Descent toward familiar terrain, a standard arrival selected from the database, and a doubt: the altitude constraint shown on the chart at the first point of the transition does not appear on the procedure page. The displayed cycle is in effect, the equipment is nominal, no notice has come through. Coding error, rendering difference, or a procedure amended since publication? Answering that requires tracing what happened between the text published by the State and the screen on the flight deck: an industrial chain in which the same content is rewritten three times.
Four links, and three translations
The procedure is born at the State. A designer applies obstacle clearance criteria, places points, sets minimum altitudes and a gradient, and the result is published in two forms intended for the same human reader: descriptive text and a chart. Neither one loads into a computer.
Then an aeronautical data provider steps in. It reads the publication and rewrites it in a standardized exchange format, where the procedure is no longer a drawing but a series of records: georeferenced points, legs, altitude and speed constraints, sequencing flags. This is the first translation, and the only one that is interpretive in the full sense — we will see why.
Then an avionics supplier — the manufacturer of the equipment or of the aircraft — adapts this database to the target computer. It extracts what the machine can handle, within the limits of its memory capacity, and produces a binary file specific to an equipment type, sometimes to a software version. Second translation. Finally the operator loads this database on board and distributes charts to the crew in parallel: third handling, and the first opportunity for the two products to fall out of sync.
Here is the structuring point. What the crew reads on the chart and what the computer executes are two products of the same source, obtained through separate chains, sometimes by separate companies. A discrepancy between the two does not prove that a fault was committed; it proves that there were two translations.
The regulatory framework governs these translations through attributes carried by the data itself rather than by a final check: accuracy, resolution, traceability, integrity classification. Three classes are distinguished — routine, essential, critical — corresponding to increasingly low probabilities of undetected corruption; the coordinates of a runway threshold fall into the most demanding of these. Added to this is a mandated reference frame, the World Geodetic System: a publication left on a local datum would introduce a systematic offset invisible on board.
When a procedure becomes a sequence of legs
The chart says: climb on heading 070 to a given altitude, then turn left direct to the next point. A computer cannot read that: it needs to know what defines the path and what terminates it. The exchange format answers with a pair, a path and a termination — here two legs, a heading held to an altitude then a direct-to on a point.
This pair concentrates most of the coding risk, because several pairs can translate the same published text without being equivalent in flight.
| What the chart says | Corresponding coding | What determines the path | What changes if the coding differs |
|---|---|---|---|
| Follow the track between two published points | Track between two fixes, TF |
Two coordinates: path fully determined | Nothing, except an error in one of the coordinates |
| Intercept a course then follow it to a point | Course to a fix, CF |
A course and an ending point; the origin is not fixed | The intercept path further upstream, hence the lateral protection |
| Direct to the next point | Direct to fix, DF |
A single point; the path depends on position and speed at the trigger moment | The arc flown, the distance remaining, the obstacle margin |
| Climb on heading to an altitude | Heading or course to an altitude, VA or CA |
An altitude, not a position: the leg’s end is mobile | The point where the next turn begins, very sensitive to weight and wind |
| Constant-radius turn between two points | Constant-radius arc, RF |
A radius, a center, an ending point: path fully imposed | The very ability to fly the procedure: there is no degraded equivalent |
| Hold until instructed | Hold with manual termination, HM |
A point, a leg, a direction; exiting is a crew action | The sequencing: with automatic termination the hold exits on its own, with manual termination it is held until the crew acts — the two surprises are symmetrical |
Two properties compound the issue. The coding does not determine the entire path: the transition between two legs is computed on board, so two units receiving the same coding do not fly the same arc — the constant-radius leg exists precisely to remove this indeterminacy. And the way a point is crossed, overflown, or anticipated is itself part of the coding, while reading poorly on a chart.
Then there is the question of magnetic courses. The database carries magnetic values referenced to a declination, and the computer may apply its own model. A displayed course can therefore differ from the chart by about a degree without any error existing. Symmetrically, when a State updates the magnetic orientation of its runways, the courses of all associated procedures change even though nothing has moved geometrically — and the two products, chart and database, do not necessarily switch over on the same cycle.
