Why a control valve stops responding to downstream pressure — called Choked Flow — and what the speed of sound has to do with it. Is valve cavitation the same phenomenon, or something else entirely?
Every instrument measures a condition at a point. But before a measurement can mean anything, the physical condition must travel from where it changes to where it is sensed. In a fluid system, that journey has a speed limit — and at high enough flow velocities, the limit is crossed. What follows is not instrument failure. It is physics.
Place two pressure transmitters in a pipe — one at the inlet, one at the outlet — with fluid sitting perfectly still between them. No flow, no valves open, nothing moving. Introduce a pressure change at the inlet. How quickly does the outlet transmitter respond?
Almost immediately. Not because the fluid rushed from one end to the other, but because a pressure wave did. A pressure wave is not a moving mass of fluid — it is a pattern: compression followed by relaxation, propagating through the fluid via molecular collisions. A disturbance at one point energises a layer of molecules; they surge forward, compress their neighbours, transfer momentum, and relax. Those neighbours do the same. The pattern ripples outward.
This mechanism is identical to the transmission of sound. A pressure wave is a sound wave. Its speed — the speed of sound in that medium — is set by how stiff the medium is (how readily molecules rebound after compression) and how dense it is (how much mass each collision must accelerate). That speed varies considerably by medium:
| Medium | Speed of sound | Signal travel — 100 m pipe |
|---|---|---|
| Water | ~1,500 m/s | 0.07 seconds |
| Hydraulic oil | ~1,300 m/s | 0.08 seconds |
| Steam | ~500 m/s | 0.20 seconds |
| Air at ambient | ~340 m/s | 0.30 seconds |
In a static or low-velocity system, both transmitters respond to a pressure event almost simultaneously. The entire fluid volume behaves as a single, causally connected domain. What one end feels, the other end knows.
This is the only mechanism by which one part of a fluid communicates with another. There is no other carrier. If the pressure cascade cannot reach upstream, upstream remains ignorant — regardless of what any transmitter downstream is reading.
Return to those two transmitters, pipe closed at both ends. Open the inlet to a continuous high-pressure source. The first compression wave reaches the far transmitter at sound speed — both sensors register the change almost at once. But the source continues feeding energy. The reflected wave bounces back upstream, meets incoming waves, and the superimposed compressions accumulate. With no outlet, the energy converts entirely into compressing the fluid further — raising its density, raising its pressure.
This continues until pipe pressure equals source pressure. The driving gradient vanishes. Both transmitters now read the same value. The static fluid has absorbed continuous energy input and converted it entirely to pressure — elastic potential energy stored in the compressed volume. No flow, no gradient, one unified pressure domain.
This is the baseline: a fully connected system where information propagates freely in both directions and pressure equilibrates completely. The two transmitters will never disagree for long.
Now introduce bulk flow. Open the outlet. Fluid moves from high pressure to low through the pipe and its restrictions — a narrowing, a valve seat, an orifice throat. The inlet transmitter reads supply pressure; the outlet transmitter reads a lower downstream value. A measurable, stable gradient exists between them.
The molecular collision cascade still propagates at sound speed 'c' relative to the local fluid — but the fluid itself is now moving at velocity 'v'. A pressure wave trying to carry news upstream — that downstream pressure has dropped — travels at a net speed of:
Where velocity is low (large cross-section, far from restrictions), c dominates. The wave reaches upstream easily. The outlet transmitter's world is still legible to the inlet. The system remains connected, responsive, controllable.
But as fluid enters a narrowing, velocity rises. In a gas, the local sound speed also falls as the fluid expands and cools with the pressure drop. The two values close on each other. At the throat — the narrowest cross-section, the point of maximum velocity — v approaches c.
When v = c, the net upstream speed is zero. The molecular cascade is swept downstream exactly as fast as it tries to travel upstream. No pressure information crosses this plane in the upstream direction.
The fluid at the throat is not blocked by metal or obstruction. It is isolated by its own velocity.
Beyond this line, the downstream world goes silent to everything upstream of the throat. Any pressure reduction further downstream — a vessel emptying, a back-pressure dropping — generates a wave that travels toward the valve, reaches the vena contracta, and stops there. The upstream fluid never receives the message. What the outlet transmitter reads becomes causally severed from what the valve passes.
This is choked flow: not a blockage in the physical sense, but an information blockade. The downstream world has fallen silent to the throat.
