Why does it take half a year to decode an airplane’s black box?


In light of the recent plane crash in Pakistan, reports suggest that it will take 6-7 months to decode the black box.
The company that made the black box surely knows how to decrypt their encryption, so why would it take so long?
Also, assuming the encyrption is super-complicated, what sensitive data would warrant such encryption? Is it just voice recordings, or something more?

In: Technology

Normally they are slightly damaged due to the conditions of the crash, then it isn’t the decoding that is a problem it is interpreting the data, the data tells you what happened and when but not all the other things that happened. So an engine failed at 11.03 20 seconds later the plane goes into a steep dive. Is it crashing or are the pilots trying to dive to restart the engine or even put out a fire?

Im no black box expert, but from what I know, it’s basically a log of all of the aircraft’s sensors and all of the things the pilots did. Imagine a massive excel file with hundreds of columns(vertical) . Each column is a sensor, be it altitude, aircraft nose angle, air speed, engine power, amount of fuel, door closed sensors, etc, anything you can imagine that an aircraft might have. Then every second, the value of all the sensors fills out a row(horizontal) in that excel file. Well, that’s a shitload of data, and looking through an hour or more of that pure data is going to take a lot of time. You’d be looking for abnormalities in sensor readings or weird combinations of sensor values, sensors that stop working, pilot actions that didn’t result in the expected sensor response, etc. It’s probably just a ton of crap to go through, and unless you find a real “smoking gun” figuring out what exactly happened would take a lot of time a knowledge of the entire aircraft. That make sense?

It’s probably inaccurate reporting.

Most likely they can read it right away, but it’s the full analysis what takes time. Sure, the black box says what each sensor reported, but what does that actually mean? Eg, did the black box record low fuel because there was too little to start with, or a sensor malfunctioned, or there was a leak, or some unexpected condition caused the airplane to burn fuel faster than normal?

They may need to collect all the wreckage and look at the remaining bits, and interview survivors.

They have to be careful with these things, and make sure not to blame the wrong thing just because that can cause additional harm.

No expert, but decode in this sense doesn’t mean it’s encrypted for security reasons and they’re thinking “oh shit, we forgot the password” it’s more to do with the complexity of the task at hand.

It is also extremely important to make sure all the data is correct and none of it is corrupted (bear in mind, this box went through a hell of an impact) and if something is reading the wrong figure, it could screw up the whole investigation. So they have to error check and back up (safely) millions of lines of code reporting everything that happened, from thousands of sensors, millisecond by millisecond. Not an easy task.

It depends on the type of aircraft and the operator but usually there are two “black boxes” — one is the CVR (cockpit voice recorder) and the other is the FDR (flight data recorder). The FDR records panel settings, alarms, fuel flow, and similar. If they are damaged in a crash it can be difficult to extract the information. Sometimes it requires the expertise of the regional transportation safety board and sometimes it even requires the original manufacturers to extract meaningful information. All of this needs to be methodically analyzed in an attempt to understand what happened leading up to, during, and sometimes even after a crash. Sometimes it’s obvious and sometimes its a major puzzle.

When people use the word decode in contexts like this, they’re talking about taking raw data, packed for efficient storage, and decoding it into a form that is understandable by humans, and then working out an interpretation of events.

Normally extracting the data is easy, but if the recorder is damaged, it could require specialized skills and tools to retrieve and decode.

Normally the vast majority of the time is just spent piecing together events, and making sure you got it right and every anomaly in the data is explained.

Imagine someone’s car crashed and you want to understand what happened. You have a recording of the crash, but it’s not a video – it’s a recording of the exact speed each wheel was doing, what the brake pressure was, what the position of the accelerator was, the engine temperature, the air con setting, and so on.

You’d eventually be able to understand a lot about the crash, but you’d need an expert to look at all that data and tell you what it meant.

Now imagine that for an aeroplane except there are several hundred times more pieces of data to look at, and the ultimate cause of the crash may be buried deep in one obscure sensor reading. That’s what you’re ‘decoding’ with a black box.

