29

So, right off the bat, I want to acknowledge that the data on these cards is likely lost to time. However, I am hoping someone may know some best practices to check for sure.

I recently recovered some data loggers that were sitting at 40 m depth in salt water. They were completely flooded. These units were powered by reusable lithium batteries, which were also ruined due to flooding.

I'd love to see if I can salvage ANY of the data from the microSD cards. Are there any best practices for cleaning them, or checking their status without risking damaged my computer or SD reader?picture of rusted and ruined microSD cards

So far I have just lightly cleaned them with a paper towel. I've thought of getting some distilled water and perhaps a mild cleaner, like the type used for glasses or phone screens. I also have no plans to use these cards again. I just want to see if I can get the data that was on them before the unit flooded.

10
  • 17
    I would be surprised if there were any problem reading these, beyond the corroded contacts. Silicon is stable and durable, and (unlike the lithium batteries) these have no significant stored energy to contribute to corrosion.
    – Sneftel
    May 25 at 10:28
  • 2
    @Sneftel I'd be more concerned about the metals in there than the Si - though there's a good chance they're reasonably corrosion-resistant (contacts to Si are often tungsten, and the wire bonds are likely to be gold. The contacts on the one I cut up look like Cu under Au
    – Chris H
    May 25 at 12:44
  • 8
    @JW0914 That's all true on bigger cards/older designs, but in micro SD there's no PCB. Just bond wires from the silicon wafer to the pins. Also a de-ionised water wash is common in PCB manufacture and better than salt water. I'd use DI because I can get it easily. How do you propose minerals get into distilled water?
    – Chris H
    May 25 at 12:46
  • 4
    Definitely micro - I have a bag of both sizes right here. As for water: "consumable" has nothing to do with it - you're not drinking it. Both distilled and deionised have 4-5 orders of magnitude less dissolved solids than the seawater we're trying to get rid of, and 5-6 orders of magnitude less conductivity - remember the card has been soaking in saltwater, so minor differences in purity of lab supplies are tiny in comparison to leftovers if the rinse is imperfect. Tap water is somewhere in the middle.
    – Chris H
    May 25 at 14:04
  • 2
    @JW0914 if it's distilled, the minerals are down to ppm levels, and shouldn't be sold as food. Some bottled water is distilled then has minerals added back in, so it's no longer distilled water. You can buy actual distilled water for topping up car batteries and filling steam irons, those these days it's more common for it to be deionised. The car parts chain round here has it, even the biggest supermarkets - but not sold with beverages
    – Chris H
    May 26 at 8:05

6 Answers 6

40

If you truly want to attempt to recover the data, take them to a company that specializes in data recovery. If there is anything left, they will be the best chance to get anything off the drives. They will clean the cards and attempt to recover the data. If that doesnt work, they might remove the memory and read it from a specialized hardware directly.

If you want to try yourself, drop them in 90% isopropyl alcohol (rubbing alcohol) to displace any water in or under any parts. Dry them off and then let them air dry. Then dip a toothbrush in some distilled white vinegar and scrub the contacts. Vinegar is a mild acid and will help loosen the corrosion on them. Wipe them with alcohol again and let them dry. Then you can try reading them. It all comes down to how far the corrosion has spread.

15
  • 9
    Technicians typically use higher concentration isopropyl alcohol than 90%. If your doing it yourself, go for the higher concentration, it could make the difference. You basically want it to evaporate quickly. distilled water is probably a bad idea.
    – Ramhound
    May 24 at 12:36
  • 18
    its already been to the bottom of salt water, 10% distilled thats in rubbing alcohol isnt going to make a big difference. Also, you arent going to find much over 90% Maybe 95% alcohol can be found in stores.
    – Keltari
    May 24 at 14:22
  • 9
    @Keltari depends on the country, in Poland one can easily buy 95% and 98~99% with some effort. 99.9% is hardest to get and most expensive, about 3 days wait and $20 for 5 liters / over a gallon. In my experience, 99% evaporates way, way faster than 90%. Sadly, it also catches fire much easier. But for a task like this I would get with strongest IPA for another reason. Lithium and products of its reaction tends to react more readily with water than with isoprophyl. Amount is probably negligible but why risk it?
    – Mołot
    May 24 at 20:53
  • 2
    @etgriffiths you are pretty much likely to read all the data that was written prior to flooding.
    – fraxinus
    May 24 at 22:15
  • 16
    Rubbing alcohol may have oils in it (castor oil, methyl salicylate) as well as significant amounts of water. It's best to use pure isopropyl alcohol at the highest concentration you can get. (When you're trying to drive out water, using something with more water in it is less effective.) 99% IPA is widely available in many countries; I just get it directly from Amazon here in Japan, and amazon.com appears to sell several different brands as well (though they are all out of stock right now).
    – cjs
    May 25 at 0:19
26

