10
\$\begingroup\$

All data recovery companies, regardless of skill, unanimously say that if the memory chip of a device has just a hair line crack, data recovery is impossible. Not unlikely, not expensive, but impossible. One company even stated that even the FBI can't retrieve the data. Is this true?

Why is this? I find it hard to believe if just a tiny section of an extremely common chip has a tiny crack, all of the data is completely gone.

I would have thought some talented person somewhere would be able to patch up the area of the chip and get some of the data back...

Is it something to do with the charge? I know flash memory uses transistors to store its ones and zeroes in the form of an electrical charge. If the chip is cracked, do the transistors "short-out", turning them all to zeroes, something like that? Is the data gone rather than irretrievable?

All I want to get are some awesome holiday videos back. Thought they were gone for good, then I learned about data retrieval, thought I had a good chance of getting them back and then I realised there's not really a chance at all if the memory chip is cracked.

How much would retrieval be? Hundreds? Or thousands? A million, as RedGrittyBrick says? If were to hold onto the memory card, in a few years do you reckon the price of such an advanced retrieval could come down? Or is this just being unrealistic?

We're talking about a 256mb sd card here.

I suppose technologies are moving away from SD cards and more to integrated memory and then goodness knows what else...atomic memory, DNA memory..You don't see people coming out with new advanced procedures regarding cassette tapes today, do you? Should I just bite the bullet and give up?

Also, I'm not even a hobbyist in this field, however I am interested in general by how things work so if someone could explain simply the problem I would appreciate it.

\$\endgroup\$
9
  • \$\begingroup\$ given that it is a lattice of nanoscale nand gates your tiny hairline crack is a grand canyon on that scale... no hope \$\endgroup\$ – Paul Sullivan Apr 13 '14 at 22:00
  • \$\begingroup\$ I don't know about the FBI's capabilities, but if the information is still there (and it should be there, stored on floating gates, except where the physical damage is) it should be retrievable. At a cost that is probably prohibitive. xkcd.com/538 Why? Do you have $100m worth of bitcoins on a cracked chip? \$\endgroup\$ – Spehro Pefhany Apr 13 '14 at 22:00
  • \$\begingroup\$ regardless of the directly damaged area it is all about how the surrounding areas are accessed... think about how bits either side are accessed (without directly observing/probing eac bit \$\endgroup\$ – Paul Sullivan Apr 13 '14 at 22:02
  • \$\begingroup\$ @PaulSullivan You can probably wire a section of a working chip to the old chip to retrieve rows of information. This stuff is possible with ion mills and such like. Just very expensive. They can re-wire individual chips. \$\endgroup\$ – Spehro Pefhany Apr 13 '14 at 22:13
  • \$\begingroup\$ How sure are you that the actual IC in your 'storage device' is physically damaged? Sometimes (often?) the actual circuitry does not fill the card's case completely so that a damage at the edge for example may be purely cosmetic/mechanical. Damaged contacts should also be fixable. \$\endgroup\$ – JimmyB Apr 14 '14 at 13:42
14
\$\begingroup\$

All I want to get are some awesome holiday pictures back.

Lets be frank. They are not worth $1M to you are they. For that sort of money you could go on that holiday again a few times and recapture the same photos or something equally awesome.

I would have thought some talented person somewhere would be able to patch up the area of the chip

The technologies used to manufacture flash memory do not lend themselves to repairing cracked ICs. The manufacturing approach is to simply test and discard defective dies. There is no established technology capable of making repairs.

A silicon fab capable of manufacturing ICs costs $1bn. Any plant capable of repairing ICs is likely to be similarly expensive - it would need high volume usage to be economical. There just isn't that demand, most people probably find it cheaper and easier to copy photos onto a few $50 hard disks than to hope for sci-fi movie tech to rescue them.

Your talented person might need millions of dollars worth of equipment, a university-grade set of research labs, a large team of postgraduates and years or decades of funding.

and get some of the data back.

Some of the data may still be present in undamaged parts of the chip but conventional retrieval is likely to be completely dependent on damaged portions.

We're talking about a 256mb sd card here.

Imagine examining 2,000,000,000 individual grains of sand one at a time under a microscope. That's the scale of task involved. Grains of sand are much bigger than transistors of course. Transistors on flash-memory chips are far too small to see.

in a few years do you reckon the price of such an advanced retrieval could come down?

In a few hundred years?