Edit: thank you for the gold and awards! My first ever!

AliHB brings up a good point. If the company that makes them can’t read them, how do they know they’re working? Why “invent” a way to read them after the fact? Whichever system they use to record the data, there should already be a user friendly system to read the data.

this is bad reporting. The FDR will be repaired and all the data will be pulled. it’s raw data, but from that they can build a pretty good picture of what was happening to the aircraft’s systems leading up to the crash. That data will be compared with crash data from other aircraft and compared with findings from air accident investigators who have examined the wreckage and any eye witnesses/survivors/video footage and the Cockpit Voice recorder which has all the pilot and copilot’s conversations right up until they die.

At the 6-7 month stage the press release will be “we’re pretty sure the plane crashed because of XYZ reasons”

Basically a black box regards all the data that is occurring, think of the information as words. The words are then put on the page in just every which way. The information isnt in any sort of order or make any sort of sense.

It is up to the engineers to take those words and put them into a paragraph so that it can be understood.

There are two parts –
1) Flight data recorder which records hundreds of parameters for last 25hours (new models). It does that in raw data (timestamps, numerical parameters etc).
2) Cockpit voice recorder which records the boom headset microphones and a dedicated microphone channel along with timestamps upto 2hours.

Now in such cases the post crash fire and the impact itself damages the structure and maybe data ports etc. So the technical team of the manufacturer have to somehow extract the data as much as possible and verify the integrity if the data, match the the timestamps of both the recorders and produce the files in readable format for the investigators.

All this effort takes quite a while aa you can imagine.

The final output which investigators use to create an analysis video looks like [this.](https://youtu.be/DpZzwcDyaDY)

As others have mentioned, the analysis is the long part. But ensuring integrity of the data is a huge challenge as well. Plane crashes are violent affairs, and getting to the storage media, and getting the data off in a safe way is a forensics and computing nightmare.

It’s not 6 months to computationally decrypt the data from the device. But the data is pretty raw. When they make a plane they don’t know what might go wrong with it so they record as much as possible of every instrument possible. The data is raw and there’s a lot of it. Piecing through that forensically to reconstruct what happened and actually pinpoint the cause of the crash, and then cross check that against all other available data so you can be confident you know the story of what occurred… that takes time.

This process isn’t performed a hundred times a week, either. It’s a rare event and so it’s not like pumping out an electric car which people have been tuning a factory to do, making small improvements over and over day in and day out until it’s fast and reliable. Every crash is different.

The delay you are talking about isn’t really the delay in downloading the data. Usually the data download is done rather quickly. As others have said, it’s the analysis that takes awhile. Aside from that, NTSB (here in the US) hearings and reports can become slightly political and it takes quite a bit of time to get all of the involved parties to discuss/analyze/vote on probable cause and contributing factors.

As a side note, these flight data recorders (the infamous “blackbox” as it’s referred to) have other uses as well. Airline safety departments regularly download the data and company/pilot analysts audit the flights to see where there are potential higher risks in the operation. At my previous airline, I was one of these analysts. It was very fascinating and we were able to improve many of of procedures based on recommendations from those findings.

Finally a question I can answer. I’ve been an avionics tech for about 10 years.

A Cockpit Voice Recorder (CVR) and Digital Flight Data Recorder (DFDR) are what is commonly revered to as the “black box.” Fun fact, they are in fact not black, but international orange and typically consist of two different systems. One used to record pilot voice and the other used to record aircraft flight surface data.

I am not sure at all where this idea come from that it takes 6 months to “Decode” that data as a recent crash only took us 10 days to decode. Sometimes it can take longer to decode for a few reasons however.

1. It’s destroyed – the CVR and DFDR are in a very tough case we simply call the “Crash Survivable storage unit.” (Real original I know.) Even so, this case may not survive an impact into a mountain. In this unfortunate case we simply may not be able to retrieve the data.

