Pump Characteristic Curve

One pump, one speed, one diameter — one head per flow. How the H–Q curve is built, and what its shape tells you about stability and motor sizing.

The idea

A centrifugal pump at fixed speed and impeller diameter is a one-flow machine: for every flow it delivers exactly one head, draws exactly one power, and converts the difference with exactly one efficiency. Plot head against flow and you have the characteristic curve — the pump's fingerprint, the first thing a vendor publishes and the last word on what the pump will actually do once installed.

Three landmarks organize the curve. The shutoff point at zero flow, where the pump only churns liquid; the best efficiency point (BEP), where hydraulic losses bottom out; and the runout end, beyond which vendors stop publishing data — required NPSH climbs steeply and performance is undefined.

Curve shape is not arbitrary. It follows from the impeller's exit geometry, and it evolves through a family of canonical shapes as specific speed changes.

The equation

Strip away slip and losses and the ideal (Euler) head is a straight line that falls with flow:

$$H_{e}=\frac{U_{2}^{2}}{g}-\frac{U_{2}\,V_{m2}}{g\tan\beta_{2}}$$
SymbolMeaningUnits
\(H_e\)ideal (Euler) head — before slip and lossesft (m)
\(U_2\)peripheral speed at impeller OD, \(U_2=\pi D_2N/60\)ft/s (m/s)
\(V_{m2}\)meridional (radial) exit velocity, \(V_{m2}=Q/(\pi D_2 b_2)\) — proportional to \(Q\)ft/s (m/s)
\(\beta_2\)blade angle at exit from the tangential direction; backward-curved means \(\beta_2 \lt 90°\)°
\(g\)gravitational acceleration32.17 ft/s² (9.81 m/s²)

From the straight line to the real curve

Real pumps fall short of the line for three reasons. Finite blade count causes slip: the fluid leaves with less whirl \(V_{u2}\) than the blade angle implies, tilting the line down. Passage friction grows with \(Q^{2}\), sagging the curve further at high flow. Off-design incidence (shock loss) penalizes every flow except the one the inlet vanes were sized for. Near shutoff, recirculation loops liquid back through the impeller and adds churn. The net result is the familiar drooping curve: efficiency is zero at shutoff, peaks at BEP, and power follows one of the canonical shapes below.

The normalized family — shape follows specific speed

Normalize both axes with the BEP values,

$$q=\frac{Q}{Q_{BEP}},\qquad h=\frac{H}{H_{BEP}},\qquad p=\frac{P}{P_{BEP}}$$

and every pump collapses onto one of a handful of canonical curves. The classic seven-band family, in US units of \(N_s=N\sqrt{Q}/H^{3/4}\):

Curve\(N_s\) (US)\(n_q\) / \(\Omega_s\)Normalized H–Q shapePower behaviour
1≈ 90017 / 0.33rising toward shutoff — unstable drooppower rises steeply with flow
2≈ 150029 / 0.55slight rise toward shutoffpower rises with flow
3≈ 220043 / 0.80flatpower nearly flat
4≈ 300058 / 1.10continuously falling — stablepower nearly flat
5≈ 400077 / 1.46falling, steeperpower flattens toward shutoff
6≈ 5700110 / 2.09steeply fallingpower peaks near shutoff
7≈ 9200178 / 3.37very steep, with a low-flow saddlepower maximum at shutoff

Curves 1–2 are the rising (unstable) family, 3 is flat, 4–7 are the steeply falling (stable) family. Curves 1–5 are typical of double-suction impellers, 6–7 of single-suction wheels.

