« Next year's books ... | Main | Not dead »

Marking time

Right now I'm up to my elbows in the guts of "The Fuller Memorandum", doing a final draft (with editorial feedback) and writing the afterword. It's due on my editor's desk at the end of the month, and so takes priority over, er, blogging. When I start posting here frequently again you can probably take it as a sign that I've broken the back of the job.

In other news: given that we're beginning to see in-flight internet access trailed on airliners, how long it is going to be until we get real-time streaming of airliner telemetry data to save us from this sort of wild goose chase?

(Yes, I know there's some hope of pulling the flight data recorders out of the mid-Atlantic. Yes, I know in-flight internet via satcom is still a new area and global coverage won't be available for a while. Yes, I know that losing airliners over open water is relatively rare. Nevertheless, if that A330-200 had real-time streaming telemetry, the air accident investigators could well be on the way to issuing interim findings by now. And if you've got passengers willing to pay for in-flight internet access to stave off the boredom, why not tunnel the telemetry over the link?)

|


41 Comments

user-pic
1:

They could do real time streaming of telemetry tomorrow, but who's going to pay for it?

Aviation is very conservative, and airline owners are very cost-concious. Find a way in which it can save money or make pilots look good and the momentum would build overnight.

2:

If you're going to go to all the trouble of streaming that data, then you could also archive it for future reference and/or data-mining. I'm sure the engineers and aircraft designers could find some uses for a very large amount of real-world telemetry data.

3:

Ian: that's what I was suggesting -- let the customers pay for satellite internet, and piggyback the telemetry stream on top of that. Done right, they ought to be able to turn a profit on it (or at least break even).

4:

There is realtime telemetry. It's not incredibly expressive, though, being designed back in 1978 when a 64Kb/s mp3 stream of screaming passengers would have been considered science fiction (in addition to bad taste, I suppose).

Anyhoo -

http://www.airdisaster.com/cgi-bin/view_details.cgi?date=06012009®=F-GZCP&airline=Air+France

At 0210 GMT, a series of automated ACARS messages began to arrive at Air France, indicating problems with the airplane's flight control and electrical systems. The first indicated that the aircraft's autopilot disengaged and the fly-by-wire system had changed to alternate law. Between 0211 and 0213 a flurry of messages regarding ADIRU and ISIS faults arrived, at 0213 PRIM 1 and SEC 1 faults were indicated, and at 0214 the last message received was an advisory regarding cabin vertical speed.

Details of the ACARS system at wikipedia: http://en.wikipedia.org/wiki/Aircraft_Communication_Addressing_and_Reporting_System

5:

I worked on the implementation of some of the inflight Internet services - it was a hoot!

Anyway, most of them (though not all) are implemented on cellular networks, rather than satellite. That clearly wouldn't work over the Atlantic.

user-pic
6:

Clearly some of them use satellite...I had the pleasure of Internet access on a Lufthansa flight from San Francisco to Frankfurt a few years back. (Unfortunately, I heard they canceled the service.)

7:

It's not as if it'd be much data overall, just a slow steady trickle.

Anyway, have pre-ordered the Wireless ebook :)

8:

Here's an interesting write-up of the weather during the crash...

http://www.weathergraphics.com/tim/af447/

user-pic
9:

How hard can it be? Every plane transmits its own 4096bit key every n seconds. Spoofable? Sure. But when its the only way of locating the plane, who's going to be spoofing it?

10:

Aside from there already being such a system in place as noted above, IIRC the FAA (I'm not sure about regulators in other countries) has a very strict policy of not allowing passenger networks and the plane's systems anywhere near each other, to the point of not even being allowed to use the same transmission equipment. IIRC Airbus got into trouble with their latest plane because somewhere along the line the two networks passed through the same router. Different physical and logical interfaces with no connections between them, but shared hardware violated the rules.

11:

What they need to do is the conversion from radar to GPS for guidance, then you would know exactly where the plane went down instead of this wild goose chase. Right now all planes guide by radar. This is brutally inefficient and costs fuel beyond reason as well because planes don't go in a straight line from takeoff to destination.

12:

Couldn't they just make the flight data recorders buoyant and ejectable? I do like the idea of full real-time telemetry; the folks in non-destructive testing would go nuts for it. Then again, with the dark ages we're heading into, passenger air service, especially transoceanic service, will be dead anyway.

user-pic
13:

Leper@2: Researchers are already developing systems for detecting anomalous events in telemetry data. The vision would be to do this in real time by streaming the data back to a computing cluster.

See e.g. http://www.nesc.ac.uk/talks/ahm2006/657.pdf, or for an older project, http://www.cs.york.ac.uk/dame .


