Two aircraft from the same operator fly the same departure two minutes apart: the first cuts sharply inside the turn point, the second swings wider around it, and both are compliant. A hundred miles further on, on an approach in constrained terrain, the same discrepancy would no longer be acceptable. The difference has nothing to do with piloting technique and everything to do with the type of leg coded: in one case the FMS invents the curve at the moment it flies it, in the other it receives one it has no right to depart from. What remains is to know, looking at the magenta line, which of the two you are following.
The FMS does not fly the chart, it flies a sequence of coded legs
The navigation database contains no image. It contains a sequence of legs, and each leg carries a two-letter path terminator that answers two questions: how the path is constructed, and what it ends on. The vocabulary is finite. TF joins two fixes along a great-circle arc. CF joins a fix while holding a magnetic course. DF goes directly to a fix from a position not known in advance. RF imposes a constant-radius arc terminating on a fix. FA, CA and VA terminate not on a fix but on an altitude. AF follows a constant distance from a beacon.
The chart you read is a rendering of this sequence, and the magenta line a rendering of that rendering. Nothing requires the three to coincide everywhere: the chart publishes what the designer intended, the coding retains what the underlying standard is able to express, the display shows what the FMS has managed to reconstruct with the functions it has available. Performance-based navigation specifications work precisely on this vocabulary: they authorise only a subset of path terminators, exclude those whose ground track is not reproducible, and treat the constant-radius arc as an optional function an aircraft either has or does not have — never as a consequence of an RNP value.
The question « where is the aircraft relative to the path? » therefore does not mean the same thing depending on the active leg. On some, the path exists before the flight, it is purely geometric, and the deviation can be measured against it. On others, no reference exists at all: an intention, an algorithm, a result.
The FMS doesn’t follow a picture of the route. It follows coded instructions that tell it how to fly from point to point. Some paths are fixed, like a straight line between two points. Others are flexible, like turning based on wind or speed. The magenta line on your screen is just the FMS’s best guess at what the chart means.
An Airbus A320 and a Boeing 737 fly the same SID with a constant-radius turn. The Airbus starts the turn 1.5 NM early due to its bank angle logic, while the Boeing starts 0.8 NM early. Both stay within protected airspace, but their exact tracks differ. On an RNP AR approach in the Alps, the same flexibility would be unacceptable, and the procedure would require a coded arc instead of an anticipated turn.
Three families of path, and only one where lateral deviation means something
The first family groups the legs whose geometry is entirely fixed by the database: TF, RF, CF, AF. Two different aircraft, two different FMS manufacturers, the same ground track. It is here, and only here, that the displayed lateral deviation is a usable quantity, that containment can be reasoned about, and that path monitoring has operational meaning.
The second family is not in the database: it is the transition an FMS builds for itself between two consecutive legs when the fix separating them is a fly-by. Nobody has published this curve. It is computed on the fly, from the day’s ground speed, the change of course, the bank angle limit adopted by the manufacturer, and the wind estimate.
The third family groups the legs whose path does not exist before it is flown: the DF, which starts from an unknown position, and the heading-held legs, where it is the wind that writes the ground track. A heading leg does not define a course, it defines an aircraft orientation.
| Family | Typical path terminators | Who sets the geometry | Usable lateral deviation | Track reproducible from one aircraft to another |
|---|---|---|---|---|
| Database-defined path | TF, RF, CF, AF |
The designer, then the coding | Yes, over the whole length | Yes |
| Transition computed in flight | Anticipated turn between two legs | The FMS algorithm | No, a value with no published reference | No |
| Path undetermined before flight | DF, heading legs |
The starting position and the wind | No, during the leg | No |
The anticipated turn: a geometry nobody has published
Approaching a fly-by fix, the FMS determines a turn initiation point located upstream. The anticipation distance depends on the change of course, through a trigonometric function of its half-angle, and on the radius the aircraft is going to hold. This radius is not in the database: it follows from the estimated ground speed and the bank angle authorised in the active mode. Since it grows with the square of the speed, the distance cut inside the fix grows likewise, and grows further still as the change of course approaches a right angle.
Add the roll-in and roll-out rates, the reduced-bank logics active at altitude, the modes that cap earlier than others: two aircraft on the same departure will not describe the same curve, and neither is at fault. Inside an anticipated turn, the displayed lateral deviation compares to nothing published. What is to be judged is the recapture of the next leg, not the track through the bend.
The designer, for their part, pays for this uncertainty in surface area: they do not protect a curve but a turn area built on an assumed speed and bank angle, widened by a wind spiral, because they do not know where the aircraft will begin to roll. This area is wide, and in a constrained corridor it becomes unaffordable.