Engineering criteria

SituationWhat it meansVerdict
Shutoff head ≥ 1.15–1.20 × BEP head, no hump anywherecontinuously rising to shutoff — stable at any duty and safe for parallel operation🟢
Shutoff head 1.0–1.15 × BEP head, still monotonicrising but shallow — verify static and dynamic stability against the actual system curve🟡
Flat or drooping segment (hump, shutoff head below peak)operate only to the right of the hump; never run two such pumps in parallel on the drooping branch🟡
Duty point in a positive-slope regionflow hunts, and can surge against large gas cushions🔴
Pump curve slope ≥ system curve slope at the crossingstatically unstable; humped curves make twin intersections and the low-flow one always collapses or runs away🔴
System holds a large gas volume or free surface, curve flat or risingstatic stability is not enough — dynamic surge is possible even on a zero-slope curve🟡
\(N_s \gtrsim 5000\), single-suctionshutoff power exceeds BEP power — size the motor for shutoff torque and start against an open or bypassed valve🟡
\(N_s \lesssim 5000\)shutoff power below BEP power — closed-valve start is permissible and the driver can be sized for normal duty🟢
Operation beyond runout (published curve end)NPSHr and vibration escalate steeply; performance is undefined🔴
Continuous operation below MCSF (minimum continuous stable flow)suction and discharge recirculation: low-frequency pulsation, cavitation, fatigue — forbidden for continuous duty🔴
Pressure pulsation peak-to-peak reaches the stage static pressure risesignature of deep low-flow damage — check for impeller-OD pitting and shroud bulging🔴
Wearing-ring clearance ≥ 2 × as-new valuereplace or restore the rings; with abrasive particles the gap grows fast and BEP efficiency bleeds away🔴
Semi-open impeller tip clearance > 0.060 in (1.5 mm)efficiency degrades markedly; tightening from 0.060 to 0.010 in can recover up to 10 points of efficiency🟡
Impeller trim beyond 10–20 % of diameterrarely done; blade overlap is lost and the affinity-style prediction fails — confirm by test🔴
Impeller mounted backwards (vanes forward-curved)BEP efficiency collapses to roughly 47–71 % of normal while power can rise up to 1.8× — a field-diagnosis classic🔴
Blade/vane count combination with multiple difference m = 0 or m = 1all blades pass the cutwater in phase — torque pulsation (m = 0) or acoustic resonance (m = 1); check to at least the 3rd order🔴
Gap B (impeller–diffuser radial gap) < 4 % of impeller radiuspressure-field interaction — blade, vane and cutwater vibration with structural risk; normal band is 4–15 %, high-energy pumps take the upper end🔴
Volute discharge cone angle > 7° or exit area > 2 × throat areadiffusion losses grow — the low-loss volute design window is exceeded🟡

Rules of thumb — quick estimates

NameRule / formulaApplies to
Shutoff head coefficient\(\psi_{s/o}=gH_{s/o}/U_2^{2}\approx0.585\); typical radial pumps \(\psi_{s/o}\gt0.5\)radial impellers at \(Q=0\)
Shutoff power ratio\(P_{s/o}\approx0.44\,P_{BEP}\); grows with \(N_s\) and with \(b_2/D_2\)radial pumps
Shutoff power coefficient\(\hat P_{s/o}=P_{s/o}/(\rho\omega^{3}r_2^{5})=f(b_2/D_2)\); \(b_2/D_2=0.1325\rightarrow\hat P\approx0.047\)empirical chart estimate
Stability shutoff risedesign for \(H_{s/o}\ge(1.15\text{–}1.20)\,H_{BEP}\)conservative stable-curve design
Slip factor\(\mu=V_s/U_2\approx0.1\text{–}0.2\)conventional closed impellers
Collector pressure recoverystatic rise in volute/diffuser ≈ 1/3 of the impeller's own static rise; collector recovers 20–25 % of the stage totalconventional geometry
Impeller OD pressure\(p_{OD}\approx75\text{–}80\,\%\) of the stage pressure rise above inletaxial-thrust estimates
BEP vs shockless flow\(Q_{BEP}\approx0.9\,Q_{SE}\) — blade blockage pulls BEP about 10 % below the shockless-entry flowinlet design checks
Radial-vane curve shape\(\beta_2=90°\): H–Q flat from shutoff to ≈ 75 % of \(Q_{BEP}\), steep beyond; single-stage heads to ≈ 8000 ft (2400 m)small high-speed radial-vane pumps, \(D\lt6\) in, up to 30,000+ rpm
Volute throat velocity\(r_T V_T\approx(0.9\text{–}0.95)\,r_2 V_{u2}\); radial diffuser throat \(\approx0.8\,r_2 V_{u2}\)BEP collector sizing
Volute discharge coneexpansion angle 7°; exit area ≤ 2 × throat arealow-loss volute outlet
Blade blockagevane thickness ≈ 2 % of \(D_2\) (halved near the leading edge); exit passage openness ≈ 85 % (up to 90 % on large wheels)closed impellers
Inlet incidencepositive incidence 2–3° at the rms radius; shroud blade angle ≈ 1° below the flow angleinlet vane design
Inlet decelerationmean relative velocity \(W\) drops ≥ 10 % entering the blade row — minimizes incidence + friction loss at BEP and lowers NPSHr above BEPradial impellers
Speed-versus-efficiency gain15,000 → 30,000 rpm: ≈ +15 efficiency points; 1240 → 1880 rpm on a small pump: ≈ +1 pointhigh-speed machine trade-off

