HVAC
Pump cavitation and NPSH diagnosis field guide for HVAC
Read the gravel sound, measure the suction, work out whether NPSH available fell below NPSH required, and fix the suction side before the impeller is gone.
Direct answer
Pump cavitation is the collapse of vapor bubbles inside a pump, which happens when the suction pressure falls below the water's vapor pressure and the water flashes to vapor at the impeller. It sounds like gravel, erodes the impeller, and means NPSH available has dropped below NPSH required. Fix the suction side, not the pump.
Key takeaways
- Pump cavitation is vapor bubbles collapsing at the impeller when suction pressure falls below the water's vapor pressure, and it erodes the metal.
- The cavitation rule is NPSHa greater than NPSHr plus margin; when NPSHa drops to NPSHr the water flashes and the pump cavitates.
- NPSHa equals absolute suction pressure plus static height, minus suction friction loss, minus the water's vapor pressure at its temperature, all in feet.
- The gravel-and-marbles sound with vibration is cavitation until readings say otherwise, and the cure is on the suction side, not the pump.
- Fix the suction cause before replacing a pitted impeller, seal, or bearing, or the new part erodes the moment the pump runs.
Cavitation, and the suction pressure that stops it
Pump cavitation is vapor bubbles forming and collapsing inside a running pump, and it happens for one reason: the pressure at the suction dropped below the vapor pressure of the water. Water boils when its pressure falls low enough, not just when it gets hot enough. Drop the pressure at the impeller eye below the vapor pressure and the water flashes to vapor right there, makes bubbles, and then those bubbles collapse a fraction of an inch downstream where the impeller has raised the pressure again. Each collapse is a small implosion against the metal.
The plain version is this: the pump is not getting enough suction pressure. The trade name for that suction pressure margin is NPSH, net positive suction head, and cavitation is what you get when it runs out. A pump does not cavitate because it is worn out or undersized. It cavitates because the suction side is starving it, and the cure is almost always on the suction side, not in the pump.
This guide is the diagnosis. Picking the pump, setting the suction piping, and starting it clean are the companion pump installation guide's subject, and the plant-level pumping schemes sit in the chilled water pumping guide. Here the pump is already in and it is making the gravel sound, and the job is to find why the suction pressure fell and put it back before the impeller is eaten.
What happens inside a cavitating pump?
Inside a cavitating pump the water changes state twice in the space of an inch. As water accelerates into the impeller eye, the local pressure drops, the way pressure always drops where a fluid speeds up. If that local pressure falls below the water's vapor pressure, the water flashes to vapor and a cloud of bubbles forms at the lowest-pressure point, which is the leading edge of the impeller vanes. The impeller then does its job and raises the pressure, and the moment the pressure climbs back above the vapor pressure the bubbles cannot survive. They collapse.
The collapse is the damage. A vapor bubble in a high-pressure region does not pop softly. It implodes, the surrounding water rushing in to fill the void and slamming against itself and against any metal nearby. Each implosion drives a microjet of water at the vane surface hard enough to fatigue and tear away metal a few atoms at a time. Multiply that by thousands of bubbles a second and the vane face erodes into a rough, pitted, sponge-looking surface.
So cavitation is not the bubbles. It is the collapse of the bubbles against the impeller. That distinction matters in the field, because it is what separates true cavitation, which eats metal, from air passing through the pump, which is noisy but does not implode the same way. Hold the suction pressure above the vapor pressure and the bubbles never form, which is the whole point of NPSH.
Why does my pump sound like gravel?
A pump that sounds like it is pumping gravel or marbles is cavitating until proven otherwise. That rattling, crackling noise is the sound of thousands of vapor bubbles collapsing against the impeller every second, and once you have heard it on a pump you do not mistake it for anything else. It is the first and most reliable symptom, and it usually shows up before anyone has measured a thing.
The noise comes with vibration you can feel through the casing and the piping, because the collapses are hammering the impeller unevenly. The performance falls off at the same time: the flow drops, the head drops, and a discharge gauge that should be steady wanders, because vapor in the impeller passages means the pump is no longer moving solid water. On a hydronic loop the building feels it as zones going short, a chiller losing flow, or a pump that cannot hold its design point no matter what the drive does.
