AA77 FDR Data, Explained

[beachnut]

Indeed. That proves that the last frame is at least impact-3 seconds.

if the system works the way you think it does based on the data you see

but the plane can not crash at the last data point of 173 PA

so yes at least 2 or 3 seconds are missing, not recorded yet or being processed when the crash took place

 

[apathoid]

I was trying to say the position could be off by what we see, yet the gyros are good.

My navigator load the our INS with 17 south instead of 7 south, we were, 600 miles off, he got pissed and dumped it. he could of left it running and used it as a rough offset to back up his cel nav,,,,,,,,,

so if it is possible to have the nav system align offset and stay, and the pilots either entered the beginning start point 20 miles in error, but they are not using the internal navigation but the vor radio nav aids which they have to use anyway, why waste time on your computer nav this would account for the 20 mile offset

I know the nav system can be set up at the wrong initial point, my crew did it, now is that possible with this system?

Somewhat possible, but not 20 miles. When you align the the IRUs, the latitude is sensed(by the earths rotation - thats why you cant align an IRU in flight) automatically. After alignment, the present position is entered via the FMS. If the entered position is off by more than one latitude degree from either the last known position, or the sensed latitude - you'll get an error.

I understand what you have on the go no go, and you are correct but it does no preclude setting up the wrong inial point in the nav computer, the gyros do not care, that much

Each departure gate has an IRS lat/long to make life easier. But, if the pilots fat fingered the keypad - and the entered position is off by less than a minute - the IRUs wont know or care. How far is one latitude minute. I thought it was about a mile, but maybe its more.

our INS need 20 minutes running in one spot to work best, and yes, when you set it up 600 miles wrong, there are some problems and error modeling in the software i imagine will not work very well but the ins gyros are still good for attitude.

Yep, the principles are bacially the same between INS, AHRS, and IRS. The difference is mainly in the gyros(laser gyros over mechanical gyros) and general moderization. IRS still needs updating from navaids, as it can drift greatly - just as INS does. The vertical gyros for the ADI do not need aligment and can still be used even if the IRU has failed("ATT" mode I described earlier).

and I think it bears more time to investigate, and I do agree with your chart, it is the chart and I do not believe it precludes the pilots setting or accepting a 20 mile offset error in the nav system

does this make sense

You also have to consider that the 757 EHSIs have moving map displays with the navaids/airports/runways overlaid. The pilots wouldve noticed Dulles being 20 miles away while they were parked at the gate!

 

[Shrinker]

What would be the chances of a Pentagon gas station employee reporting exactly what the plane did from the Navy Annex on, being exactly the same thing that the plane did in the NTSB animation did?

Anti-sophist you seem to have it all figured out, maybe you can figure out those numbers.

Just curious have you spoke with any witnesses who saw the plane?

Did you know about the witnesses who described a plane that looked nothing like an American Airlines? Have you spoke with them?

I take it you've never bothered to research the accuracy of eyewitness testimony?


http://www.rpi.edu/~verwyc/oh13.htm

However, there is only a small positive correlation between self-confidence and accuracy of report.
Large scale analysis of eyewitness accuracy indicates that eyewitnessess are only about 80% accurate under good circumstances.

The Misinformation Effect: Questions asked can often contribute to the distortion of a memory.

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=2708267&dopt=Abstract

Confidence is neither a useful predictor of the accuracy of a particular witness nor of the accuracy of particular statements made by the same witness.

http://www.findarticles.com/p/articles/mi_m2405/is_1_130/ai_98709930/pg_1

It is interesting to note that the previous literature predicts that most of the eyewitness testimony should have accurately recalled the central details of the Titanic's final plunge. However, both the Senate and the Board of Trade investigators found results that were in direct contrast with current forensic evidence. ... Most of the eyewitnesses' testimony (15 eyewitnesses out of 20) is consistent with current forensic evidence...

 

[beachnut]

You also have to consider that the 757 EHSIs have moving map displays with the navaids/airports/runways overlaid. The pilots wouldve noticed Dulles being 20 miles away while they were parked at the gate!

I would be using the old junk

but then I did skip the new computer in the tanker too, but it did not save time

right, there would be errors not modeled in the filters, and the nav system could be upset.

you have the book answers for today, my ideas are using AF INS, from the late 70s

Still need to know where the position comes from.

So you plotted about 20 miles offset, that is about where it ended up with me.

I agree with the map stuff, some thing has to line up any integration of the HSI display,,

The errors would not be as great with an east west error in present position as would be a north south error, just thinking about errors in nav, but even that is my bs shot at nav,

I still think it is possible to set it up 20 miles off, and I agree with your stuff, wonder if you can have an old fashion HSI on the monitor and use the jet like a navy guy would!