The navigation database is like a translation of the paper charts into computer language. But because different people translate it at different stages, small mistakes can creep in. The pilots see one version, the plane’s computer sees another, and they might not match perfectly. That’s normal, but big differences mean trouble.
An Airbus A320 is flying into Paris Charles de Gaulle (LFPG) on the ILS RWY 27L approach. The chart shows a “heading 070 to 3,000 feet” leg, but the FMS shows “course 070 to 3,000 feet.” The difference is subtle but important: a heading is relative to the plane’s position, while a course is a fixed magnetic direction. If the plane’s magnetic variation model differs from the chart’s, the actual flight path could be slightly off.
Where the discrepancy enters, link by link
Each link has its own error signature, and above all its own observability. Confusing the two leads to searching in the wrong place, usually on board, where the discrepancy was not detectable.
| Link | Typical nature of the discrepancy | Observable from | Who can catch it |
|---|---|---|---|
| Design and publication by the State | Internal inconsistency of the procedure, constraint missing from the publication, non-conforming reference frame | Nowhere on board: chart and database say the same thing | Design review, operational feedback reported to the State |
| Coding by the data provider | Questionable leg termination, omitted or reversed-sense constraint, misresolved duplicate point identifier | Chart-to-database comparison, procedure by procedure | Provider’s quality assurance, entry by the operator |
| Adaptation to the equipment | Content truncated by memory capacity, unsupported function rendered in a degraded form | Database accompanying notices, ground selection test | Avionics supplier, operating restriction |
| Loading and distribution by the operator | Non-conforming cycle, partial database, charts and database on different cycles | Cycle and coverage check before flight | Operator, before departure |
| Selection and insertion by the crew | Duplicate procedure name, wrong transition, duplicate point, manual entry | Cross-check, track-and-distance readout | The crew, immediately |
The last line deserves a remark: the crew is a coding link, not merely a reader. Five-letter point identifiers are not guaranteed unique across a worldwide-coverage database, and a manually entered point can resolve to a homonym far removed from the expected position. The track and distance readout to the inserted point is the only check that catches this.
The AIRAC cycle: what the effective date cannot carry
The Aeronautical Information Regulation and Control cycle has a periodicity of twenty-eight days, with effective dates common to all States, at zero hours Coordinated Universal Time. Its function is industrial before it is regulatory: to synchronize every chain so that publication, coding, adaptation, and loading switch over at the same instant. The lead times required upstream — on the order of several weeks before the effective date, and notably extended for major changes — have no other purpose than to let this chain run its course.
Three consequences follow, and all of them are operational.
- The database is a snapshot. Everything decided after a cycle’s freeze date cannot appear in it. That is exactly the function of the notice to airmen — and such a notice never enters the database.
- A flight can span a cycle change. Depending on the installation, the active database can switch at the effective time, in flight. A procedure briefed at departure may be coded differently on arrival.
- Two cycles often coexist on board. Checking which one is active, and that it covers the whole rotation, return included, is a check distinct from « is the database current. »
The trap. A database does not go stale: it stays internally consistent. A procedure suspended or amended by notice to airmen still retains, in the computer, a complete coding, constraints, a gradient and a final segment, and can be selected without the slightest warning. The cycle shown as current dates the database’s production, not its agreement with the state of the terrain. The only safety net is reading the flight folder, done before selection and not after.
What the crew detects, and what it will not detect
The chart-to-database comparison is the chain’s last check, and it is a gross-error check: it catches what shifts the geometry, not what offsets it. It still helps to know what it usefully covers.

- The sequence and identity of the points, the transition selected, the runway identifier and the procedure suffix: two identically named procedures on the same runway are distinguished by the transition or by the required equipment.
- The track and distance between consecutive points — the telltale sign of a coordinate error.
- The altitude and speed constraints, and above all their sense: mandatory, minimum, maximum, or a window between two values.