Open up a pump that has been cavitating and the impeller tells the rest of the story. The vane surfaces, usually on the low-pressure side near the leading edge, are pitted and eroded, eaten back into a rough honeycomb. Run it long enough and the vibration takes the mechanical seal and the bearings with it, so a cavitation problem often arrives at the shop disguised as a seal leak or a bearing failure. The gravel sound is the tell. Trust it and chase the suction.
What is NPSH?
NPSH is net positive suction head, the suction-side pressure margin that keeps water from flashing to vapor as it enters the pump. It comes in two values, and the whole problem lives in the gap between them. NPSH available, NPSHa, is what the system actually delivers at the pump suction, expressed in feet of head above the water's vapor pressure. NPSH required, NPSHr, is what the pump itself needs at a given flow to keep the water liquid through the impeller eye, and it comes off the pump manufacturer's curve.
The rule is short. NPSHa must exceed NPSHr, with margin, or the pump cavitates. NPSHa greater than NPSHr means the suction delivers more pressure than the pump needs, the water stays liquid, and the pump runs quiet. Let NPSHa fall to NPSHr and the water starts flashing. The margin is the cushion you build on top so that a hot afternoon, a partly fouled strainer, or a load swing does not push you over the edge.
How much margin is a manufacturer-and-application question, not a fixed number. The Hydraulic Institute treats NPSH margin as a ratio or an added head over the published NPSHr, and the right figure depends on the pump, the fluid, and the duty. The safe field practice is to read NPSHr off the actual pump curve at the actual operating flow, then confirm NPSHa beats it by a real margin rather than by a hair. NPSHr also climbs as flow climbs, so a pump pushed far out to the right on its curve needs more suction head than the same pump at its design point.
NPSHa > NPSHr + margin- NPSHa
- Net positive suction head available; the suction pressure the system delivers above the water's vapor pressure, in feet
- NPSHr
- Net positive suction head required; the suction head the pump needs at a given flow, read from the manufacturer's curve
- NPSH margin
- The cushion NPSHa is held above NPSHr; set per the manufacturer and the application, not a fixed number
What sets the NPSH available?
NPSHa is set by the system, and it is a sum of what helps minus what hurts. In feet of head, NPSHa equals the absolute pressure on the water at the suction source, plus the static height of water sitting above the pump, minus the friction loss in the suction piping, minus the vapor pressure of the water at its actual temperature. Build it up term by term and you can see exactly which lever moved when a pump that ran fine for years starts to cavitate.
The absolute pressure is atmospheric pressure on an open system, about 34 ft of head at sea level and less at altitude, or the system fill pressure on a closed hydronic loop. Static height helps when the water sits above the pump, a flooded suction, and hurts when the pump has to lift water up to itself. Friction loss is everything the water fights getting to the suction: the strainer, the suction valve, the fittings, the pipe run. The vapor pressure is the term temperature drives, and it is the one that surprises people, because hot water has a high vapor pressure that eats straight into the margin.
Every HVAC cavitation cause is one of these terms going the wrong way. The fill pressure fell, so the absolute pressure term dropped. The strainer clogged, so the friction term grew. The water got hot, so the vapor pressure term grew. The pump ended up above the water instead of below it, so the static term went negative. Find which term moved and you have found the cause.
NPSHa = Pabs + Hstatic − Hfriction − Pvapor- P abs
- Absolute pressure on the water at the suction source, in feet: atmospheric on an open system, fill pressure on a closed loop
- H static
- Static height of water above the pump suction in feet; positive on a flooded suction, negative on a suction lift
- H friction
- Friction head lost in the suction piping, strainer, and valve before the pump inlet, in feet
- P vapor
- Vapor pressure of the water at its operating temperature, in feet; it rises sharply with temperature
What causes a pump to cavitate in an HVAC system?
On HVAC water systems cavitation almost always traces to a short list of suction-side faults, and they rank by how often they actually bite. Run the list in order and you find most cases before you have unbolted anything.
Hot water is the first. Heating water and condensate carry a high vapor pressure, so the same suction conditions that are fine on chilled water can cavitate a boiler or condensate pump, because the temperature has already spent most of the margin. Low system pressure is the second, and it is usually one of two things: the fill or makeup pressure has fallen, or the pump is piped on the wrong side of the expansion tank so the circulator pulls the suction into a vacuum. A clogged suction strainer or a partly closed suction valve is the third, starving the suction by piling on friction. High suction lift is the fourth, and it belongs to open systems, the cooling tower and condenser water pump that has to lift water up out of a basin.