14:

Chris: I seem to recall reading that the Boeing 787 passenger network and flight systems use the same wires (as in, shared hardware).

Maalekar: what if the GPS network goes down? (It's a major target in event of war between developed nations, and the GPS signal can be switched off or degraded by the US military on command -- just as GLONASS is controlled from Moscow, and Galileo -- if it gets off the ground -- from Brussels.)

Haar: they're not ejectable, but designed to detach in event of a crash. Buoyand ... a lot harder: buoyancy doesn't usually go with structural strength, unless you make them rather large.

user-pic
15:

Well meteorological information is already regularly transmitted from aircraft, very important for the mid-ocean areas where there is no met radar coverage.

There is current move for the transponders to move to ADS-B which will include additional information such as altitude, GNSS derived position etc. Again this will provide more detailed information in tha areas that do not have radar coverage, as opposed to the pilot reporting their position over radio.

The improved accuracy will enable better routing of aircraft (i.e. more fuel efficient) which is an incentive for the airlines to implement it, but piggybacking it on a customer service to spread the cost would seem to ease the transition

user-pic
16:

The airline industry really is stuck in a timewarp over communication technologies.

ACARS is Telex based, right down to the way that the principal backbone firms charge by the character. While the bandwidth could easily be increased there's no reason to do so while the two carriers (ARINC and SITA) have monopolies in their particular territories.

Also, nothing safety critical goes over ACARS: it was designed AFAIK to help airline honour US airline pilot pay scales. In the 1970s, pilots and cabin crew were paid more for time in the air than time on the ground, so knowledge of the precise time of takeoff and landing was important. Gradually the protocol was extended to support other non-safety critical messages such as gates at a destination airport, passenger lists, and so forth.

Given that most airlines are bankrupt, or virtually so, there's no incentive for them to do anything better than what they do at the moment. Not without legislation, and it's been a long time since a flight data recorder hasn't been recovered.

17:

Short answer: harder than you think. Satellite uplink bandwidth is expensive; and one of the reasons why we're able to use it to put GSM BTSs in the middle of nowhere is the introduction of IP on demand on the satellites, which starts the uplink on request. Continuous uplinking is more costly. Yes, you can cache it and burst it, but then 'snot real time is it?

Also, most of the loss scenarios for this aircraft involve things being quite normal up to some point after the last ACARS message. Even the ultimate Boeing-funded wank fantasy that teh airboos computers killed them all! falls down, as losing the electrics would lose you ACARS very early on, long before anything useful.

user-pic
18:

Just check twitter. It's only a matter of time until the first tweets from a crashing plane turn up...

19:

@18 er, dude it already happened

user-pic
20:

I actually use to work on planes, never could get over how these engineers trust,
a bolt and safety wire to hold things together. thats why everytime i fly i have to have a few.

21:

Harr@12

To make the Flight Recorders ejectable they'd need at least to be packed in a mini launch-tube. The recorders would need to be projected clear of the aircraft - either by a blast of compressed air, or by a rocket motor.

They'd need covers that would either break when the recoders were fired through them - or pop off due to the blast of compressed air.

Both the ejection and the creation of debris near the aircraft could make a situation where an aircraft might be able to recover or have a survivable crash much more unlikely. The ejectable recorders also introduce a very dangerous failure scenario: what if they fire when they're not supposed to?

Also - when to eject the recorders? How do you design the recorder to launch when the plane is definitely lost? You could have it fire at a certain altitude and sink rate - thus losing vital data - or after impact, which would imply a much more complex design.

Charlie@14

Buoyancy could be provided by gas-inflated flotation bags. Then again - that would provide another dangerous failure mode: flotation bags inflated in cabin. All that said - under most circumstances I don't think you'd want the recorders to be buoyant: god only knows where the currents would take 'em.

22:

IIRC before GPS existed airliners navigated by a combination of inertial navigation and radio beacons, recalibrating the inertial gyros from the beacons, because they tended to drift significant amounts in the time frame of 1 or 2 hours. Do they not even carry inertial systems anymore?

23:

They use Laser Gyros nowadays.

user-pic
24:

Wow! This reminds me of the old RISKS-L list back in the day in bitnet still ruled. Does there exist a currently active permutation anywhere? Or did it fragment and specialize - I can't believe it just faded away. There were some seriously knowledgeable people commenting there.

25:

ScentOfViolents: google for comp.risks; it's still going, available via email or as a moderated usenet group.

26:

I was thinking about making them float last night, actually.

Making them buoyant shouldn't be that hard; little inflatable gas cannisters and some flotation aids -- there are liferafts in little pods which you throw overboard and automatically inflate themselves. You could do it with tech like that.

