Flight data recorder flash memory

[jaydeehess]

As explained to me;

Flash is written in blocks... data is buffered until an entire block is ready to be written. Blocks are typically multiple kB in size. 256 kW/s, at 4B/W, would be 1kB/s. Assuming an 8kB block, that's 8 seconds worth of data.

In a FDR such as www.l-3com.com/products-services/docoutput.aspx?id=696 would this mean that several seconds of data could be lost when power goes out if it was still being collected in the write buffer?

[Ian Pearse]

Yes, unless they fit an internal battery to cover that.

[PhantomWolf]

If this is in reference to Flight 77, approximately 6 secs of data is missing. This isn't unusual and occurs in a lot of accidents.

[Ian Pearse]

A bit of a pain if it's THE vital 6 seconds though ;D

[stutefish]

I wonder if there are actually any scenarios in which the final 6 seconds would truly be "vital", in the context of a skilled investigation by subject-matter experts...

I mean, are there hypothetical scenarios or real-world examples, where the final 6 seconds of FDR data might hold some necessary information, without which it is impossible to determine the nature of the event or its root cause?

[PhantomWolf]

Well that depends.... As far as Flight 77 goes, it has allowed a gap for CTs, especially those that claim to be pilots, and yet seem unable to fly a kite let alone a plane.

The last FDR data point is still about 1/2 mile from impact, so the FDR doesn't have any information on the period where 77 hit the lamposts, some are using this as evidence that it never did.

[jaydeehess]

Yes it is the Flt 77 FDr that I refer to.

I am trying to get a handle directly on the , up to 6 seconds of lost data.

I am not that familiar with flash memory though.
However I know what the PfT are going to say.
The data rate is 64 words per second, the write buffer on the specific flash chip is 64 , 16 bit words (write block is 64 words) therefore there should be no more than 1 second of data still in the buffer when power goes down.
If there were a spike that caused an erase command then 128KB get erased all at once as the erase block (erase block is 128 KB)
Obviously there is a logical explanation such as transient destruction of the last area of the die to be written to which would include the area proximate to the last block being accessed.

Anybody know someone at L3 Communications that might speak to this issue? It was an L3 FDR that was in that flight.

[dumbtechie]

I wonder if recording right up to the very end- when there isn't an airplane any more but only the wreckage of one- is a requirement or even a reasonable expectation of an FDR, in the light of its purpose, which is to help investigators sort out the causes of a crash. Would data from the last few seconds before impact be of real value to investigators, or would they be more concerned with understanding the process by which the plane came to be a few seconds away from smashing into the ground?

[stutefish]

Sept 26, 2008 14:48:20 GMT -4 PhantomWolf said:

Well that depends.... As far as Flight 77 goes, it has allowed a gap for CTs, especially those that claim to be pilots, and yet seem unable to fly a kite let alone a plane.

The last FDR data point is still about 1/2 mile from impact, so the FDR doesn't have any information on the period where 77 hit the lamposts, some are using this as evidence that it never did.

Well, sure. CTs will always find a gap of some kind. Even if there is no gap, they'll invent or imagine one.

But from a crash-investigation point of view, is anybody saying, "well, without those final six seconds, we really don't know what happened to that plane, or even if it really crashed at all"?

Seems like in almost all cases, what the crash investigators actually end up saying is things like "well, we lost the final six seconds of the FDR data, but the light poles, airplane debris, and building damage do a pretty good job of explaining what happened next, so we're good, thanks".

[LunarOrbit]

I guess if the accident was caused by something sudden like a midair collision there wouldn't be any sign of trouble until the very final moments, but usually in those cases there are other clues to the cause (ie. the wreckage from two planes instead of just one).

[PhantomWolf]

Seems like in almost all cases, what the crash investigators actually end up saying is things like "well, we lost the final six seconds of the FDR data, but the light poles, airplane debris, and building damage do a pretty good job of explaining what happened next, so we're good, thanks".

Rational people say this, irrational ones say the plane pulled up over the building at the same time a massive explosion was set off to hide it and scatter preplaced plane parts all over the place.

[jaydeehess]

I understand that it is naive to assume that showing the Ct's a perfectly reasonable explanation for the missing (up to) 6 seconds of data will cause them to give up their ridiculous contentions.

However, I am personally interested in knowing how data gets corrupted in this fashion.

In the case of Flight 77 one does not need those 6 seconds to understand what happened to the plane. LO however brings up a situation in which it would be reasonable to assume that there would be some use for those final seconds. Why did the planes collide? Did the FDR or CVR show that the pilots were asleep ( no conversation right up to impact ) or that they used the controls to try and avoid collision but the aircraft did not respond?

[PhantomWolf]

Okay, in this case we're talking flash memory and packets that are being written over time. The system stores them until it has the entire packet and then writes it. This means that at the time of the crash any system not physically stored is lost. That's part of the issue. Thsecond R. Mackey explains here

An overvoltage or unstable power during a write operation, caused in this case by collision leading to arcing in power supplies, can easily cause damage to nearby cells through parasitic capacitance.

So the main cause of the loss of data is a power spike caused by the crash itself overloading the electrics as the plane shatters arcing through the cells of the memory near the one being written to at the time of the crash.

[jaydeehess]

That is exactly what I said in my own poists in the same thread. That transients damaging the cels being accessed at the time would be the most susceptible to destruction and the next most susceptible would be those cells physicall residing on the die, proximate to the cells being accessed and that those cells would be most likely to be the last cells written to.

The other matter concerns the data in the buffer which would be lost completely as it had not been written to flash yet anyway.

The way I read the data sheet on this particular falsh memory chip that would be about 1 second's worth of data. ( a write block looks like it equals one frame of data)

I am not sure I am reading everything correctly though. It could be 4 frames worth of data which would mean 4 seconds and the DAU seems to output 12 bit words whereas the chip stores 16 bit words. The extra 4 bits would be , I suppose, either ignored or used as check bits which would have to be generated before being stored.

DAU , as I understand it, outputs 64 (12 bit) words per second but a block may consist of four 64 word frames.

[jaydeehess]

A poster known as McgyverS2000 from the eng-tips forum has responded to the question regarding corruption of data in an EEPROM such as that in the FDR. He has stated unequivocably that he finds it in no way anomolous that up to 6 seconds of data could be missing from the DFDR on Flight 77.

Since the thread did not adhere to the forum rules and since the questioner was not a practicing engineer the thread was removed. Mcgyvers2000 then posted for a short time at JREF in order to continue the discussion.

Naturally this is unconvincing to the idgits at PfT who posed the question at eng-tips.