Final approach at three miles, full configuration: speed drifts a few knots below the target, the glidepath sags, and the reflex fires on its own — a small pull on the stick. The aircraft climbs for a second or two, then sinks faster than before, speed still falling. Nothing is broken, no system has lied. The aircraft is working on the left branch of its drag curve, the one where the control that makes you climb is no longer the stick.
The drag curve, and the point where it turns over
Drag splits into two families that move in opposite directions: parasite drag grows as the square of speed, induced drag falls as the inverse of that square at constant weight. Their sum traces a bucket-shaped curve, whose minimum falls where the two families are equal — the minimum-drag speed, which also gives the maximum lift-to-drag ratio. To the right of this point, slowing down reduces the thrust required for level flight: the first regime, where intuition works. To the left, slowing down increases it: the second regime, or region of reversed command.
Imagine a bucket shape: on one side, slowing down makes it easier to stay level; on the other side, it makes it harder. The second regime is the side where slowing down makes the plane sink faster.
One point governs everything that follows: this reversal has nothing to do with the stall. The minimum-drag speed sits well above it, the flow is fully attached, no alarm triggers, and no marker on the speed tape shows where the boundary is crossed. You enter the second regime in an aircraft whose every indication is normal.
A Boeing 737 on final approach at 130 knots (Vref + 5) in landing configuration. If the speed drops to 125 knots, the pilot pulls back to maintain the glidepath. The aircraft initially climbs slightly, but then sinks faster because the increased angle of attack raises drag beyond available thrust.
Minimum drag, minimum power
The limiting quantity depends on the type of propulsion, so the boundary doesn’t fall in the same place. A jet delivers thrust that is fairly insensitive to speed: its equilibrium is written in terms of thrust, and the boundary falls at minimum drag. A propeller delivers power, the product of drag and speed, whose minimum sits lower — on the order of three-quarters of the minimum-drag speed for a parabolic drag polar. For an equal polar, a turboprop therefore has a wider stable range, and its entry into reversed command happens closer to the stall.
Two equilibria for one thrust setting
In stabilized flight at a shallow angle, gamma = (T - D) / W and Vz = V (T - D) / W: the excess of thrust over drag, divided by weight, alone decides whether you climb or descend. Set an intermediate thrust, and the equality between thrust and drag admits two solutions, one on each side of the bottom of the bucket. A single throttle setting therefore permits two level-flight speeds, which behave in entirely different ways.
On the fast solution, a loss of speed drops drag below the thrust set: the excess that appears reaccelerates the aircraft. The equilibrium defends itself — this is speed stability, and it is what makes cruise flying so forgiving. On the slow solution, the same deviation makes drag rise above the thrust set: the deficit slows the aircraft further, which deepens the deficit. The gap does not close, it opens, and faster and faster. Nothing brings it back: what’s needed is thrust, or altitude to spend. It is this asymmetry, not proximity to the stall, that defines the second regime.
| What you observe | First regime | Second regime |
|---|---|---|
| Loss of speed: effect on drag | It decreases | It increases |
| Speed’s response to a deviation | Returns to equilibrium: stable | Deviation widens: divergent |
| Pitch up at fixed thrust: settled effect | Climb, speed stabilizes lower | Transient climb, then increased sink |
| Control of flight path | Pitch attitude, thrust follows | Thrust, the only lever that raises the path |
| Control of speed | Thrust | Pitch attitude, through angle of attack |
| Time pressure | Low, margin exists | Set by the spool-up time |
| Alarm at the boundary | Not applicable | None: the boundary is not instrumented |
Why pulling back makes you descend
The contradiction comes from two time scales that are easily confused.
On the scale of a second, pitching up increases the angle of attack and the lift coefficient: lift momentarily exceeds weight, and the flight path curves upward. This climb is real, but it is paid for in kinetic energy. The action therefore seems to confirm the very mental model it contradicts, which is what makes it so dangerous.
On the scale of ten seconds, the added angle of attack has increased induced drag. Thrust has not changed, T - D becomes more negative, and the sustainable path angle in steady flight drops: shallower than before the action. As speed keeps decreasing, drag keeps increasing and the path keeps sagging. The final state is a faster sink, at a lower speed, with less margin than at the start.
Reading it in energy terms puts the roles back in order. Pitch attitude redistributes energy between kinetic and potential form; it does not change the total, which only the difference between thrust and drag can alter, at the rate V (T - D) per unit mass. In the first regime, this redistribution goes unnoticed because speed stability corrects it on its own. In the second regime, it is all the stick has to offer, and the account is empty.
What moves the boundary
The minimum-drag speed varies as the square root of the load the wing carries: it is not a fixed constant of the aircraft. What remains is to separate what moves the boundary from what eats into the thrust reserve above the curve.

Weight acts on both. The boundary rises as the square root of weight; minimum drag, meanwhile, grows in direct proportion to weight, since it equals weight divided by the maximum lift-to-drag ratio. Since the reference speed follows the same square root, the ratio between the two is roughly preserved — but the thrust required increases, so the reserve narrows.
