Comment by quietbritishjim | Hacker News Reader

Comment by quietbritishjim | Hacker News Reader

quietbritishjim

Another good write up of this incident is on the excellent Admiral Cloudberg blog:

https://admiralcloudberg.medium.com/the-long-shadow-of-war-t...

Another article in that series is also related to a UX mistake: Air France Flight 447.

This crashed, in part, because the inputs from two control sticks (one for pilot and other for copilot) were averaged if they disagreed, unlike on a Boeing (at the time at least) where they're physically connected so you can't have contradictory inputs in the first place, and you'd feel the other pilot fighting you. When the plane stalled, one pilot correctly pushed down to come out of the stall (after which they would be able to pull back up) while the other pulled up instead (which is wrong but does feel like the instinctively correct thing to do). The inputs cancelled out so had almost no effect. By the end both pilots were pulling up, but that hadn't been the case earlier on when the problem could have been resolved.

https://admiralcloudberg.medium.com/the-long-way-down-the-cr...

The captain had been on a rest brake and only re-entered the cabin at the last moment. He finally figured out the problem but it was too late to do anything.

> “Go on, pull,” Dubois said. Was this comment a sardonic resignation to fate?

> unlike on a Boeing (at the time at least) where they're physically connected so you can't have contradictory inputs in the first place, and you'd feel the other pilot fighting you.

Note that it’s not quite that flagrant: because Boeings don’t normally autotrim (famously aside from the 737 MAX, whose MCAS was not documented and trained for — I think the more recent Boeings like 777 and 787 are FBW and do autotrim) it’s possible to fight your co-pilot while thinking you’re not trimmed, or even that you have excessive air load:

> excessive air loads on the stabiliser may require effort by both pilots to correct miss-trim. In extreme cases it may be necessary to aerodynamically relieve the air loads to allow manual trimming.

  because Boeings don’t normally autotrim

Yeah they do. On the 737/757/767 they have a mach trim system to account for mach tuck. On the 737, at least, they also have a speed trim system to keep the plane under control at lower speeds when autopilot is off.

> unlike on a Boeing (at the time at least) where they're physically connected so you can't have contradictory inputs in the first place

I don't know if it's true, but I recall reading somewhere that this physical connection is a breakable link, so if one of the controls gets stuck the other control can still be used to fly the airplane (after some application of strong force to break the connection).

Yes it is true. The force needed is pretty high.

It also caused one accident, where both pilots thought the controls were jammed because the other pilot was pulling in the opposite direction. This is no better than averaging. At least Airbus gives an alarm.

Which accident was that? Thanks!

Don't remember, sorry.

I couldn't find any evidence of it happening either, which makes sense, a system which gives feedback there's a problem intuitively seems better than one which doesn't.

  a system which gives feedback there's a problem intuitively seems
  better than one which doesn't.

Except that your intuition isn't always correct. Airbus throttle levers don't move on their own, Boeing throttle levers (at least pre-FBW) were mechanically linked to the control system and would move in response to computer inputs. Sounds safer, right?

Look at the Sriwijaya Air 737 crash (among others). Even with the throttle levers showing wildly asymmetrical thrust, when the automation gave up the pilots didn't notice, didn't react properly, and crashed the plane.

Sticks (not throttles) mechanically linked to each other do indeed give feedback indicating there is a problem when the two sticks conflicted, compared to, it seems, the Airbus concept, which did not.

I see no reason the intuition here isn't correct: it's better to know about this problem when flying than to not know about it. Stories about throttles or automation are a distraction from what we're talking about: mechanically linked sticks.

On that note (back on topic), I still couldn't find any evidence of the crash mentioned above, but maybe I'm just searching wrong. Does anybody know about this crash supposedly caused by Boeing sticks being mechanically linked?

So-called "side stories" about throttles that provide feedback are entirely relevant because they address the issue at hand: the supposed benefits of physical feedback in the controls. The Sriwijaya crash is additionally relevant as it happened during a phase of flight where someone's hands were likely to be on the throttle quadrant in the first place.

