Realism
Overview
Twelve areas the simulation models, where the numbers come from, and where they do not come from anywhere yet.
Twelve areas the simulation models, and where the numbers in each of them come from.
| Area | What is modelled | Where it comes from |
|---|---|---|
| Aircraft types | 213 types, each with its own record: speed envelope, V-speeds, climb and descent rates by altitude, drag polar, thrust, masses. A descent to a cleared altitude is flown at the shallower of the tabled rate and the rate the type can hold its speed at with idle thrust, from its drag polar. With the speedbrake out, the 737, 757, 767, 777 and 787 families and the A320 come down steeper than that | Manufacturer flight manuals, type certificates and published performance data; the idle descent rate from that record's drag polar and the idle thrust model. For the Airbus A320 and the Boeing 737-800, idle thrust comes from the engine figures manufacturers supply to EASA's aircraft noise and performance database, and is checked against Airbus's and Boeing's own idle descent rates. The 737-900 uses the 737-800's figures |
| Aerodynamics and thrust | A force balance in every step of the calculation. Thrust falls away with altitude and speed, configuration drag is added on top, and it falls again as the air warms: jets past the temperature their engine is rated to, turboprops and unsupercharged piston aircraft steadily, both of which a controller sees on a hot day. How fast an aircraft climbs comes out of that balance rather than off a table: it uses the thrust left over once drag and weight are paid for, with the published rates for its type as a ceiling over the result, so a climb responds to weight, temperature and how hard the aircraft is being asked to accelerate | Jets fly a published gas-turbine model, scored against the manufacturers' own thrust figures. Propeller aircraft work from the shaft power on their certificates, because a propeller has no thrust rating at all. Turboprops lose that power with altitude by the relation two aircraft performance textbooks publish, checked against a worked figure one of them prints; piston engines by a relation fitted to engine test data in the 1930s and 1940s. How the leftover thrust is split between gaining height and gaining speed follows EUROCONTROL's published aircraft performance model |
| Atmosphere | Density, speed of sound, and the conversions between indicated, true and Mach, with a temperature deviation | The International Standard Atmosphere |
| Mass | One mass per type. Where it is researched, it is built from an operating mass, a payload and the fuel for a typical sector, so the aircraft flies its approach at the weight it lands with | Manufacturer weight data and certification limits |
| Configuration | Flaps by two limits: the manoeuvring speed of the configuration the aircraft is in, and the placard it must stay under. Landing gear on height above the airport, under the speed its type may operate the gear at | Certificated placards and gear speeds. The manoeuvring speed is computed from the type's own drag polar |
| Approach guidance | An energy plan worked backwards from being stabilised at 1000 ft above the airport, against the configuration the aircraft will have there rather than the one it has now: only flap the type may legally select at that speed, wheels placed at whichever comes later, the point an approach path reaches gear height or the point the aircraft's own descent does, and no speed above the type's own certificated maximum. Outside the final approach the deceleration is planned on the descent the aircraft flies while it slows, at idle thrust and a comfortable rate of slowing, rather than on the direct path to the runway; inside it, from the point where the procedure's platform altitude meets the glidepath, on the glidepath itself | Flight Safety Foundation stabilised-approach criteria, and each type's certificated speed and flap limits. The approach path is the runway's own published glideslope where the navigation data carries one, and the ICAO standard three degrees where it does not. The descent while slowing comes from each type's own drag and idle thrust, and is never steeper than the type's idle descent rate table. The method resembles the one Airbus and Thales flight management systems use, a fixed rate of slowing with the descent following from it, but the angles are not checked against theirs. Real systems differ in this. Boeing plans a flatter descent of about 500 ft per minute and some older systems slow down level, which the simulator does not yet tell apart. The point where the glidepath takes over is each procedure's own platform altitude from the navigation data, with ICAO Doc 8168's ten nautical miles as the fallback where a procedure carries none |
| Navigation | A full AIRAC cycle: departures, arrivals, approach transitions, holdings, navaids. Magnetic variation from the world magnetic model, approach distances computed on the earth ellipsoid rather than a flat map | Navigraph, NOAA, WGS-84 |
