Manchester is the hand-over point — the Oxford Street drop-off, or the post. From there, every drive walks the same benches by the same route, whether it arrived from Manchester, Burnley or Bolton. Consider this the tour: bench by bench, what each piece of kit is, what it does, and why it settles which recoveries work.
Intake write-blocking, imaging done in hardware, firmware repair, and a clean-air corner for the mechanical jobs — each job runs through the stations it needs and skips those it doesn't.
Roughly three nanometres, and it explains most of what this work involves.
A modern hard drive’s read/write head flies about three nanometres above the platter — a gap thousands of times smaller than a human hair, maintained by an air cushion generated by the disk’s own rotation. That single figure explains almost everything about mechanical recovery.
It explains why a drive cannot be opened on a desk: ordinary room air carries particles hundreds of times larger than the gap, and one of them under a head gouges the surface your data is written on. It explains why dropping a running drive is catastrophic and dropping a parked one usually is not. And it explains why a drive that has been clicking for a fortnight is a different proposition from one switched off after ten minutes — because every rotation with a compromised head passes the whole track underneath it again.
The most common question, and the honest answer to it.
Plugging a failing drive into a computer and dragging files across fails for a specific reason: the operating system was designed for healthy hardware. Hit a bad sector and Windows retries, waits, retries again, and eventually aborts the whole copy — hours of work discarded because of one unreadable block near the end. Meanwhile every one of those retries is more stress on a drive that has limited reads left in it.
Purpose-built imaging hardware inverts that. Timeouts are set in milliseconds rather than seconds, so a bad sector costs a fraction of a second instead of a minute. Unreadable areas are logged and skipped rather than retried indefinitely. Reads can be targeted at one head at a time, so an undamaged surface is captured completely while a scored one is worked around. And the whole thing is strictly read-only — the source cannot be written to even by accident, which no ordinary operating system will guarantee.
Which is why a mechanically perfect drive can be completely invisible.
A hard drive is not a passive lump of magnetic material. It runs firmware — a real program, with its own processor and memory — and that firmware lives in a reserved service area on the platters rather than in a chip. It holds the defect map, the translator converting logical addresses to physical positions, and adaptive calibration values unique to that individual drive.
When part of that becomes unreadable the drive cannot finish starting. It spins perfectly, sounds healthy, and reports zero capacity or a blank model string, because it cannot describe itself. Your data is entirely untouched behind a piece of software that will not load. Manufacturer diagnostic modes let that region be read and repaired directly — modules rebuilt or substituted from a matched donor, translators regenerated from surviving defect lists, ROM adaptives carried across when a board must be replaced. SSDs have the equivalent problem and the equivalent route in.
The reconstruction stage, and the principle behind it.
When a RAID or NAS reaches us, the one component that has already proved unreliable is the thing that used to describe the array. So its own account of itself is treated as a hint rather than a fact.
Instead the layout is derived from the data. Stripe size shows up as a repeating structural period across members. Disk order is established by testing arrangements until file-system structures resolve into something coherent rather than noise. Parity rotation is confirmed by checking that reconstructed stripes actually compute. Every candidate is tested against what is physically on the disks, and the correct answer announces itself by producing a mountable volume where the wrong ones produce nothing.
Above that sit whatever layers the vendor added — LVM pools, Synology’s SHR, QNAP thin provisioning, Drobo’s undocumented BeyondRAID — each unwound in order, because an error at one level surfaces as nonsense several levels up. All of it happens against read-only images, entirely outside the hardware that failed.
You can, and on a healthy drive it works. On a failing one the operating system retries each bad sector for seconds at a time and eventually aborts the whole copy — discarding hours of work and putting real stress on a drive with limited reads left.
The distance a hard drive’s head flies above the platter, maintained by an air cushion from the disk’s own rotation. It is why drives cannot be opened in ordinary room air, and why a dropped running drive is so much worse than a dropped parked one.
It cannot load its own firmware. Drives store that in a reserved service area on the platters rather than in a chip, and if part becomes unreadable the drive cannot describe itself — so it reports zero capacity while your data sits untouched behind it.
No. It is the component that already failed, so its description is treated as a hint. The layout is derived from the data itself — stripe size, disk order and parity rotation all tested against what is physically on the members.
No, and that is not the goal. Repairs go only as far as reading the data once. A drive with donor heads or a transplanted platter pack is a recovery vehicle rather than a disk to trust. You get your data back on fresh media.
No. Work is carried out in-house by our own engineers rather than subcontracted, and data does not leave the UK at any point.
The first station is free: a diagnostic within 48 hours, then a written quote before any recovery begins.