Project: Physically Damaged USB Drive Repair & Data Recovery

A while back, on 16th July, I was contacted by a fellow Sydneysider, and a fan of my blog with a request for help. While, ordinarily, I don’t respond to these, there was a rather long story behind this particular request. To put it in a sentence, the person in question had experienced an act of goodwill, and in return, offered to return the favour which resulted in him being tasked with the job of repairing a USB flash drive that was physically damaged. The drive itself belonged to a uni student, who was apparently anxious to get some of his hard work back, as he needed it for the upcoming semester.

Rather wisely, the reader contacted me with photos of the drive, and had some doubts as to his abilities to repair the drive. He was right – it wasn’t as simple as it appeared.

Rather than trusting someone else to screw it up, I offered my services free of charge, to do my best (but not accepting any liability) to repair and recover the drive provided that he covered the postage to me. I also made it clear, up-front, that I wouldn’t be accepting any liability and that he should consult with professional recovery firms if he wanted some level of assurance. In the end, he accepted my offer of assistance, and hence, this post was born.

The Item

While I was generally in a good mood, despite my constant busyness, and eager to get cracking for a stranger (as I know semester started this week), sadly a string of circumstances led to the drive reaching me just this afternoon.

delivery

As requested, the item was posted with Registered Post, to safeguard against loss and provide some level of tracking.

20150731-1345-2582

Inside, the item was well packed inside anti-static bag and an old SODIMM container, which was great as this helped the package survive the mailing trip and reduced any possibility of electrostatic discharge causing further damage to the components.

The Patient

20150731-1348-2587

The patient was a Sandisk Cruzer. This drive seems to have suffered some violent physical impact on the USB connector, resulting in the complete removal of the connector from the printed circuit board (PCB). Other components on the board are intact.

Initially, the reader thought that the job would be a simple one, soldering the connector back onto the board. It isn’t. Look closely.

A printed circuit board is made from a thin copper sheet, laminated to a fibreglass resin base (FR4). The connections are made by etching this copper sheet into traces, and then soldering the components to these traces.

In the case of this drive, when the connector pulled away from the board, the pins lifted the traces of the bottom three pads (D-, D+, GND). This means it’s not a simple job to just solder it together because there’s nothing to solder to. The top pad (Vcc) is still intact, as the solder joint itself failed, letting the pin “lift free” of the board.

20150731-1348-2585 20150731-1348-2586

The trauma is clear from the connector, with bending of the pins from the rear view (left image). From the underside (right image), you can see the brown colour of the lifted and torn copper traces still attached to three of the four pins.

20150731-1350-2589

The underside of the drive had one flash component, of 8GiB. Interestingly, the drive also has a polyfuse on the rear (self-resetting fuse) to protect against overcurrent (which may happen if the drive gets wet with dirty water, or receives overvoltage or reverse polarity) which improves the chances that the drive might survive some mistreatment. It’s not something you find on lower quality drives.

The Repair

When I had received the images of the drive, I had already developed a game plan. But just for those who are following at home, key points to note are the following:

  • USB 1.1 and 2.0 consist of four connections – Vcc and GND (which provide 5v and 0v, i.e. power) and D+ and D- (which provide the differential communications channel). All of these must be connected for the drive to function.
  • The Vcc pad is fine, we can still use that. But D-, D+ and GND are missing.

What is needed is to find ways to connect the D-, D+ and GND connections. Looking at the PCB, the D- and D+ routing on this board is straightforward – they are two parallel traces on the top side with no vias (through-board) connections. The GND connection can be made to the connector shell pad as most boards bridge those connections together. Other than that, the “exposed copper bar” on the edges of the PCB also seem to be connected to the ground plane which would work just as well.

Now that the connection point candidates are identified, it’s important to remember that the green colouration of the board is due to a coating known as solder resist or mask. This stops you from soldering to the rest of the copper trace/sheet in order to prevent solder from running rampant across the copper and making it easier to solder and manufacture. It also helps insulate the board from contaminants which might cause conductivity between traces. Unfortunately, this means we can’t get to the remaining traces without removing this.

20150731-1354-2590

In order to do this, I’ve decided to use a simple box cutter knife to scrape away at the solder resist layer and expose the copper trace underneath. Care needs to be taken not to cut the trace (especially if you don’t have much of it available to get to). Remember, the copper layer is very thin – we just need to expose enough of it to solder a credible connection.

The next step is to heat that up with some flux core solder and tin the copper. I salvaged some thin AWG28 wire from an old USB cable, prepared the ends by tinning them, in order to do the connecting using an “air-wire” method.

20150731-1414-2592

After some very careful manipulation, this is the result. The connections to the D- and D+ traces were the most tricky, requiring steady hands and patience. AWG28 wire is not thick at all and yet, it was a challenge to ensure they didn’t short together when soldering.

20150731-1414-2591

We’re all good to go to the next stage of recovery.

The Recovery

In all, the unit was intact to the point that when it was plugged in, it came up immediately with no problems.

Sandisk-Cruzer-Detected

The data was still intact. A full image was taken, then an archive of the files made. That’s now being uploaded and sent back to its original owner. Success!

Drive-and-VolLbl

At this point, I can’t share anything more with you guys – all of the data is kept private.

