Data Recovery Case File · Cameras, Drones & Cards · Where It Stopped, Not How Much It Got
Ninety-Five Images Is Probably Where the Reading Failed
This enquiry reports a partial result and reads it as a percentage. A card that failed during a shoot — "after approximately two hours of shooting my camera said it was unable to read the card, and when I got home my computer wouldn't recognise it. I have had it run through a recovery programme which has salvaged 95 images, probably 10% of what was on there." Recovery software does not sample a card — it works through it, so what came back is very likely a contiguous run ending exactly where reading stopped.
| Media | SD card failing during a photographic session — not recognised by camera or host; prior software recovery returning a limited contiguous set |
| Reported situation | Card in use during approximately two hours of shooting · camera reporting inability to read the card during the session · card not recognised by a computer afterwards · recovery software run against the card · 95 images returned · estimated at a small fraction of the total |
| Fault class | Read failure with a defined boundary — recovered set indicating the point at which access ceased rather than a proportion of readable content |
| Equipment used | Prior output examined for sequence and timestamps before any conclusion · extent of prior scanning established · device addressed through hardware with imposed timeouts · imaged write-blocked with marginal regions re-read across passes · chip-level read past the controller where enumeration failed |
The decode: what a partial result usually means
Why the figure is not a proportion: recovery software reads a card from beginning to end. It does not select ten per cent of the images — it takes what it can reach in order, and stops when it cannot proceed. So what came back is almost certainly everything up to a point.
How to confirm that in a minute: look at the timestamps on the 95 images. If they are consecutive and end abruptly, they are the first part of the shoot, and the boundary is where reading failed. If they are scattered across the whole session, the picture is different and worse.
Why the distinction matters so much: a contiguous run ending at a boundary means the remaining images are behind a specific obstacle that has not yet been worked around. Scattered survivors would mean damage distributed across the medium, which is a harder proposition.
What the camera's message during the shoot tells us: the card failed while writing. It was accepting images and then could not, which places the failure mid-session and means the directory was very likely being updated at the moment it stopped.
Why the software could not go further: consumer recovery tools read through the operating system and inherit its behaviour when a device stops answering. They do not impose their own timeouts, cannot reset a stalled controller, and cannot read past it — so they stop where the device stops rather than working around it.
What that scan has already cost: a full-medium read attempt on a card that had already failed once. Every scan is sustained work on a failing device, and a second run will produce the same 95 images at the same cost.
What can be done differently: the card addressed through hardware that imposes short timeouts and resets the controller between attempts, with marginal regions re-read across several passes rather than abandoned on first failure. Where the controller cannot be brought up at all, the memory is read directly past it.
What must not happen: no further scanning, no formatting, and no reinsertion into the camera. The camera will offer to prepare the card, which is the one action that would end this.
On the bench
Prior output was examined for sequence and timestamps before any conclusion — recovery software reading a medium in order rather than sampling it, so a partial result is characteristically a contiguous run terminating where access ceased, which timestamps confirm. Extent of prior scanning was established, each scan constituting a full-medium read on an already-failing device. The card was addressed through hardware with imposed timeouts and imaged write-blocked with marginal regions re-read across passes.
The outcome
The prior output read for sequence rather than proportion, the card addressed through hardware with imposed timeouts, and marginal regions re-read across 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: ten per cent is where it stopped, not how much survived. Check the timestamps on those 95 images — if they run consecutively and end abruptly, the rest is behind a boundary rather than gone.
Recovery software that returned only part of a card
Look at the timestamps on what came back — that tells you whether you have a boundary or a scatter. Recovery software reads a medium in order rather than sampling it, so a partial result is usually everything up to the point where it could no longer proceed. Consecutive images ending abruptly means the rest sits behind a specific obstacle. Images scattered across the whole session would mean damage spread across the card, which is worse. Don't run the scan again either: it will return the same files at the cost of another full read on a failing device. And keep the card out of the camera, which will offer to format it.
Check the timestamps first — then call Oxford Data Recovery on 01865 593000; output read for sequence rather than proportion, addressed through hardware with imposed timeouts, marginal regions re-read across 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.