- The final course, the vertical path angle, the threshold crossing height.
- The first legs of the missed approach, up to the first level-off or holding point.
- The presence of a constant-radius leg, and whether the equipment supports it and is approved for it.
One nuance conditions the usefulness of the exercise: distinguishing the artifact from the discrepancy. A degree on a course, a fraction of a mile on a distance, a rounded constraint value are matters of resolution and of the declination model; they are small, systematic, and shift nothing. A genuine discrepancy changes a point, the sense of a constraint, a leg type, or a gradient. Confusing the two produces either unactionable reports or a habituation to differences that ends up masking the one that matters.
| Discrepancy | Caught? | Observable sign | Course of action |
|---|---|---|---|
| Missing or extra point in the sequence | Yes | Number and identity of the displayed points | Do not fly the procedure from the database |
| Constraint omitted or reversed in sense | Yes, if the page consulted displays it | Procedure page and vertical profile | Constraint reinstated outside the final segment, or a different approach |
| Large-magnitude coordinate error | Yes | Abnormal track or distance between two points | Procedure abandoned, immediate report |
| Small-magnitude coordinate error | No | None on board | No in-flight remedy: falls to the upstream chain |
| Declination model different from the chart | Not applicable | A degree-order discrepancy, systematic and constant across the whole procedure | Recognize the artifact, do not report it as an error |
| Procedure amended or suspended by notice | No | Nothing in the computer | Read the flight folder before selection |
| Database internally consistent but offset from the source | No | None: everything agrees on board | Provider acceptance, audit, operational feedback |
| Constant-radius leg not supported | Yes | Procedure absent, incomplete, or an equipment message | Verify declared capability before the flight |
What escapes the comparison falls into three families: an error too small to shift a displayed track but large enough within a confined segment, a database that is wrong yet perfectly self-consistent so that everything agrees on board, and an omission that the chart shares because the error originates upstream.
Hence a rule that looks rigid and is in fact mechanical: the coded final approach segment is not modified — no point insertion, no coordinate entry, no constraint edit. If it does not match the chart, the procedure is not flown from the database. For vertical minima supported by satellite-based augmentation, the data block for this segment is further protected by a cyclic redundancy check: any alteration is rejected, not executed. That is the technical reason this segment cannot be repaired in flight.
The right reflex. Compare the procedure as loaded, not as published: sequence of points, tracks and distances between consecutive points, sense of the constraints, final course and gradient, then the first legs of the missed approach. Do it before top of descent. The missed approach is the part least compared and the one most often flown under heavy workload.
Reporting a discrepancy: the loop that closes the system
A discrepancy found and not reported is a discrepancy that will return at the next cycle, over the same terrain, for another crew. The correction chain is the exact reverse of the production chain, and it does not start on its own.
The operator has a designated function for navigation data management. It is this function that qualifies the discrepancy, escalates it to the provider, and through the provider to the State if the origin lies in the publication. An entry in the aircraft technical log reaches no one who can correct the data: that document deals with the aircraft, not the data.
The scheme rests on formal acceptance rather than on re-verification. An aeronautical data provider is recognized by the authority at two levels: one covers generic data processing, the other adaptation to a specific piece of equipment, and both rely on the industry standard for aeronautical data processing. The operator therefore does not have to re-verify the entire database — that would empty the acceptance of its meaning. It must, however, be able to demonstrate that the loaded database is the one it ordered, that it processes reported discrepancies, and that it issues a restriction when a discrepancy is confirmed before the correction arrives.
The time constant is that of the cycle. Between the report and the correction, the safeguard is documentary: an instruction that prohibits the procedure or prescribes a particular check. This is the most vulnerable moment of the whole system, because protection no longer resides in the data but in a hurried crew’s reading of an instruction.
What to find in your own documentation
The mechanism is universal; its local translation is not. Three families of documents carry the answers.
