Why the Suction Side Decides Whether a Pump Works
Why this matters
Two shops call the same complaint in the same week: the pump will not make flow. One of them replaces the pump and the new one makes flow. The other replaces the pump and the new one does exactly what the old one did, and now there is a second pump on the same ticket and no explanation. The difference between those two jobs was decided before anyone touched a wrench, by a single gate that costs three readings and one curve lookup. The gate does not tell you what is wrong. It tells you which end of the machine to work on, and that is the decision the money and the callback both hang on.
Before you read a discharge tapping
Discharge pressure is the pump's full developed head, and a tapping opened at it sprays. Fit a gauge only through a tapping that has its own isolation valve, open the valve slowly with the gauge already made up, and do not crack a plug on a line in service. If the liquid is above scald temperature, wear a face shield rather than glasses and stand out of the plane of the joint.
Do not throttle the suction valve to see what happens. Closing a suction valve on a running pump drives it straight into cavitation and, if it is closed far enough, into a condition where the pump is doing shaft work on a liquid it cannot move; the liquid in the casing heats fast, and on hot service it can flash the moment anything is opened. If the suction side has to be varied, it is varied by the process, not by your hand on the valve.
Do not close the discharge valve to read shutoff head unless you already know how long that pump is permitted to run there. A centrifugal at shutoff puts nearly all of its shaft power into the small volume of liquid in the casing, and the temperature rise is measured in tens of degrees per minute rather than per hour. Everything below is read with the pump in normal service.
The asymmetry, stated once
A centrifugal pump is a discharge-side machine. Its impeller adds energy to liquid it already has, and the curve describes how much head it will add at each flow. Everything the pump was bought to do lives on the discharge side.
The suction side does not add anything. Its only job is to keep the liquid at the impeller eye above the pressure at which it boils, and it does that job or it does not. There is no partial credit and no gradual version: below the required condition, the liquid flashes at the eye, the vapor collapses inside the impeller, and the pump stops behaving like the curve says it should. So the discharge side sets what the pump is asked to do, and the suction side sets whether it can do anything at all.
That asymmetry is why almost every field complaint that presents as insufficient flow needs to be split at the suction side first, even though the symptom, the customer's description, and the pressure gauge everyone looks at are all on the discharge.
The gate: three readings and one curve lookup
- Suction pressure, at the inlet tapping, with the pump running in normal service.
- Discharge pressure, at the outlet tapping, at the same moment, with the same gauge moved between the two.
- Flow, measured. A clamp-on ultrasonic meter, a plant flow meter you have reason to trust, or a timed catch on a drain. A number off a control screen is often a calculated value derived from valve position or pump speed, and in that case it is the thing you are trying to test, not evidence about it.
Convert the difference between the two pressures into feet of the liquid at the temperature it is actually at (psi times 2.31, divided by specific gravity at that temperature). That is the pump's developed head. Take it to the pump's own curve for the impeller diameter and speed the pump actually has, and read off the flow the curve says goes with that head.
Then compare that curve flow against the flow you measured:
- Curve flow and measured flow agree. The pump is on its curve and is doing exactly what it was built to do. The system is asking for more head than anyone expected, and the work is on the discharge side.
- Curve flow is much higher than measured flow. The pump is producing less flow than its own head implies, which means it is not on its curve. Now, and only now, is the suction side worth measuring.
Why one gauge, moved, beats two gauges
The number the gate uses is a difference, and that changes which parts of the instrument's error survive.
A gauge's fixed zero offset is a systematic error. Move one gauge between the two tappings and that offset appears identically in both readings, so it cancels exactly in the subtraction. If the gauge reads 0.4 psi high, it reads 0.4 psi high at both ends, and the difference is untouched. What does not cancel is the proportional term, and that leaves you that percentage of the difference rather than of either reading, which is a much smaller number.
Use two gauges and you have two independent errors instead of one shared one. If each is specified as a worst-case bound, worst-case bounds add linearly, so two 0.75 psi bounds give 1.5 psi on the difference. If instead you have measured each gauge's repeatability as a random spread, independent spreads combine in quadrature, so two equal terms multiply by the square root of two and 0.75 becomes about 1.06 psi. Either way it is worse than one gauge moved, and it is worse for a reason you can state on the ticket.
