A jet engine is a machine for keeping air moving in one direction. When it briefly stops doing that, the result is spectacular: a bang like an explosion, a sheet of flame out of the tailpipe, sometimes out of the intake as well, and an aircraft that has just lost a large fraction of its thrust.

Passengers reach for their phones. Controllers occasionally report an engine fire. Crews have mistaken the noise for a burst tyre or a bomb and rejected a take-off above V1.

What has actually happened is a compressor stall, and the physics behind it are the same physics that stall a wing.

A compressor blade is a wing

This is the sentence that makes everything else make sense. Each blade in an axial compressor is an aerofoil, and its effective angle of attack is the vector sum of two velocities: the axial speed of air entering the stage, and the rotational speed of the blade itself.

Change either one and the incidence angle changes. Reduce the axial airflow while the shaft keeps turning at the same rpm, or spool the shaft up faster than the airflow can follow, and the incidence rises until the blade does exactly what a wing does when you pull too hard.

The consequence, though, is nothing like a wing stall. A stalled wing stops producing lift. A stalled compressor stage stops producing the pressure rise that holds back everything downstream of it, and there is a great deal of very high-pressure air downstream.

Rotating stall and surge are not the same thing

Engineers separate two failure modes that pilots experience as one event.

In rotating stall, a localised cell of stalled flow forms and then travels circumferentially around the compressor annulus. NASA Glenn measured a single cell in its Low Speed Axial Compressor moving at 50.6 per cent of rotor shaft speed, with genuinely reversed flow inside the cell. Published values elsewhere range from about a fifth to four-fifths of rotor speed depending on the compressor and the type of stall, so treat any single figure with suspicion.

Surge is the violent one. Here the whole annulus lets go and the flow reverses axially: the compressed air behind the stall escapes forward, back out through the compressor and out of the intake. That is the bang, and that is why flame can appear at the front of an engine as well as the back.

Which of the two you get is not random. It is governed by a dimensionless ratio known as Greitzer’s B-parameter, essentially a Helmholtz-resonator relationship between the compressor and the volume it is pumping into. Above a critical value the machine surges; below it, it settles into rotating stall. The boundary is reported at approximately 0.8.

Mentour Pilot walks through the phenomenon from the flight deck perspective above.

How little it takes

The numbers here are the genuinely surprising part. NASA tested the TF30, the engine in the early F-14, and found that inlet temperature distortion amounting to only 14 to 20 per cent of the face-average temperature was enough to produce complete compressor stall. Halving the Reynolds number index halved the distortion required.

Throttle handling does it too. NASA testing on the same engine showed that rapid slams between Military and Maximum power produced stalls beginning in the fan hub which then, in the report’s words, quickly propagated and produced complete stall in the high-pressure compressor.

Designers buy margin against all of this. Rolls-Royce states that low-pressure compressors are typically given around 15 per cent surge margin and high-pressure compressors around 20 per cent, with up to half of that reserved purely to absorb transient excursions during acceleration and deceleration. The working line has to stay below the surge line on the compressor map with enough room for a bird, a gust, a slam and a worn engine simultaneously.

The defences are mechanical. Variable inlet guide vanes sit ahead of the rotor blades and variable stator vanes behind them, rotating with engine speed to keep incidence within limits. Bleed valves dump air from intermediate stages during acceleration. On a modern three-spool engine the control unit schedules all of this continuously.

What it looks like when it happens

The flame is short. Rolls-Royce fuel-ingestion testing conducted for the Concorde investigation confirmed a forward flame lasting 80 to 100 milliseconds on the Olympus 593. Long enough to photograph, and long enough to convince an observer that the aircraft is on fire.

FAA training material sets out the instrument signatures with useful precision. A single self-recovering surge drops EPR to 1.05 in two tenths of a second, with N1 falling 16 per cent in the first two tenths and another 15 per cent over the next three. In a multiple surge the engine returns to 98 per cent of its pre-surge power between bangs. A non-recoverable surge shows EPR falling at 0.34 per second and EGT climbing at 15 degrees Celsius per second, continuing for a full eight seconds after the thrust lever has been pulled to idle.

The gas-turbine technician behind the AgentJayZ channel answers the engineering questions in more depth above.

Four events worth knowing

Airborne Express, Kansas City, 4 March 1999. A DC-9 freighter ingested snow geese into both engines. The number one engine surged roughly once per second. The captain found that reducing power slowed the cycle enough to keep flying.

SAS 751, Gottröra, 27 December 1991. Clear ice shed from the wings into both engines. The first right-engine surge came 25 seconds after rotation, at 1,124 feet. An automatic thrust restoration system, which the crew did not know was fitted, kept re-advancing the throttles into the stalling engines. The right engine failed 51 seconds after its first surge.

Concorde F-BTSC, 25 July 2000. Engines one and two surged on hot gas ingestion, with a later surge on engine one recorded at 13 degrees angle of attack. Engines three and four surged from airflow distortion attributed to the roll and high angle of attack.

US Airways 1549, 15 January 2009. Two birds of roughly eight pounds each entered each engine. The NTSB found that destruction of all the high-pressure compressor variable guide vanes caused the loss of directional control of the airflow into the compressor, causing it to stall continuously with no recovery possible. A detail rarely mentioned: the same aircraft had a right-engine stall at 17,000 feet two days earlier.

What the crew is supposed to do

The published answer is unglamorous and consistent across sources. The FAA is explicit that the desired pilot action is to retard the thrust lever until the engine recovers, and that the flight crew should then re-advance it slowly. The word slowly is capitalised in the original.

The airframe side matters too: reduce angle of attack and increase airspeed, because a stalled compressor is often a symptom of distorted air arriving at the intake rather than a fault inside the engine. For ice-crystal events specifically, Boeing guidance is to bring the throttle to idle, and notes that cycling the fuel switch may be required to clear some stalls.

Ice crystals are the modern version of the problem. Boeing documented at least 100 engine power-loss events attributed to ice-crystal icing since 1990 in its 2007 accounting, most above 22,000 feet and around 60 per cent in Asia, caused by crystals as small as 40 microns. Later literature quotes higher totals.

Why it still happens

Every one of these defences is a compromise against something else. Wider surge margin costs efficiency. More bleed air costs thrust. More variable geometry costs weight, complexity and reliability. Engine designers spend the margin they can afford and no more.

So the compressor stall survives as a failure mode not because it is poorly understood, but because the fix is expensive and the trigger is usually something the designer cannot control: a goose, a hailstone, a sheet of ice, or an aircraft pointed somewhere its intake was never designed to cope with.

It looks like a catastrophe. Most of the time it is an engine telling you, loudly, that the air arriving at its face is no longer acceptable.

Sources: FAA, Airplane Turbofan Engine Operation and Malfunctions; NASA/TM-2006-214270; NASA TM X-2921 and TN D-6839; NTSB CHI99FA102 and AAR-10/03; Swedish Accident Investigation Authority report on SAS 751; BEA report on Concorde F-BTSC; Rolls-Royce patent US11326526B2; Boeing AERO Q4 2007.