I am just brainstorming about that 20 mile error, I know you are telling me I can't but gee I have told my kids the same and sure enough they prove me wrong and do it, red flags and all!

so I can imagine my 20 mile error, and not caring about it, or a pilot using radio nav aids, he has to anyway

you do not need the map, you do not need the computer nav

you can use the radio aid you have to use to do the departure, and if you have radar vectors you just need a compass

and even if you were 20 miles off you could fly INS direct to LA or SF from DC and center would never know you had messed up your ins, cause you would be on vectors for arrival or radio nav aids well before you were 100 miles out, and we are only 20 off

the pilot could know the airport, no need for computer crap

just my stupid pilot ideas, sorry for the crap but I feel even stronger on the 20 mile error on takeoff was a mistake or known, or ignored, or not known as a possible explanation on the error in the first place (the old INS would never get on the real position if you started that far off you would have to do some work helping it)

then thinking about KAL 007 messing up their way points or nav system and my own nav off, as in a real person as the nav, by 600 miles, it is 600 17 south vs 7 south, or so

sorry to go on, guess you did not say it couldn't happen just unlikely

I bet I am wrong but I want to know more before I let it rest

so can you use the HSI display without the nav map, and just set it up with vor course and bugs and lubber lines etc

and like you said INS errors would be detected by the software cause you are not really on the part of the earth you say you are and things happen different but I was thinking that those errors are more evident in a N-S mistake

thanks for the book answers, good job

 

[MRC_Hans]

*snip*
How far is one latitude minute. I thought it was about a mile, but maybe its more.*snip*

It is exactly one nautical mile. And a nautical mile is about 1900 meters. A US mile is about 1500 meters.

Hans

 

[apathoid]

I am just brainstorming about that 20 mile error, I know you are telling me I can't but gee I have told my kids the same and sure enough they prove me wrong and do it, red flags and all!

I didnt say you can't! Its just that this is something I've dealt with before and have hands-on experience with. I even know a sooper special trick that the pilots dont know about that 'fixes' 99% of the align disagree errors. However, my trick only works against the last known position and not the sensed latitude.

then thinking about KAL 007 messing up their way points or nav system and my own nav off, as in a real person as the nav, by 600 miles, it is 600 17 south vs 7 south, or so

I'm not familiar with this incedent, but if it was an overwater flight, INS/IRS is susceptible to drift over time. Depending on latitude - I think the max error is can be as much as 10 longitude minutes and 5 latitude minutes. Thats pretty substantial.

sorry to go on, guess you did not say it couldn't happen just unlikely

Quite correct.

so can you use the HSI display without the nav map, and just set it up with vor course and bugs and lubber lines etc

Yep.

and like you said INS errors would be detected by the software cause you are not really on the part of the earth you say you are and things happen different but I was thinking that those errors are more evident in a N-S mistake

If you take a 757 and move it along a parallel of latitude 5000 miles or so, you can still align the IRS 5000 miles off, but you have to stay within 1 minute!

thanks for the book answers, good job

Notta prob.

ETA: Well, I've read up on this a little and the acceptable align disagree is 1 latitude minute and/or 1 longitudinal minute for the last known(stored) position while IRS is in alignment *AND* (cosine of entered lat - cosine of computed lat) > .01234
(sine of entered lat - sine of computed lat) >.15 ..for the sensed position after alignment finishes.

Time to crunch some numbers

 

[apathoid]

It is exactly one nautical mile. And a nautical mile is about 1900 meters. A US mile is about 1500 meters.

Hans

Ah. Thanks!

 

[Anti-sophist]

Did I understood this correctly? Nobody knows (not even NTSB) when the last frame was recorded prior to impact? They just adjusted the timescale in FDR-data to match T-1 or T-0 before impact?

If this is true then there can be even 5 seconds missing from FDR and we nor JDX can't say or analyse anything for the last altitudes, accelerations etc. So the whole FDR-thingy is pointless for any argument?

The :44 frame was the last completed frame, that's why it was included in the CVS file (remember, the CVS file was meant to generate plots). The :45 and :46 frames were both partial frames. According to the NTSB, at some point during the :45 frame, there was a temp. power loss, and when it came back up, it regained synch to start the :46 frame, before then stopping completely.

JDX and them would like you state that the :46 frame is impossible because the NTSB said the impact time was at "9:35:45". We have no idea how accurate or precise that time is, nor how accurate and precise the CVS times are. They seem to think that the :44 frame started at :44.00000 according to Naval Observatory. That is, of course, gibberish.

Relative to the CVS file (which shows each frame starting on each new second), the likely time of impact is somehwere in the :46-:47 range.

 

[uk_dave]

You mean to say.....after 8 pages of impressively dense and detailed and incomprehensible (to me) posts, it comes down to an arguement over 1 SECOND?

 

[UnderTow]

Firstly, I would like to apologize to Anti-Sophist for dragging him about over and over again over and over. I'm sure it was a test of his patience and thanks Anti-S for just not telling me to STFU.

