Data Recovery Case File · Solid State & Flash · Charge Does Not Wait Forever
Flash Memory Left Unpowered for Years Loses Charge, and Files Fail One by One
His enquiry describes a pattern that is diagnostic in itself. A memory stick handed over as part of a wedding photography package nine years ago, unused since, where "there are a bunch which are ok, but the majority are corrupted." Some working and most not is exactly what long unpowered storage produces — and it is a property of the medium rather than anything he did or failed to do.
| Media | USB flash memory stick issued as a photographic deliverable approximately nine years earlier — retained unpowered throughout; majority of images failing to open, a minority intact |
| Reported situation | Memory stick supplied as part of a wedding photography package · approximately nine years elapsed · stick not used in that period · files opened recently · majority of images reporting as corrupted · a proportion opening normally · repair of the remainder sought |
| Fault class | Charge retention loss across prolonged unpowered storage — cell drift exceeding error correction on affected blocks; controller and device otherwise functional |
| Equipment used | Failure distribution assessed against retention behaviour rather than device fault · device addressed through hardware with imposed timeouts rather than the host stack · memory read past the controller with error correction applied in software · marginal blocks re-read across multiple passes and results compared · recovered images validated by opening at full resolution |
The decode: why time alone does this, and why some files survived
How flash memory holds a value: as a quantity of electrical charge trapped in a cell. The cell is insulated, but no insulation is perfect, and charge escapes slowly over time.
Why powering the device matters: a drive in use refreshes what it holds as it works, and its controller relocates data showing signs of drift. A device that is never connected never gets the chance, so nine years of leakage accumulates untouched.
Why a proportion of files still open: the loss is per cell rather than per device. Every read is corrected for a small number of wrong values as a matter of routine, and blocks whose cells drifted less remain within what the correction can repair.
What a corrupted file therefore represents: a block where drift exceeded that margin. The correction could not repair enough of the errors and the data returned did not match its own checksum, so the file is rejected rather than absent.
Why the ratio he reports fits so well: retention loss produces a spread rather than a cliff. Some blocks fail, more sit near the boundary, and a minority remain comfortable — a mixture, which is what he found.
Why this is worth saying to him plainly: nothing he did caused it. The stick was stored carefully and in a drawer is exactly where it should have been; the medium simply does not hold indefinitely without power.
Why the delivery format is the real issue, and it is worth stating for anyone reading: a memory stick is transfer media rather than archival media. Content received on one should be copied elsewhere on the day it arrives, and ideally to more than one place.
Why the position is nonetheless better than it looks: the device works and the failure is at cell level. Reading the memory past the controller allows far more computational effort per block than a stick spends in normal operation, and correction applied in software recovers a great deal the device gave up on.
Why repeated reading helps here specifically: a drifted cell does not return the same wrong value every time. Multiple reads produce slightly different results, and comparing them across passes reconstructs the intended value — the opposite of the situation on a mechanically failing drive.
What must not happen: nothing written to the stick, and no repair utility run against it. Every write to a device in this condition consumes margin that recovery depends on.
On the bench
The failure distribution was assessed against retention behaviour rather than device fault — flash cells holding values as trapped charge that escapes slowly through imperfect insulation, with connected devices refreshing and relocating drifting data while an unused device never does. Loss occurring per cell produces a spread rather than a cliff, so blocks within correction margin open normally and those beyond it fail verification. Memory was read past the controller with correction applied in software across multiple compared passes.
The outcome
Failure distribution assessed against retention behaviour, the device addressed through hardware, and marginal blocks re-read across compared passes. Free assessment, one fixed written figure including VAT; where a chip has to be removed, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode: nothing you did caused this. Flash holds values as trapped charge that leaks over years, and a device never connected never gets to refresh itself — so some files stayed inside the correction margin and most did not.
Files supplied on a memory stick
Copy them somewhere else on the day you receive them, and to more than one place — a memory stick is transfer media rather than archival media. Flash holds each value as trapped charge that escapes slowly, and a device in use refreshes itself and relocates data showing drift, while one left in a drawer never gets the chance. That's why files fail in a mixture rather than all at once: the loss is per cell, so blocks that drifted less stay within the correction every read applies routinely. If it's already happened, don't write anything to the stick or run a repair utility.
Don't write to it — call Manchester Data Recovery on 0161 871 0788; failure distribution assessed against retention behaviour, memory read past the controller, marginal blocks re-read across compared passes.
Request a quote online →
Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.