation history, there were navigators and radio operators. (BOAC even had a "flight clerk", equipped with a typewriter, on some flights.) As crew size dropped, the pilots picked up more system management duties. This remains a problem. Cockpit automation generally means that the systems take care of themselves unless something goes wrong. Then the pilots have to devote a lot of attention to dealing with the problem. Crashes have occurred because the pilots were dealing with some system problem and not focused on driving.
Consider reading Breaking the Mishap Chain by Peter Merlin [1]. It's a pretty fascinating series of case studies where human error was a primary factor. Chapter 5 discusses the crash of M2-F2 during testing for NASA and the role that task saturation had in the incident.
Both pilots became so preoccupied with a burnt-out landing gear indicator light that they didn't notice the autopilot had become disengaged, and crashed into the Florida Everglades.
The recent crash at SFO was caused by mistakenly thinking the airplane was working a different way than it was actually working. The pilots thought the autothrottles were engaged when they weren't and were fixated on trying to land the plane manually; something they hadn't done in a long time.
Which is the sort of problem we'll likely see more of as "mostly autonomous" systems become increasingly common. For example, it's easy to imagine that it will become possible (whether or not we elect to go down that route) to design cars that can be self-driving most of the time but which will require a human to take over in some scenarios. The good news is that those scenarios will presumably be mostly low speed ones--the handoff timeframe pretty much has to be of minute rather than second magnitude--but you could still end up with an increasing number of drivers who have very little actual driving experience.
Honestly, that's not entirely true, and not very respectful to the pilots in that crash.
AF447 was caused (partially) by the pilot's misunderstanding that holding the control stick all the way back was gonna make the aircraft go up at the best rate of climb, and not put the control surfaces all the way back.
The AP (autopilot) was off for quite some time when the crash happened. What was supposed to be on though, as always, is the Airbus Fly-by-wire system in normal law, that is in between the pilot's controls and the actual commands sent to the control surfaces (ailerons, elevators...). This system was off (in it's direct control law, more precisely) due to the pitot tubes sending non accurate data, because of the icing problem. The pilot didn't notice that.
Yes, sorry, that's what I meant. They had turned the system to direct control law and one of the pilots was pulling back, and they were ignoring the "stall" warning blaring in the background.
Honestly, first of all, who averages two sticks? Like, when is averaging the input of the two sticks ever a reasonable thing to do? This is a major design flaw, in my opinion.
Second of all, why was the "stall" warning even on when the system was in normal law, and thus impossible to stall? All that did was desensitize the pilots, who learned to just ignore the warnings.
And this is why I'll go up to my IFR and maybe dual prop but nothing more. That guy is the kind d of person I want flying my plane, maybe. Only if he handles context switching well...
Don't be afraid of all that complexity, most of that swithches are used sparingly or at all during each flight. The ones that get more use are the common flight controls and navigation aids (FMC, autopilot, etc...). I'm airbus 320 pilot by the way.
Yes exactly. It's the same way in fighters. The vast majority of switches are used maybe once. There were some F-16 switches that I never touched in 11 years. In fact, some switches are actually secured with copper wire to be used only in certain emergencies.
Believe me, you can handle it. IFR is all about cross-check, preparation and staying ahead of the jet. Not the number of switches available to you.
I remember seeing this posted on Quora a while ago and it still stands out as one of the best answers I've ever read. And the biggest takeaway for me was (and still is):
"Redundancy. Aircraft manufacturers are really, really (really) big on redundancy."
That and off switches, overrides and manual backups. It's a different mindset when every survivable emergencies involve "keep flying for at least 15-30 minutes to reach an airport and land."
This is a true takeaway. Redundancies are even layered - for example, there may be redundant computing modules, running redundant (and possibly independently developed software). These computing modules may be multi-processor, have multiple banks of RAM and flash, etc ... but not used in the typical COTS way. Instead of being devoted to increasing computing power, they are devoted to ... you guessed it, redundancy. In highly critical hardware, you could physically remove one of the CPUs from the board while it was running, and the system would log a fault and keep on ticking without so much as a timing jitter.
