How to Work Out NPSH Available at the Pump You Are Standing At
Why this matters
You can calculate available suction head from a drawing, and the answer will be wrong in the one direction that matters, because a drawing does not know about the strainer basket somebody left half loaded, the suction isolation valve that is two turns short of open, or the flex connector that collapsed inward. One gauge at the pump inlet already contains all of it. That makes this a measurement rather than a calculation, and it means the whole job comes down to filling in a record correctly. Two of that record's eight fields are the ones that go wrong: which barometric number you used, and which moment you took the reading at. In the case below the second one moved the answer by 10.8 ft and cut the margin ratio by more than half between breakfast and the end of shift, on a pump nobody had touched.
Before you take the reading
The reading is taken with the pump running and the piping intact, which is the safe part. The unsafe parts are getting an instrument onto the line and getting a temperature.
Use an existing gauge tapping with its own isolation valve. Do not crack a plug on a line in service; a suction line under vacuum draws air and debris inward rather than spraying out, so you get no warning and you contaminate the pump. If a tapping has to be added, isolate the pump, lock and tag the disconnecting means under 29 CFR 1910.147 rather than switching it off, relieve trapped pressure at a high point, and watch a gauge for a full minute to confirm the isolation held before the first thread moves.
Get liquid temperature from a thermowell or from a surface reading on bare pipe, not by opening anything. If the line is above scald temperature, an insulated surface reading taken with an infrared thermometer from arm's length is the correct method; disturbing the insulation to reach bare metal on a hot line is a burn and, if the insulation is of unknown age or type, a respirable-fibre exposure that needs the material identified before it is disturbed.
With the pump running, the coupling guard stays on, sleeves stay buttoned, no rings, and nothing goes near the shaft. Everything on this list is reached with one hand on a gauge valve.
The record, and what a blank field costs
Eight fields. Each one is here because a blank one changes the answer or makes it unusable later.
- Time, date, and what the plant was doing. Blank, and nobody can tell later whether you caught the worst condition or the easiest one, which makes the whole record unusable as a baseline. Suction friction climbs with roughly the square of flow, vapor pressure climbs with temperature, and static head falls as a tank draws down; a 9 a.m. reading at half demand on a full tank is the easiest ten minutes of the day.
- Flow, measured, at that same moment. Blank, and you cannot look up the required value, because required suction head rises with flow. A reading with no flow attached is not comparable to anything on the curve.
- Liquid temperature at the inlet, at that same moment. Blank, and you cannot subtract vapor pressure or pick a specific gravity, and both terms move fastest exactly where margins are tightest.
- Suction gauge reading, with the gauge's type and accuracy grade. Blank on the grade, and you have a point estimate with no error bound. A dial gauge is graded on percent of span, so its bound is a fixed number of psi that does not shrink as the reading shrinks, and near the vacuum end of a compound dial it can exceed the whole margin you are trying to prove.
- Barometric basis, with the site elevation. Blank, and somebody will default to 14.7. See below; this is the field that is most often filled in wrong rather than left empty.
- Gauge height above or below the pump datum. Blank, and the sign gets guessed. A gauge above the impeller eye reads lower than the eye sees, so its height is added; below the eye, subtracted.
- Suction pipe size at the tapping. Blank, and velocity head cannot be computed. It is a small term at sensible suction velocities and it stops being small the moment the line is undersized.
- Required suction head off the curve at the measured flow, with the curve identified. Blank, and the measurement has nothing to be measured against. Required value is defined at the flow AND at the impeller diameter and speed the pump actually has, so a curve for the wrong trim is not a smaller error, it is a different pump.
On field 5: the pressure a weather app or an aviation report gives you is corrected to sea level. That is the altimeter setting, and it is not the pressure at your pump. You want station pressure, which is roughly 1 in Hg lower for every 1,000 ft of elevation near sea level. At a 700 ft site, a reported 29.92 in Hg is a station pressure of about 29.22 in Hg, or 14.36 psia against 14.70. That is 0.34 psi, about 0.8 ft of head you do not have, and it is free to get right.
The arithmetic, in six lines
Take the gauge reading and, in this order, add the station barometric pressure to make it absolute, subtract the liquid's vapor pressure at the temperature you measured, convert the remaining psi to feet using 2.31 divided by the specific gravity at that same temperature, add velocity head (pipe velocity squared, divided by 64.35), and apply the datum correction with the sign from field 6. Then compare against field 8.
Nothing in that sequence is optional and none of the terms is large by itself. What kills the answer is dropping two of them and letting the errors cancel.
