Home / Case Studies / It Died Mid-Run, With No Knock At All
Power, Boards and Liquid · case file

It Died Mid-Run, With No Knock At All

A research group at Loughborough had a 4TB Seagate portable on the desk while a batch of measurements was being processed. It stopped, mid-job, and never came back: no light, no sound, no response on any machine. Nothing had been knocked, dropped or moved. The postdoc who owned it spent the afternoon asking colleagues whether anyone had a copy of the raw set and getting the answer people usually get.

External DriveElectrical / PCBDead / No Power
// the job in four lines
MediaA 4TB Seagate 2.5-inch external drive that failed instantly during use with no impact of any kind, and now shows no sign of life on any machine or supply.
What was reportedSeagate 4TB external, died suddenly while data was being processed on it, no knocks or drops, no light and no response since, raw measurement data needed, no confirmed backup elsewhere.
Fault classElectronic. A protection component on the drive's own circuitry had failed short, which is a designed response to an out-of-specification supply and stops the drive dead. The mechanism, the heads and the recorded data were all untouched.
On the benchExamined under a stereo microscope with the failure located by current draw on a current-limited bench supply, the failed protection component removed and replaced and the original circuitry retained so the drive kept its own calibration data, powered under supervision, imaged on DeepSpar behind PC-3000 UDMA, the volume parsed in R-Studio and the measurement files opened in the software that reads them before being written to fresh media.
// reading the fault

Reading the fault

Instant and total is an electrical signature. Mechanical faults announce themselves: noise, slowness, files that read once and not twice. Electronics fail in a moment. A drive that was working normally one second and showed nothing the next, with nothing having touched it, is describing a circuit rather than a mechanism, and that distinction sets both the work and the odds.

What usually fails is the part designed to. Small drives carry a component whose entire job is to short itself if the incoming voltage goes out of range, sacrificing itself to keep the rest of the circuit intact. When a supply, a hub or a machine's port delivers a spike, that component does exactly what it was fitted to do. The drive stops, which looks like catastrophe and is closer to a fuse blowing.

Fitting a second-hand circuit board is the common mistake. Boards from the same model are easy to find and rarely work, because each one carries calibration data unique to the mechanism it was fitted to. Bolt on a stranger's board and the drive will often spin and then read nonsense, and any attempt to correct that on a live drive can write to areas that should never be touched. Either the original is repaired, as here, or its memory chip is moved across.

None of this reaches the data. The recorded measurements sit on the platters, which have no electrical connection to the failed part beyond the fact that both live in the same box. Once the supply reaches the mechanism again the drive behaves as it did the moment before it stopped. The group's real problem was never the failure; it was that a single external drive had become the only copy of a term's work.

// bench work

Bench work

The circuitry was inspected under the microscope, where the failed part was discoloured but the surrounding traces were intact. A current-limited supply confirmed the short by where the current went rather than by anything burning, and the diagnosis took minutes rather than hours because the fault was exactly where the design says it should be.

The failed component was removed and replaced, and the original circuitry went back on the drive so its own calibration data stayed with it. Powered under supervision it spun up, readied and identified itself normally. Imaging ran on DeepSpar behind PC-3000 UDMA with no weak regions, R-Studio parsed the volume from the image, and a selection of measurement files were opened in the analysis software that had written them, because a file that loads is proof and a file that merely exists is not.

// what went back

What went back

The full raw set came back, along with the intermediate output that had been half-written when the drive stopped. The group now keeps the raw data in two places, which is what should have happened before, and the drive went home working.

A single disk is £300 + VAT. Repairing electronics is named in the published exclusion, so the guarantee did not run: 50% was payable upfront and the remainder once the group had confirmed the data loaded.

A drive that dies instantly under use

Disconnect it and leave it disconnected, and be wary of the instinct to try it on a different supply. If the fault was caused by an out-of-specification voltage in the first place, the next attempt can take the surviving circuitry with it, and a shorting drive can also stress whatever it is plugged into. Do not buy a matching board from an auction site and swap it over: boards carry calibration belonging to the drive they were fitted to, and a mismatched one usually turns a repairable fault into a much longer job. Say what the drive was plugged into when it died, including the hub or the supply, because that often identifies the component before anything is opened.

Is this your drive as well? Every file on this page began with somebody describing a noise or a message, and the route in is the same each time. Nothing is charged for finding out what has happened. That examination finishes two working days after the parcel is logged in — never sooner, because rushing it is how evidence gets destroyed — and it ends in one number, written down, which does not move afterwards. Where a single drive is recoverable it is generally back with you two to four working days later. How you pay turns on which kind of fault it is, and the wording does not vary across this network: no fix, no fee all jobs except electronic and mechanical failures, chip level work, DVR and Forensic jobs. So a deletion, a broken file system, a lost partition, a firmware or service-area repair all keep the guarantee — nothing comes back, nothing is owed. Opening a drive under laminar flow, rebuilding a board, lifting a memory chip, or anything that arrived wet or burnt does not: half the agreed figure is taken before the work, and the balance only if your files come home. The published figures run: £250 + VAT for a memory card, a USB stick or a pen drive; £300 + VAT for any one hard drive or SSD, NVMe included; £400 + VAT for a recorder disk or a BitLocker volume; from £500 + VAT for an array, a NAS or a server; and £800 + VAT for a forensic investigation that ends in a written report. Five bands, all of them public, and the only one outside the guarantee is the forensic work. On packing: if the storage is screwed inside a laptop, a tower, a Mac or a server, take it out and send the bare unit, because dismantling equipment is not work this bench does. Three things go the other way and travel complete — an external drive in its own case, a NAS, and a WD My Passport or My Book with its cable, since on those the key sits on the bridge rather than the disk. A Fusion Mac needs both of its drives, labelled. A RAID or a server sends its member disks only, numbered by bay, with the front of the chassis photographed first. Flash soldered onto a mainboard, as on Apple Silicon Macs and many thin laptops, cannot be taken in at all, and neither can phones or tablets. Post it tracked and insured to Cambridge Data Recovery, Compass House, Vision Park, Chivers Way, Cambridge CB24 9AD, or put your own courier on it; the Cambridge reception takes devices across the counter too, Mon–Fri 9:00am–5:30pm. Nothing is collected, and there is no shopfront in the county to call at. Freephone 0800 689 0668. Packing notes and the shipping form live here.
Start a free diagnostic

Each of these is a real enquiry, taken from a household or a business somewhere in Leicester or the surrounding county, with the name of the customer, the firm and anything else that would identify them taken out. What is left is the reasoning: what turned up, what had actually failed inside it, the method that fault required and the equipment it was worked on. One of the thirty-six is a job that was refused, and it belongs here every bit as much as the successes do.

// nearby on the shelf

Three more files along the same lines

See every case file →

Is this your drive as well?

Examining it is free and takes 2 working days from arrival, the figure goes to you in writing and does not move afterwards, and a logical fault carries no fix, no fee. Start online or ring the freephone.