Notice what is not on the list: the pump. A pump in good order, sized right and turning the right way, does not cavitate on its own. It cavitates because one of these suction-side terms moved. So when a pump goes to gravel, the question is never what is wrong with the pump first. It is what changed on the suction side. The next sections take the common causes one at a time.
Hot water and the vapor pressure trap
Temperature is the lever most people forget, because it does not touch the piping at all. As water gets hotter its vapor pressure climbs, and vapor pressure is a straight subtraction from NPSHa. Cold water near room temperature has almost no vapor pressure to speak of, so it barely costs you anything. Heating water at 180 to 200 degrees F has a vapor pressure high enough that it can swallow a large share of the suction head, and water at or near its boiling point has nothing left to give.
This is why the same suction arrangement behaves differently on different loops. A pump that runs dead quiet on 45 degree F chilled water can cavitate on a 200 degree F heating loop with the identical piping, identical fill pressure, identical everything else, purely because the hot water's vapor pressure ate the margin the cold water never touched. Boiler pumps, primary loop pumps on a hot system, and condensate pumps moving near-saturated water are the classic victims, and condensate pumps in particular live close to the edge because the water arrives almost ready to flash.
The defense against hot-water cavitation is suction pressure and flooded suction. Keep the system fill pressure up, keep the pump below the water with positive static head, and on a boiler system respect the manufacturer's minimum suction pressure for the loop temperature. When a pump only cavitates once the system comes up to temperature, and runs fine cold, the vapor pressure term is the one that moved, and no amount of work on the strainer or the valve will fix a temperature problem.
Low fill pressure and the pump on the wrong side of the tank
The single most common cavitation cause on a closed hydronic loop is a suction-pressure problem at the expansion tank, and it shows up two ways. The first is simple: the fill or makeup pressure fell, the relief leaked, the fill valve failed, or the system lost water, so the whole loop is sitting at less pressure than it was designed for and the suction term in NPSHa dropped with it. Check the fill pressure against the design cold-fill before anything else.
The second is a piping mistake that survives for years: the pump is on the wrong side of the expansion tank. The point where the expansion tank ties into the loop is the point of no pressure change, the spot whose pressure stays the same whether the pump runs or not. When the pump is piped so it pumps toward the tank, the suction sits on the low-pressure side, and the circulator's own differential pressure subtracts from the suction, pulling it down and sometimes into a vacuum. Run that arrangement on hot water and the pump cavitates as soon as it starts.
The fix is the pump-away rule, and it is worth getting right because it costs nothing to pipe correctly and a fortune to live with wrong. Pipe the pump so it pumps away from the expansion tank, with the tank connection close to the pump suction. Then the circulator's differential pressure adds to the suction instead of subtracting, the suction stays at the higher pressure, air clears more easily, and the pump gets the NPSHa it needs. The expansion tank placement and pre-charge are their own subject, but the rule for cavitation is short: pump away, and keep the fill pressure where the design put it.
The clogged strainer and the throttled valve
A clogged suction strainer starves a pump exactly like a half-closed valve, and on a new system it is the cause that arrives right on schedule. The strainer sits in the suction line to catch debris, and every bit of friction it adds is friction subtracted from NPSHa. A clean strainer costs a little head you accounted for. A strainer loading up with construction debris, mill scale, or sediment costs more and more head as it plugs, until the suction pressure across it falls below the vapor pressure and the pump goes to gravel.
The startup version of this is the temporary commissioning screen, the fine mesh fitted in the suction or the suction diffuser to protect the pump through the dirty first hours. It does its job and then becomes the problem, because it catches so much debris that it turns into the restriction that starves the pump it was meant to protect. A pump that started clean and went noisy a day or two later is very often a startup screen nobody pulled and cleaned. That sequence, flush, screen, then remove the screen, is the install guide's territory, and the cavitation lesson is the consequence of skipping it.
A partly closed suction valve does the same thing with no debris involved. A valve someone throttled to balance the wrong side, or a valve never fully opened after service, or a butterfly disc that looks open but is not, all pile friction onto the suction. The first move on a cavitating pump is to confirm the suction valve is fully open and the strainer is clean. It is the cheapest check on the list and it catches a surprising share of the calls.
