The Autopilot - ATPL 022 — Instrumentation

The autopilot is the previous lesson's closed loop given an aircraft to fly. This is the longest lesson of the module, because the syllabus wants the machine known from every side: its channels and types, its parts, its modes and their traps, the ways a pilot flies with it, through it and against it, and the manners of letting it go.

Three channels, three types

An autopilot works the aircraft through up to three basic control channels, one per rotation the previous lesson placed at the CG:

pitch channel — driving the elevator.
roll channel — driving the ailerons.
yaw channel — driving the rudder.

How many channels are fitted defines the type:

single-axis autopilot — commands roll only, the simple wing-leveller of light aircraft.
two-axis autopilot — commands pitch and roll, enough to hold altitude and fly headings, and is the common working fit.
three-axis autopilot — adds yaw, completing the set.

The parts of the system

The loop's blocks from the previous lesson now get their aircraft names:

Flight control unit (FCU) or mode control panel (MCP) — or equivalent: the panel on the glareshield, the shelf above the instrument panel, where the crew dials the targets (speed, heading, altitude, vertical speed) and pushes the mode buttons. It is the input-signal side of the loop.
Flight mode annunciator (FMA) — the strip that reports which modes are actually armed and engaged. Dialling a target is a request; the FMA is the receipt, and the flight-director lesson later in this module reads it properly.
Autopilot computer — the error detector and signal processor in one box: it compares the measured state with the targets and computes the surface commands under its control laws.
Actuator — the servo (the powered actuator that moves a control surface) turning the computer's commands into surface deflection.

Around that core flows a defined set of inputs and outputs. In:

attitude information — the inner loop's raw material, from the inertial systems of the previous module.
flight path and trajectory information — where the aircraft is actually going, for the guidance layer.
control surface position information — feedback that a commanded deflection actually happened.
airspeed information and aircraft configuration information — gear and flap state, both of which scale the control response, as the next section explains.
the FCU/MCP selections — the targets.

Out: the surface commands, and the FMAs announcing what the system is doing about it all.

Control laws, and how hard to push

The control laws for the pitch and roll modes serve the purpose the previous lesson gave them: they decide how strongly and how quickly the autopilot answers a given error, so that the response is firm enough to correct and gentle enough not to overshoot. The right answer is not fixed. The autopilot uses speed, aircraft configuration and flight phase as a measure for the magnitude of its control inputs: at high speed a small deflection produces a large response, so the laws command less; with flaps and gear out, or close to the ground on approach, the aircraft answers differently, so the laws adjust again. Scaled well, this preserves both precision and stability across the whole envelope. Scaled badly, the previous lesson's failure mode waits: too much authority and the loop overcompensates into oscillation, too little and it follows its targets sluggishly.

Lateral modes

HDG / TRK — heading or track: fly the heading dialled in the window, or the track (the path over the ground, drift already accounted for). The plainest guidance there is: point where told.
VOR / LOC — capture and track a radio beam: a VOR radial, or the localizer (the lateral beam of the instrument landing system (ILS), marking the extended runway centreline). The aircraft intercepts the beam and then holds it, correcting for wind.
LNAV / managed NAV — lateral navigation: follow the route in the flight management system (FMS), leg by leg. Boeing says LNAV, Airbus says managed NAV; both mean the FMS's plan is steering, not a single dialled number.

Vertical modes

V/S — vertical speed: hold the selected rate of climb or descent, whatever speed results. The mode flies the rate and lets the airspeed fend for itself, which is a danger this lesson returns to.
FPA — flight path angle: the same idea expressed as a slope rather than a rate, holding, say, a three-degree descent path.
LVL CHG / OP CLB / OP DES — level change (Boeing), open climb and open descent (Airbus): fly the speed with pitch while thrust sits at a fixed setting, climbing or descending to the new level. Speed is protected by construction, because speed is what pitch is flying.
SRS — speed reference system: the take-off and go-around pitch mode that flies a safe reference speed with pitch while thrust is at its rating, holding the climb-out honest.
ALT hold — hold the captured target altitude.
VNAV / managed CLB or DES — vertical navigation: the FMS's computed profile flies the climbs, cruise and descents, altitudes and speeds together, rather than one dialled target at a time.
G/S — glideslope: capture and track the instrument landing system's vertical beam down the approach, the vertical partner of LOC.