| Operating rules | 250 knots below FL100 with a proper high-speed clearance, holding speeds capped by altitude band, and the intercept angle within which an approach will capture | European and US air traffic regulation, ICAO Doc 8168 for holding. 30° is the ICAO standard for vectoring onto final; 35° is our own cutoff, and no standard states a maximum |
| Traffic | Real flights seen landing on real days: callsign, type, origin and the route each one filed, drawn only from flights seen landing in the direction you start. Eight airports have such traffic today | Records of the flights and of the flight plans filed for them, from days between 1 and 3 October 2026. The route is the one the operator filed, not one the server worked out. Only the arrival route at its end is changed, to the one for the runway direction you start, so the strip and the runway in use always agree. Where it enters the sector comes from the handover agreements IVAO Germany and Austria publish for the area controller to approach, at 23 of the 25 airports the server serves. Each flight enters before the handover point its own flight plan crosses, as early as IVAO Germany has the frequency change at the latest (two minutes or ten miles, whichever comes later), or at the distance an airport's own page gives. The Austrian fields use the same rule. The distance is counted from the handover point rather than from the sector boundary, which the server does not know. At Memmingen, where no agreement covers the main flows, and at any other field with arrival routes, it comes from the level and speed the arrival routes publish at their first fix, raised to FL130 where they publish no level or a lower one, the level closest on average to the agreements that are known |
| Crew | An instruction is not executed the instant it arrives. A dial turn takes about a second, something that means programming the flight management system about two | A modelling decision, not a document. Both figures are ours |
| Emergencies | A radio failure a trainer starts: complete, transmitter only, or unnoticed by the crew. The aircraft goes silent and keeps its last clearance. Unless the crew has not noticed, it squawks 7600 | SERA.14083, ICAO Doc 4444 8.8.3 and CAP 413 for what the crew does. The half minute to two minutes until 7600 is our estimate, since no source gives a time |
| Weather | Wind and temperature reach ground speed, climb and descent performance, and the drift an aircraft flies to track a fix | NOAA GFS, optionally in place of the standard profile |
Where it is not finished
The Where it comes from column is the point of this page, and so is this list.
- Nearly three in five types fly a mass that is a fraction of their maximum rather than a researched one: 124 of 213. Each record says so, and the type list marks which.
- The plateau a flat-rated engine holds is not modelled for a propeller. A real one delivers its full rated power up to an altitude and a temperature, and only falls away past that corner. Ours falls away from the ground up, so those types are given slightly less power than they really have rather than more. The slope past the corner is published in manufacturer operating manuals rather than anywhere we can read.
- Supercharged and turbocharged piston aircraft hold their full power below the altitude their maker gives, whatever the temperature. A real one loses a little.
- 16 types, helicopters among them, still fly a thrust law with no published source behind it. They are marked in the type list.
- Idle thrust uses manufacturer figures on three types only. On the engines we have such figures for, idle at descent speeds gives almost no thrust, and slightly less than none as the speed rises. The A320, the 737-800 and the 737-900 use them. Other jets follow one rule for the whole fleet that gives a little thrust at every speed, and propeller aircraft get none at all, where a real propeller at idle gives a small braking force. Those types descend at whatever their drag figures and that rule produce, which is not always the published rate.
- How much a speedbrake adds comes from Boeing's figures for a few families only. Boeing prints how fast the older 737s, the 757, the 767, the 777 and the 787 descend at idle with and without it, and those types use that figure. No table covers the 737 MAX, so it is assumed to behave like the 737 it derives from. We found no comparable figure from any other manufacturer, or for the 747, so every other type with a speedbrake carries an estimate nothing published confirms. The A320's emergency descent is the one Airbus figure we have. It names no speed, so it cannot settle the A320's value. Flown at the A320's maximum speeds, it suggests our A320's speedbrake is somewhat too strong. Only on those Boeing types and the A320 does the speedbrake take a descent past the type's usual rate. On every other type it never does, because its estimate would have some business jets coming down far faster than any real one.