In the operations manual: who manages the database, how a discrepancy is reported and in what form, what processing time is announced, how the cycle changeover is handled on flights that span it. And a question rarely asked: do the chart and the database come from the same provider? The answer determines the frequency of benign discrepancies, and hence a crew’s reasonable alert threshold.
In the avionics and manufacturer documentation: which leg terminators the equipment supports, what it does with an unsupported leg, which pages actually display the constraints — a constraint can exist in the database without appearing on any page consulted routinely — and the rule for the final segment, as the manufacturer states it.
In the approval documents relating to performance-based navigation: which operations are made contingent on database integrity, what comparison is required before which procedure, and what conditions accompany the use of constant-radius legs.
One gray area remains that nobody settles in flight. Two equally defensible codings of the same published text produce two different paths, and nothing on board says which one the designer had in mind. The memory capacity of older computers still forces content trade-offs whose list the crew does not have. And the underlying asymmetry remains: the crew is the last barrier while seeing the least of the chain, since it compares two products without ever accessing the source. For anyone preparing a type rating, the expected answer is not the list of termination codes: it is the course of action to hold when the loaded procedure differs from the chart on the final segment, minutes from top of descent, with a distant alternate — what to decline to fly, what to substitute, and what to write down afterward. Threads from the FL330 community regularly supply concrete cases of this.
The applicable text, the aeronautical information publications of the States concerned, and the operator’s approved documentation take precedence over this article, whose purpose is limited to explaining the mechanism.
Questions fréquentes
Does an up-to-date database guarantee that a procedure is valid?
No, it only guarantees that the database was produced for the current cycle. Everything decided after the cycle’s freeze date is distributed by notice to airmen, and a notice never enters the database: a procedure that has been suspended or amended remains present, internally consistent, and selectable there, with no warning at all from the computer. Operational validity is therefore checked in the flight folder, before the procedure is selected, not on the page that displays the cycle.
Is a one-degree discrepancy between the chart's course and the one shown by the computer a coding error?
Usually not. The database carries magnetic courses referenced to a declination, and the computer may apply its own declination model, which produces discrepancies on the order of a degree, small and systematic. A genuine discrepancy is recognized differently: it changes a point, a leg type, the sense of a constraint, or a gradient. Treating the two the same way leads either to unactionable reports or to a habituation that ends up masking the discrepancy that matters.
Can a missing constraint in a loaded procedure be restored in flight?
Outside the final approach segment, inserting a constraint is possible subject to the operator’s procedure and a cross-check. Within the final approach segment, no: this coded segment is not modified, neither by point insertion, nor by coordinate entry, nor by constraint editing. For vertical minima supported by satellite-based augmentation, the data block for this segment is protected by a cyclic redundancy check, so that any alteration is rejected rather than executed. If the loaded procedure differs from the chart on this segment, a different procedure is being flown.
Why can the chart and the database diverge without anyone having made a mistake?
Because they are two different products drawn from the same source. The State publishes text and a chart intended for a human reader; an aeronautical data provider then rewrites the same procedure in an exchange format intended for a computer, and an avionics supplier then adapts it to a given piece of equipment. These chains are separate and sometimes held by separate companies, and they do not necessarily switch cycles at the same time when a runway’s magnetic orientation or a constraint is updated.
Who should a coding discrepancy be reported to, and with what expected effect?
To the function designated by the operator for navigation data management, which qualifies the discrepancy, escalates it to the provider, and through the provider to the State if the origin lies in the publication. An entry in the aircraft technical log reaches no one able to correct the data, since that document deals with the aircraft. The correction arrives at the earliest at a later cycle; in the meantime, protection takes the form of an operating instruction that prohibits the procedure or imposes a particular check.
What happens if a flight spans a cycle change en route?
Depending on the installation, the active database can switch at the effective time, that is, in flight, and a procedure briefed at departure may be coded differently on arrival. Two cycles frequently coexist on board, which requires two distinct checks before departure: which one is active, and whether the coverage selected spans the entire rotation, return included. The course of action for affected flights must be described in the operations manual rather than improvised en route.

Leave a Reply
You must be logged in to post a comment.