Devices · flash drives and USB sticks
Nobody chooses to keep anything important on a flash drive. They do it anyway: a year of invoices that exist nowhere else, or a copy made for one afternoon that quietly turned into the only one. Plugs break off. Controllers go quiet. Either way the memory itself has usually come through, and that is where we start. A Hungerford dealer's stock photographs, a dissertation off the Whiteknights campus, the quotes a Tilehurst sole trader carries between jobs — a few reach the Reading bench every week, at £250 + VAT with the look before it free.
An engineer answers, and looking is free
0800 6890668
| What you are seeing | What is likely wrong | What happens next |
|---|---|---|
| Windows says only: Unknown USB Device (Device Descriptor Request Failed) | Nothing at all. The controller has stopped answering | Stop trying ports |
| USB device not recognized | Enumeration never completes | One more port, then stop |
| Windows asking to format it before it will open | The file system is broken; the files are not | Turn down the format offer |
| It mounts at 0 bytes, then asks you to insert a disk into removable disk | The controller cannot see its own memory | That one is chip-level work |
| 0x80070570 every time you open it — corrupted and unreadable, says Windows | The FAT or exFAT structures are corrupted | Still recoverable while nothing further is written |
| Windows was unable to complete the format | The controller has stopped permanently | That refusal is actually helping you |
A case number goes on your device the day it reaches us, and an engineer then finds the real fault — free, and before anything else happens. You are told plainly what will come back and what will not, with one fixed figure in writing beside it. Nothing chargeable starts until you have read that and said yes.
On an older stick the memory package can be taken off its board and read separately. A monolith gives you nothing to lift: memory and controller are cast into a single lump, and the way in is a line of test pads a fraction of a millimetre across.
Wear levelling moves blocks about the die, and interleaving spreads them over the planes for speed. A raw read brings all of that back mixed together, and it is unpicked in software, one layer after another, until named files with real dates come back.
You get the list of files before anything else happens. Nothing is billed until you have accepted it, and everything goes back on fresh media with return postage paid by us.
You see the file list first, then decide. Nothing is invoiced until you have said yes, and on most jobs there is no fee at all if the data does not come back. What we pull off goes onto media bought in for your job, and the postage home is ours. The case stays open on the bench until you tell us the files open on your own machine.
Plain speaking about the statistics: no failure rate for consumer flash exists that is worth repeating. USB sticks are not deployed in fleets, and a data centre counts hard disks rather than these, so nobody ever gathered the figures. All we can honestly describe is what reaches this bench, and the pattern holds year after year: the physical failures are nearly all a connector separated from the board, and the ones that stop without notice are usually unbranded, or were handed out at an exhibition.
A stick trodden into the floor of a site office carried seven years of invoices, the CIS records alongside them, and a return falling due. The connector had come away from the board. On a monolith there is nothing left to plug into, so the controller's pinout was identified and the data taken directly off the test pads. That return was filed by the deadline.
Rarely. The shell does nothing except carry contacts, and the files live in the memory package behind it. A stick in two pieces will normally still be readable, provided the package and the resin around it came through uncracked.
No. The controller stopped answering. The memory under it did not. Go to that memory directly and whatever the controller thinks stops being relevant.
Cheap NAND, run by a cheap controller. Some go further and misreport capacity altogether — a small die presented as a large one, behaving perfectly right up to the moment you exceed what is physically on board. Fine for moving a film from one machine to another. Never fine as the only copy of anything.
Only one: a memory die split right through it. A stopped controller, a board in several pieces, mapping tables in disarray — that is an ordinary week here.
You pay nothing for the look. What comes back is a list — the files that opened, the ones that did not, and one fixed figure to finish the job. Leave it switched off until you have that in writing.