A film collection built up over twenty years, about 3.5TB of it, sat on a 4TB My Book in a house in Formby. The drive would come up, run for a few minutes, and drop off. He had bought a replacement and had been trying to copy across for weeks, getting a bit further some nights and no further others.
| Media | A WD My Book 4TB external drive holding roughly 3.5TB, which mounts and works for short periods before dropping out, preventing any complete transfer. |
| What was reported | WD My Book 4TB, around 3.5TB in use, works for short periods then stops, repeated attempts to copy to a replacement drive all stall part way, data still wanted. |
| Fault class | Logical, at sector level. A large and growing area of weak sectors was stalling the drive long enough for the host to give up on it and drop the connection. The mechanism was sound and the board healthy. |
| On the bench | Read behind a hardware write blocker so nothing could be written, imaged read only on DeepSpar across several controlled sessions with readable areas captured first and weak regions revisited in short attempts under a health watch, filesystem parsed in R-Studio from the completed image, files validated and written to fresh media. |
Dropping out is a timeout, not a power failure. When a drive stalls on a sector it cannot read, it stops answering. The host waits, decides the device has gone, and drops it. The drive is still spinning. It is just not talking.
Copying is the worst possible read order. A file copy reads in whatever order the filesystem hands out, which means it walks into bad regions early and repeatedly. Every stall is minutes of running time spent achieving nothing, and it is running time the drive has a finite amount of left.
Imaging takes the easy ground first. The bench captures everything that reads cleanly in one pass, then comes back for the difficult regions in short bursts. On a drive with limited working life left, the order in which you read it is the single biggest factor in how much you get.
Weeks of retries had not made it worse, but they had not helped. Each stalled attempt re-read the same early files and then hit the same wall. He was not doing anything wrong. He was using a tool built for healthy drives.
The drive went behind a hardware write blocker and was imaged read only on DeepSpar. The first pass skipped anything slow, capturing the clean majority of the surface without ever letting the drive sit grinding.
Weak regions were revisited across several later sessions in short controlled attempts with head health watched throughout, and the drive was rested between them. The filesystem was parsed in R-Studio from the completed image, files validated by playing samples across the collection, and the whole set written to the replacement drive he had already bought.
Effectively the entire collection came across, with only a small number of titles affected by the worst of the weak region and identified by name so he knew exactly what to replace.
One drive, £300 + VAT. The work was a read only image with nothing opened and no board repair, so it ran on no fix, no fee terms.
Stop trying to copy. Every stalled transfer spends the drive's remaining working life on the region least likely to give anything back, and you only get so many hours out of a drive in this state. Do not run a check disk and do not let Windows repair the volume, because both write. Do not leave it plugged in overnight to see whether it gets further. Unplug it, and if you have already rescued part of the data, note what you have so the bench can concentrate on the rest. Imaging in controlled passes is a different job from copying and it gets a different result.
Every case file here is written up from a real enquiry handled for households and businesses across Liverpool, Merseyside and the wider North West, anonymised so that nobody is identifiable. Each one gives the reasoning behind the diagnosis, the method the fault actually called for, and the equipment it was worked on.
A large drive that will not register has a firmware, board or mechanical cause, and only a direct read separates them
Clicking is a head fault and a rattle after a knock is usually loose hardware, and both are settled in one clean-air visit
Write errors are an early warning, and the right response is one careful image, not weeks of retries
The diagnostic is free and takes 2 working days from arrival, the figure is fixed in writing, and logical faults run on no fix, no fee. Start online or ring the freephone.