But there's another problem with buoyant;

A lot of planes go down in shallow coastal water. If you make the boxes float, by the time your rescue mission turns into a recovery mission, the boxes have been hammered against the coastline by the waves for half a day.

So what you really want is to dangle them on 50ft of cable below a float. If you're on the coast the box'll ground before the waves smash it into the rocks.

If you're in the ocean, the float keeps the box near (but not ON) the surface giving you a chance to go grab it..

27:

How about not streaming the telemetry data out, but keeping the last 10 or 20 minutes of it in a circular buffer, and bursting it out if an emergency occurs (system failure message, pilot declares emergency, etc.), then continuing to stream until the transmission fails? In normal operation you just send status updates every minute or 10. That way you're only using channel capacity when you really need the data. And you get the data for some time before the emergency is declared, giving some chance of seeing the cause(s) in the data.

28:

One good feature of sinking is that it is going to tend to be pretty reliable.

If the things sink they are going to be on the bottom pretty near the crash. So you know where to look and you know you have time to do it.

If they are supposed to float maybe they float maybe they don't. Maybe there are currents. Maybe they go with the surface current maybe they get into a deeper one. Maybe they are driven by the wind. Maybe they are dashed on the rocks.

The boxes are following the old rule of being lost in the wilderness about staying put and waiting to be found rather than wandering about making the search more difficult.

29:

The boxes are following the old rule of being lost in the wilderness about staying put and waiting to be found rather than wandering about making the search more difficult.

This is sense.

user-pic
30:

As I recall, the flight recorder is designed to survive the crash, including fire, and remain attached to a large lump of wreckage. So it's much easier to find. You're not looking for something small.

But it's tricky to find battleships in deep water, never mind aircraft debris. You may have a more accurate location, but the currents aren't often well-known.

31:

ScentOfViolets @ #24, Charlie Stross @ #25:

Also in a near-real-time web form, http://catless.ncl.ac.uk/risks (last issue was this Monday).

32:

"But it's tricky to find battleships in deep water, never mind aircraft debris."

Yeah, but the recorders are at least putting out easily detectable "find me" noises.

The problem in this case is that we don't even seem to know closer than about +/-1000 miles where the plane went down...

33:

Katie@31: Also, just how easily detectable are the "find me" noises when the box is 1000 metres down? Are the IFR boxes even rated for a depth like that?

34:

@32: You clearly underestimate the situation. The depth there is more like 4000 meters.

Meanwhile there is even speculation that the airplane might have been brought down by a meteorite ... (Ok, they have about 12km less atmosphere to speak for them.)

user-pic
35:

speculation that the airplane might have been brought down by a meteorite

Yeah, 'cause it's not like that's improbable or anything...

36:

My favourite crazy-arse theory is that it collided with a cocaine-running AN-12 doing the Westbridge route from northern Brazil to west Africa, operating without transponder or lights for obvious reasons. The only problem is that there's no way one of those crates would get up to FL410.

But as you know, Bob, a laden A330-200 at that altitude is near its aerodynamic limits, and the max turb and green-dot speeds are very close together; so if it hit a big Cb, it would have to descend to recover. Somewhere in the middle of the storm, there's Ivan, trying to climb over it...

37:

Although - having checked - an Il-76 could indeed get up there.

38:

Yeah, 'cause it's not like that's improbable or anything...

"Really improbable" multiplied by 40,000 flights per day, times many years, equals "not as improbable as you might think." There was a near-miss just a couple of years ago, for example.

For a sufficiently large meteor, an actual hit isn't even necessary. A few hundred tons of explosive force would create a bright flash for anyone more than a few miles away (which is what some reports have claimed), and could take an airliner out of the sky pretty easily. A Hiroshima-sized blast happens about once a year worldwide at high altitude, supposedly. An airliner hitting a shock wave from such an event would come apart about as fast as hitting a brick wall.

Even a small rock would take a big airliner out of the air just as effectively as a large antiaircraft missile.

Then, of course, there's Ann Hodges. She was sitting in her living room when a meteorite hit her house, bounced off of her radio, and struck her in the abdomen, bruising her badly. How improbable is that?

39:

I heard it was space aliens!

40:

which is what some reports have claimed

Unfortunately the aircraft in question was 2,000 km away.

41:

Way back to the continuous telemetry question: What available spectrum will support that? 40,000 simultaneous channels pumping out a minimum of 88 parameters (US federal regulation), multiple times per second. If we assign, say, 96 bits per parameter, with no redundancy, and assume 5 entries per second, we're looking at 41kbps or so. Plus unique plane ID, plus encapsulation, plus, plus, plus...