- An arrival is taken off the scope on short final and never lands. It goes at about 200 ft above the runway, still short of the threshold. There is no flare, no rollout and nothing on the ground, so an approach is modelled up to the point where a tower controller would take it and no further.
- An aircraft that gives up on an approach does not go around by itself. It tells you why on frequency and then waits for you. One that is too high to be stable in time keeps following the localiser at the altitude it has; one that has lost the localiser flies straight ahead on its present heading. Neither climbs away, neither flies the published missed approach, and neither comes back on its own. You can order a go-around and it flies the missed approach then, but only once you have cleared it for the approach.
- An arrival configures earlier and more deeply than a crew would, because while it is slowing down the placard governs rather than what the aeroplane needs. What that looks like.
- A speed you assign on final can be overridden without anyone saying so. Below 1500 ft above the airfield, an aircraft that is following the localiser and has been given a speed more than 20 kt above what it can still fly slows to what it can fly instead. Nothing about that reaches the frequency, so a speed being ignored and a speed being flown look the same from the scope.
- The plan is made in still air and flown in wind. An aircraft works out where to start slowing down as though the air were still, then flies that plan in the wind it actually has. On downwind that mostly cancels itself, because a tailwind there becomes a headwind on final. What costs distance is a tailwind on final, which the choice of landing direction normally avoids; where it happens, published figures for one airliner on a shallow approach put the cost at about a fifth more room at 30 kt of tailwind and a quarter at 40.
- Some figures a jet flies on are ours, not a manufacturer's, among them the tolerance at which an aircraft counts as having left the glideslope, and the height at which the gear goes down. The gear one is worth naming precisely, because what a manufacturer publishes is not a height at all: Boeing's own training manual puts the wheels down when the glideslope comes alive on an ILS, and about 2 nm before the final approach fix on everything else. Our aircraft have no glideslope signal at the point the plan is made, and have to work at airports whose procedures are not loaded, so they use 2000 ft above the airport instead. The one published height nearby is a limit rather than a cue: the gear is down by 1000 ft on an instrument approach, because the aircraft has to be in landing configuration by then.
- Two aircraft of the same type fly almost identically. There is no turbulence, and no difference in skill or temperament between crews. What does vary is the moment an instruction takes effect: about a second for a single change, and about two for a longer clearance or anything that means programming the box, with a little scatter either side. A go-around is the exception and starts at once, because a crew does not think about that one. Past that, two of the same aeroplane given the same clearance do the same thing. For sequencing practice, which lives on crews being different, that is the difference from reality you are likeliest to feel.
Why that list exists
Any simulator can produce an aeroplane that descends and slows down. The question is whether the numbers it uses came from anywhere.
The rule here is that a value which changes what the simulator does has to name its source in the code, next to the value: a document and a section, not "matches what we observed". Where no source can be found, the field is left empty and says so rather than being filled with something plausible. A page claiming everything was sourced would be easier to write and would fall over the first time somebody checked.
It has consequences you can watch. The speed at which an aircraft may put its wheels down and the speed at which it may fly with them down are two different certificated limits that print in the same table, and getting them the wrong way round is the easiest mistake in this domain. It is why a business jet you have held fast on final will not extend its gear.
How it is checked
More than 300 automated scenarios run against the real flight model before anything ships, plus seven further suites covering the protocol, the navigation data and the session machinery, and an endurance run that flies every type in the fleet continuously and fails on any speed, climb rate or position out of bounds.
That catches regressions, which is the cheap half. The harder half is checking the model against something outside itself, because a test written from the same understanding as the code cannot see a shared mistake. So the clean speed of the Airbus narrowbodies is pinned to the manufacturer's published rule of thumb, the idle descent of the A320 and the 737-800 to the manufacturers' own descent figures, engine thrust at altitude to the manufacturers' ratings, magnetic variation to the eight values NOAA publishes for the purpose, and air density to the standard atmosphere tables. Where those disagree with us, we are wrong.