How do you diagnose a cavitating pump?
Diagnosis runs from the easy and free to the involved, and you rarely have to reach the far end. Start with your ears. The gravel-and-marbles sound with vibration is cavitation until the readings say otherwise, and it points you straight at the suction side. While you are there, feel the pump and the suction pipe for the vibration that rides with the noise.
Then measure. Put a gauge on the suction, ideally a compound gauge that reads vacuum, and watch the suction pressure with the pump running under real load. Convert that suction pressure to absolute, and compare it against the vapor pressure of the water at its measured temperature. If the suction pressure is close to or below the vapor pressure, the water is flashing and you have confirmed cavitation by the numbers, not just the noise. Take the water temperature while you are at it, because the vapor pressure you compare against depends entirely on it.
With the noise and the readings agreeing, check the four suction-side terms in order: the strainer, the suction valve, the fill or system pressure, and the water temperature. Is the strainer plugged? Is the valve fully open? Is the fill pressure at design? Has the loop come up to a temperature that ate the margin? One of those four has almost always moved. The trap to avoid is condemning the pump or the impeller before you have checked the suction, because replacing a pitted impeller without fixing the suction cause just feeds a fresh impeller into the same grinder.
Measuring and calculating NPSHa in the field
When the noise and the gauge are not enough and you need to settle it, you can work out NPSHa from field readings and put it against the pump's published NPSHr. Read the suction pressure at the pump with the pump running at its real flow, and read the water temperature at the same point. Those two readings carry most of the answer.
Convert the suction gauge pressure to absolute by adding atmospheric pressure, then express it in feet of head. Subtract the vapor pressure of the water at the measured temperature, also in feet, taken from a steam table or a vapor-pressure chart. What remains is the NPSHa the system is actually delivering at that operating point, measured rather than estimated, which folds the real fill pressure, the real friction, and the real temperature into one number. Account for the height difference between the gauge and the impeller centerline if it is significant.
Now read NPSHr off the manufacturer's curve at the flow the pump is actually passing, not at the design flow, because NPSHr rises with flow. Compare the two. NPSHa comfortably above NPSHr with margin and the suction is healthy, so look elsewhere. NPSHa at or below NPSHr and you have proven the cavitation and identified that the suction cannot supply what this pump needs at this flow. The published NPSHr and the right margin are the manufacturer's numbers, so treat the curve as the authority and the measured NPSHa as what your system gives it.
How do you fix a cavitating pump?
Fixing cavitation means raising NPSHa back above NPSHr, and you do it by moving whichever suction-side term went wrong. The fixes run cheapest first, and the cheap ones solve most calls. Clean the suction strainer and pull any startup screen that is loading up. Open the suction valve fully and confirm it. These cost an hour and fix the largest share of cavitation complaints.
If the pressure is the problem, raise it. Bring the system fill or makeup pressure back to the design cold-fill, repair the fill valve or the leak that let it drop, and confirm the expansion tank is holding its pre-charge and is not waterlogged. If the pump is piped on the wrong side of the expansion tank, re-pipe it to pump away from the tank so the circulator's differential pressure adds to the suction. On a hot loop, lower the water temperature where the system allows it, since dropping the temperature drops the vapor pressure and hands the margin straight back.
When the suction piping itself is the limit, fix the geometry. A suction line that is too long, too small, or full of elbows close to the pump throws away head you need, so shorten it, go up a pipe size to cut the velocity and the friction, and keep a straight run into the suction nozzle. The strongest move of all is a flooded suction: arrange the source so water sits above the pump and feeds it by gravity, which buys positive static head instead of fighting a lift. Match the fix to the term that moved, and confirm the gravel sound is gone under real load before you call it done.
Suction piping and the eccentric reducer
The suction piping is where NPSHa is won or thrown away, because every foot of friction on the suction side comes straight off the margin while the discharge side forgives almost anything. The rules are short and they all serve the same end. Keep the suction line short, keep it as large as or larger than the pump nozzle, and keep it straight into the inlet so the water enters the impeller eye evenly. An elbow slammed right onto the suction nozzle throws the flow to one side, loads the impeller unevenly, and can drag the local suction pressure down enough to start cavitation on its own. The common target is several pipe diameters of straight run ahead of the suction, often cited around five to ten diameters, with no flow-disturbing fittings in that run.