EDIT: Also, for those of you who don't mind working for a massive corporation instead of a startup, avionics is a fascinating niche of computing. It's what I do, and I love it :) They pay is a bit worse, but hey, I have a pension. It also helps that many aerospace companies are in lower cost of living areas than startups. More importantly, the problems of next-gen avionics are completely unique, unsolved, and engaging.
Boeing or LM? I worked for Tupolev - Tu-214D, auxiliary fuel system control module, which actually did D in the name (which means Dalnii, Long Range in Russian).
My famous takeaway from these times was
---
ARINC Report 431: No Fault Found – A Case Study
provides the final report of AMC Task Group 116 formed to
discover the causes of “No Fault Found” in avionics
equipment during test. This standard identifies sources and
provides recommendations for improvement.
---
Neither ... it's a subcontractor. Also your mention of ARINC brings to mind another thing - the business environment is way different than most of tech. Patent lawsuits are basically unheard of. Many of the companies within the field are not only competitors, but typically also partners in one or more ventures. The reason ARINC jogged that memory loose is that Rockwell Collins bought ARINC a while back, and then spun off the standards making division so as not to damage their credibility as the standards-creator for aviation. When I think of, say, mobile ... I just don't see those companies behaving that way.
Exactly, I felt the same. I believe it (cooperation, attention to detail, much less hype and lawsuits, etc) is because of the responsibility. If I sue competing startup and they will get distracted and miss their scaling problem and begin to return 502 errors - herd will call me a "fighter", some call me "this jerk, who's smarter in the courtroom than at the keyboard", etc, VCs will like me, anyway, I win. If I sue someone making a competing ARINC interface, they get distracted, miss a race condition and then 502 people die... I also get distracted, and another 404 people die.
Well, it's a little bit worse than 502 error. And there's so many training and so many manuals, books, anecdotes around how many people may die if we don't "put enough assert()s in our code" that you just stop thinking about hype and lawsuits and gold miners.
Numerous factors. I work in the midwest, where the cost of living is dramatically lower (in fact, in terms of salary/cost of living, it may actually be better). You also trade in some potential returns for a large reduction in potential risk - I know that the company I work for will be around 10 years from now, probably even 100 years from now. Other benefits are better - paid time off, retirement benefits, etc. Many of the companies in the field are also somewhat prestigious - similar to landing a job at google or apple - so they have their pick of newgrads (which in itself doesn't account for the pay difference).
Really, I think it all comes down to location. You can get a ~1500-2000sf house on 10 acres here within 30 minutes of work for $200-$300k. Crazy cheap compared to the west coast.
I worked once on the design of the auxiliary fuel control system for a long range version of jet commercial aircraft. This single job taught me more about "Proper System Design"(tm) than my whole previous experience.
which was pretty awesome (it wasn't used for crew training, it's a popularization of what's involved in operating a Shuttle). It also takes you through a pretty substantial number of control switches and indicators, even while simplifying things.
I wonder if there's a way to compare the complexity of the body of knowledge that jet pilots have mastered (this explanation doesn't even go into aerodynamics or flying skills, or the purpose or physics of some of the systems it alludes to, or rules and procedures, or abbreviations or jargon) to what doctors know about bodies, pathology, diagnosis, and medical interventions.
NASA released many old manuals to the public, including several Space Shuttle PDFs.
Flying the space shuttle is more like programming a flight computer and checking the flight data. There are some manual overwrites, that's the challenging part.
I read on HN that Quora pivoted and joined Ycombinator, more for for exposure I guess? It was basically a StackOverflow alternative with real names. Will be interesting to see how their piviot to something else turns out.
The first quora comment is long and interesting, but the "photos" are from a flight simulator, probably Flight Simulator X from Microsoft. And flying a plane in that simulator is not that hard, something one can learn in a few hours. And all the knobs have tooltips too.
They make extremely high detail planes, with incredibly in-depth simulations. Many real commercial pilots use this software to train (although they can't earn training hours with it).