Reference data tables

Effect of wearing-ring clearance growth on performance

Values are percentages of the as-new BEP performance at the stated ring clearance (percent of new):

\(N_s\) (\(\Omega_s\))Design head ft (m)Clearance, % of new\(Q\) %\(H\) %\(P\) %\(\eta\) %Shutoff \(H\) %Shutoff \(P\) %
2100 (0.77)63 (19.2)17810098.398.999.497.0100
2100 (0.77)63 (19.2)35610097.599.098.593.698.2
2100 (0.77)63 (19.2)68810096.098.997.191.294.8
2100 (0.77)63 (19.2)137510094.397.496.888.892.5
3500 (1.28)65 (19.8)35410090.099.190.885.096.2
4300 (1.57)41 (12.5)72706265.581.749.844.3106
4800 (1.76)26 (7.9)52209678.889.284.878.283.3

The higher the specific speed, the faster clearance growth eats the curve — at \(N_s\approx4300\) a worn ring can cost half the BEP efficiency.

Impeller mounted backwards — measured penalties

Six pumps re-tested with the impeller reversed (vanes effectively forward-curved); values are percentages of normal BEP performance:

Stages\(N_s\) (\(\Omega_s\))Shutoff head %Head %Flow %Power %Efficiency %
2828 (0.303)861116510471
21024 (0.375)821128814568
11240 (0.454)7510538.568.559
11430 (0.523)8210669.713853.5
12570 (0.940)74.51176213852.5
12740 (1.003)77.513861.518047

Common mistakes

⭐ Deep-data appendix

Field-grade numbers behind the rules above: minimum-flow limits by service, starting and standstill criteria, the bookkeeping formulas for torque, acceleration time and temperature rise, plus measured trim-correction and vane-underfiling data.

Additional engineering criteria

SituationWhat it meansVerdict
Minimum flow vs suction-recirculation flow \(Q_{SR}\)water service, ≤ 2,500 gpm and ≤ 150 ft: \(Q_{min}\ge50\,\%\,Q_{SR}\) continuous / 25 % intermittent; hydrocarbons: 60 % / 25 %; the higher the energy level, the closer \(Q_{min}\) must stay to \(Q_{SR}\)🟡
Pumps in series, second-stage stuffing boxthe box sees the first stage's discharge pressure — a special high-pressure box or leak-back to the first-stage suction is mandatory🔴
Hot-liquid pump start-upwarm the pump to working temperature first (unless designed for fast start); never run a hot pump at shutoff — a metered-orifice bypass is required🔴
Restart into backspin (discharge valve left open, no check valve)the pump may be reverse-running at runaway speed; starting now means a long driver overload and a protection trip — confirm zero reverse rotation first🔴
Synchronous-motor drivestarting torque between 90–100 % of rated speed must stay below the pull-in torque; the motor must pull the train into step from ≈ 95 % speed within ≈ 0.2 s (watch siphon-discharge propeller pumps)🟡
Pumps in parallel at reduced combined flowthe operating split must never force any single unit below its own anti-recirculation minimum flow🔴
Cavitation surge (1–6 Hz oscillation)a recirculation–vapor-lock cycle at low flow; avoided by keeping the operating flow above \(Q_{min}\) — rare on low-\(N_{ss}\) impellers🟡
Double-suction single-stage pump at low flowaxial thrust at part load can become the factor that sets the minimum flow🟡