If you're going to double down on the idea that somehow feedback in a control column is a wildly different affair than feedback in the throttle quadrant, Air France recently had a go around where the two pilots provided opposing inputs without realizing it. The inputs were forceful enough to un-synchronize the controls, so yeah, physical feedback is less helpful than one's intuition suggests.

With Air France 447, there should've been an annunciator that conflicting input was being provided (as well as the ability to lock out one set of controls).

However, when you're task saturated it doesn't matter what you're flying — there are plenty of things you're going to miss. Were there strong benefits to feedback (or lack thereof) on FBW controls there would be some sort of mandate from the FAA or EASA one way or another.

It was an air france accident, it happened twice (both times to air france), one time it resulted in an accident, then it happened again later but it only resulted in an extremely stressful go around

>unlike on a Boeing (at the time at least) where they're physically connected so you can't have contradictory inputs in the first place

Air France seemed to manage ithttps://www.avweb.com/aviation-news/air-france-crew-fought-e...

Also glad to see that airbus are planning to put force feedback side-sticks in future.

> Air France seemed to manage it [on a Boeing]

Hence my "at the time" bracket. Since then, newer Boeing planes have actually removed the physical connection between the sticks! Absolutely bonkers decision given the background. (I put "at least" at the end of my bracket because I wasn't 100% sure I rememebered correctly.) But maybe there is a good reason - e.g. like a sibling comment suggested, if one gets stuck.

AFAIK the 777 still has mechanically linked yokes. It has a breakout mechanism in case one of the controls gets jammed (allows freeing the other one), but it's not been removed.

As the sibling comment points out the 777 has mechanically linked yokes. The 787 does too. With the Air France go around, the pilots were pulling hard enough that the mechanical connection between the yokes was broken. Both pilots continued to provide contradictory inputs.

On an Airbus this would trigger a "dual input" annunciator, and there's a "priority takeover" button that would lock out one set of controls.

On the Boeing once the torque tube splits each yoke controls one side of the aircraft.

The synchronization and feedback intuitively seem like a good idea, but the reality is that at the point where you need to notice that the other pilot is doing something wrong you're already up shit creek and you're no more likely to notice an increase in force required to move the yoke than you are an annunciator. Linked controls aren't a substitute for CRM.

Thanks for this, very interesting.

The tragic AF447 flight comes up somewhat regularly on the site — see this recent conversation [0].

[0] https://news.ycombinator.com/item?id=37089363

CFIT is a particularly spectacular form of failure.

<https://en.wikipedia.org/wiki/Controlled_flight_into_terrain>

It's not something that can be changed on airbus cockpits, since the nature of side sticks makes them impossible to be mechanically linked, unless there was some very complex motor system built into each stick that provided force feedback equal in strength to a pilot tugging hard on it.

> unless there was some very complex motor system built into each stick that provided force feedback equal in strength to a pilot tugging hard on it

It does not have to be “equal in strength to a pilot tugging hard on it” since there is normally no significant load applied by the stick to the hand (only the centering springs). So most every feedback should be noticeable.

And force feedback is being deployed right now on commercial planes. The Irkut MC-21 was supposed to be the first airliner featuring them, but the invasion of ukraine and subsequent sanctions nixed that (as the provider of the sticks is the french company Ratier-Figeac). Gulfstream’s 7th gen (GVII) also have active sticks, provided by BAE.

Yes, but then no one would consider the feedback to be the equivalent of a mechanically linked system, like what Boeing uses.

Of course they would: the problem is that there is a lack of feedback when incompatible input is entered.

Audio feedback was added since AF 447 but as well known sound is one of the first thing to be blanked out in high stress situations. Physical feedback provides a second stimulus, and one which does not get ignored as much.