To fly it correctly, it takes more than a few hours to learn.
The embedded videos shows a scenic Innsbruck city airport in the alps: Innsbruck Airport is well known for having a difficult approach due to surrounding terrain, prohibiting certain aircraft types from operating at the airport.[3] The approach and descent is a very complicated process—the Alps create vicious winds and currents, which the pilot has to deal with throughout the process. Because there are mountains all around, the plane usually circles the airport, enters a pattern, to decrease both speed and altitude.http://en.wikipedia.org/wiki/Innsbruck_Airport
Most of the quora answers seem to ignore the significant shift in avionics that has occurred with the general adoption of glass cockpits.
In older aircraft, you would've been staring at a wall of gauges, lamps, and buttons i.e., C130[1]. These, for the most part, have been consolidated into large multi-function displays (MFDs) like the C130 AMP [2] or C-5M [3] with the remaining analog controls are either there for emergency use or for the pilot's benefit like the yoke.
The MFDs generally reduce pilot workload by consolidating relevant data into task-oriented screens. However, the main driver was more likely the significant weight reductions and costs associated with manufacture and maintenance of the aircraft - fewer cockpit components and simpler wiring.
Ultimately, the types of controls are similar between the older and new aircraft. The notable exception is the keyboard(?!) in the A380s [4].
As a complementary curiosity, an Airbus A320 has 575 buttons and knobs (that includes alphanumerical pads and double knobs), and 364 circuit brakers.
I had so many kids asking how many buttons there are at the cockpit, that I finally spent a Madrid - Rome flight counting them (under the surprised look of the captain).
I guess that a long haul model like a A 380 has more or less the same, maybe some dozens more due to the increased system complexity (more engines, extra hydraulic and electric sys..) but the brakers are located at the instrument bay under te cockpit and that makes it look a bit les cluttered).
37 comments
[ 3.6 ms ] story [ 80.7 ms ] threadThe description of the controls is long, but there are two main divisions - flying the aircraft, and systems management. One of the big problems in cockpit design is keeping the pilot from spending too much "head down time" on systems management.
The 737 was the first jet transport designed for a flight crew of two. Previous jets had a flight engineer as well. Further back in aviation history, there were navigators and radio operators. (BOAC even had a "flight clerk", equipped with a typewriter, on some flights.) As crew size dropped, the pilots picked up more system management duties. This remains a problem. Cockpit automation generally means that the systems take care of themselves unless something goes wrong. Then the pilots have to devote a lot of attention to dealing with the problem. Crashes have occurred because the pilots were dealing with some system problem and not focused on driving.
[1] http://www.amazon.com/Breaking-Mishap-Chain-Development-Aero...
I definitely remember reading about others but can't find them right now.
Both pilots became so preoccupied with a burnt-out landing gear indicator light that they didn't notice the autopilot had become disengaged, and crashed into the Florida Everglades.
http://en.wikipedia.org/wiki/Asiana_Airlines_Flight_214
AF447 was caused (partially) by the pilot's misunderstanding that holding the control stick all the way back was gonna make the aircraft go up at the best rate of climb, and not put the control surfaces all the way back.
The AP (autopilot) was off for quite some time when the crash happened. What was supposed to be on though, as always, is the Airbus Fly-by-wire system in normal law, that is in between the pilot's controls and the actual commands sent to the control surfaces (ailerons, elevators...). This system was off (in it's direct control law, more precisely) due to the pitot tubes sending non accurate data, because of the icing problem. The pilot didn't notice that.
Honestly, first of all, who averages two sticks? Like, when is averaging the input of the two sticks ever a reasonable thing to do? This is a major design flaw, in my opinion.
Second of all, why was the "stall" warning even on when the system was in normal law, and thus impossible to stall? All that did was desensitize the pilots, who learned to just ignore the warnings.
http://n631s.blogspot.com/2011/11/children-of-magenta-line.h...
Believe me, you can handle it. IFR is all about cross-check, preparation and staying ahead of the jet. Not the number of switches available to you.