Additional rules of thumb

NameRule / formulaApplies to
Shaft torque\(M=5252\,P/n\) lb·ft (\(P\) in hp); \(M=9549\,P/n\) N·m (\(P\) in kW); add ≈ 10 % when acceleration torque mattersdriver sizing
Water horsepower\(\text{lhp}=Q\cdot H\cdot SG/3960\) (gpm, ft); \(P\,[\text{kW}]=9.797\,Q\cdot H\cdot SG\) (m³/s, m)hydraulic power
Acceleration time\(\Delta t=I\,\Delta n/[k\,(M_m-M)]\); \(k=307\) (US: lb·ft², lb·ft) / 9.549 (SI: kg·m², N·m); requires \(M_m-M\) ≈ constant over the intervalstart-up studies
Breakaway torque≈ 15 % of rated with sleeve bearings + packing, ≈ 10 % with rolling bearings; falls roughly linearly to near zero at 15–20 % of rated speedstarting-torque curves
Parallel combined efficiency\(\eta=(H\cdot SG/k)\cdot(\sum Q/\sum P)\); \(k=3960\) (US: gpm, hp) / 0.1021 (SI: L/s, W)parallel operation
Series combined efficiency\(\eta=(Q\cdot SG/k)\cdot(\sum H/\sum P)\); same \(k\) as aboveseries operation
Pump temperature rise\(\Delta T=gH(1-\eta)/(g_oC_pJ)+\Delta T_c\); US: \(g/g_o=1\), \(J=778\) ft·lbf/Btu; SI: \(J=1\)minimum-flow / bypass design
Compression heating \(\Delta T_c\)hydrocarbon fuels ≈ 3 °F/1000 psi (0.24 °C/MPa); 350 °F feedwater ≈ 1.6 °F/1000 psi (0.129 °C/MPa); cold water negligible — subtract it when inferring efficiency from temperature risethermodynamic efficiency tests
Temperature-rise limitsΔT ≤ 100 °F (56 °C) general-purpose cold-liquid pumps; ≤ 50 °F (28 °C) modern boiler feed pumps — beyond that a minimum-flow bypass is mandatorylow-flow protection
Best-efficiency specific-speed bandhighest BEP efficiency at \(N_s\approx2000\text{–}3000\) (\(\Omega_s\approx0.73\text{–}1.10\)); total-loss minimum near \(N_s\approx2500\) (\(\Omega_s\approx0.91\))double-suction single-stage, ≥ 12 in discharge
Exit-vane underfilingfiling the back of the exit vanes adds up to ≈ 10 % flow and often improves peak efficiency; overfiling changes almost nothingcapacity tuning
Two half-size pumps, one motortwo pumps in parallel on one driver can run ≈ 40 % faster than one double-capacity pump — the cheaper high-speed motor offsets much of the second pumpdriver-cost trade-offs

Impeller trim — empirical corrections vs the bare ratio

Worked corrections for a radial impeller starting at \(D=16\) in: the target diameter from the bare \(Q\propto D,\ H\propto D^{2}\) ratio is corrected upward by the empirical trim-correction chart, and an independent chart (\(k\approx0.6\) at \(N_s\approx1950\)) lands between the two. Prediction error against test grows with the depth of the cut (BEP flow error −2.9 % at the first trim to +7.3 % at the deepest):

Target \(D'\)Target \(D'/D\)Corrected \(D'/D\)Corrected \(D'\)Independent chart \(D'\)
15.125 in (384 mm)0.9270.93515.25 in (387 mm)15.60 in (396 mm)
14.000 in (356 mm)0.8580.87414.26 in (362 mm)14.93 in (379 mm)

Mixed-flow wheels behave far worse: trimming \(D\) 16 → 15⅝ → 15 in (mean diameter \(D_m=\sqrt{(D_o^2+D_i^2)/2}\): 13.17 → 12.89 → 12.62 in) produced measured flows off the prediction by −11.3 % to +17.3 %, and the second cut removed most of the blade overlap — \(N_s\cdot D_m\) is not conserved on mixed-flow impellers.

Exit-vane underfiling — measured results on nine pumps

Changes relative to the pre-filing test, after filing the back (low-pressure) side of the exit vanes (\(d_F/d\) = diameter ratio after/before filing):

\(N_s\) (\(\Omega_s\))Stages\(D_2\) in (cm)\(d_F/d\)Shutoff head Δ%BEP head Δ%BEP flow Δ%BEP power Δ%BEP η Δ pts
862 (0.315)111.5 (29.2)1.13+4+1.5+40+9 / −5*
945 (0.346)28⅜ (21.3)1.05+5.50+3+4−0.5 / +7.5*
1000 (0.366)49⅛ (23.2)1.055+5.50+5.50+3 / +2.5*
1080 (0.395)29⅞ (24.9)1.08+10+2.5+100+6.8 / +3*
1525 (0.558)210½ (26.7)1.035+3.2+3+0.5+4.5+3.5 / +1.2*
1950 (0.714)122 (55.9)1.02+1.5+1+1.50+1.5 / 0*
3300 (1.208) dbl.112 (30.5)+7.8+2.5+7.80+8.5 / −0.6*
3450 (1.262) dbl.130½ (77.5)+6.5+0.5+6.50+7.8 / −2.2*
4300 (1.573)141¼ (104.8)+5+50+0.5+4.5 / —*

*Where two values appear, the pump was tested at two speeds. The consistent pattern: flow and shutoff head gain a few percent, efficiency usually improves — and the gains fade as \(N_s\) rises past about 3000.