Pump A: high differential, low flow
Cold water, 80 F, specific gravity 0.997, so 2.317 ft per psi. One gauge, whose bench check showed a fixed 0.4 psi high offset.
| Reading | Value |
|---|---|
| Suction, at the inlet tapping | +2.0 psig |
| Discharge, same gauge, same moment | +38.0 psig |
| Fixed 0.4 psi offset, present in both | cancels in the difference |
| Differential | 36.0 psi |
| Converted at 2.317 ft per psi, not 2.31 | 83.4 ft |
| Proportional term, 1 percent of the difference | differential known to within 0.8 ft |
| Curve flow at 83.4 ft, this trim and speed | about 110 gpm |
| Flow, measured with a clamp-on meter | 105 gpm |
The pump is on its curve. It is making 83.4 ft of head and delivering the flow the curve says goes with 83.4 ft of head. Nothing is wrong with the pump. Something on the discharge side is making the system demand 83 ft where it was designed to demand less, and the pump has simply ridden up its own curve to meet it: a coil fouled, a balancing valve closed in, a filter loaded, a check valve half stuck.
Replacing this pump with a bigger one buys a small increase in flow at a much larger increase in head, which is a separate article's subject. The point here is that the gate closed at the second reading and the suction side never needed to be measured.
Pump B: low differential, low flow
Hot water, 190 F, specific gravity 0.966, so 2.391 ft per psi. Vapor pressure at 190 F is 9.34 psia. Station barometric 14.3 psia at this 700 ft site. Same single gauge. Suction line is 4 in schedule 40, and the suction tapping is 0.5 ft above the shaft centerline.
| Reading | Value |
|---|---|
| Suction | -2.0 psig |
| Discharge, same gauge | +21.0 psig |
| Differential | 23.0 psi |
| Converted at 2.391 ft per psi | 55.0 ft |
| Curve flow at 55.0 ft | about 225 gpm |
| Flow, measured | 95 gpm |
The pump is producing about 40 percent of the flow its own developed head implies. It is not on its curve, so the second branch of the gate applies and the suction side gets measured. Working the inlet reading the way the sibling card sets out:
| Line | Value |
|---|---|
| Suction gauge, static | -2.0 psig |
| Absolute, using station barometric 14.3, not 14.7 | 12.30 psia |
| Minus vapor pressure at the measured 190 F | 12.30 - 9.34 = 2.96 psi |
| To feet at 0.966, so 2.391 ft per psi | 7.08 ft |
| Plus velocity head, 2.39 ft/s at the measured 95 gpm | +0.09 ft |
| Plus datum, gauge 0.5 ft above the eye | +0.50 ft |
| Available | 7.67 ft |
| Required, off the curve at the flow this pump should be at | about 14 ft |
Available 7.67 ft against roughly 14 ft required is not marginal, it is well past the documented onset. Note which flow the required value was read at: at the 95 gpm it is actually managing, required would be much lower, and the pump would appear to pass. That is the trap in a cavitating pump, because cavitation is what dropped the flow in the first place. The required value has to be read at the flow the pump is supposed to be delivering, or the fault hides behind its own symptom.
The 190 F liquid is doing most of it. The absolute pressure at the inlet, 12.30 psia, is worth 29.4 ft of this liquid; vapor pressure at 190 F takes 22.3 ft of that back before the pump sees anything, which is 76 percent of it, and the datum and velocity terms give back 0.6 ft. Everything else in the installation is fighting over the remaining 7 ft.
How to verify you got this right
Re-run the gate after the repair and require both halves to move in the directions you predicted. On pump A, opening the restriction should move the differential down and the flow up, because the pump slides back down its own curve. If differential and flow both fall, you did not open a restriction, you introduced a suction problem. On pump B, restoring suction margin should move the differential up and the flow up together, because the pump returns to its curve.
Then check one thing that is easy to get backwards: motor current. A centrifugal that has been starved and is now pumping properly draws more current, not less, because it is moving more liquid. A tech who expects a repaired pump to run quieter and lighter has the direction wrong, and will read a rising amp draw after a correct repair as a new fault.
References
- 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before any work that opens the pump or its piping
- 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
- Pump manufacturer's performance curve for the installed impeller diameter and speed, which owns both the head-flow relationship and the required suction head
- See related: How to Work Out NPSH Available at the Pump You Are Standing At; Reading Pump Curves Reference; Why a Bigger Pump Does Not Fix a Restriction