Norfolk gets photographed more than most counties and backed up rather less. Weddings in the Lanes, boatyards at Wroxham, the light coming off the sea at Cromer and Sheringham — a great deal of it travels home on one small card, and cards are the least protected storage anybody owns. They arrive here from working photographers, dashcam owners and families who have mislaid a fortnight abroad, and the pictures usually go back.
Every memory card job is diagnosed free. The quote follows in writing, fixed, before a screwdriver is picked up.
No fix, no fee all jobs except electronic and mechanical failures, chip level work, DVR and Forensic jobs. Full pricing is on the data recovery cost page.
The first job on any memory card is matching the symptom to the fault — and after twenty-odd years, these thirty account for very nearly everything that comes through the door.
Camera, reader and laptop all shrug at once, which points at the controller or its firmware rather than the memory. The pictures are sitting in the NAND behind the silence, entirely unaffected by the fact that nothing can currently ask for them.
A damaged file table, not lost photographs. Agreeing to that prompt is the one dependable way to turn an easy job into a hard one, so take the card out and leave it alone.
Frames stay in the flash until something new is written on top of them. Put the card in a drawer instead of carrying on shooting and most of a deleted folder walks back out.
Cameras perform a quick format and nothing more thorough, so the shoot underneath is generally intact and lifts out whole. The one thing that spoils it is carrying on using the card afterwards.
Cards give way in slots, in wallets and in cheap adapters. Where the fracture has missed the internal tracks, we probe the die itself and read straight off it, which is why a snapped card is worth sending rather than binning.
That is the controller's fallback identity after a crash, and it is a firmware state rather than a measurement. From here the work is chip-level and thoroughly routine.
A dying controller locks itself; a nudged slider on the side of an SD card looks identical from the computer's point of view. One of those takes ten seconds to fix, so it is the one we check first.
Cards survive their accidents better than anyone expects. Cleaned, stabilised and read at chip level, a card that went through a wash cycle usually gives up everything on it.
Fakes fold new writes back over one small genuine chip as soon as it fills. Anything recorded before that fold-over is recoverable; everything after it went nowhere, and no lab can retrieve what was never stored.
Half-moved files and a directory cut off in the middle of a sentence. Rather than trusting what is left of the file table, the frames are carved out of a raw capture and rebuilt from their own headers.
Continuous recording spends the flash in a couple of years. Carving the blocks that are still readable brings the footage back, and the clip that prompted the call is usually among it.
Ultras and Extremes fail at the controller over healthy memory more often than any other pattern in our card log. It is well-practised chip-level work and the results are good.
Everything is sealed into a single slab, so there is no package to remove. Test points are mapped against our pinout records and the memory is addressed directly. It takes longer and it works.
Type B cards in mirrorless bodies and drones fail hard at cinema bit-rates, and the Broads wildlife and wedding crowd push them hard. They are PCIe internally, so the NVMe side of the bench deals with them.
Bent, cracked and interrupted mid-write, all at once: physical damage layered over logical mess. Worked strictly in that order, and the endings are better than the state of the airframe suggests.
An Extreme PRO microSD sealed inside a waterproof housing on the coast all summer ages quickly. Heat and vibration bring the read errors forward by a year or two.
Kingston's budget SDCS2 line freezes into a mute controller state and stops answering. Once that is bypassed, the memory underneath reads without complaint.
Bent socket pins and a shorted interface follow immediately. At this end it gets read on a rebuilt socket, and the bent pins are somebody else's problem.
Counterfeit Samsungs corrupt at the point where writes wrap past their true size. The genuine span carves out cleanly and the sticker on the front fools nobody here.
Long continuous sequences can hang the 633x and 1066x lines and scramble the tail end of a shoot. Rebuilt from a raw read rather than from whatever the card says about itself.
The power failed before the index got written, so the file carries no header telling a player where anything is. The video data is all there; we rebuild the structure around it from the stream itself.
Wear-table corruption produces a card full of zero-byte ghosts with the right names and dates. The genuine contents are still down in the flash, waiting on a rebuilt mapping.
Cold, dry day, card passed from one person to another, and the controller takes the discharge. Instant, invisible, and repairable at chip level.
No-name readers run cards outside their voltage tolerance and drop them mid-write on a loose plug. The card ends up carrying the cost of the saving, and chip-level reads settle the account.
Formatted as internal storage means encrypted against that one handset and useless without it. That is said plainly rather than dressed up, and phones are outside what we take on in any case.
A card that lives between two bodies and a reader makes several thousand connections in its life, and the gold plating is only so thick. Intermittent mounting and files that read once and not twice are the symptoms. Read on a fresh contact block, most of them behave perfectly.