Conclusion

From all of this, I think we can come to several rather disparate lessons:

  1. It’s always better to realize your own limitations and ask for help when uncertain. The worst we can do is give you either the cold shoulder, or a nasty reply, in which case, you should take it on the chin and move on. This is better than forging ahead with no game plan and potentially making things worse for yourself. This applies to me too – I wouldn’t ever accept a job if I didn’t feel I was suitably qualified to undertake it.
  2. Evaluate your options. How much is your data worth? Most data recovery companies are quite expensive, but it’s a much better guarantee of safety. If you are definitely not going to spare the expense for that, remember that many accidents could be recovered from with some skill. Do some research, find the right software, get cozy with Linux. It’s all worth it.
  3. Prevention is better than cure. Back-up as frequently to reduce the devastation from actually losing physical media. It can always fail unexpectedly, become damaged or lost.
  4. Resist the temptation to leave USB drives plugged into laptops while moving around. I don’t know exactly how this one came to be damaged, but I’d have to guess that this is probably the leading cause. Not only do you risk damage to the drive which can be difficult to repair, you also risk damaging the port itself.
  5. USB is tolerant of hacky cabling to some extent, and that’s partly why low quality cheap unshielded and untwisted cable seems to work (at least for a while). It also means that you can make such “air-wiring” fixes for temporary recovery, but do keep it short and tidy if possible. The worst that can happen is that you end up not getting a link or you get a slower USB 1.1 connection but still might get your data back.
  6. Store your devices properly. If you don’t, you can risk secondary damage from electrostatic discharge that could further jeopardize your data.
  7. Practice and experience are important, as are having the right tools. With skills such as soldering, practice helps you better understand and improve your technique, which helps you target finer pitch joints, and helps you produce more quality joints that don’t fail under mechanical stress. But if you don’t have a fine-tipped iron, or decent solder, you might never be able to get the right result no matter how hard you try.

Needless to say, the data is on its way back to its rightful owner, and I’ve ended up helping out two people along the way – one directly, and one indirectly. Maybe with this post, I’ve ended up helping others … indirectly.

That being said, this is not an invitation for you to start e-mailing me your stories and expecting me to do things for you for free … I will use the silent treatment where necessary!

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Computing, DIY, Electronics and tagged , , , , , , . Bookmark the permalink.

10 Responses to Project: Physically Damaged USB Drive Repair & Data Recovery

  1. Mark B. says:

    The flash chip propably was 8 GB, not GiB, as the drive shows a capacity of almost 8 GB on the screenshot above. Please fix it!

    • Mark B. says:

      **correct it

    • lui_gough says:

      I’m sorry, but you’re the one in the wrong here. EVERY flash chip ships with the BINARY gigabyte value PLUS extra space for storage of ECC codes, this has been standard practice within the industry as the chips themselves will have a certain number of defective cells (usually <1%) from the factory.

      If you look up a flash list (e.g. http://www.usbdev.ru/databases/flashlist/flcbm93e98s98p98e/), you will find the SDTNQCAMA-008G is specified as a 72bit/1K meaning it’s actually 8GiB, but for every 1KiB block, there are 72 extra ECC bytes. Unfortunately as Sandisk datasheets are generally NDA – I can’t find the official specs, but here’s a similar example from Micron Technology’s NAND 101 (https://www.micron.com/~/media/documents/products/technical-note/nand-flash/tn2919_nand_101.pdf?la=en) Page 6 where it shows you a 2GBit (256MiB) NAND device. Internally it is a 2048 x 64 device, with 2112 byte pages – 2048 bytes data + 64 bytes spare area. As a result, its real size is 2048 x 64 x 2112 = 276,824,064 bytes = 270,336 kiB = 264 MiB. The user (in this case, controller) accessible storage is 2048 x 64 x 2048 = 268,435,456 bytes = 262,144 kiB = 256 MiB.

      The reason you don’t see the whole binary value is because of overprovisioning in the controller. The controller reserves the difference between a DECIMAL Gb and the BINARY GiB value (and sometimes more) to enable effective wear levelling. This is often referred to as 7% overprovisioning – think of the 256GB SSD that has 256GiB of actual flash. This requires room to store the indirection table, as well as spare cells to be used in future writes so they can better plan and erase/write cells in blocks.

      So, I am not wrong and nothing needs to be corrected. As a rule, I don’t usually modify articles once published but I will discuss any issues in the comments.

      – Gough

  2. Neat Job! you have done a great job, a real pain in the a$$ :D, you must have a steady hand for that solder.

  3. Ramiro says:

    Hello, first of all very informative and educational post, i have a broken pen drive that need to be soldered back to the usb connector, and wanted to know how would you know which one are GND D+ D- and +5V on the chip???? i have no clue of which one is which. is there any visual clue on the board???
    Thanks in advance for any help!
    Salutations

    • lui_gough says:

      If you want to solder directly to the chip, which is not as easy, you will have to look up the chip’s datasheet if it is available. Some chips have their pin-outs shown on the datasheet, so you can identify which pin. If the pads, or the traces near the USB connector are still intact, it is probably easier just to use those and use the standard USB connector pin-out and follow that. You can easily find that by Googling it.

      – Gough

  4. Barbara W says:

    Does it matter what gauge wire you use? Glad you could help them, I am trying to fix one too right now but mine (Lexar 128G) doesn’t seem to have the copper used in the same manner as a sheet so not sure if I can.

    • lui_gough says:

      Technically speaking, it doesn’t really matter which gauge of wire is used, although thinner is probably preferable (28AWG or so) mainly because it’s easier to solder to thin traces, and it has less capacitance which will reduce the burden on the drivers pushing the high speed signals down the line. For recovery purposes, as long as it’s kept short, almost any wire will do if you can get it on.

      – Gough

Error: Comment is Missing!