Combustion instability — low-frequency (chugging), high-frequency (screaming): Question bank

Every "answer" side below gives the reasoning, never a bare yes/no — that is the whole point of a concept trap.

Symbols used on this page — read first

Before the traps, every symbol below is pinned to plain words and units, so no reveal uses anything undefined.

Definition Symbol glossary

$p^{'}$ — the fluctuating part of chamber pressure, i.e. the wobble around the mean: $p_{c} (t) = \overset{p}{ˉ}_{c} + p^{'} (t)$ . Units: pascals (Pa). It is what "swings" in the swing analogy.
$q^{'}$ — the fluctuating heat-release rate per unit volume, the wobble around the mean burn rate. Units: W/m³. It is the "push".
$ϕ$ — the phase lag (in degrees or radians) by which the heat wobble $q^{'}$ trails the pressure wobble $p^{'}$ . $ϕ = 0$ means they peak together; $ϕ = 18 0^{\circ}$ means opposite.
$ω$ — the angular frequency of the oscillation, in rad/s. It relates to ordinary frequency $f$ (in Hz, cycles/s) by $ω = 2 π f$ : one full cycle is $2 π$ radians of phase.
$τ$ — the combustion/transit lag: the time (seconds, typically ms) between propellant entering and its heat actually being released. It is the slow clock of chugging.
$Δ p_{inj}$ — the injector pressure drop, $p_{f ee d} - p_{c}$ : how much harder the feed pushes than the chamber pushes back. Units: Pa. A large drop = a "stiff" injector.
$\overset{m}{˙}_{in}$ — the propellant mass flow into the chamber through the injector (kg/s). The dot means "per second". It grows with the injector drop: $\overset{m}{˙}_{in} \propto Δ p_{inj} = p_{f ee d} - p_{c}$ .
$\overset{m}{˙}_{o u t}$ — the mass flow out through the nozzle (kg/s). For a choked throat it is $\overset{m}{˙}_{o u t} = p_{c} A_{t} / c^{*}$ , so it rises with chamber pressure.
$A_{t}$ — the nozzle throat area (m²), the narrowest cross-section where flow chokes. Typical: a few cm² to tens of cm².
$c^{*}$ — the [[Characteristic Velocity c|characteristic velocity]]* (m/s), a propellant-and-chamber property measuring how efficiently combustion raises pressure; typical $\sim 1500$ – $1800$ m/s. It appears in $\overset{m}{˙}_{o u t} = p_{c} A_{t} / c^{*}$ .
$k$ — the feed-coupling gain: a dimensionless number saying how strongly a chamber-pressure wobble $p^{'}$ feeds back into a heat-release wobble through the injector. Large $k$ = tight coupling = easy chugging; drive $k \to 0$ and the loop cannot close. It shrinks as $Δ p_{inj}$ grows because a stiff injector barely changes $\overset{m}{˙}_{in}$ when $p_{c}$ wobbles.

Deriving $f_{c h ug} \approx 1/ (4 τ)$ — the quarter-cycle rule

Figure — Combustion instability — low-frequency (chugging), high-frequency (screaming)

Formula From phase to frequency

A pressure dip makes extra propellant flow, but that heat only arrives after the lag $τ$ . For that late heat to land in phase with the next pressure rise — the swing-pushing condition — the delay must equal about a quarter of a cycle. A full cycle is $2 π$ radians of phase; a quarter cycle is $π /2$ radians. The phase accumulated in time $τ$ is $ω τ$ (angular frequency $\times$ time). So the marginal condition is $ω τ \approx \frac{π}{2} .$ Substitute $ω = 2 π f$ : $2 π f τ = \frac{π}{2} .$ Why divide both sides by $2 π τ$ ? We want $f$ alone on the left; $f$ is currently multiplied by $2 π τ$ , so dividing by exactly that factor cancels it on the left and applies the same operation to the right (keeping the equation true): $f = \frac{π /2}{2 π τ} = \frac{1}{4 τ} .$ That is the chug frequency: $f_{c h ug} \approx \frac{1}{4 τ}$ . For $τ = 2.5$ ms this gives $f = 1/ (4 \times 0.0025) = 1/0.01 = 100$ Hz — squarely in the chugging band.

True or false — justify

A rocket engine can be unstable even if the total heat released per cycle is exactly the average design value.

