Jump to content

Hard Drive Maintenance and Repair

From RetroTechCollection
Revision as of 00:22, 25 September 2026 by Josh (talk | contribs) (P4 verification: link IBM's PS/2 CD-ROM and SCSI-A technical references, hosted here but previously unlinked from any article)
IBM 3.5-inch hard disk drive

Mechanical hard drives from the early 1980s to the mid 1990s fail for a small number of well-understood reasons: perished rubber, hardened lubricant, stuck heads, and dead electronics. Most of what actually gets a vintage drive working again is done without opening it.

This page covers what is realistically repairable at a bench, what is not, and how to avoid turning a recoverable drive into scrap. It deliberately gives no rail-voltage tolerances, spin-up times or service intervals of its own: those are per-drive figures that belong in the drive's own OEM specification.

First: decide what you are trying to save

Three different jobs get confused with each other, and they need opposite approaches.

  • Recovering the data. Every power-up is a risk. Make an image at the first opportunity, then work on the drive afterwards. If the data genuinely matters, stop and consider a professional; a hobby attempt that fails usually removes the professional's chance too.
  • Getting the drive working as a drive. More latitude: you can retry, swap boards and experiment.
  • Keeping the machine period-correct without relying on a 35-year-old mechanism. Image the original drive, put it on a shelf, and fit an emulator — BlueSCSI, SCSI2SD, an IDE-to-CF or IDE-to-SD adapter. This is what most restorers end up doing, and it is not a defeat.

Do not open the drive

The heads fly on an air bearing a fraction of the thickness of a smoke particle. A drive opened in ordinary room air is contaminated the moment the lid comes off; the filter inside is designed to clean recirculated air, not to recover from an open-air exposure.

  • The breather hole (usually a small labelled port with a filter behind it) equalises pressure and must not be blocked.
  • Opening a drive with any intention of using it again afterwards is not a realistic plan outside a clean environment.
  • Do not clean platters. Wiping platters with alcohol on a swab is not a hobby procedure: platter surfaces carry a lubricant layer measured in nanometres, and wiping them removes it and scores the surface. If a platter is visibly contaminated the drive is already finished as a drive.

The one legitimate reason to open a vintage drive is to deal with degraded head bumpers (below), on a drive you have already accepted you may lose.

Faults you can fix without opening the drive

Stiction

Drives that have sat for years can have the heads stuck to the platter surface, or the spindle bearing lubricant set. The symptom is a drive that clicks, whines, or draws current and does not spin up.

The standard technique is a sharp rotational jolt of the whole sealed drive about the spindle axis — a firm twist of the wrist, in the plane of the platters, while the drive is powered and attempting to spin up. It breaks the head-to-platter bond without opening anything. It also risks the heads, so:

  • Do it once or twice, not repeatedly.
  • Have the imaging set up before you try it, so that if it spins you capture the data immediately.
  • If two or three attempts do nothing, opening the drive will not help either.

Controller board failure

Symptoms: no activity at all, or spin-up with no recognition.

  • Inspect the board for the usual suspects: shorted tantalums, a blown TVS diode across the supply rails, corroded solder joints, and burnt traces near the motor driver. A shorted TVS diode is a protection device that did its job — fitting a replacement without finding what over-volted it just repeats the event. See Capacitor Failure Symptoms.
  • A donor PCB can work, but it must be the same model and the same board revision, and on many drives the original board carries drive-specific data — a ROM or serial EEPROM holding calibration or adaptive parameters — that must be transferred to the donor. On older stepper-motor drives without adaptives, a straight board swap is more likely to work than on later voice-coil drives.

Interface faults

  • SCSI — check termination (one terminator at each physical end of the bus, and nowhere else), termination power, unique target IDs, and the parity setting. A single mis-terminated bus makes every drive on it look faulty.
  • IDE/ATA — check master/slave/cable-select jumpering against the drive's own label, and the cable orientation.
  • MFM/RLL (ST-506) — two cables per drive: a shared 34-way control cable and a dedicated 20-way data cable per drive. Getting the data cables swapped between two drives produces confusing, drive-specific symptoms.

