What NPSH Available Actually Is and What It Is Measured Against

Why this matters

A starved pump does not report itself as a suction problem. It reports low flow, a seal that keeps weeping, a bearing that got loud, a motor pulling less current than you expected. Every one of those sends a tech to the discharge end of the machine. Net positive suction head available is the number that settles which end to work on, and the reason it settles so few arguments in the field is that it is not a pressure you read off a gauge. It is a pressure you read off a gauge and then correct four times. In the case worked below those four corrections are worth, individually, between a quarter of a foot and two and a half feet, and the dial gauge's own error bound is worth 3.5 ft, against a required value of 11.0 ft. Get the frame wrong and you are arguing about a margin nobody actually measured.

Before you put an instrument on a suction line

A suction line can sit under vacuum, so a tapping opened in service pulls air and debris inward instead of spraying out and the failure is silent. Use an existing gauge tapping that has its own isolation valve, and do not crack a plug on a line that is in service. If a tapping has to be added, that is work on a depressurized line: 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 confirm the isolation held by watching a gauge for a full minute before the first thread moves.

Where the liquid is above scald temperature, let the section cool before breaking a joint, drain to a routed hose rather than to the floor, stand out of the plane of the joint, and wear a face shield rather than safety glasses. Where the SDS names a vapor or aerosol hazard, the control is respiratory rather than a glove, and it puts you inside the employer's program under 29 CFR 1910.134, which requires fit testing before first use and at least annually under 1910.134(f)(2). The reading itself is taken with the pump running, so the coupling guard stays on, sleeves stay buttoned, and the only things your hands reach are the gauge and its isolation valve.

The definition, in one sentence

Net positive suction head available is the total head of the liquid at the pump's suction connection, referenced to the pump datum, in absolute terms, expressed in feet of the liquid you are actually pumping, minus the head equivalent of that liquid's vapor pressure at the temperature it is at right now.

Four of those bolded phrases are the four corrections. Total rather than static means velocity head gets added. Referenced to the datum means the gauge's height above or below the impeller eye gets applied. Absolute means the local barometer gets added, not 14.7. Feet of this liquid means the conversion from psi uses the specific gravity at operating temperature. Vapor pressure is not a correction, it is half the definition: NPSH available is a margin above boiling, not a pressure.

The reference plane is the impeller eye

For a horizontal end-suction pump the datum is the shaft centerline, which is where the impeller eye sits. Your gauge is almost never there. It is on a tapping in the suction flange or a few inches upstream, and it is usually above the shaft.

      tank liquid surface
   ============================
              |
              |   static head, source above pump
              |
              |        gauge tapping
              |             o
              |             |  gauge sits above the
              |             |  eye, so pressure at
              |             |  the eye is higher
   shaft =====+=========[ impeller eye ]=========
              |
   -------------- floor --------------

Pressure rises with depth in a standing column, so a gauge mounted above the eye reads lower than the eye sees, and its height gets added. A gauge mounted below the eye reads higher than the eye sees, and its height gets subtracted. That is a sign, not a magnitude, and getting it backwards on a 1.2 ft offset moves the answer by 2.4 ft, which on a tight installation is the whole margin.

Vertical turbine and submersible pumps use a different datum, usually the first-stage impeller, and it is several feet below anything you can reach. On those, the manufacturer's own drawing owns the datum and there is no way to derive it from the floor.

Absolute, local, and feet of this liquid

Absolute, and local. A compound gauge reads relative to whatever the atmosphere is doing where it is standing. Standard atmosphere is 14.7 psia at sea level, 14.3 psia around 700 ft, 13.66 psia around 2,000 ft, and roughly 12.2 psia around 5,000 ft. Using 14.7 at a 2,000 ft site adds 1.04 psi that is not there, which for warm water is about 2.5 ft of head you do not have.

Feet of this liquid. Head in feet is pressure in psi times 2.31, divided by specific gravity. Cold water at 1.00 gives 2.31 ft per psi. Water at 150 F is around 0.980, giving 2.36. Water at 180 F is around 0.97, giving 2.38. The error is small on warm water and it is not small on a hot or a heavy fluid, and it always runs the same way: using 2.31 on a light liquid understates the head you have.

Velocity head. The gauge reads static pressure. The definition wants total head, which is static plus velocity head, and velocity head is the square of pipe velocity divided by 64.35. That term is small at sensible suction velocities and it stops being small when a suction line is undersized, because it grows with the square of flow.

What it is measured against

It is measured against net positive suction head required, taken off the pump's own curve at the flow you were actually running, not at design flow, because the required value rises with flow. Two things about that comparison decide whether your reading means anything.