True — instability cares about timing, not amount; the same average heat, if bunched when pressure is high, satisfies Rayleigh Criterion and grows the oscillation.

Chugging and screaming are governed by different physical laws.

False — both obey the identical Rayleigh condition

\oint p^{'} q^{'} d t > 0

; only the time-delay reservoir differs (feed lag in ms vs acoustic transit in μs).

Adding heat exactly in phase with pressure (

ϕ = 0

) is the most destabilising case.

True — since the driving overlap scales as

cos ϕ

, zero phase lag gives the maximum positive

\oint p^{'} q^{'} d t

, so the swing is pushed hardest at its forward peak.

If heat release lags pressure by

18 0^{\circ}

, the oscillation still grows, just slower.

False — at

ϕ = 18 0^{\circ}

cos ϕ = - 1

, so

\oint p^{'} q^{'} d t < 0

; heat arrives when pressure is low, actively damping the mode rather than feeding it.

A perfectly stiff injector (infinite

Δ p_{inj}

) removes the chugging feedback loop.

True — infinite drop makes

\overset{m}{˙}_{in} \propto p_{f ee d} - p_{c}

nearly independent of

p_{c}

, so the feed-coupling gain

k

collapses to zero and the delay loop can no longer close. See Injector Design & Pressure Drop.

The first tangential (1T) mode is dangerous mainly because it is the highest-frequency mode.

False — it is dangerous because its pressure sloshes side to side, scrubbing hot gas against the wall; it is actually one of the lower transverse modes, not the highest.

Baffles on the injector face are a good cure for chugging.

False — baffles disrupt transverse acoustic modes (screaming); chugging lives in the feed system and is untouched by them. The chug cure is injector

Δ p_{inj}

Raising chamber pressure

p_{c}

always improves stability.

False — what matters is the ratio

Δ p_{inj} / p_{c}

; raising

p_{c}

without raising the drop makes the injector "softer" and can worsen chugging.

Nozzle radiation and viscous losses are what set the final steady amplitude of an instability.

True — Rayleigh growth is unbounded linearly; nonlinear damping (nozzle radiation, viscosity) is what caps it into a finite limit cycle.

A feed system can only ever chug at the single frequency

1/ (4 τ)

False — any delay of

(2 n + 1) π /2

radians also lands heat in phase, so higher tongues at

f \approx (2 n + 1) / (4 τ)

(

3/ (4 τ)

5/ (4 τ)

, …) are possible harmonics.

Spot the error

"Chugging is at kHz because the feed lines are long, so sound takes a while to travel them."

Two errors: chugging is low frequency (10–400 Hz), and its clock is the combustion/transit lag

τ

, not acoustic travel — kHz belongs to screaming and chamber acoustics.

"To kill screaming, just increase the injector pressure drop like you would for chugging."

Wrong reservoir — screaming feeds through chamber gas acoustics, not the feed system; the cure is acoustic damping (Injector Baffles, Helmholtz Resonator cavities), which adds loss to that mode.

"The Rayleigh integral

\oint p^{'} q^{'} d t

uses the total heat

q

, so a bigger engine is always more unstable."

q^{'}

is the fluctuating heat-release rate, not the mean

q

; a large steady heat with no fluctuation contributes zero to the integral.

f_{c h ug} \sim 1/ (4 τ)

, so shortening the feed line shortens

τ

and always stabilises."

τ

is the combustion + transit + ignition lag, dominated by chemistry and residence, not just line length; and shifting

τ

only moves the resonant frequency, it doesn't guarantee the phase leaves the growth window.

"Sound speed in the chamber uses ambient air,

a \approx 340

m/s."

The gas is combustion products at

\sim 3000

K, so

a = γ R_{g a s} T_{c} \sim 1000

–

1200

m/s — roughly three times higher, which is exactly why screech lands in the kHz band. See Acoustic Modes of a Cylindrical Cavity.

"A choked nozzle means

\overset{m}{˙}_{o u t}

is fixed, so it can't participate in the chug loop."

For a choked nozzle

\overset{m}{˙}_{o u t} = p_{c} A_{t} / c^{*}

(throat area

A_{t}

fixed,

c^{*}

a propellant constant) still rises with $p_{c}$ — the mass flow tracks chamber pressure, providing the restoring/emptying term in the mass balance.

