← All resources

History & heritage

Thirty years in service: keeping airborne equipment alive

An aircraft flies for twenty, thirty, sometimes forty years. The memory chip inside its recorder may leave its manufacturer's catalogue after three. Between the two calendars sits a discipline nobody mentions at the kick-off meeting, and it decides whether an installation can still be maintained in year fifteen. Here is how it works, with examples from our own fleet.

6 min read·Updated October 2026

Two calendars that never match

An airframe is designed for decades of service, and the equipment installed in it is expected to follow. Electronic components live on another calendar. A processor, a converter or a flash memory is produced as long as it sells, then announced at end of life, usually with a last order date a few months away. The equipment manufacturer sits between the two calendars, and its job is to keep the first one running while the second one keeps resetting.

scroll

One installation, several generations of parts — orders of magnitude: AIRCRAFT EQUIPMENT production support of the fleet COMPONENTS end of life → last-time buy, equivalent or redesign MEMORY 0 5 yrs 10 yrs 15 yrs 20 yrs 25 yrs 30 yrs The equipment has to outlive every part inside it.
Orders of magnitude, not a schedule: over the life of one installation, the equipment meets several generations of components, and many more generations of memory.

This is not specific to recorders. Every long-lived industry — aviation, rail, energy, space — deals with it under a name of its own; in aerospace it is often called DMSMS, for diminishing manufacturing sources and material shortages. What changes from one product to another is how much of the problem has been anticipated when the first schematic was drawn.

What disappears first

  • Memory, first. Flash generations follow the consumer market, and the smallest capacities are the first to go, because nobody else wants them any more. On one of our removable cartridges, the flash disk of the smallest capacity reached the end of its production, and the generation that replaces it starts eight times larger: same connector, same form, but a different part, which has to be qualified before it flies.
  • Programmable logic and processors, next. A component family is eventually retired, and its successor is rarely pin-compatible: what was a part change becomes a board change.
  • Connectors and passive parts, less often. But a discontinued connector series can be harder to replace than a chip, because it reaches into the aircraft harness.
  • And sometimes a regulation. In 2020, the rules on rechargeable lithium batteries changed enough for us to withdraw an internal battery option from the catalogue and move the energy reserve to aircraft level. Nothing had failed: the world around the product had changed.

Three answers, three costs

When a part goes, there are only three answers, and none of them is free.

scroll

AnswerWhat it costsWhen it is the right one
Last-time buyStock bought once and stored for years, under controlled conditions for parts sensitive to moisture; money tied up long before it is usedIdentical spares are needed over a known horizon, the fleet is small, and requalifying an equivalent would cost more than the stock
Form, fit and function equivalentA requalification against the original definition, or a statement of similarity when the change is limited and the evidence carries overThe part has a successor that fits the existing board and performs the same function
RedesignA new board, new qualification evidence, sometimes a new part numberEquivalents are exhausted, or the change is the occasion to bring improvements the fleet will use

The choice is rarely made once. A product that lives twenty years will usually see all three: a last-time buy to bridge a gap, an equivalent qualified on the next batch, and eventually a redesign that resets the clock for its components.

Sizing a last-time buy

A last-time buy is a bet on the future, and it is sized like one: with the fleet, its usage and its failure rate. Take 40 units flying 300 hours a year, with 15 years to go: 180,000 operating hours. With an MTBF of 50,000 hours, about 3.6 failures are expected over the period. Six spares cover it with a probability of about 93%; four would cover it only about seven times out of ten.

Consumables are sized separately. A removable memory cartridge is handled at every flight, dropped, lost, worn: its consumption follows the operations, not the failure rate, and only the operator knows its rhythm. The two figures together give the quantity; a margin for growth of the fleet gives the order.

Equivalent does not mean identical

"Form, fit and function" is a precise promise: the unit keeps the same dimensions, the same interfaces and the same behaviour, so that it stays interchangeable with those already flying. Keeping that promise takes evidence. The equivalent is tested against the original definition, on the same benches; when the change is limited, a statement of similarity explains why the earlier qualification still applies, and says which tests were repeated.

This is why test benches and reference units are kept alive as long as the product: a requalification in year fifteen needs the benches of year one. And it is why configuration control matters — the part number of a unit, and its serial number, must still say years later exactly what was delivered.

The data has to outlive the equipment

A recorder exists for its data, and the data often matters long after the flight: an incident reviewed years later, a fatigue analysis over a whole fleet's life, a certification file reopened. The media changed four times in our own history — tape, disk, SSD cartridge, network stream — and a recording made on each of them must still open today.

  • Open formats help: an IRIG 106 Chapter 10 file can be read by any compliant tool, now and later.
  • Proprietary formats must stay documented and convertible: the D.T.MUX format converts without loss to Chapter 10.
  • Replay software has to keep reading old recordings when it evolves, on computers that did not exist when they were made.
  • Readers for old media must outlive the media: a cartridge that can no longer be read is a lost recording.

What it takes

  1. Write the programme horizon into the specification. Twenty years of service is not specified like a six-month campaign, and the sensitive points are cheapest to handle while they are still design choices.
  2. Prefer components with long industrial life cycles, and keep the design in house: a redesign should not depend on someone else's availability.
  3. Follow every unit by its serial number, from manufacture to the fleet it serves. Our records still identify units delivered in 2006 in service today.
  4. Keep the test benches, the reference units and the documentation of every definition alive as long as the fleet.
  5. Publish the changes that affect ordering or operation, with a last-time-buy date when there is one; keep the invisible ones invisible.