Cavitation and suction-side criteria — extended

SituationWhat it meansVerdict
\(NPSH_A \ge NPSH_i\) (inception NPSH)no cavitation at all — cavity length is zero; inception sits far above the 3 % point, typically \(NPSH_i \approx 2\text{–}5 \times NPSH_{3\%}\) with most observations near 5×🟢
\(NPSH_{3\%} \lt NPSH_A \lt NPSH_i\)extensive cavities already exist even though head has not dropped 3 % — erosion and pulsation are already at work between the two limits🟡
\(NPSH_A \le NPSH_{3\%}\)head drop begins — performance is no longer guaranteed🔴
Suction specific speed \(S = N\sqrt{Q}/h_{sv}^{3/4}\) (double-suction: half the total flow)\(S \le 6000\) (\(\Omega_{ss} \le 2.2\)) is the conservative band for large machines; 6,000–8,500 normal service; the standard recommendation is 8,500 and most pumps fall in 7,500–11,000 (\(\Omega_{ss}\) 2.7–4.0); above 11,000 (\(\Omega_{ss} \gt 4\)) only inducers or special designs belong🟢 ≤ 6000 / 🟡 mid / 🔴 > 11,000
High-energy pump (high inlet-tip speed, high head per stage)the 3 % point says almost nothing about damage — specify the damage NPSH that guarantees 40,000 h of impeller life (\(NPSH_{40,000h}\)), or "no / minimal visible cavity" on an inspection test; life ends when erosion has eaten 75 % of the blade thickness: \(\text{Life} = 0.75\,t/\text{MDPR}\), \(t\) = blade thickness (mm), MDPR = mean penetration rate (mm/yr)🟡
Entrained gas at the impeller eye (GVF)commercial industrial pumps tolerate only a gas volume fraction \(\lesssim 0.03\); beyond it head, flow and power sag until suction is lost (only special aerospace designs go higher)🟢 ≤ 0.03 / 🔴 above
Inducer + impeller running above rated flowthe combination's \(NPSH_r\) climbs steeply past the rated flow — constant-pitch inducers must not be run oversize on flow; variable-pitch inducers relax the limit🔴
NPSH reduction credited for hot water / hydrocarbonscredit at most \(\min(50\,\%\times\text{cold-water }NPSH_r,\; 10\text{ ft} / 3.0\text{ m})\); forbidden outright when dissolved gas can come out of solution at low suction pressure, when the system sees strong transient pressure/temperature swings, or for liquids outside the tested reduction chart — add margin instead🟢 within cap / 🔴 if any disqualifier
Suction piping layouttwo 90° elbows set at 90° to each other seed bad inlet flow and cavitation damage; 3–4 radial anti-swirl fins extending ≈ ¼ of the inlet diameter suppress preswirl at far less NPSH cost than full-length vanes🔴 bad layout / 🟢 fins
High-Δp labyrinth wearing ringcan excite large-amplitude, long-period shaft vibration; enlarging the second (inboard) clearance relative to the first calms it — at the price of more leakage🟡