The mandatory flyover fix — circled on the chart — answers the same problem from the other end: it forbids anticipation because the upstream geometry cannot support it — an obstacle gradient, adjacent airspace, or a track that must not be left too early. Its turn area is wider still, since roll dispersion begins at the fix. And what follows it is often a DF, because no coding can promise where the aircraft will be at the end of the turn.
The imposed arc: the radius is fixed, the bank angle follows ground speed
The constant-radius leg reverses the logic. The arc is a complete object in the database: a centre, a radius, a turn direction, two endpoints, and a tangency constraint with the preceding and following legs. The FMS does not invent it, it tracks it. Lateral deviation regains its full meaning there, over the whole length of the curve, including in the middle of the turn.

What the arc imposes on the flying can be read in the mechanics of flight, with no regulation involved. Holding a given radius at a given ground speed requires a turn rate equal to the ratio of speed to radius, hence a bank angle whose tangent is proportional to the square of the speed divided by the radius. With the radius fixed by the procedure, only one variable remains: ground speed. Doubling it quadruples the tangent of the bank angle required. The speed limitation shown on the chart is therefore not a comfort margin, it is the value beyond which the required bank angle leaves the design assumption — an assumption that typically remains on the order of twenty degrees or so in the approach environment, and less at altitude.
Three effects add up. Wind: on an arc, the longitudinal component changes continuously, so that a curve entered in a crosswind sometimes ends in a tailwind, and the required bank angle peaks where ground speed peaks. Density altitude: for the same indicated speed, a hot, high airfield gives a higher true airspeed, hence more bank angle for the same arc. Finally, the go-around: when an arc appears in the missed approach, the natural acceleration of the manoeuvre works against holding the radius, and that is where the bank angle limit is reached first.
When that limit is reached, nothing dramatic happens: the aircraft banks as much as it is allowed to, drifts toward the outside of the arc, and the only indication is the lateral deviation opening up. Any logic that lowers available bank angle produces the same effect with no dedicated alert. Hence the requirement, in many approved flight manuals, for an engaged autopilot or at least a flight director to follow an arc.
The right reflex. Before entering an arc, check ground speed rather than indicated speed, and verify that no reduced-bank logic is active. If ground speed exceeds what the chart limitation assumes, the reduction must be decided before the tangency point: once the arc is engaged, slowing down no longer brings the aircraft back onto the curve, it only stops it from drifting further away.
What obstacle protection assumes of each regime
The difference in how the designer treats each case explains why the arc exists. Around a constant-radius arc, the protected area keeps the same half-width as the adjacent straight legs: no wind spiral, no widening, since the aircraft is assumed to be tracking the curve. This is what allows a path to fit where a conventional turn area does not. The space saved is bought against a path-keeping requirement, an aircraft eligibility, and a speed limitation.
| Point of comparison | Anticipated turn between two direct legs | Constant-radius leg |
|---|---|---|
| Origin of the geometry | Computed in flight, unpublished | Coded in the database, tangent to the neighbouring legs |
| Protected area | Widened turn area, wind spiral included | Corridor of the same half-width as the straight portion |
| Speed assumption | Assumed by the designer, not binding on the crew | Limitation shown on the chart, structural |
| Effect of excess speed | Wider turn, generally still inside the area | Required bank angle outside the assumption, exits the corridor |
| Aircraft eligibility | No particular function | Declared function, independent of the RNP value |
| After a guidance interruption | The next leg is rejoined without difficulty | Recapture of the arc is not guaranteed |
The manoeuvres that erase the arc without a word
A coded arc is fragile because it is only a line in a flight plan, and any action that rewrites that flight plan erases it. The common case is the direct-to shortcut toward the fix that ends the arc: the FMS replaces the arc with a direct leg and the aircraft flies the chord instead of the curve. On a tight arc, the chord cuts well inside the curve, hence outside the protected corridor: the maximum deviation, reached at the middle of the curve, is a fraction of the radius that grows fast with the angle swept, so that a long, tight arc is exactly the case where the shortcut does the most damage. Inserting a fix or an open discontinuity upstream produces the same erasure, with a replacement geometry that is even less predictable.
Switching to heading mode or to vector guidance leaves the arc, and the recapture is not symmetrical: some FMS cannot recapture an abandoned arc and sequence to the next leg, so that the return to lateral guidance rejoins a straight line toward the downstream fix. This behaviour is known type by type and standard by standard; it is not something to improvise in flight.
The trap. A shortcut clearance received during an arc looks harmless because it names a fix that is already on the path. Accepting it nonetheless destroys the arc and substitutes a chord for it: the aircraft leaves the corridor with no alert to signal it, and on a procedure whose protection stops at the containment boundary, there is nothing behind it. This shortcut should therefore not be accepted as such — it calls for a refusal, or a heading with a new clearance and an explicit recapture path.