Load factor acts abruptly. At thirty degrees of bank it is about 1.15; since induced drag varies as its square, it grows by a third at the same speed, and the boundary rises by roughly seven percent. A turn on final consumes a margin you thought you had, at the moment your attention is elsewhere.
Configuration, for its part, blurs the picture. Gear and flaps extended sharply raise parasite drag: the bucket rises, flattens, and its minimum shifts toward lower speeds — but the speed actually flown drops too. What matters is not the direction of the shift, it is the shape: a flat bottom gives very weak speed restoring force. In landing configuration, speed stability is no longer something you can count on, and the drag level to be compensated for is the highest of the flight.
Altitude, finally, barely moves the boundary in indicated airspeed but shaves the thrust available above it. Near the aerodynamic ceiling, the thrust-available curve comes to touch the thrust-required curve: the excess reaches zero and the stable range shrinks to a single point. Below this tangency speed, maximum thrust no longer reaccelerates the aircraft at that altitude, and the only lever left is altitude itself — an emergency descent obeys the same equation as the final approach, settled in a different currency.
| Factor | Minimum-drag speed | Thrust reserve | Consequence in flight |
|---|---|---|---|
| Higher weight | Rises as the square root of weight | Shrinks: minimum drag proportional to weight | Ratio to reference speed preserved, less thrust in reserve |
| Bank angle, load factor | Rises as the square root of load factor: roughly seven percent at thirty degrees of bank | Shrinks sharply: induced drag as the square of load factor | A turn on final consumes a margin you thought you had |
| Gear and flaps extended | Drops, but the speed flown drops too | Shrinks: highest drag of the flight | Flattened bucket bottom, less speed restoring force |
| Icing, degraded surface condition | Shifts, position poorly known | Shrinks | Usable band narrows, thresholds computed for a different aircraft |
| High altitude | Barely shifts in indicated airspeed | Shrinks with thrust available | Near the ceiling, the two branches meet |
How you get there on approach
No one decides to enter the second regime: you get there by ordinary paths, and it is their ordinariness that makes it a problem.
Deceleration in configuration, first. Gear down, flaps extending, thrust levers at idle: drag rises while thrust is at its minimum, and the only lever able to take back control is precisely the slowest one on the aircraft. Calling for thrust when the needle reaches the target means calling for it already late by the spool-up time.
Wind shear with decreasing performance, next: a decreasing headwind component removes indicated speed without any control having moved. At low height, on an already thin reserve, the crossing is suffered, not committed.
Level flight held in dirty configuration — after a descent flown late, during a spacing hold — is exactly the unstable equilibrium described above, and the stick is the worst tool available there. Correcting for a path intercepted too high belongs to the same family: you get out of drag to come back down, it works beyond what was expected, and the energy state collapses under the path. The last path, more subtle, lies in the automation itself.
The trap: the mode that pays in speed. A vertical guidance mode with a geometric target — vertical speed or flight path angle — commands pitch attitude to a path target. If the thrust set is not enough to hold it, the automation still holds its target and takes the difference out of speed, with no limit of its own, until a protection or the crew intervenes. In landing configuration, on a thin reserve, this is the shortest path to the left branch, and the quietest: the flight path stays perfectly held right to the end. It is this impeccable flight path that delays the diagnosis — you’re watching the path deviation, which is zero, while the quantity that is degrading isn’t displayed in the same place.
Thrust becomes the flight-path control
The whole technique follows from gamma = (T - D) / W. With configuration fixed, the only free quantity able to raise the path is thrust: angle of attack can only move drag, and on the left branch it can only increase it. The assignment of controls therefore reverses: thrust commands the flight path, pitch attitude commands the speed, that is, the angle of attack.
The reversal comes at the cost of a time constraint, because the lever that has become indispensable is the one with the longest time constant. Thrust requested from idle takes several seconds to establish, and during those seconds the divergence keeps running. Hence the rule of anticipation: thrust is set not on the value of speed but on its rate of change, and in a single sufficient movement — a series of small increments repays the spool-up delay at every increment, without ever making up the lost lead.
Next comes the order of actions in a recovery: reduce drag before pulling anything. Rolling the wings level removes the extra load factor, and so the third of induced drag a thirty-degree turn had added. Retracting the gear, when the situation allows, removes parasite drag without touching maximum lift coefficient — which is not the case for retracting flaps, which raises the stall speed at the worst possible moment. The benefit of the gear is nevertheless not instantaneous: on many types, the doors opening add drag during the cycle, so the maneuver costs before it pays — it completes a thrust increase, it does not replace it, and it is not attempted if the gear may still be needed in the seconds that follow. Pitch attitude only comes into play afterward, not to climb, but to hold the angle of attack while thrust is establishing.
One decision remains. Trading altitude for speed is a valid lever, and the only one that is always available, but it requires altitude to spend. On short final, there is none: exiting the second regime is then no longer a flight-path correction, it is a go-around — the only maneuver that restores thrust, drag reduction, and height all at once.