First, the required value is not a pass line. It is defined at the point where the pump's developed head has already dropped by 3 percent, so the pump is cavitating there by definition. A sibling card works through the four-term budget that produces the available number and what happens as each term drifts; the short version is that equality is the documented onset, not a margin, and common specification practice runs a ratio of roughly 1.2 to 1.5 times required, with the pump manufacturer's own guidance governing where it differs.

Second, and this is where field measurements quietly fail: an error figure needs its basis and its character before it goes anywhere near this comparison. A dial gauge is graded on percent of span, not percent of reading, so its bound is a fixed number of psi that does not shrink as the reading shrinks. ASME B40.100 grades dial gauges this way, and it binds through the gauge's own listing or through a specification that calls it out rather than on its own; a Grade A gauge is 1 percent of span through the middle half of the dial and 2 percent over the top and bottom quarters. Character matters too: that bound is a worst case, so it is reported as a bound and never as a plus-or-minus interval, and it does not cancel here because you are computing an absolute level rather than a difference between two readings of the same instrument.

Worked example: one gauge reading, corrected line by line

An end-suction transfer pump lifting from an open tank at a 2,000 ft site. Local barometric 13.66 psia. Compound gauge on the suction flange tapping, span 30 in Hg vacuum to 60 psi, Grade A. Suction pipe is 4 in schedule 40, inside diameter 4.026 in. The pump is running at 300 gpm, measured. The gauge tapping is 1.2 ft above the shaft centerline. Liquid is water; the thermometer in the same tapping well reads 150 F, where vapor pressure is 3.72 psia and specific gravity is about 0.980, giving 2.36 ft per psi. The curve gives 11.0 ft required at 300 gpm.

The gauge reads minus 6.0 psig.

Point estimate: 11.39 ft available against 11.0 ft required, a ratio of 1.04.

Now apply the instrument. Minus 6.0 psig sits in the bottom quarter of a 74.7 psi span, where the grade allows 2 percent of span, so the bound is 1.49 psi, about 3.5 ft of this liquid. It is a worst-case bound, so it is written with one inequality sign: NPSH available is not better than 11.39 minus 3.5, so NPSHa is at least 7.9 ft and may be as good as the point estimate. Against 11.0 ft required, this measurement cannot establish that the pump has any margin at all.

That is the honest answer, and it is more useful than the false one. The instrument, not the pump, is the thing to change: an electronic gauge or transmitter specified in percent of reading gives a bound that shrinks with the reading, and at these pressures that turns a 3.5 ft uncertainty into a fraction of a foot.

What the four corrections were individually worth here: local barometer instead of 14.7, 2.45 ft; specific gravity instead of 2.31, 0.25 ft; velocity head, 0.89 ft; datum, 1.20 ft. Notice that they partly cancel. A tech who skipped all four would have landed at 11.50 ft, which is within about 0.1 ft of the corrected 11.39 and looks like proof that the corrections do not matter. They cancelled at this site, at this temperature, in this pipe. Move the same pump to sea level and the barometric term flips from a 2.45 ft penalty to nothing, leaving only the three terms that add, so the same shortcut lands about 2.3 ft LOW rather than dead on. Move it to a 5,000 ft site instead and the barometric penalty grows to about 5.9 ft, and the same shortcut runs about 3.6 ft optimistic. Neither site is the one the cancellation happened at, and the two errors point opposite ways.

The failure mode, and it is the common one: read minus 6.0 psig, call it a vacuum of about 12 in Hg, decide that is normal for a lift, and go look at the discharge side. Nothing in that reading is wrong. It just was not converted into the units the pump's own curve is written in, so it was never compared to anything.

How to verify you got this right

Write down the flow, the liquid temperature, the tank level and the gauge's height above or below the datum on the same line as the number. A suction head figure with no stated conditions is a note, not a measurement, because every term in it was evaluated at a condition that moves.

Then run one sanity check that catches sign errors: predict the gauge reading from the physical layout before you look at it. A flooded suction should read positive at low flow and a lift should read negative, so a positive reading on a pump that is lifting means either the gauge is wrong or the line is not doing what the drawing says.

Finally, if the point estimate and the instrument bound land on opposite sides of the required value, as they did above, say so on the ticket in those words. "Cannot establish margin with this instrument" is a finding. "About 11 ft, looks fine" is a guess wearing a number.

References

• 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before adding a tapping or opening a pump
• 29 CFR 1910.134, respiratory protection, including fit testing before first use and at least annually under 1910.134(f)(2), where the SDS names a vapor or aerosol route
• ASME B40.100, in the edition the gauge's listing or your specification calls out, for the percent-of-span accuracy grades used above
• 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
• See related: How Suction Conditions Fail in Practice; Reading Pump Curves Reference; What Cavitation Is and How It Announces Itself