Eligibility, coding, magnetic variation: the mismatches invisible in flight
The ability to fly a constant-radius arc is a declared function, found in the approved flight manual or the type’s operational suitability data, and tied to a specific avionics standard. It cannot be inferred from any navigation specification. The mismatch becomes concrete when the database contains an arc the installation cannot fly: depending on the case, the procedure is rejected, or it displays and the path followed is no longer the one that was protected.
The leg held to a magnetic course carries a slower drift. The coded course depends on the magnetic variation value adopted at the time of coding; the field itself keeps moving. Over successive cycles the discrepancy builds up, and this is one of the reasons performance-based navigation prefers the fix-to-fix leg over the course-held one.
Two inconsistencies can finally be spotted by reading. An arc placed immediately after a mandatory flyover fix contradicts the coding rules, since the tangency required at the arc’s entry is incompatible with a turn that only begins at the fix: if the chart appears to show this, it must be re-read, and the most likely explanation is that the fix is not the one you think it is. And cross-checking the loaded flight plan against the chart — sequence of fixes, courses, distances, constraints, actual presence of the arc — remains the only moment at which the crew can notice that the coding does not say the same thing as the approved document.
What is left to check in your own operation
Four points deserve to be looked up in your own documentation. Is the constant-radius arc function declared across the whole fleet or only part of it, and to what avionics standard. What limitation does the flight manual attach to following an arc regarding autopilot engagement and flight director, and what does it say about manual flying. What reduced-bank logics exist on the type, and can any of them activate without crew action where the arc needs it not to. Finally, is the operator’s rule on accepting a shortcut during an arc written down, or only assumed.
Grey areas remain, and it is worth knowing which one is yours: mixed fleets where the same procedure is not flown the same way from one aircraft to another, arcs appearing in a missed approach, high-density-altitude airfields, the exact behaviour of the FMS after leaving an arc. For anyone preparing a type rating or joining an operator serving demanding airfields, the subject shifts the technical question: not « do you know the leg path terminators », but « what happens to your arc if you accept a direct-to », and « at what point do you slow down ». This kind of subject is worth working as a module in its own right, in the spirit of the FL330 training programmes.
Applicable regulations, current charts and the operator’s approved documentation take precedence over this article, which describes a mechanism and replaces neither the approved flight manual, nor the operations manual, nor the navigation database documentation.
Questions fréquentes
Does an aircraft approved for a demanding navigation specification necessarily know how to fly a constant-radius leg?
No. The ability to follow a constant-radius arc is a separate function, declared in the approved flight manual or the type’s operational suitability data, and tied to a given avionics standard. A navigation specification never implies it. The check must be made installation by installation, and on a mixed fleet the result can differ from one aircraft to another for the same type.
Why is speed limited on an arc when it is not limited on the direct leg that precedes it?
Because the radius is fixed by the procedure. Holding that radius requires a turn rate equal to the ratio of ground speed to radius, hence a bank angle whose tangent varies as the square of ground speed. Beyond the speed assumed by the designer, the required bank angle leaves the design assumption and the aircraft drifts toward the outside of the protected corridor.
What actually happens if a direct-to clearance toward the fix ending an arc is accepted?
The FMS replaces the arc with a direct leg and the aircraft flies the chord instead of the curve, which takes it inside the arc, hence outside the protected corridor, with the deviation maximal at the middle of the turn. No dedicated alert signals the substitution. On some FMS, an abandoned arc can no longer be recaptured: the return to lateral guidance rejoins a straight line toward the downstream fix.
Does the displayed lateral deviation carry the same value in an anticipated turn and on an imposed arc?
No, and this is the most useful distinction to remember. On an arc, the reference path is coded before the flight: the deviation can be measured and monitored over the whole curve. In an anticipated turn between two direct legs, the curve is built in flight by the FMS algorithm and has been published by nobody; what is then to be judged is the recapture of the next leg, not the track through the bend.
Why must some fixes be flown over rather than flown by?
Because the upstream geometry cannot support anticipation: an obstacle gradient, adjacent airspace, or a track that must not be left too early require that the turn not begin before the fix. These fixes are circled on the chart and their turn area is widened, with roll dispersion starting at the fix. What follows is often a direct-to leg, because no coding can predict the aircraft’s position at the end of the turn.
Can a constant-radius leg be hand-flown?
Rarely without a guidance reference: many approved flight manuals make following an arc conditional on an engaged autopilot, some accepting a flight director, and it is this document that decides for a given type. What makes the exercise demanding by hand is the continuous variation of the required bank angle, as the wind component changes throughout the curve. The design does not assume a visual flying of the curve: the only indication of drift is the lateral deviation, and it reads late.

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