Where the suction pipe is larger than the pump nozzle, the reducer that steps it down has to be an eccentric reducer installed flat side up. The flat on top keeps the top of the pipe level into the pump so air cannot collect in a high pocket against the suction and feed bubbles straight into the impeller. A concentric reducer, or an eccentric one installed flat side down, builds a high spot on a horizontal suction that traps air, and that trapped air is a cavitation and air-binding complaint waiting for the first low-flow morning. Flat side up on a horizontal suction is one of the first things an experienced eye checks on a noisy pump.
No air pocket is the whole intent. The suction has to deliver solid, even, debris-free water to the impeller, and any high point that traps air or any restriction that adds friction works against that. The piping detail itself sits in the companion pump installation guide. The cavitation point is that bad suction geometry shows up later as gravel, and the eccentric reducer flat side up is the cheap detail that prevents one common version of it.
- Eccentric reducer
- A reducer offset to one side; on a horizontal suction it is installed flat side up so air cannot pocket against the pump
- Flooded suction
- An arrangement where water sits above the pump and feeds it by gravity, giving positive static head and the most NPSHa
Is it air or is it cavitation?
Air in the system and true cavitation both make a pump noisy and both knock down the flow, so they get confused, but they are different faults with different fixes. Air binding is air carried into or trapped in the pump: bubbles that were already in the water, or air pulled in through a leak or drawn down from a vortex on an open suction. Those bubbles pass through the pump or sit in the volute. They are noisy and they hurt performance, but air entrained in the water does not implode the way vapor does, so it does not erode the impeller the way cavitation does.
Cavitation is the water itself flashing to vapor at the impeller because the pressure fell below the vapor pressure, and then those vapor bubbles collapsing violently. That collapse is what eats the metal. The practical difference is the cause and the cure. Air binding is fixed by getting the air out and keeping it out: vent the casing and the high points, fix the leak letting air in, get the eccentric reducer right so no pocket forms, and let the air separator do its work. Cavitation is fixed by raising the suction pressure: strainer, valve, fill pressure, temperature, suction geometry.
The tells help you sort them. Air often comes and goes as slugs of air pass through, and it follows fill events, venting, or a system that keeps swallowing air. Cavitation is steadier under a steady condition and tracks with the suction-side term that is wrong, getting worse as the loop heats up or as the strainer plugs. When in doubt, vent thoroughly first, because clearing the air is quick and free. If the gravel sound survives a proper purge, you are looking at cavitation and the suction pressure is the problem. The air separator and venting details live with the install and balancing work.
The pitted impeller, and why you fix the cause first
Pull a pump that has cavitated and the impeller shows it. The vane surfaces are pitted and eroded, usually worst on the low-pressure side near the leading edge where the bubbles collapse, chewed back into a rough, porous, eaten-away texture. Severe cavitation can carve material out of the vanes and the shroud, unbalance the impeller, and on a bad case hole it through. A pitted impeller does not recover. It gets replaced.
The mistake that turns one repair into a string of them is replacing the impeller and stopping there. The impeller did not erode because it was defective. It eroded because the suction pressure fell below the vapor pressure and the water flashed, and that suction condition is still sitting there waiting for the new impeller. Drop a fresh impeller into an unfixed suction problem and you have bought a part and a labor call and changed nothing. The new one starts eroding the moment you energize it.
So the order is fixed. Find why NPSHa fell below NPSHr, correct the suction-side cause, prove the gravel sound is gone, and then, if the old impeller is too far gone, replace it. The replacement is the last step, not the first, and a pump that keeps eating impellers is a pump whose suction problem was never found. Fix the cause, then the part.
The seal and bearing damage cavitation leaves behind
Cavitation does not stop at the impeller. The vibration it sets up shakes the whole rotating assembly, and that shaking works on the mechanical seal and the bearings the entire time the pump runs noisy. The mechanical seal depends on a thin, stable film of water between two lapped faces; pound it with the vibration and pressure pulses of cavitation and the faces chatter, the film breaks down, and the seal starts to weep and then to fail. The bearings carry the cyclic load of the collapses and fatigue ahead of their rated life.