Mixed modes

Three phases of flight demand lateral and vertical work as one package, and the syllabus calls their modes mixed modes:

Take-off — pairs SRS-style pitch guidance with wings-level or runway-track lateral guidance.
Go-around — does the same job upward from an abandoned approach: pitch to the reference speed, thrust up, lateral guidance out of the missed approach.
Approach (APP) — arms both halves of the instrument landing system at once: LOC laterally and G/S vertically, flown together down to the runway.

Armed, engaged, and the capture sequence

Modes have two live states, and the difference does real work:

engaged — the mode is flying the aircraft now.
armed — the mode is waiting for its conditions: not every mode has an armed state, a mode arms only if certain criteria are met, and an armed mode engages, replacing the previously engaged mode, when its criteria are met.

Arming LOC does not turn the aircraft; it tells the system to watch for the beam, fly the old mode meanwhile, and swap the moment the capture criteria are satisfied.

The engagement itself unfolds in three phases:

initial phase — the autopilot changes attitude to set up a new trajectory toward the new parameter, the rate of closure sized on the gap between the old parameter and the new one.
capture phase — the aircraft follows a predefined rate of change of trajectory, easing off so the parameter arrives without overshooting or undershooting.
tracking or hold phase — the aircraft maintains the captured parameter until the next change is asked for.

When the mode changes itself

Sometimes the FMA changes without the crew touching anything: automatic mode reversion. The syllabus names its typical situations:

The current mode may run out of suitable data — as when LNAV or managed NAV meets a flight plan discontinuity, a gap in the route with no leg to fly.
A parameter may change during its own capture — as when the crew moves the altitude target while the aircraft is mid-capture (the phase annunciated as ALT ACQ or ALT*), leaving the capture nothing to finish.
A mode may be mismanaged into the envelope protection — as when an excessive vertical speed is selected and speed decays until the protection (a later lesson of this module covers it) takes over.

In each case the system falls back to a mode it can honour, and the FMA is where the change shows.

Two mismanagement patterns are named as dangers outright:

V/S with no speed protection of its own — select a climb rate the thrust cannot sustain, or an excessive descent rate, and the mode will faithfully hold the rate while the airspeed runs away, an uncontrolled loss or gain of airspeed; FPA shares the trap, flying its angle with the same indifference to the speed (contrast LVL CHG and OP CLB/DES, which fly the speed by construction).
arming VOR/LOC or APP outside the protected area of the localizer or ILS — outside the coverage the beams are not guaranteed honest, and the system may capture a false or side-lobe signal and steer confidently along it.

Engaging, flying through, letting go

Engagement has its own courtesy. The synchronisation function matches the autopilot's commands to the aircraft's current attitude at the moment of engagement, so the controls do not snatch toward some stale target. And the autopilot should be engaged when the aircraft is in trim: engage it out of trim and it inherits a standing control force it must hold from the first second, working against the very servos that should be making fine corrections.

Engaged autopilots also differ in what they do with the flight-deck controls, and the syllabus names two configurations:

the flight-deck controls move with the control surfaces — while the autopilot is engaged, the column and wheel are back-driven, so the pilot can see and feel what the autopilot is commanding, a continuous cue to its activity, though anything resting against the controls then resists the autopilot itself.
the flight-deck controls remain static — while the autopilot is engaged, the surfaces move, the controls do not, and the flight deck is quieter for it, but the pilot gets no control-movement cue at all; awareness of what the automatics are doing must come entirely from the instruments and the FMA.

Comparing the two, the moving configuration gives visibility with bustle, the static one a calm that demands vigilance.