This condition has a precise expression. The ratio of bulk flow velocity to local sound speed is the Mach number:
Choking occurs when M = 1 at the throat. At that point, c − v = 0, and no further reduction in downstream pressure can increase throat velocity — there is no mechanism by which that instruction could arrive. For a gas flowing through a restriction, this corresponds to a critical pressure ratio across the valve. For air, that ratio is approximately 0.53 — downstream absolute pressure must remain above roughly half the upstream absolute pressure to avoid choking. Below that threshold, the throat is already at Mach 1. The downstream pressure is simply irrelevant to what happens there.
The critical pressure ratio is not an empirical rule. It is where the physics says the messenger stops.
A messenger swimming upstream in a river. The swimmer moves at a fixed speed relative to the water — that is the sound speed, the molecular cascade. The river current is the bulk flow velocity. When the current matches the swimmer's speed, the messenger is stationary relative to the bank. No message travels upstream, regardless of how urgent the news from downstream.
The throat is exactly that point. The current has matched the swimmer. The outlet transmitter may be registering a falling pressure with every passing second. The throat cannot hear it.
As liquid accelerates through a valve restriction, local pressure drops. If it falls to the vapour pressure of the liquid at that temperature, the liquid flashes into vapour bubbles at the vena contracta. The vapour pocket caps the velocity — further reduction in downstream pressure no longer increases flow. The mechanism is phase change rather than Mach 1, but the outcome is identical: the throat becomes deaf to downstream conditions.
The bubbles do not persist. Downstream of the restriction, pressure recovers and the vapour collapses back into liquid — violently, asymmetrically, in microseconds. The implosion drives a micro-jet into the surrounding metal at pressures that exceed the yield strength of most trim materials. Repeated billions of times per second, the result is not wear. It is pitting — metal eaten from within.
A cavitating valve announces itself: it sounds like gravel passing through the body, vibration follows, and trim inspection reveals a cratered surface that recalibration cannot fix. The response is resizing — raising inlet pressure, selecting a trim that staggers the pressure drop across multiple restrictions, or reducing the pressure ratio enough to keep the vena contracta above vapour pressure throughout the operating range.
The F_L factor in IEC 60534 quantifies exactly that margin — how close a given valve geometry brings the vena contracta pressure to vapour pressure, and therefore how much headroom remains before the liquid decides it has had enough of being a liquid.
| Static system | Flowing — subsonic | Flowing — choked | |
|---|---|---|---|
| Bulk velocity | Zero | v < c | v ≥ c at throat |
| Upstream signal speed | c | c − v (positive) | ≤ 0 |
| Information domain | Fully connected | Connected, with drag | Split |
| Downstream influence on flow | None (no flow) | Full — both ends coupled | None — throat isolated |
The same molecular collision cascade governs all three states. The difference is whether the wave can outrun the flow. When it can, the system is connected. When it cannot, the system is split. Choked flow is not a special case — it is the direct, inevitable consequence of the speed limit embedded in the medium itself.
You size a control valve. You set downstream pressure low enough. Flow increases — then plateaus. You reduce downstream pressure further. The flow does not respond. You suspect a faulty positioner. You check for debris. You re-stroke the actuator.
The valve is fine. The physics has drawn a line at the vena contracta, and nothing downstream of that line reaches the controller, the actuator, or the upstream fluid. The flow is not stuck. It is deaf.
Understanding this is not academic preparation — it is the difference between a correct valve sizing calculation and one that underestimates maximum flow, between a confident commissioning decision and a frustrating fault-finding exercise that ends where the physics always said it would.
The practical tools for working with choked flow in valve sizing — the liquid pressure recovery factor FL, the critical flow factor Cf, and the IEC 60534 sizing procedure that incorporates both — are built directly on this foundation. Those are the subject of the valve sizing brick.
Notice what the choked condition reveals about measurement itself. The outlet transmitter continues to read faithfully — it reports a real pressure, at a real point, with no fault. But its reading has become causally disconnected from the throat. Two instruments, both working perfectly, whose readings no longer share the same physics.
This is the deeper lesson. A signal is only as meaningful as the physical path connecting it to the condition it represents. When that path is severed — not by wiring, not by calibration drift, but by fluid mechanics — the instrument is correct and irrelevant at the same time. The I&C engineer's job is to know when that condition exists before the control loop discovers it the hard way.