Faults that require opening the drive

Degraded head bumpers

Many drives of this era park the heads against a moulded rubber or foam bumper. Over decades the material breaks down into a sticky tar that glues the actuator in place; the drive then reads as a seized actuator or a stiction failure that will not clear.

This is the one fault where opening the drive is routinely worthwhile, because the alternative is a drive that will never work again. Remove the degraded material completely with a lint-free swab and isopropyl alcohol, taking care that nothing falls onto a platter, and replace the stop with a non-degrading substitute — a small piece of Kapton tape, a nitrile O-ring or heat-shrink tubing are all used. Accept that the drive is now contaminated and treat it as a data-recovery attempt rather than a returned-to-service drive.

Seized spindle bearing

If a jolt does not free the spindle, the bearing lubricant has set or the bearing has failed. Some drives allow the spindle to be freed by hand through hub access holes. Realistically, a seized spindle usually means a donor drive — and on a drive with the platters removable as a stack, transferring the stack is a clean-room job, not a bench job.

"Click of death"

The phrase properly belongs to the Iomega Zip drive, whose failure mode misaligns the head and then damages every cartridge inserted afterwards. Applied to a hard disk it is used loosely for any repetitive seek-retry noise, which can be:

  • Heads stuck or damaged
  • The drive failing to find its servo or track-zero reference
  • A supply rail sagging under the drive's start-up current
  • A failed preamplifier on the head stack
  • Firmware or system-area corruption on later drives

Only the third of those is fixable at a bench. Diagnose the supply before assuming the drive.

Head alignment

On stepper-motor drives (ST-506 era), the head position is open-loop and can drift, with the classic symptom that the drive reads only what it wrote itself. Correcting it needs the manufacturer's alignment procedure and a servo track reader or equivalent; it is effectively not achievable at a hobby bench. On voice-coil drives the servo information is embedded on the platters and there is nothing to align.

Related: a stepper drive of this era can usually be low-level formatted to re-establish its track and sector structure, which sometimes recovers a drive with a corrupted format. A voice-coil drive with embedded servo cannot be low-level formatted by the user — running a utility that claims to do so either does nothing or writes zeros to the data area.

Electrical checks

Measure the supply rails at the drive's own connector, with the drive connected and spinning up — the current drawn during spin-up is the highest the drive will ever take, and a marginal supply or a corroded connector shows up there and nowhere else.

Compare the readings against the tolerance printed in your drive's OEM specification. No general figures are given here, because the tolerance is a per-drive specification. For a worked example of a real manufacturer's figures for a drive of the era, Tandon specified +5 V ± 0.25 V and +12 V ± 0.6 V, with under 100 mV peak-to-peak ripple on each, for its 5.25-inch drives.[1]

Also check the drive chassis is properly earthed to the machine. Tandon's manual makes the same point for its drives and it applies here: a floating drive chassis produces intermittent faults that look like media or head problems.[1]

Storage

  • Store drives dry, dust-free and at stable temperature, on a shelf rather than in a machine.
  • Store them flat, in the orientation they ran in, unless you know the model was rated for another.
  • Label known-good drives and donor drives, and record the model and board revision on the donor — that is the information you will want later.
  • No annual exercise cycle is prescribed here. The advice to spin a stored drive up once a year to prevent stiction is widely repeated but has no manufacturer figure behind it, and each power-up of an ageing drive is itself a risk. The safer policy is to image what matters once, properly, and then leave the drive alone.

References

  1. ↑ 1.0 1.1 Tandon TM100-1 / TM100-2 48 TPI Operating and Service Manuals, Tandon Corporation, manual part number 179022-001 Rev. B, July 1983 — hosted on this wiki as File:Tandon TM100-1 TM100-2 48TPI Operating and Service Manuals 1983.pdf, Table 2-1. Cited as an example of how a drive manufacturer of the period states supply tolerances (+5 V ± 0.25 V at 600 mA average, +12 V ± 0.6 V at 900 mA average, under 100 mV peak-to-peak ripple), and for the requirement that the drive chassis be earthed. This is a floppy drive specification and is offered as an example of the form such a specification takes, not as a figure for hard drives.
Retrieved from "https://wiki.retrotechcollection.com/index.php?title=Hard_Drive_Maintenance_and_Repair&oldid=13322"