This is why a cavitation problem so often shows up at the shop wearing a different label. The call comes in as a leaking seal or a noisy bearing, the part gets replaced, and the pump is quiet for a while and then fails again, because the cavitation that wrecked the seal the first time is still running. A seal or a bearing that keeps failing fast on the same pump is a strong hint to stop replacing parts and go look at the suction.
The maintenance tie is the same one the install guide makes: the seal and the bearings are the parts the owner inherits, and cavitation is one of the things that shortens both. The seal and bearing replacement procedure is in the pump installation guide. The lesson here is that quieting the cavitation protects them, and replacing them without quieting it does not.
The no-cavitation check at startup
The cheapest cavitation to fix is the one you catch on the day the pump starts, before it has run long enough to eat anything. Confirming no cavitation at startup is a standard part of commissioning a pump, and it takes minutes. Bring the pump up under real load, then stop and listen. A pump running clean is smooth and steady. A pump making the gravel-and-marbles rattle is cavitating, and you catch it now, with a clean impeller, instead of three weeks out with a pitted one.
Back the ears up with the suction gauge. Read the suction pressure under load and confirm it sits well above the vapor pressure of the water at its temperature, and on a system that will run hotter than its startup condition, think ahead to where the vapor pressure goes once the loop is up to temperature. A pump that is quiet cold but has no suction margin to spare will cavitate the first hot afternoon, so a startup check on a cold system is not the whole story for a heating loop.
This is the moment the startup screen comes out, too, because a pump commissioned clean can go noisy a day later when the temporary screen finishes plugging. Listen at startup, read the suction, pull the startup strainer, and confirm the pump stays quiet across the conditions it will actually see. The full startup sequence belongs to the install guide; the cavitation piece of it is the listen-and-read check that proves the suction is healthy before the pump runs for keeps.
Cooling towers and open-system suction lift
Open systems change the NPSH problem, because the water is exposed to atmosphere instead of sealed in a pressurized loop, and the pump may have to lift it. The condenser water pump drawing from a cooling tower basin is the case you meet most. If the pump sits above the basin water level, it is on a suction lift, the static term in NPSHa goes negative, and the margin is tighter from the start than any closed-loop pump enjoys.
That tighter margin means the open-system pump has less room to absorb the other faults. A strainer starting to plug, a basin level running low, a long or undersized suction line, or warm condenser water on a hot day can each push it over where a closed-loop pump would have shrugged it off. The defense is to keep the static lift as small as the layout allows, ideally a flooded suction with the pump below the basin level, keep the suction line short and large, keep the strainer clean, and hold the basin level up. Confirm NPSHa against the pump's NPSHr the same way as on a closed loop, just with the atmospheric pressure and the basin level standing in for the system fill pressure.
The plant-side condenser and chilled water arrangements sit in the chilled water pumping guide. The open-system cavitation point is narrow: a suction lift spends margin before the pump even starts, so the suction-side housekeeping that a closed loop can be sloppy about, the open system cannot.
Chilled water pump cavitation in a data center
In a data center the cost of a cavitating pump is not just the impeller, it is the cooling the impeller was supposed to deliver to a hall that never stops needing it. The pumps run around the clock at part load most of the time, and the plants are built redundant, so a standby pump that cavitates the moment it is called is a redundancy that exists only on the single-line drawing. The standby has to be commissioned to the same no-cavitation standard as the running pumps, with its suction proven, or it is not really backup.
Two data-center habits raise the cavitation stakes. Variable primary flow and aggressive staging push pumps far out on their curves at times, and NPSHr rises with flow, so a pump that has margin at its design point can lose it when the controls drive it hard. And the move toward higher chilled water temperatures to gain economizer hours nudges the vapor pressure up, trimming the margin that warmer water leaves. Neither is a reason to avoid those strategies, both of which the chilled water pumping guide covers, but both are reasons to confirm NPSHa across the full operating range rather than at one design point.
The discipline is the same as anywhere, applied harder: clean strainers, held fill pressure, suctions that are not starved, and standby pumps that have actually been run and listened to. A cavitation problem found on a quiet commissioning afternoon is a strainer cleaning. The same problem found when the lead pump drops on a hot day is an outage.
