Disk Health Assessment

What we found

• 1 sector(s) would not read on the first attempt but were recovered by re-reading the captured signal with different decoder settings. Nothing was lost, but this is an early warning sign rather than a clean result.
• 13 sector(s) could not be verified at all, affecting 6.5 KB of the disk. Where those sectors belong to a file, that file is listed below as damaged.
• This capture recorded a single revolution of each track. A multi-revolution capture would allow us to measure how repeatable each sector is, and would very likely recover more of any marginal areas.

Prospects for the unread sectors

Of the 13 sector(s) that could not be read, 13 look worth a further attempt. These are estimates from what this capture shows, not guarantees. The reasoning is straightforward: a sector that returns different data each time still has a readable signal being lost intermittently, and more passes frequently resolve it. A sector that fails identically every time is failing for a structural reason that more reading will not fix.

What is on the disk

38 file(s) in 0 folder(s), totalling 330.2 KB. The disk is labelled 10705210.05. Full checksums are in the technical appendix.

Capture provenance

Detected format

Nothing in a flux capture states the physical format. The identification below is inferred from what the decoder found: cylinder count, sector layout, data rate and measured spindle speed. A 1.2MB drive turns at 360 rpm and everything else at 300, which is what separates the two high-density formats.

How the score was reached

Signal timing

Our process records when each magnetic transition occurred: how far each transition landed from where the encoding says it should be. On healthy media the histogram below shows three narrow spikes at two, three and four bit cells. As media ages those spikes broaden and begin to merge.

Sector status by cylinder

The same information as the surface map, in the coordinates you would use to re-read a specific area. Rows are cylinders, columns are sectors.

• Read cleanly
• Needed a retry
• Failed checksum
• Header damaged
• Not found

Retry analysis

Where a sector fails its checksum, the captured flux is decoded again against another revolution of the disk and, if that fails too, with different phase-locked-loop settings. Nothing is re-read from the physical media.

No single tuning wins everywhere: a track whose speed drifted wants a loop that tracks aggressively, while a track with localised dropouts wants one that holds its lock and coasts through. The count below is how many sectors each tuning rescued that earlier ones could not.

Affected sectors

Angle is measured from the index pulse, which is what makes physical damage identifiable: a scratch or a mould ring produces failures clustered at similar angles across many cylinders.

Filesystem integrity

The filesystem was parsed directly from the assembled image without mounting it. The kernel driver exists to make damaged filesystems look usable — it prefers the first FAT silently, skips entries it dislikes and reports none of it — which is precisely the information this section is for.

File inventory and checksums

SHA-256 is computed over each file's recovered contents, truncated to the length recorded in its directory entry. Files marked other than intact contain at least one sector that failed verification, or have a damaged allocation chain.

Method and limitations

What was done

The capture was decoded from raw flux timings. A software phase-locked loop converts the intervals between magnetic transitions into a bit stream; sector address marks are located in that stream and each sector's header and payload verified against its own CRC. Where a checksum failed, the same flux was decoded again with different loop settings and against other captured revolutions before the sector was declared unreadable. No data was re-read from the physical disk at any point in this analysis; every figure here comes from the single capture named above, which is why the same capture can be re-analysed indefinitely without further wear on the media.

What this cannot tell you

• Magnetic signal strength is not measurable from this equipment. A Greaseweazle records transition timing only. Every quality figure in this report is derived from timing accuracy and read repeatability, which correlate with signal health but are not a direct measurement of it.
• The overall score is a weighted heuristic, not a measurement. Its components are shown individually above so that the verdict can be checked against the underlying numbers. Calibration version 2026.08-r1.
• A single-revolution capture cannot measure repeatability. Sectors that read correctly once may not read correctly again.
• Physical media size cannot be determined from flux alone and is inferred from cylinder count, data rate and spindle speed.
• A sector that passes its CRC is almost certainly correct, but CRC-16 is not a cryptographic checksum. Roughly one damaged sector in 65,536 would pass by chance.
• This assessment describes the condition of the media at the moment of capture. Magnetic media continues to degrade in storage.
• A raw sector image cannot record that a sector is unknown; it can only record what a sector contains. The list of unreadable sectors in this report and its JSON companion is the authoritative record, not the image. Where an unambiguous artefact is needed, an ImageDisk (.imd) copy carries per-sector error and unavailable flags directly.