Why questions

Why does the product

p^{'} q^{'}

appear in Rayleigh's criterion rather than the sum?

Because a product measures overlap in phase: it is large-positive only when both signals are high together, capturing "heat added while pressure is already high" — the swing-pushing condition.

Why is

τ

the only slow clock that sets the chug frequency?

Compare timescales: acoustic transit across a

\sim 0.1

m chamber at

\sim 1100

m/s is

\sim 0.1

ms, and chamber fill/empty follows the fast gas dynamics too, whereas the combustion lag

τ \sim 2

–3 ms is

\sim 20 \times

slower — so

τ

is the bottleneck delay the pressure cycle must sync with, giving

f \sim 1/ (4 τ)

Why does a quarter-cycle (

ω τ \approx π /2

) mark marginal chug instability?

A quarter-cycle phase shift brings the delayed extra burning to arrive right as pressure is climbing again, putting heat back in phase with pressure and closing the positive feedback.

Why is screech a thousand times faster than chugging?

Its clock is the acoustic transit time across the chamber (

\sim R / a

, microseconds), whereas chugging's clock is the combustion lag (

\sim

milliseconds) — a factor of

\sim 1000

difference.

Why can a Helmholtz cavity be tuned to absorb one target frequency?

A Helmholtz Resonator has a single natural frequency set by its neck and volume; at that frequency the gas in its neck oscillates a quarter-cycle out of step with chamber

p^{'}

, so the flow it draws does negative work on the mode — heat/energy leaves exactly where

p^{'}

is largest, flipping the local

\oint p^{'} q^{'} d t

negative.

Why does putting the flame at a pressure antinode maximise screaming risk?

The antinode is where

p^{'}

swings largest, so heat released there has the biggest possible

p^{'}

to multiply against — maximising the Rayleigh overlap and drive.

Edge cases

If the combustion lag

τ \to 0

(instant burning), what happens to chugging?

The delayed heat term collapses onto the present pressure; with no phase lag to reach the growth window, the classic

1/ (4 τ)

chug band pushes to infinite frequency and the low-frequency loop effectively vanishes.

If the flame sits exactly at a pressure node of an acoustic mode, can that mode scream?

No — at a node

p^{'} \approx 0

, so

p^{'} q^{'} \approx 0

and the Rayleigh integral gets no drive from the flame; that particular mode cannot be pumped from that location.

At the marginal-stability boundary where

\oint p^{'} q^{'} d t = 0

exactly, what is the engine doing?

Neutral oscillation — energy fed in per cycle exactly equals energy lost, so a disturbance neither grows nor decays; it hovers at constant amplitude and any small change tips it either way.

What if

Δ p_{inj}

is made huge to kill chugging — is there a hidden cost?

A very stiff injector demands much higher feed/pump pressure (heavier turbopumps, more mass and power), so stability is bought at a real system-mass and performance price. See Injector Design & Pressure Drop.

If damping (nozzle radiation, viscosity) were exactly zero, what would the limit cycle look like?

There would be no limit cycle — with nothing to cap growth, a positive Rayleigh integral drives the amplitude up until nonlinear/thermal failure (the metal melts), which is why real caps come from thermoacoustic losses.

At extremely high mean chamber pressure with proportionally scaled

Δ p_{inj}

(constant ratio), does stability change?

To first order no — since the feedback gain

k

depends on the ratio

Δ p_{inj} / p_{c}

, holding it fixed keeps the chug margin roughly constant regardless of the absolute

p_{c}

Recall Fast self-test

The single rule behind both chugging and screaming ::: Rayleigh's criterion, $\oint p^{'} q^{'} d t > 0$ — grow when heat is added in phase with high pressure. The one thing that differs between them ::: The time-delay reservoir: feed/combustion lag $τ$ (ms) for chugging vs acoustic transit (μs) for screaming. The chug cure vs the screech cure ::: Chug → raise injector $Δ p_{inj}$ ; screech → acoustic damping (baffles, Helmholtz cavities). The quarter-cycle chug rule and where it comes from ::: $f_{c h ug} \approx 1/ (4 τ)$ , from $ω τ \approx π /2$ with $ω = 2 π f$ . The higher chug harmonics ::: $f \approx (2 n + 1) / (4 τ)$ from $ω τ \approx (2 n + 1) π /2$ , since adding whole cycles keeps heat in phase.