What to document
A cavitation call you do not write up is a cavitation call someone repeats. The value of the record is that it ties the symptom to the cause to the fix, so the next person who hears gravel on that pump, or on its twin across the plant, starts from what was already found instead of from zero. It also catches the pattern where a pump keeps cavitating because the real cause was never corrected.
Record what the pump sounded like and showed, the suction pressure and water temperature you measured, the NPSHa you worked out and the NPSHr off the curve if you went that far, which suction-side term was wrong, the fix you made, and whether the noise cleared under real load afterward. Note the condition it cavitates under, cold or hot, full load or part load, because that points at the term. The table below pairs the common cause with its symptom and its fix as a field reference.
| Cause | Symptom or tell | Fix |
|---|---|---|
| Hot water, high vapor pressure | Quiet cold, cavitates once the loop heats up | Raise fill pressure, flooded suction, lower water temp where possible |
| Low fill or makeup pressure | Whole loop below design pressure, suction near vacuum | Restore fill to design cold-fill, fix the fill valve or leak |
| Pump on wrong side of expansion tank | Cavitates on start, suction pulls into vacuum, air trouble | Re-pipe to pump away from the tank, tank near pump suction |
| Clogged suction strainer or startup screen | Went noisy hours or days after a clean start | Clean the strainer, pull and clean the startup screen |
| Throttled or partly closed suction valve | Gravel sound, low suction pressure, easy to miss | Confirm the suction valve is fully open |
| Suction lift on an open system | Tower or condenser pump, low basin level, tight margin | Lower the lift toward flooded suction, hold basin level, larger suction |
| Air binding (not true cavitation) | Noise comes and goes, follows fill or venting, no pitting | Vent the casing and high points, fix the air source |
Common mistakes
- Running a cavitating pump instead of stopping it, so the gravel sound grinds the impeller and shakes out the seal and bearings.
- Letting the system fill pressure fall, or piping the pump on the wrong side of the expansion tank so the suction pulls into a vacuum.
- Leaving a clogged suction strainer or an unremoved startup screen to starve the suction and drop NPSHa.
- Treating a hot-water loop like cold water and leaving no NPSH margin for the vapor pressure the temperature adds.
- Mistaking air binding for cavitation, or cavitation for air, and chasing the wrong fix.
- Replacing a pitted impeller, a leaking seal, or a noisy bearing without finding and fixing the suction cause that wrecked it.
- Confirming NPSHa at one design point and missing that NPSHr rises with flow when the controls push the pump out on its curve.
Field checklist
Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.
Standards and references
The pump manufacturer is the controlling authority on the numbers that decide cavitation, and that is not a hedge, it is where the values live. The NPSHr curve, the required margin, the minimum and maximum flow, and the minimum suction pressure for the service all come from the manufacturer's curve and the installation and operation documents. NPSHr is read at the actual operating flow because it rises with flow, and the margin over it follows the manufacturer's guidance for the pump and the fluid. When a habit disagrees with the curve, the curve wins.
The Hydraulic Institute, HI, publishes the rotodynamic pump standards the trade works to, including the NPSH margin practice and the suction-piping guidance such as the eccentric reducer flat side up and the straight run into the suction. ASHRAE covers the hydronic and chilled water systems the pumps serve, with the handbooks and Standard 90.1 setting the system and energy basis, and the system design fixing the fill pressure, the temperatures, and the expansion tank arrangement that determine NPSHa. The vapor pressure of water at temperature comes from standard steam tables.
Name the document that governs the point and confirm the edition, because these revise on their own cycles. Cite the manufacturer for the NPSHr and the margin, HI for the pump and suction-piping practice, ASHRAE and the project's mechanical design for the system conditions and the chilled water or boiler basis, and let the project specification and the engineer of record control when they are tighter than common practice. The rule that does not move is the physics: NPSHa must exceed NPSHr, the cure is on the suction side, and pumping away with the fill pressure held is how a closed loop keeps the margin it needs.
Units, terms, and conversions
Cavitation and NPSH carry their own vocabulary and a couple of unit systems, so the same quantity reads differently across a pump curve, a steam table, and a metric drawing.