Now to clarify that I know what you've been saying and that we agree on the "implied time stamp" of a 717 frame I will make up an example.
If a 64wps frame is being recorded and starts recording at 0.000 seconds with the first word/parameter being recorded is Time (and let's say the value for this Parameter is 12:00:00 ).
A parameter which is placed in the 8th word of this frame has an implied timestamp of +0.125 seconds from the beginging of the frame. Let's say this parameter was Vertical Acceleration. Therfore, VertAcc was recorded to the FDR at 12:00:00.125 FDR Time.
Continuing, every 1/8 of a second, a new value (updated) is placed in the frame but still correlates to the value of the first word (which was Time 12:00:00) until the 64 word frame is complete and a new updated value for Time is placed in word one and the process starts over.

Now, given a word/value/parameter which we do not know the location of within the frame. It could be placed anywhere within that second. Therefore all we can say is that during that second, value X was recorded.
Also, given two different value/parameters like Altitude and Speed, even if they are placed next to each in adjacent words within the frame, thier time of recording within the FDR would still be 1/64 seconds apart. But they still were recorded within/during the time of the overvall frame of 1 second (maybe at 3/64 or maybe 60/64).

Is all of this agreeable?
If so and given an understanding that a parameter required to be recorded 8 times a second must be placed evenly throughout the frame and that a parameter recorded only once per second could be anywhere within that second.
Would you agree that the following table is correct?

(Additional Note: For each subsecond record, they must differ by at least 1/64 of second since 2 parameters can not occupy the same word unless thier resolution is small enough to only need part of the whole 12bit word. Therefore, if VertAcc is exactly recorded at 2/8ths, then LongAcc/LatAcc would both be either before or after this value/parameter/word)

 

[Anti-sophist]

That table is correct as long as you understand that each sample has a time error range equal to to it's sample rate. In other words, the altitude parameter is placed in 0.000-0.124 row, but the altitude's real error range is 0.000-0.999 (1 second range, equal to the sample rate). Similarly, the first V-Acc could have been recorded anytime between words 1-8 (0.000 to ~0.124), and all the subsequent V-accs would shift accordingly.

It can be misleading, in this format, to most people, who assume that the altitude, for example, was recorded at 0.000.

With that understanding, that explaination is correct.

 

[CurtC]

The :44 frame was the last completed frame, that's why it was included in the CVS file

This error has been propagating for a while, so I'll say something. It's "CSV" file, as in "comma-separated values."

 

[Anti-sophist]

This error has been propagating for a while, so I'll say something. It's "CSV" file, as in "comma-separated values."

Yes, I'm well aware. The problem is I used CVS(http://en.wikipedia.org/wiki/Concurrent_Versions_System) every day at work, and so I'm repeating it out of rote habit. I've editted alot of my posts to fix it, but I've missed some.

 

[Darth Rotor]

It is exactly one nautical mile. And a nautical mile is about 1900 meters. A US mile is about 1500 meters.

Hans

On nm = 2000 yards = 6000 feet. One statute (US) mile = 1609 meters.

DR

 

[CurtC]

The way that I always remember it is that a nautical mile is about 15% longer than a statute mile.

And an arc-minute of latitude, or of longitude around the equator, is right at one nautical mile. For longitude not on the equator, you multiply by the cosine of the latitude.

 

[Anti-sophist]

You mean to say.....after 8 pages of impressively dense and detailed and incomprehensible (to me) posts, it comes down to an arguement over 1 SECOND?

Or two. Yes.

Unfortunately, 1 or 2 seconds is an absolute eternity when we are talking about events that happened 1200 feet apart on an object traveling 500mph.

 

[beachnut]

It is exactly one nautical mile. And a nautical mile is about 1900 meters. A US mile is about 1500 meters.

Hans

nautical mile converts to:

• 1.852 km (exact)
• 1.150779 mile (statute)
• 2025.372 yard
• 6076.1155 feet
• 1 arc minute (within 0.2 percent)

I know we use 6000 feet for a nm

guess it depend on who we are

 

[UnderTow]

I just want to get this out.

I have worked the FDR binary to find repeating sync words which mark the beginging of a 1 second frame. (I can explain this later)

If my anaylsis is right, I found the following whole flight periods in the data file.

Working backward in time from the Pentagon flight the flight periods are:
86min 20sec (PentFlight)
273.6 min
260.6 min
233.13 min
251.9 min

There are more previous to this but I stopped there for now.
And OMG what happened to the color scheme here :O

 

[Anti-sophist]

So you found the 4 sync words you expected? Are the frame lengths as expected? The time of the flight fits.

Beyond that, I have no idea what you hope to accomplish w/o the data frame descriptor. Even reverse engineering a single parameter would be a borderline miracle.

 

[DavidJames]

And OMG what happened to the color scheme here :O

phew, I thought it was me , I'm glad it's back to normal.