Data Recovery Case File · Cameras, Drones & Cards · The Current Frontier
The Chip Can Be Read and the Reading Is Not the Problem
His enquiry is the most technically specific in this archive and it names a real limitation. A physically damaged card, where "a data recovery lab analysed it and informed me that recovery would require a chip-off procedure, but the memory uses an error correction scheme which they currently cannot decode." That is an honest answer describing the hardest problem in flash recovery — and it deserves explaining rather than second-guessing.
| Media | Monolithic SD card with physical damage — memory and controller integrated in a single package; modern soft-decision error correction in use |
| Reported situation | Card physically damaged · assessed by a specialist laboratory · extraction identified as requiring direct memory access · memory employing an error correction scheme the laboratory reported itself unable to decode · no recovery achieved · contents required |
| Fault class | Monolithic construction with soft-decision error correction — extraction and correction as separate problems; correction the limiting factor |
| Equipment used | Construction confirmed as monolithic before any extraction planning · technical point contacts located against the package layout · raw cell data read with multiple threshold voltages where the package permitted · correction attempted against the controller's implementation · position stated honestly before any charge |
The decode: two separate problems, and which one is hard
The first problem — getting the data out. On an ordinary card, the memory is a discrete chip that can be removed from the board and read on dedicated equipment. On a monolithic card there is nothing to remove: memory and controller are a single package, and access is through technical points on the package itself, located against a known layout. Difficult, and established work.
The second problem — making sense of what comes out, and this is the one they named. Flash memory does not store data reliably enough to be read directly. Every device applies error correction, and what the memory physically holds is the data plus a substantial block of correction information. Nothing is readable until that correction has been applied, using the same scheme the controller used to write it.
Why older schemes were tractable: they worked on hard decisions — each cell was read once and reported as one value or another, and the correction mathematics was well understood and reproducible outside the controller.
Why the current generation is different: modern devices use codes that operate on probabilities rather than certainties. The correction needs to know not just what each cell read as, but how confident that reading was — which requires reading the same cells repeatedly at several different threshold voltages and combining the results. It is a fundamentally different kind of operation, and it is tied closely to the specific controller's implementation.
Why that makes it the current limit of the field: the raw cell contents can frequently be obtained. Turning them back into files requires reproducing a correction scheme that the manufacturer never documented, and for many controllers that has not been achieved by anybody.
Why the laboratory's answer was a good one: they identified the exact obstacle rather than attempting the work and failing. "We cannot decode this" is a more useful and more honest answer than an expensive attempt, and it tells him precisely what would have to change.
What is nevertheless worth establishing: which controller, and whether its scheme has been resolved since. This is an area where capability changes, and a card that could not be decoded last year is occasionally decodable now.
The honest position: monolithic construction combined with modern correction is among the least recoverable combinations there is, and that should be said plainly before any money changes hands.
On the bench
Construction was confirmed as monolithic before any extraction planning — memory and controller integrated in a single package, so access is through technical points on the package rather than by removing a discrete chip. Raw cell data was read with multiple threshold voltages where the package permitted, modern correction operating on read confidence rather than fixed values and therefore requiring repeated reads at differing thresholds. Correction was attempted against the controller's implementation, and the position stated honestly before any charge.
The outcome
Construction confirmed, raw cell data read at multiple thresholds and correction attempted against the controller's own scheme. 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 — and no charge where the position is genuinely no. The decode: the laboratory named the real obstacle. Getting the data out is one problem and making sense of it is another — modern correction works on read confidence rather than fixed values, and reproducing it outside the controller has not been achieved for every device.
Told your card uses correction that cannot be decoded
That's an honest and specific answer, and it describes the hardest problem in flash recovery rather than a lack of effort. Getting data out of a card is one job; making sense of it is another. Flash doesn't store data reliably enough to be read directly, so every device applies error correction, and what the memory physically holds is your data plus a large block of correction information — nothing is readable until that's been applied using the scheme the controller used. Older schemes read each cell once and were reproducible. Current ones work on probabilities, needing repeated reads at different voltages, tied to a specific controller nobody documented. Worth asking which controller, since capability does change.
Ask which controller — call Oxford Data Recovery on 01865 593000; construction confirmed before extraction planning, raw cells read at multiple thresholds, and an honest position before any charge.
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.