A card put away full in 2016 and not looked at since has been leaking charge the whole time. The card may still mount and hand back files riddled with errors. Patient repeated reads with ECC given time to work recover far more than one hopeful copy.
Sustained 4K or high-bitrate video needs a write speed the card must hold continuously, not in bursts. When it cannot, the camera's buffer overruns, clips are truncated and the directory is left describing files that stop early. The footage that did land is carved out; the frames that never reached the card are gone.
A flat battery halfway through a format leaves the card with half its old structure and half a new one, and most hosts then refuse it altogether. It is a good outcome as these things go, because a format that did not finish did not finish erasing either.
Free tools that offer to repair a card in place, or to save recovered files onto the same card, overwrite the very blocks they are reading from. It happens weekly. Stop at the first sign of trouble and send the card as it is, because every extra write costs something.
A card is a thin piece of NAND and a small controller inside a plastic shell, with no shock mounting, no redundancy and nothing whatever held in reserve. So the working rule is to copy first and investigate afterwards: a sector-level duplicate is made before any file is examined, which spares the original a second knock and is why an attempt that went wrong somewhere else is usually still salvageable. Every common format behaves the same way under that treatment — SanDisk Extreme Pro and Ultra SDXC, Lexar Professional, Kingston Canvas Select Plus, the Samsung PRO Endurance cards living in dashcams and doorbells, Sony TOUGH, older CompactFlash out of professional bodies, Sony XQD and the CFexpress Type B cards in current mirrorless kit. Speed class, V30 or V90, A1 or A2: none of it alters the method, only the hours.
Broken cards go to the chip bench instead. In a microSD there is nothing to separate, because controller, memory and contacts are moulded into one monolithic part, so the way in runs through the test points on the body: probe those, take a raw dump, and rebuild that dump into files a computer will happily open. Whether it works turns on whether the break spared those internal tracks, which the free assessment establishes rather than guesswork. Full-size SD cards are kinder, since controller and memory are separate components on a small board and a cracked shell frequently means nothing at all. Salt air near the coast and a summer in a camera bag do their own slow work on the contacts, which cleans up more often than not.
Card work is microsurgery over bare flash and runs on the same specialist rigs as sticks and single-slab devices:
Reads the memory chips directly and rebuilds in software the arithmetic the dead controller used to do. When a stick or card answers nothing at all, this is where the conversation starts again.
Hot-air and infrared profiles set so a memory package leaves a ruined board while the charge stored in it stays where it is. The whole skill is temperature discipline and patience.
Some cards have no separate chip to lift. Test points are identified against our pinout records and wired out by hand under magnification: slow work, and the only work that gets those cards read.
The XOR, ECC and wear-levelling puzzle worked through until a raw dump reads back as named folders. Unglamorous arithmetic with a good ending.
Lifted pads, torn tracks and sheared plugs are all rebuilt under magnification. A fair share of sticks written off as dead need this and nothing more.
Every stick and card is read behind a blocked interface, so the original memory goes home in exactly the state it arrived in.
The Extreme PRO family and the Ultra SDSQUAB dominate the tray, and nine times in ten the controller has failed over memory that is in perfect condition. CFexpress Type B is the part of the trade that keeps growing: wildlife photographers working the Broads and the coast at Cromer and Sheringham, and wedding shooters covering venues across Norfolk, all run those cards at rates that leave no margin, and when a Type B card fails it does so without any warning at all. The Lexar Professional 633x and 1066x lines stall on long bursts; the Canvas Select Plus goes mute at the controller; and the EVO Select and PRO Plus arrive as often in counterfeit form as genuine, corrupting neatly at the point where writes wrap past the real capacity. Dashcam and doorbell cards deserve their own note, because looping wears flash out and nothing else, and the footage comes back by carving whichever blocks held on.
Most of what reaches this bench arrived by tracked, insured post. It is the steadiest way to move a poorly drive, and a parcel posted in Norfolk is usually on the bench the next working day.
Is the drive still bolted inside a laptop, desktop, MacBook, iMac, server or CCTV / DVR recorder? The hard drive or SSD needs to come out first, and only the bare drive travels — taking drives out of machines is not something we do here. Storage soldered to a motherboard (Apple Silicon Macs, one or two very thin laptops) is the single thing beyond us: if it will not come out, it cannot come in.
↓ Print the shipping & booking-in form (PDF)
Mark the parcel for the attention of Cambridge Data Recovery. From Norwich it is about an hour and twenty down the A11, then two minutes off the A14 at Junction 32 — or next working day by tracked post. You hear from us as soon as it is booked onto the bench.
Unsure what to put in the box? Ring 0800 689 0668 before you seal it, or run the free online diagnostic.
Diagnosis free, one figure written down, most work under no fix no fee. Start online, or ring us.