2. It got “lost” – this is pretty typical in underwater scenarios but we have underwater locator beacons that operate at a depth that I cannot disclose. Just know that it’s pretty deep. Sometimes though, it may exceed this depth and it makes it difficult to late.

3. Some maintainer didn’t calibrate the damn thing correctly. – this suck because it’s our own people’s fault that we can’t quickly calculate what happened. People usually go to jail for this stuff though so there is that.

The data is generally decoded quite quickly. The meaning of the results is tested and analyzed by many groups before the results are released.

It’s not about getting the data off, it’s about analyzing why the throttle was in a high power position, and the engines were spinning slowly. That can happen from fuel problems, engine damage, control problems, or other things.

So they don’t say “engines were not making power, we don’t know why”, but don’t publicly reveal the results until the meaning and cause is known.

Pilot here.

**TL;DR… The time is in the full investigation, not just the box data.**

If a Flight Data Recorder is intact, getting the data is relatively quick and easy once found. There may be some damage to components or incomplete information that requires interpretation. Forensic techniques can get some of that data back. Relative to the whole endeavor, it’s a small portion of the effort.

That data is correlated to other evidence to figure out what happened. That evidence is gathered physically and through interviews and records.

Here’s a non-comprehensive list of things an investigation will be looking for:

* Aircrew Background: Mannerisms before flight. Diet. Illness. Fatigue. Alcohol. Medication. Stress. Major life events. Reputation. Training history and performance. Identified human factors in all of this history. Personality (Excess/Lack of Confidence, attitude toward bending/breaking rules, communication skill etc). Group Dynamics. Social barriers (see Korean flight that crashed in SF). Conflict. 24hr(or more) history to include what they did day of, night before, day before, even week before. Family Dynamics. Things they’ve said not just in the cockpit, but out of the cockpit.

* Aircraft Operation: within SOP/Regulation/Training/Aircraft Limits. Deviations from normal operations and reason. Acceptance of failed equipment (there’s always failed equipment on a plane, there is guidance on what is safe to take flying), unusual control inputs (too aggressive, not aggressive enough, autopilot usage), systems setup (navigation used, parameters such as Baro Bug set correctly)

* Environment: weather, history of aircrew flying in said weather, equipment and qualification suitable for weather, aircraft/aircrew/airport/maintenance/operational/navigational complications due to weather, airport/control operations in general (pace, safety, same considerations for controllers as there are for aircrew), company culture, dispatch helpfulness, clarity in communication.

* Physical/Engineering: Part exceeded limits. Part failure due to random/user error/malice/engineering/maintenance/unforeseen incompatibility. Personnel access to those parts same considerations for maintainers/engineers as for aircrew and controllers and their companies and training.

* Training: Adequate. Culture…

Just fill in the gaps already because I’ve only got like half of them. All of this has to be strung together to figure out and rule out causal factors to an incident, and come up with recommendations to prevent further related incidents, and do so with a high degree of certainty that all of this is correct.

Assuming, like most others here, that “decoding” refers to the publication of a final repost rather than just reading the data.

The reposts can be some hundred pages of dense technical detail to understand

1. What happend
2. What [chain of events](https://en.wikipedia.org/wiki/Swiss_cheese_model) lead to it happening
3. Which measures need to be taken to avoid this happening in the future

Take your pick of [accident investigation authorities](https://www.icao.int/safety/AIA/Pages/default.aspx), you’ll find loads of reports, not only on fatal crashes, but all sorts of incidents that shouldn’t have happened (and there are [loads every day](https://avherald.com/))

Then there is also the matter of who (i.e. which state’s authorities) actually [take part in the investigation](https://www.icao.int/about-icao/FAQ/Pages/icao-frequently-asked-questions-faq-10.aspx). In this case it’ll be Pakistan (because it both happened there and the aircraft was registred there; were those different, add another), France definetly (because Airbus; [CFM](https://en.wikipedia.org/wiki/CFM_International) [engines]), US probably (see: CFM), unlikely UK, Germany, Spain (because, again, Airbus), people from Airbus, CFM, whoever made the Landing Gear, FDR, and whatever else may be found to have occured.