I would start with distilled or dionised water, before high purity isopropanol. The alcohol won't do anything to remove salt deposits, and salt is hygroscopic, pulling enough water out of humid air to form a conductive film. While micro-SD cards are encapsulated, I wouldn't trust them to be sealed so I'd assume there's salt-water inside - and you can't open them to dry them.

In the interests of science curiosity I sacrificed a micro-SD card:

Opened up micro-SD

The colourful shiny bits are the actual silicon on which the data is stored (the patterns of the transistors act as a diffraction grating hence the colours). The plastic formed over the bond wires out to the connectors is probably good against short term immersion, but I wouldn't trust it long term when there might be mechanical action too (though inside a data-logger it should be fairly well protected.

Actual Si inside micro-SD The structure of the silicon inside, viewed under a microscope; the total field of view is 1--2mm. The inset shows the diffraction pattern of a 630nm diode laser from the structures we can see, after removing all but one fragment of the Si. Click for full size

But you do want to be sure it's completely dry before trying to read - warm air, displacement with isopropanol, dessicant etc.

Compared to the data, an external USB card reader is cheap and easy to replace - so don't worry about damaging that, which is anyway unlikely. Using an external reader will offer essentially complete protection for the computer against electrical faults.

You definitely want to grab a clean copy as quickly as you can once you plug it in. There are various read-only tools, that will get quite a lot back even if the file system is badly damaged. I've successfully used PhotoRec (cross-platform, open-source) to recover all sorts of files from a corrupted SD card. If the file system is intact a simple copy may work. However to be on the safe side, a tool that's designed to copy data from damaged drives (such as ddrescue, described in more detail in other answers) would be a good place to start; PhotoRec can be used on the copy.

To be honest I suspect you'll either be able to read everything, or nothing (or the read might start successfully then the card die but this seems less likely). It's hard to envisage a mechanism that could damage some of the storage and not other areas, unless it was being written to when it got wet.

If an initial read after a DI water then isopropanol rinse and thorough drying doesn't give anything at all (e.g. not recognised as storage even by data recovery tools) I'd try (in this order):

