@ BJ-ENG
It's also interesting that the BEA allowed Honeywell to dig into the electronics. In section 2.4.1 of the EUROCAE document: MINIMUM OPERATIONAL PERFORMANCE SPECIFICATION FOR CRASH PROTECTED AIRBORNE RECORDER SYSTEMS, it states:
NOTE 1: Microscopic examination of the surface characteristics of memory
devices is not acceptable as means of data recovery.
NOTE 2: The repair of individual memory devices is not permitted.......
Seems a rather restrictive policy considering that all measures need to be applied to discover the truth - maybe it's a legal issue...
I suspect that this is a legal statement: the FDR design has to meet the specifications (that you can read out the data after up to 30 days underwater etc) without requiring those techniques. They want the design to be readable through the normal ports. But if for some reason that fails, they will go to any extent required.
I agree with you that there is some risk of exposure to water, however there are a lot of layers of defense: the CSMU casing, presumably some kind of coating on the PCB, the chip package.
In the end, IMHO, they are pretty likely to be able to read it, even if it requires extensive measures like removing the die from the package and using the equivelant of a wafer probe tester to read it out.
@ mm43
That is my understanding, and as a data state change can only occur with an applied voltage, any potential short circuits can be ignored.
I am sure that if the initial download in France fails, the CSMU will be flown to the USA to enable Honeywell to access the substrate and interrogate the chips directly.
You are correct: a flash memory chip requires a programming voltage. It is unlikely to have lost the contents although not impossible.
There are several ways to read the chips out: on the circuit board (not sure I'd try powering it up at this point, at least not without checking it carefully), by removing the memory chips and reading individually, and (at the extreme) removing the chip die from the plastic package and probing the bare die.
I couldn't see any mention in the FDR product description link, but there might be some amount of error correction/redundancy built in (something like storing 2 copies of the data in case one is corrupt, although there are more clever ways to do it). Anyone know if the solid state FDRs do this?