Looks like PIA isn’t doing their own [A320 training](https://www.piac.com.pk/corporate/training/flight-simulator) (website only shows B747 and B777 sims), possibly they’ll want to interview who trained the crew (did the have a history of becoming [task saturated](https://en.wikipedia.org/wiki/Eastern_Air_Lines_Flight_401#Cause_of_the_crash), not follwo procedures, etc.)

So you get people from all over the world together to figure this out. Let’s say the landing gear guys sort through the bdebris, find all ‘their’ parts and figure out why it didn’t work. That’s one (important) part of the investigation, but maybe it doesn’t really tell you why the aircraft crashed (that stuff [just happens](https://en.wikipedia.org/wiki/LOT_Polish_Airlines_Flight_16) [occaisionally](https://en.wikipedia.org/wiki/JetBlue_Flight_292)), because it did not impact any other systems. Then (or probably in parallel) you’re looking at the engines (which again, [somewhat optional](https://en.wikipedia.org/wiki/List_of_airline_flights_that_required_gliding)).

Depending on how much lab work needed to be done on parts to check for failure modes (and distinguish them from damage due to the crash itself), you’ve spend a couüle of months at least and still have only parts of 1) and 2) and very little of 3).

That stuff [takes time](https://www.bfu-web.de/DE/Publikationen/Untersuchungsberichte/untersuchungsberichte_node.html) (list of German accident reports, first colunm is the date of the incident, the last colunm is the date the report was published; some have taken multiple years).

And when it comes to “reports”, look who’s doing it (and how much they know about aviation), remember “[Boeing 777 will struggle to maintain altitude once fuel tanks are empty](https://external-preview.redd.it/w4ciy2Lb81hGCvDVMFiju8f3YuiDUZimJkyvF3Fc2xM.jpg?auto=webp&s=f1bcafb149054c90a52586b3c7056fc6bfbd9d44)” was a thing. The best thing to do in case of aviation accidents is to forget about them for a year or so), then read the official report. Everything else is pointless speculation.

I hope this is somewhat coherent, I should be sleeping for a while now, [but . . .](https://xkcd.com/386/) 😉

Decodeing the flight data recorder and the cockpit voice recorder dont ussually take very long.

BUT. An accident investigation is a VERY detailed investigation that involves more than just the data recorded by the aircraft. Investigators dont usually reveal the results of any part of the investigation until ALL parts of the investgation and its reviews are complete.

This may take many months.

Hey! I work on this equipment! Here is my input. It doesn’t take this long at all to get the data or even put it into a video format for peoples viewing. This process takes a few weeks from recovery of the black box to someone sitting and seeing a finalized video. A download of the data takes at most 1 hour on most different “black boxes”.

Fun Fact the actual color is “International Orange”. Nothing black about it.

But like others have pointed out, it take awhile to figure out what exactly happened. The public will not see any of the data/video until experts have agreed what happened or at least come up with a finalized report.

Some things in the comments that I feel need to be addressed for correction or clarification is:

Most DFDRs (blackboxes) only record from last power on to last shutdown or a limited amount of time like 12 hours. For crash investigations you typically only need to analyze the last few minutes to see what went wrong. The CVRs (cockpit voice recorders) are usually very limited on recording time(some 4 hours) due to the Pilots Unions because not all of the cockpit lingo needs to be brought up and become public information for privacy reasons. Imagine a pilot who is now dead being heard on takeoff about how he hasn’t gotten laid in weeks from his wife but thank god for his mistress in New York. Unions and crash investigation teams work hand in hand to make sure laws and union agreements are kept in mind.

That movie with Denzel Washington (Flight) shows good examples of Unions being a big part of these investigations. Someone asked the question about the data on the blackbox after normal flights. This is usually just dumped after each flight by the next flight recording over or sometimes frequently downloaded and sent to engineers to analyze the data to make sure the plane is recording this information accurately just in case it does one day crash.