Should I just bite the bullet and give up?

Unless you are a billionaire with nothing much else to do who can fit this into some kind of bigger plan.

\$\endgroup\$
5
  • \$\begingroup\$ Do you know, this tough answer makes me feel a bit better actually. The other answers left me in purgatory a bit. $1M, eh? Had no idea. And by the time the price could come down, I will be (1) long dead and (2) the data will have completely corrupted away. Such a tiny piece of damage though. You can barely see it (though it does go right across the chip).. \$\endgroup\$ – Starkers Apr 14 '14 at 10:47
  • \$\begingroup\$ I don't think it will be $1M. Expensive, yes, but the equipment exists I think. For some initial research, see here: ee.ucl.ac.uk/lcs/previous/LCS2009/NEMS/Konopinski.pdf \$\endgroup\$ – RJR Apr 14 '14 at 13:12
  • \$\begingroup\$ More here proceedings.spiedigitallibrary.org/… (abstract only) \$\endgroup\$ – RJR Apr 14 '14 at 13:24
  • \$\begingroup\$ @RJR fascinating article, but sadly I think it just solidifies RedGrittyBrick's points. I wouldn't be surprised if it was a $1M procedure. \$\endgroup\$ – Starkers Apr 14 '14 at 16:55
  • \$\begingroup\$ @RJR a large degree of the complications in retrieving data from flash memory are caused by the controller chip (I think) doing all kinds of clever things such as load balancing. With a SIM card it's a lot more basic because there is no load balancing. User data is written to it once, and then read forever. SIM cards also hold about 16kb, one 16,000th that contained in a 256mb sd card. Retrieving from a SIM is far, far more simple and even that is impossible! (I think anyway, the article's conclusion seems hopeful but speculatory) \$\endgroup\$ – Starkers Apr 14 '14 at 16:56
4
\$\begingroup\$

Let's think about the scale of the problem. Silicon chips are very fragile, as are the very thin bonding wires that go from the chip to the pins/pads. They are also susceptible to environmental damage. That's why dies and wafers are handled in clean rooms, etc.

We take low cost, reliable, environmentally sealed chip packaging for granted. It actually took quite a long time to develop to this stage. I mention this because I think it helps to understand that an IC is not just a plastic box around an indestructible die. Chip packing is a big deal and provides a lot of protection that the die really needs.

\$\endgroup\$
2
\$\begingroup\$

The chip is a large casing for a silicon wafer a few mm wide and the width of a few hairs. A crack means flexing pressure and direct damage to the bonding wires at best, or to the microscopically small traces and junctions on the wafer.

Think of it like a sheet of ice. The slightest bend and CRACK. Silicon Wafers are not designed for structural integrity under pressure.

While bond wires are possible to repair, damage to the wafer is practically impossible.

\$\endgroup\$
6
  • 1
    \$\begingroup\$ I don't think "repairing" the chip is the same as recovering the data. I was told that a determined enough individual can use a scanning electron microscope to manually examine floating gates. You may not be able to get back all of the data, but in some cases even a little data is bad. This is why some organizations require electronic mediums to be completely destroyed, not just wiped or partial destroyed. \$\endgroup\$ – helloworld922 Apr 13 '14 at 22:13
  • 1
    \$\begingroup\$ I smell a quantum problem here. Wouldn't direct probing of a floating node with an electron beam cause it to change state? \$\endgroup\$ – Sean Boddy Apr 13 '14 at 23:47
  • \$\begingroup\$ @helloworld922 flash storage is not like magnetic storage in that what you get when you mount it, the file system, the inodes, the data blocks, etc, are essentially emulated by the flash controller, abstracting the physical nand from what is typically considered the "physical" layer and logical layers. Two levels of logical abstraction. This is because of the difference in how nand works, how the controller handles bad bytes/recovery blocks, wear leveling, etc. \$\endgroup\$ – Passerby Apr 14 '14 at 3:48
  • \$\begingroup\$ @helloworld922 Even then, physical damage to a wafer, even a tiny scratch or piece of dust, let alone a crack which cause plenty of microfissures, damages a huge percentage of the flash memory. It will make whatever data you recover, indistinguishable from garbage. It's not trust worthy recovery. Think of a snow flake. Tiny, microscopic. The smallest break in the crystalline structure, and you can never tell what that pattern looked like again. \$\endgroup\$ – Passerby Apr 14 '14 at 3:51
  • \$\begingroup\$ @helloworld922 Finally, we are talking about flash memory, which in common sizes, come in GIGABYTES. Hell, Terrabytes. We are talking 8589934592 bits in a single gigabyte. A determined guy with an S.E.M. will die of old age trying to recover a single 1kb text document on a 4 gb flash drive. A damaged flash is in all PRACTICAL terms, unrecoverable. \$\endgroup\$ – Passerby Apr 14 '14 at 4:02
2
\$\begingroup\$

I think this is a question of capability - as answers above indicate, a cracked IC is likely to be electrically non-functional and possibly damaged if it was powered up at the time.