Control wheel steering (CWS) is manual manoeuvring of the aircraft through the autopilot computer and the autopilot servos, using the control column or wheel. It is an autopilot mode: the pilot's force on the column becomes an input the computer translates into servo commands. The pilot, not the autopilot, is controlling the flight path; there are flight phases where CWS cannot be used (the take-off and the landing among them); the flight path and performance protections of the full modes may not be available; and the feel and response can differ noticeably from ordinary manual flight, because every input is passing through the computer's interpretation.

Touch control steering (TCS) reaches the same goal down a different path, and the differences are the point: with TCS the autopilot remains engaged but its servos are disconnected from the control surfaces while the TCS button is held, the pilot flying the aircraft directly, conventionally, for exactly as long as the button is depressed. On release the servos reconnect and the autopilot returns to its previously engaged modes. CWS flies the pilot's wishes through the computer; TCS takes the servos out of the circuit and hands them back afterwards.

One rule governs how many autopilots work at once: only one autopilot may be engaged at any time, except when APP is armed, where additional autopilots join to facilitate a fail-operational autoland, the redundant automatic landing the autoland lesson covers in full. Until then, one computer flies; the rest wait.

Failures, disengagement and trim

The autopilot lives on other systems' data, and that dependence cuts both ways. Failure of other systems may take the autopilot away (lose the attitude source and the inner loop has nothing to fly), and, more insidiously, incorrect data from other systems may lead the autopilot into an undesirable aircraft state, potentially without any failure indication at all: a subtly wrong airspeed or attitude feed gives the loop wrong truths to correct toward, and every annunciation may stay green while it does. The defence is the one the syllabus states: prompt and appropriate pilot intervention, noticing that the aircraft's behaviour and the expected behaviour have parted company, and taking over early, exactly as the previous lesson taught for the oscillating loop.

Letting go is designed as carefully as engaging. The proper procedure is to disengage with the design method, typically a disconnect button on the control wheel or sidestick, with a hand on the controls to catch whatever forces the autopilot was holding. Both aural and visual warnings announce any disengagement, and their duration carries meaning:

temporary warning — for an intended disengagement made by the design method.
continuous warning — for an unintended disengagement, or one made any other way, because a disconnect nobody asked for is exactly the event no crew should be able to miss.

On aircraft with manual trim the trim wheel and the autopilot must respect each other: the autopilot may not engage unless the controls are in trim; flown properly, the aircraft will normally be in trim when the autopilot is disconnected, because the autopilot has been trimming as it flew; and using the manual trim while the autopilot is engaged will normally disconnect it, with a risk of an out-of-trim situation handed to a pilot who was not expecting the aircraft at all.

That is the machine itself. What the crew watches while it flies, the flight director's bars and the FMA's annunciations, is the next lesson.

What the exam asks

The mode families, the armed-engaged logic and the CWS/TCS contrast carry the marks here. The overview, following the learning objectives:

Define the three control channels and the 1-, 2- and 3-axis autopilot types

Describe the components: FCU/MCP, FMA, autopilot computer, actuator

Describe the autopilot's inputs and outputs

Describe the purpose of pitch/roll control laws and the speed/configuration/phase scaling of control inputs

Explain the lateral modes: HDG/TRK, VOR/LOC, LNAV/managed NAV

Explain the vertical modes: V/S, FPA, LVL CHG/OP CLB/OP DES, SRS, ALT hold, VNAV, G/S

Explain the mixed modes: take-off, go-around, APP

Explain armed vs engaged modes and the three-phase engagement sequence

Explain automatic mode reversion and the dangers of V/S or FPA and premature VOR/LOC/APP arming

Describe the two configurations: controls moving with the surfaces vs remaining static, and their implications

Describe synchronisation, in-trim engagement, CWS and TCS, and the one-autopilot rule with its APP exception

Describe how other systems' failures or bad data affect the autopilot, and the pilot's intervention duty

Explain the disengagement warnings and the manual-trim interactions