  • Cleaning the contacts with gentle abrasives (see other answers) - you might even want to do that before the first test depending on how they look after a bulk clean.
  • A longer soak in DI water (days), again IPA rinse, dry for even longer (even in a vacuum chamber if you can, pumping down slowly).
  • Finally, an ultrasonic bath in DI water, IPA rinse, dry, perhaps keeping ~50°C for a good few hours if you can't pump on it.
13
  • 2
    @etgriffiths I'm a physicist and general hardware nerd, working from home today with a dead card in front of me. The reason I know it's dead is that I tried and failed to use it in my pocket oscilloscope the other day to measure a bike dynamo's output, so this is very much in character!
    – Chris H
    May 25 at 14:30
  • 10
    @ChrisH Regarding the reading of data: before using PhotoRec (a file carving tool) I strongly suggest making a read-only image of the data as best you are able, probably using a tool like ddrescue (gnu.org/software/ddrescue) which can cope with the (perhaps flaky) card vanishing mid-read, and offers options to copy nonlinearly so that bad blocks (that have a habit of making your OS unhappy) can be skipped. Once you've got an iso file, then is the time to try to get actual files off it. Repair the electrical interface and leave the die well alone – light alone can cause problems.
    – Landak
    May 25 at 15:40
  • 3
    @ChrisH – I agree that you won't get conventional bad blocks á la ATA-magnetised media. But (in my experience) you may find that requesting exactly a certain address causes the read to hang, forever, or the drive to spontaneously vanish. That's effectively the same as a bad block, and ddrescue will mark an area dodgy as such, before trying to recover the rest of the data – you can read it in reverse, say, and try to recover "as much as possible as quickly as possible" before it breaks forever. The mapfile is easy to edit by hand, and, in my opinion, it's a fantastic tool as a consequence.
    – Landak
    May 25 at 16:20
  • 3
    aside: I think the colours on the IC are more due to another kind of interference: there's a layer of SiO2 glass on the IC, light gets reflected both on the SiO2-air-interface and the Si-SiO2-interface and interferes. With planar flash I'd guess the structures are too small for visible light interference, could be different with 3D-flash. You can distinguish by rotating the IC about a surface normal while keeping the other angle constant: if interference changes it's caused by the rectangular transistor grid (effective lattice constant changes), if it stays constant it's due to glass. @ChrisH
    – chichak
    May 25 at 20:29
  • 3
    @chichak good point, I use intereference filters a lot working in optics. With sub-mm fragments it's tricky but I'm inclined to agree with you. 2D diffraction from a square grid would be hard to distinguish. I think I'll point a laser at it tomorrow anyway and look for a diffraction pattern. You need a structure with a period over a few hundred nm. This is an old cheap 1GB (not SDHC) so could be ~100nm cell size or more according to Samsung (blog) and the likely age.
    – Chris H
    May 25 at 21:15
6

Start by working out the value of the data. If its critical, then professional data recovery may be your best bet regardless of the cost.

If this data is required for a legal matter, push it to the lawyers to decide. Do not make any efforts on your own, this may make matters worse and you might be charged with "tampering with evidence" or similar.


If its just a "nice to have back" then by all means give it a go.

If the disk does start to be readable, you need to copy it all as fast as possible. Just that it worked once doesn't mean it will work again tomorrow.

Make sure you have enough storage space clear, ready to receive any files off each disk.

Start with the least-important card, if you can make that judgement.

Use a known-good MicroSD to USB adapter that tested okay. Expect that it may become unusable in the future due to grit or damage.

Clean the microSD card so the contacts are exposed and shiny. I would NOT use an ultrasonic cleaner, because the potential for cavitations to cause damage exists.

Ideally you want all the internal space to be dry as well. It may help to pop the top of the card off to inspect and let it dry. Have patience.


Also you need to decide if you want the files that are there, or if you want all the blocks. If the device has deleted files, you might be best imaging the whole card with something like dd then working on an image of the card.

If the files alone is okay to retrieve, then a plain copy should work.

If the filesystem is damaged, you might need a tool like ddrescue to scan the disk and pull out files it can identify. Then you can run a fsck or scandisk to fix the filesystem and try a normal read.

At the end of the day, the cards are trash and cannot be trusted in the future. Do not be tempted to keep using them for anything you care about even if they work perfectly now.

1
  • 5
    If that works, it's best to make directly a byte-by-byte copy (ddrescue) because - as also Criggie stated - it may not read it a second time. Give it a look online (e.g. in this answer ) for a procedure to minimize the risks.
    – Hastur
    May 25 at 8:24
6

The slowest connection (SPI protocol) does not use pins 1, 2 and 8. Optimally, those pins should be connected to high, low, and high levels with a resistor, but you might get away with those line unconnected. You will get away with mediocre contacts.

SPI mode is usually used when microcontrollers interface with SD cards in low-throughput applications (data loggers, for example). Ask someone who has tinkered with microcontrollers; they may also have single-contact probes that connect to corroded contacts.

Also, why not give the producer of those data loggers a call. Who knows, they might be interested to showcase how their product is used below 40m of salt water (assuming it's not their fault), or to understand how their product failed (if it is their fault).