Many data-recovery outfits rely on clever(ish) software - that is, the device still works and you just have to access it at a very low-level to tease the data out and re-assemble it into something usable.

However, there are stages below that depending on how much time and money you're wiling to throw at the problem.

There was a story on Hackaday a while back where a guy actually de-capped a gameboy ROM and read the bits out using a microscope and some image-processing software to read the set/unset bits. Now, that's not too taxing for an old ROM chip as there's a physical change visible, albeit requiring significant effort & a steady hand. For a modern flash chip, the density is insane and the change may be invisible. However, I would say it's not impossible, just incredibly hard. With suitable technology you could probe the chip die directly and probably read out some of the data, depending on the damage.

In all likelihood, this work is beyond most people's scope/budget. I'd guess that the FBI or MiB or whoever could do this if they felt it worthwhile, but I'd also guess that they wouldn't broadcast the fact to joe public, or that it may be classified or related to other classified capabilities.

Edit to add: An analogy might be a broken record; You can clean up the sound from a scratched record with clever processing, but a broken one can't be played on a turntable. It doesn't mean you couldn't get the data off, just that it's become a lot harder.

\$\endgroup\$
1
\$\begingroup\$

The data may well still be there, the problem is that there is no (reasonable) way to get an electrical signal into or out of the data cells. When you crack a chip you are cracking wires. The minimum feature size in many chips made these days is around 20nm=200 Angstroms. I suppose that if you (a) had an electron microscope, (b) knew exactly how the manufacturer had laid out the chip (what every wire means, and what signals need to be sent where to get the data), and (c) had some kind of electron microscope based soldering iron or micro-manipulator with a tip just a few atoms wide, you could laboriously and painstakingly repair all those broken connections.

But, while data recovery companies (the super duper ones) might have electron microscopes, I doubt they have either (b) any access to the manufacturer's layout or (c) a micromanipulator that could actually be used to do the repair. Also, we're talking about potentially needing to repair thousands of connections, so even if someone were able to do it, they'd probably charge tens or hundreds of thousands of dollars.

\$\endgroup\$
1
\$\begingroup\$

The problem is not so much that the flash memory cells (bits) are damaged so much as the addressing and readout electronics are damaged. Flash memory is not like a record, tape, CD, etc. where the read/write head is separate from the recording medium. In that case, damage to the medium just results in some unreadable data, the rest can still be read just fine by positioning the read/write head over the 'good' portion of the medium. However, flash memory has the 'read/write head' (row and column select wires, decoders, sense amplifiers, etc) integrated on the same piece of silicon as the recording medium (floating gate transistors). As a result, when the die is damaged, huge sections become completely inaccessible due to the damaged wiring. This wiring is built up on many layers on the chip and it is extremely small. And there are millions and millions of wires. If the die is snapped in half, it would take months inside of very expensive equipment to attempt to repair the damage, complicated by trying to not disturb the data. Note that flash memory is directly descended from UV eraseable EPROM, so once the chip is decapped, it must not be exposed to UV light (sunlight, etc.) so the data is preserved. Also, certain techniques used to examine and repair the chip such as ion beam etching could also disturb the stored charge in the floating gates.

\$\endgroup\$
0
\$\begingroup\$

It is theoretically possible on low density <500MB chips, IIRC there was a case fairly recently where someone had stored RSA keys on an old 32MB pendrive and snapped it in half as their front door was bashed in. To decrypt the files would have taken close to a decade but fortunately it was possible to recover enough consistent data from the paper fragments and the two 16MB slightly damaged Flash chips to make an educated guess at the 1024 bit key and with the combined brute force method they retrieved the files in less than a month.

I can only assume that the reason for storing them on such an old drive was data integrity, had that been a more modern multi-GB drive there would have been no hope of recovery ever.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.