The right reflex: fly the rate of change, not the value. On the left branch, the value of speed always arrives too late; what informs you is the trend vector, or the rate at which the needle is falling. Thrust is set before the target is reached, in a single ample and sufficient movement, never in successive small increments. Pitch attitude is only there to hold the angle of attack while thrust establishes, and the wings return level as soon as the flight path allows. If the recovery requires speed with no altitude to buy it, the response is no longer a correction: it is a go-around, announced and flown as such.
What the protections see, and what they don’t
Low-energy protections monitor specific quantities: speed against computed thresholds, angle of attack, sometimes a rate of energy change. Depending on the architecture, they produce an aural low-energy alert, colored bands on the speed scale, an automatic thrust command at a floor angle of attack, or a change in autopilot behavior. They do effectively catch situations the crew had not yet named.
They still have two structural limits. The first is that they work on the margin to the stall, not on the boundary of the second regime. Crossing the minimum-drag speed triggers nothing and can trigger nothing, since nothing abnormal happens there: the aircraft is flying, lift is present, angle of attack stays modest. The boundary is not instrumented, and what gets flagged later is the consequence of the divergence, not its start.
The second is that this logic has arming conditions — height band, configuration, automation status, source validity — outside of which it stays silent, and that it reasons on a model of the aircraft: an ice deposit that has changed the maximum lift coefficient, a degraded configuration, a thrust asymmetry, and the displayed thresholds remain those of the model, not those of the aircraft on that day. This is why monitoring the speed trend is not a nostalgic redundancy: it is the only available quantity that sees the boundary.
What to go check in your own documentation
Three points are worth establishing in the cold light of day. Does your operator’s documentation explicitly assign the controls on approach, or does it leave that assignment to local culture? Approach idle, which exists to shorten spool-up time, is armed under what conditions on your fleet: this is performance information, not a systems curiosity. Finally, do the stabilized-approach criteria bear on the rate of change of speed, or only on a window of values — in which case a speed still within the window but decreasing rapidly satisfies the criterion while already being on the wrong side.
Two gray areas remain. No marker shows the boundary on the speed scale, and there is no reason to expect one: the position of the minimum depends on weight, configuration, instantaneous load factor, and surface condition. Furthermore, the drag polar in non-standard configuration — degraded flaps, gear not retracted, flight controls in a reversion mode — is rarely published as a curve, even though that is precisely where the thrust reserve is hardest to estimate and where the boundary comes closest to the speed flown.
For anyone preparing a type rating or a module of performance and flight mechanics training, the question at stake concerns the direction of the mechanism and the assignment of the controls, never a value: which quantity commands the flight path, which commands the speed, and why the answer reverses on either side of minimum drag. A candidate able to redraw the bucket and write the thrust-excess relationship no longer has a rule to remember — they can re-derive it.
This article sets out a mechanism and the technique that follows from it; it is neither a procedure nor an authorization. In the event of a discrepancy, the applicable regulation, the approved flight manual, and the operator’s approved documentation take precedence over what is written here.
Questions fréquentes
Is the second regime the same thing as flying close to the stall?
No. The boundary of the second regime is the minimum-drag speed, which sits well above the stall speed, with the flow fully attached and a modest angle of attack. What changes when you cross it is not the lift available but the sign of the speed’s response to a deviation: above it, a loss of speed corrects itself; below it, it gets worse. You can therefore be in the second regime with a comfortable stall margin and no alarm at all.
How do you know you have dropped below the minimum-drag speed?
No marker shows it on the speed scale, and its position depends on weight, configuration, instantaneous load factor, and surface condition. The practical sign is the behavior of the speed: if it keeps decreasing at a constant thrust setting while the flight path is being held, drag exceeds thrust and the divergence has begun. It is the trend vector, or the rate at which the needle is falling, that reveals it — never the instantaneous value.
Why is it said that on approach thrust commands the flight path and pitch attitude commands the speed?
Because the sustainable path angle in steady flight equals the excess of thrust over drag, divided by weight. With configuration fixed, the only free quantity able to raise this path angle is thrust; pitch attitude, for its part, only moves the angle of attack, hence the drag, which it increases as soon as you are below the minimum-drag speed. Pitch attitude retains authority over how energy is split between speed and height, not over its total.
Does a turboprop enter reversed command at the same speed as a jet?
No, lower. The boundary sits at the minimum of the limiting quantity: thrust required for a jet, power required for a propeller. The minimum of power occurs at a lower speed than the minimum of drag, on the order of three-quarters for a parabolic drag polar. For an equal polar, a turboprop therefore has a wider stable speed range, but its entry into reversed command happens closer to the stall.
Why set thrust in a single movement rather than in small increments?
Because the spool-up delay is paid at every command. Thrust requested from idle takes several seconds to establish, and during that time speed keeps diverging; a series of successive small increments repeats that delay as many times as there are increments, never making up the lost lead. A single ample and sufficient movement, set on the trend rather than the value reached, is the only one that regains the initiative.
Does a turn on final really change the available margin?
Yes, and by more than one might assume. The load factor is about 1.15 at thirty degrees of bank; since induced drag varies as its square, it increases by a third at the same speed, and the minimum-drag speed rises by roughly seven percent. Rolling the wings level is therefore the first drag reduction available in a recovery, before any other action.

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