Cavitation physics — quick numbers

NameRule / formulaApplies to
Sixth-power erosion lawerosion rate \(\propto U_{t,1}^{\,6}\) (inlet blade-tip speed) — equivalently \(\propto NPSH^{3}\) at a fixed cavitation number; the higher the head per stage, the faster the damagerisk screening
Bubble lifecyclelife cycle ≈ 0.003 s; collapse pressure of order 10⁴ atm — no common material survives indefinite exposurewhy cavitation destroys metal
Water-temperature peakcavitation damage in water peaks at 100–120 °F (38–49 °C)controlled tests
Inception vs 3 % point\(NPSH_i \approx 2\text{–}5 \times NPSH_{3\%}\); the zero-head-drop band is very wide, so the 3 % figure is a performance convention, not a no-damage guaranteeacceptance testing
Typical installed \(NPSH_A\)≈ 60 % of barometric head ≈ 20 ft (6 m) in conventional installations — the basis for condensate-pump speed ceilingscondensate service
Pulsation peakcavitation pressure-pulsation amplitude peaks near \(R \approx 2\) (\(R = NPSH_A/NPSH_{3\%}\)); the frequency rises as \(R\) growsvibration diagnostics
Inducer benefitruns at ≈ 2 × the suction specific speed of a plain impeller; inducer + impeller needs only ≈ 50 % of the impeller-alone \(NPSH_r\) (up to rated flow); the inducer contributes ≤ 5 % of the total headhigh-\(S\) retrofits
Backflow recirculatora passive recirculating-vane device removes inducer cavitation instability from shutoff clear out to runoutunstable-suction cures
Inception cavitation coefficient\(\tau_i \approx 1\) at BEP for conventional blades, ≈ 0.5 for aerodynamically optimized blades; \(\tau_i\) rises at \(Q \lt Q_{BEP}\); \(NPSH_A \ge NPSH_i\) (i.e. \(\tau_A \le \tau_i\)) means zero cavity lengthdesign targets
Double-suction conventioncompute \(S\) with half the total flow (one eye)\(S\) bookkeeping

Material ranking against cavitation erosion

Mass loss in a 2 h magnetostrictive vibration test — the relative ranking of common pump materials (lower = more resistant):

MaterialMass loss, mg / 2 hRelative resistance
Rolled stellite0.6best (costly, hard to machine)
Welded aluminum bronze3.2excellent
Cast aluminum bronze5.8excellent
Welded stainless (17Cr-7Ni, 2 layers)6.0excellent
Hot-rolled stainless (26Cr-13Ni)8.0very good
Tempered rolled stainless (12Cr)9.0very good
Cast stainless (18Cr-8Ni)13.0good
Cast stainless (12Cr)20.0good
Cast manganese bronze80.0poor
Welded mild steel97.0poor
Steel plate98.0poor
Cast steel105.0poor
Aluminum124.0very poor
Brass156.0very poor
Cast iron224.0worst

Read the ranking in orders of magnitude: each step down multiplies the erosion life roughly tenfold. Combine it with the 40,000-hour damage criterion when choosing material for a high-energy stage.

NPSH reduction for hot water and hydrocarbons — worked cases

Walking the reduction chart for a pump whose cold-water \(NPSH_r\) is 16 ft (4.9 m); the credit is capped at half the cold-water value, i.e. 8 ft (2.4 m):

Liquid and temperatureVapor pressureChart reductionCreditedResulting \(NPSH_r\)
Water at 100 °F (38 °C) — chart-walking demo30 psi abs (207 kPa)≈ 2.3 ft (0.70 m)2.3 ft (0.70 m)
Propane at 55 °F (13 °C)≈ 105 psi abs (724 kPa)≈ 9.5 ft (2.9 m) — above the 8 ft cap8 ft (2.4 m)16 − 8 = 8 ft (2.4 m)
Propane at 14 °F (−10 °C)≈ 50 psi abs (345 kPa)≈ 5.7 ft (1.7 m) — below the cap5.7 ft (1.7 m)16 − 5.7 ≈ 10 ft (3.0 m)

How cavitation grows — head-drop anchor points

Cavitation coefficient \(\tau = 2g\cdot NPSH/U_e^{2}\) (\(U_e\) = eye-tip speed), walked down on a boiler-feed test stage; \(\tau_{available} = 0.41\):

\(\tau\)Cavity development stage
0.61near inception (\(\tau_i\))
0.39cavity lengthening along the blade suction side
0.29cavity closing on the adjacent-blade throat
0.203 % head drop (head = 97 % of the cavitation-free value) — the cavity has reached the throat between adjacent blades

High-energy pump suction stage — two eye designs

Same duty, two designs of eye flow coefficient \(f_e = V_e/U_e\); on the damage side the comparison reverses the 3 % comparison:

ParameterSmall eye, \(f_e = 0.30\)Large eye, \(f_e = 0.25\)
Eye diameter13.37 in (340 mm)13.92 in (354 mm)
Inlet-tip speed \(U_{t,1}\)274 ft/s (83.5 m/s)286 ft/s (87.2 m/s)
\(NPSH_{3\%}\)slightly higherslightly lower
Damage NPSH ratio \(R\) (40,000 h basis)1.692.06

For high-energy stages the small eye wins: the 3 % penalty is marginal, while the margin to the 40,000-hour damage NPSH is materially better (\(R = 1.69\) vs 2.06).

Related