Worked record: the same pump, two fills, one day
End-suction transfer pump, open break tank, 3 in schedule 40 suction (inside diameter 3.068 in, area 0.0513 sq ft). Site at 700 ft, station barometric 14.3 psia both times. Gauge tapping 0.8 ft below the shaft centerline. Water.
| Field | 09:00 fill | 16:00 fill |
|---|---|---|
| What the plant was doing | one line running | all three lines, process return hot |
| Flow, measured | 200 gpm | 320 gpm |
| Liquid temperature at inlet | 95 F | 118 F |
| Suction gauge, static | +1.5 psig | -3.2 psig |
| Station barometric | 14.3 psia | 14.3 psia |
| Vapor pressure at that temperature | 0.82 psia | 1.60 psia |
| Specific gravity, so ft per psi | 0.994, so 2.324 | 0.990, so 2.333 |
| Pipe velocity, so velocity head | 8.69 ft/s, +1.17 ft | 13.90 ft/s, +3.00 ft |
| Datum, gauge below the eye | -0.80 ft | -0.80 ft |
| Required, off the curve at that flow | 9.5 ft | 16.0 ft |
09:00. 1.5 + 14.3 = 15.80 psia absolute. Minus 0.82 vapor pressure = 14.98 psi. Times 2.324 = 34.81 ft. Plus 1.17 velocity head = 35.98 ft. Minus 0.80 datum = 35.18 ft available against 9.5 ft required, a ratio of 3.7.
16:00. -3.2 + 14.3 = 11.10 psia absolute. Minus 1.60 vapor pressure = 9.50 psi. Times 2.333 = 22.16 ft. Plus 3.00 velocity head = 25.16 ft. Minus 0.80 datum = 24.36 ft available against 16.0 ft required, a ratio of 1.52.
Available fell by 10.82 ft. Required rose by 6.5 ft. The ratio fell from 3.7 to 1.52, which is a drop of about 59 percent, and nobody touched the pump between the two fills.
What the second fill found
Three things moved and they are not equal, so it is worth saying which.
Velocity head more than doubled, and that is the finding. It went from 1.17 ft to 3.00 ft, which is exactly the flow ratio squared: 320 over 200 is 1.6, and 1.6 squared is 2.56, and 1.17 times 2.56 is 3.00. That term is small on a properly sized suction line. It is 3.00 ft here because 13.9 ft/s in a suction line is roughly double the 4 to 8 ft/s that suction piping is conventionally sized for, and the same velocity that produced the velocity-head term is producing friction upstream at the same square law, which is most of why the gauge went from plus 1.5 to minus 3.2 psig. The suction line is one size too small for the plant's actual peak, and that is a piping finding, not a pump finding.
Temperature cost less than people expect. Vapor pressure went from 0.82 to 1.60 psia, a change of 0.78 psi, about 1.8 ft. Real, and about a sixth of what was lost.
The margin is still inside normal practice and it is not comfortable. A ratio of 1.52 against required sits just above the 1.2 to 1.5 band that specifications commonly use, with the pump manufacturer's own guidance governing where it differs. But required value is defined at a 3 percent head drop, so it is the documented onset of cavitation and not a safe point. If the strainer loads up over a season, or the process return runs 10 F hotter next summer, this pump crosses.
The failure mode if only the 09:00 fill exists: 35.18 ft against 9.5 ft required is a ratio of 3.7, which reads as an installation with enormous margin, and every future complaint about this pump gets sent to the discharge side. The morning reading is not wrong. It answers a question nobody asked.
Where the method changes
Closed loops. The surface-pressure term is the system fill pressure at the expansion tank, not the barometer, so the same gauge reading has to be interpreted against where the tank connects. A waterlogged tank moves this term directly and looks correct on the fill gauge.
A gauge you do not trust. An absolute-pressure digital gauge removes fields 4 and 5 in one move: it reads 11.10 psia at 16:00 directly, so there is no barometric line to get wrong, and specified at 0.25 percent of reading its bound is about 0.03 psi, roughly 0.07 ft. Against a dial gauge's fixed percent-of-span bound at the vacuum end, that is the difference between a measurement and an opinion.
Positive-displacement pumps. The inlet condition is usually published as a required inlet pressure rather than a head, and the sensitivity is different: they tolerate low absolute pressure better and viscous fluid in a long small suction line much worse, because a fixed volume has to fill in a fixed time. Slowing the pump lowers the requirement.
How to verify you got this right
Take the second fill twice, ten minutes apart, without changing anything. Two readings that agree to a few tenths of a foot mean your temperature, flow and gauge were all captured at the same moment. Two readings that disagree by feet mean they were not, and the usual cause is a flow number read off a control screen that averages over a longer window than the gauge does.
Then check the sign of your datum term against the physical layout by walking to the gauge and looking. Field 6 is the only line in the record where being wrong flips the sign instead of shifting the magnitude, and on this pump that is 1.6 ft of swing on a 0.8 ft offset.
References
- 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before adding a gauge tapping
- Hydraulic Institute standards for net positive suction head terminology and the 3 percent head-drop definition of the required value, as adopted by the specifying engineer or by contract
- Published steam tables or the fluid supplier's property data, for vapor pressure and specific gravity at the measured temperature
- See related: What NPSH Available Actually Is and What It Is Measured Against; How Suction Conditions Fail in Practice; Reading Pump Curves Reference