NPSH, both available and required, is in feet of water in US practice and in meters in metric sources. Suction pressure is read in psi or in inches of mercury of vacuum on a compound gauge, and converts to feet of head at about 2.31 ft of water per psi for cold water. Vapor pressure is given in psia or in feet of head at a stated temperature, and it climbs steeply as the water gets hotter. Atmospheric pressure is about 14.7 psia, or roughly 34 ft of head, at sea level and less at altitude. Cavitation also goes by the gravel or marbles sound the trade names it for, and air binding is the separate fault it is most often confused with.
- Cavitation
- Vapor bubbles forming and collapsing at the impeller when suction pressure falls below the water's vapor pressure, eroding the metal
- Vapor pressure
- The pressure at which water flashes to vapor at a given temperature; it rises sharply with temperature and is subtracted from NPSHa
- Flooded suction
- Water sitting above the pump and feeding it by gravity, giving positive static head and the most NPSHa
- Suction lift
- The pump above the water level, lifting it; the static term goes negative and the NPSH margin is tighter
- Air binding
- Air carried into or trapped in the pump; noisy and performance-robbing, but it does not implode and erode like cavitation
- Pump away
- Piping the pump to pump away from the expansion tank so its differential pressure adds to the suction, raising NPSHa
FAQ
What is pump cavitation?
Pump cavitation is vapor bubbles forming and collapsing inside a running pump. It happens when the suction pressure falls below the water's vapor pressure, so the water flashes to vapor at the impeller and the bubbles then implode against the vanes, eroding them. It means the pump is not getting enough suction pressure.
What is NPSH?
NPSH is net positive suction head, the suction-side pressure margin that keeps water from flashing to vapor at the pump. NPSH available, what the system delivers, must exceed NPSH required, what the pump needs off its curve, with margin. Fall below it and the pump cavitates. NPSHr rises as the flow rises.
Why does my pump sound like gravel?
A pump that sounds like gravel or marbles is cavitating. The noise is thousands of vapor bubbles collapsing against the impeller every second, with vibration and falling flow alongside it. The cause is on the suction side: a clogged strainer, a throttled valve, low fill pressure, or water too hot for the margin. Check the suction first.
How do you fix a cavitating pump?
Raise the suction pressure above the water's vapor pressure by fixing the suction-side cause. Clean the strainer, open the suction valve, restore the fill pressure, pipe the pump away from the expansion tank, lower the water temperature where you can, and shorten or upsize the suction line. Fix the cause before replacing any pitted impeller.
How do you calculate NPSH available?
NPSHa equals the absolute pressure at the suction, plus the static height of water above the pump, minus the suction friction loss, minus the water's vapor pressure at its temperature, all in feet. In the field, read suction pressure and water temperature, convert to absolute, subtract the vapor pressure, and compare the result to the pump's NPSHr.
Why does hot water make a pump cavitate?
Hot water has a high vapor pressure, and vapor pressure is subtracted straight from NPSH available. A pump quiet on 45 degree F chilled water can cavitate on a 200 degree F heating loop with identical piping, because the hot water's vapor pressure ate the margin. Keep the fill pressure up and the suction flooded on hot loops.
Is the pump cavitating or air binding?
Both make noise and cut flow, but they differ. Air binding is air carried into the pump that passes through without imploding, so it does not pit the impeller. Cavitation is the water flashing to vapor and the bubbles collapsing, which erodes metal. Vent thoroughly first. If the gravel sound survives a purge, it is cavitation.
Why is my pump cavitating after I replaced the impeller?
Because the suction cause that eroded the first impeller was never fixed. A pitted impeller is the symptom, not the disease. The suction pressure is still falling below the vapor pressure, so the new impeller starts eroding the moment it runs. Find and correct the suction-side term, prove the noise is gone, then replace the impeller.
Which side of the expansion tank should the pump be on?
Pipe the pump to pump away from the expansion tank, with the tank connection near the pump suction. Then the circulator's differential pressure adds to the suction instead of pulling it into a vacuum, which raises NPSH available and helps clear air. A pump on the wrong side cavitates, especially on a hot loop.
Which way does an eccentric reducer go on a pump suction?
Flat side up on a horizontal suction. The flat top keeps the pipe crown level into the pump so air cannot collect in a high pocket against the suction and feed bubbles to the impeller. A concentric reducer, or flat side down, builds an air-trapping high spot that causes cavitation and air binding on low-flow mornings.