"Redundancy. Aircraft manufacturers are really, really (really) big on redundancy."
EDIT: Also, for those of you who don't mind working for a massive corporation instead of a startup, avionics is a fascinating niche of computing. It's what I do, and I love it :) They pay is a bit worse, but hey, I have a pension. It also helps that many aerospace companies are in lower cost of living areas than startups. More importantly, the problems of next-gen avionics are completely unique, unsolved, and engaging.
My famous takeaway from these times was --- ARINC Report 431: No Fault Found – A Case Study provides the final report of AMC Task Group 116 formed to discover the causes of “No Fault Found” in avionics equipment during test. This standard identifies sources and provides recommendations for improvement. ---
Really, I think it all comes down to location. You can get a ~1500-2000sf house on 10 acres here within 30 minutes of work for $200-$300k. Crazy cheap compared to the west coast.
http://www.amazon.com/Shuttle-Operators-Manual-Revised-Editi...
which was pretty awesome (it wasn't used for crew training, it's a popularization of what's involved in operating a Shuttle). It also takes you through a pretty substantial number of control switches and indicators, even while simplifying things.
I wonder if there's a way to compare the complexity of the body of knowledge that jet pilots have mastered (this explanation doesn't even go into aerodynamics or flying skills, or the purpose or physics of some of the systems it alludes to, or rules and procedures, or abbreviations or jargon) to what doctors know about bodies, pathology, diagnosis, and medical interventions.
Flying the space shuttle is more like programming a flight computer and checking the flight data. There are some manual overwrites, that's the challenging part.
e.g. http://www.nasa.gov/centers/johnson/pdf/390651main_shuttle_c...
Isn't there also a flying-it-like-a-glider part after re-entry? But I guess that's a pretty small fraction of the mission.
Source: https://news.ycombinator.com/item?id=7723278
You can read some answers as to why Quora joined YC here: http://www.quora.com/Why-did-Quora-join-the-2014-Y-Combinato....
http://precisionmanuals.com
They make extremely high detail planes, with incredibly in-depth simulations. Many real commercial pilots use this software to train (although they can't earn training hours with it).
To fly it correctly, it takes more than a few hours to learn.
That's the deep link: http://precisionmanuals.com/pages/product/FSX/ngx8900.html
The embedded videos shows a scenic Innsbruck city airport in the alps: Innsbruck Airport is well known for having a difficult approach due to surrounding terrain, prohibiting certain aircraft types from operating at the airport.[3] The approach and descent is a very complicated process—the Alps create vicious winds and currents, which the pilot has to deal with throughout the process. Because there are mountains all around, the plane usually circles the airport, enters a pattern, to decrease both speed and altitude. http://en.wikipedia.org/wiki/Innsbruck_Airport
In older aircraft, you would've been staring at a wall of gauges, lamps, and buttons i.e., C130[1]. These, for the most part, have been consolidated into large multi-function displays (MFDs) like the C130 AMP [2] or C-5M [3] with the remaining analog controls are either there for emergency use or for the pilot's benefit like the yoke.
The MFDs generally reduce pilot workload by consolidating relevant data into task-oriented screens. However, the main driver was more likely the significant weight reductions and costs associated with manufacture and maintenance of the aircraft - fewer cockpit components and simpler wiring.
Ultimately, the types of controls are similar between the older and new aircraft. The notable exception is the keyboard(?!) in the A380s [4].
[1] http://commons.wikimedia.org/wiki/File:C-130_Hercules_cockpi... [2] C130 AMP http://www.edwards.af.mil/shared/media/photodb/photos/091109... [3] C5M http://upload.wikimedia.org/wikipedia/commons/9/9f/C-5M_Cock... [4] A380 http://www.aerospace-technology.com/uploads/newsarticle/6889...
I guess that a long haul model like a A 380 has more or less the same, maybe some dozens more due to the increased system complexity (more engines, extra hydraulic and electric sys..) but the brakers are located at the instrument bay under te cockpit and that makes it look a bit les cluttered).