2
  • 6
    Oh, we've sent them an error report. These things were deployed at less than half their capibility, both pressure-wise and time-wise. We expect they will claim user error, but this was 2 units out of 10 that failed, and I've made notes to them before that their o-ring seal isn't robust enough. May 25 at 14:31
  • 1
    @etgriffiths that's only +4bar, it shouldn't be that hard for them to design it right!
    – Chris H
    May 25 at 18:02
6

Contact cleaner is magic, and your best bet here. It's designed to clean electronic contacts, and doesn't leave a residue. Isopropyl is fine and all, and probably less toxic, but this isn't quite a time for half measures. A can goes a long way, and is fairly useful to have around

Cover up the plastic bits as best as you can with tape. Do a patch test to make sure you're not accidentally going to end up with plastic goop - but contact cleaner is pretty gentle in my experience. Then give the metal parts a good spray of contact cleaner and give it a quick scrub with a Q-tip. You ought to be able to get most of the nasty stuff off. There are also folks who suggest using melamine erasers or fibreglass pencils if the corrosion is worse - and you'd want to do so in that order, as they're more abrasive.

Get a cheap SD card reader (I mean seriously - you can get one for a dollar or two) just to see if it reads, then use one you trust to actually get the data out.

1

First, let me state, that most all data is recoverable. Its just a matter of whats required to read it. If there was some evidence on those that the FBI needed to put some big corprate fruad guy in jail, best believe, the data would be recovered. The technology that they use though is not at the average persons finger tips. This is why your best bet is to hand them to a professional. Rates for that sort of work can be a lot more than you might expect, and estimates will vary from company to company. They will have tools that are far more sensitive that what you likely have.

You could always try cleaning them in a way that doesn't harm them any further, so what you get, if nothing, then hand them to a professional.

If you do decide to go all out though, and your a DIY'er at heart, which I am guilty of being, then your best bet for recovering the data would be to clean them in a drying solution. Near pure Alcohol, or Acetone, those would be your best bets (or at-least as far as I know). Afterwards you can let them sit in a bag of rice for a few days, or a week even.

At this point, try sticking them in, see if you can get a read. Its been a while since I played with disk reading software, but I know that there are much slower reads that are far better at recovering data.

If you can't get a read, you need to make sure your connecting correctly. The reason its important to ensure that you have a good connection, is because you want to know if there is a problem with the data or not. This is basic troubleshooting.

At this point you will need to get...

  1. a soldering iron,
  2. solder for the iron,
  3. and some good quality flux.

Truthfully, you probably will greatly benifit from any electonics experiance you have had in the past, I was fortunate enough to have a dad who was an electronics major when he was in college.

If you have ever soddered somthing with a shitload of pins before, like DRAM, you will probably be much better at this starting out. What you want to do is individually cover each copper connection with a thin coat of solder. You want it to be thin enough that it will still fit in your SD slot, yet you need enough coverage on the connector that you can see almost no rust.

The most important part is to ensure you don't short circut the card. In other words, soder on one connection can absolutly 100% not touch another connection. (I told you experiance helps)

If your diligant enough, smart enough, and patient enough, you can do it with no experiance soldering, but you should practicing on somthing else first.

If you still cannot get your data back at this point, and you feel confidant you correctly repaired the connecors, then your data is corruptted. If you fixed the connection correctly, depending on the software you use, the program that you try reading the card with might give you a helpful "data corupted" error message. If thats the case, you can come back here and enquire about the "Error" that you are getting.

But, I am willing to bet, at some point, you'll find that the cards are readable.

Best of luck to you.

2
  • 3
    I reckon if you can get a clean enough surface to solder to, you can get a clean enough surface to make contact with a socket. A fine wire brush may help in either case, or fine steel wool, to remove corrosion, but be sure not to leave any strands. And I reckon you should be clearer that you're talking about solder on the outside, as physically bigger cards and flash drives can sometimes be repaired internally
    – Chris H
    May 25 at 12:34
  • @ChrisH Yes on the outsit. I wouldn't hit the rust with a steal brush though. Another thing you could consider is using is vinigar and salt to clean it. Dry it, then wipe it using Sodium Bicarbonate (aka Arm & Hammer, you know the stuff you make a cake with). Mix the bicarbonate with just a tad bit of water. You want it to be like a paste. The salt and viniger, should obviously, be liquid. This is a great trick I learned from youtube 10-15 years ago.
    – j D3V
    May 26 at 23:42

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.