The planes are FBW, the side sticks have no mechanical linkage outside their control box, there is no cause for a pilot to input significant force on the stick, and thus most every feedback would be sensible. Unlike a 737, the pilots are not physically moving the control surfaces.

The moment one pilot does tug hard on the stick though, then the same problem is still there.

So it will work most of the time except when you really need to, when the other pilot is panicking and won't provide accurate answers.

> The moment one pilot does tug hard on the stick though, then the same problem is still there.

No, because the point is that the other pilot will feel the action and thus know that there are conflicting input. The sticks can even synchronise their movements, that way pilots can both see and feel that their copilot is inputting.

That is the point, making the existence of conflicting input clearer.

> So it will work most of the time except when you really need to

No.

> when the other pilot is panicking and won't provide accurate answers.

The sticks support taking priority, so a pilot seeing that their copilot is panic-inputting would be trained to press and hold the priority takeover button, deactivating their copilot’s stick.

You seem to be confused, obviously it won't work 100% of the time due to shear probability, it's not physically linked up so there will always be a possibility of one or both pilots ignoring every possible warning, shaking, vibration, mild pushback, etc... regardless of how well they've been trained on detecting these.

Whether it works the vast majority of the time, in practical situations, remains to be seen. Confusing a known, physically guaranteed, aspect of a system with an estimated likelihood of actual system performance is a common mistake.

Even with the most ideal deployment scenario, assuming it does get widely deployed one day, the real world performance reliability will likely not get anywhere near six sigma, let alone 100%.

The Cirrus SR-22 (the best selling general aviation plane every year since 2003) has mechanically linked side sticks. They aren't fly by wire like the Airbus but it shows they can be linked without a complex motor system.

> Cirrus SR-22

Are those really side sticks, though? They look much more like one half a yoke than the kind of side stick you'd see in, for example, an Airbus or an F-16.

...it's a spring and a servo. Or a brushless motor.

https://github.com/scottbez1/smartknob

...or any not-bottom-end-of-market steering wheel for driving video games.

It's not just "a spring and a servo" when we are talking about aviation. Redesigning GA aircraft and rectifying is already bad enough. An airliner, and something as crucial as flight controls?

Yeah, good luck jury rigging some servos.

> It's not something that can be changed on airbus cockpits

You'd need to run cables between them and redesign the sticks to have mechanical hooks, and it'd be a very ugly hack, but it's hardly impossible.

Impossible for it to still be viable for a commercial airliner competing on costs, especially long term upkeep costs.

Not impossible for an unlimited budget.

I think the biggest no-nos on this hack would be that it'd need to be certified and since it fundamentally changes the UI for the pilots, everyone would need to be retrained.

The hack itself would be a small detail line on the overall budget for the changes.

The retraining would barely be a concern. It would be a new input similar (but not identical) to the stick shaker, or a new audio alert (as was literally introduced for the dual-input issue after AF447).

The hack is the entirety of the cost, it would require completely redesigning the side-stick, updating their entire maintenance procedure (currently side-stick can trivially be unscrewed and swapped out), as well as redesigning the cabin cell to provide routing for the cables such that they keep tension and don't get damaged.

Stick manufacturers have been working on active sticks for more than two decades, and they're getting released right now. While force feedback on PC joysticks was introduced in 1997.

I believe Airbus is planning to add force feedback to the side sticks for this error case. But I don't know that.

Many aircraft controls already have force feedback in the form of stick pushers [1], a device to violently shake the control yoke when the aircraft is at risk of stalling. Those have been around since the 60's.

So this is not only possible, but it's been around for a long time though Airbus had chosen to use a verbal warning only instead of a stick shaker.

is there really no mechanical connection between the stick and the control surfaces on an airbus? that seems really dangerous.

There really is no mechanical connection. And the Airbus A320 (first airliner with no mechanical backup) has been flying this way since 1988, so it's not anywhere near as dangerous as you think.

Reply on news.ycombinator.com