💧 Fluid Flow & Pump Sizing

HeadLoss

Darcy-Weisbach Friction Loss, Total Dynamic Head & Pump Operating Point

Every pipe friction calculator asks you for a flow rate and hands back a head loss. That is the easy half, and it answers a question nobody has. The question people actually arrive with runs the other way: I have this pump, this pipe and these fittings — what flow do I actually get? Head loss rises with flow, and a pump's head falls with flow, so there is exactly one flow where the two agree. That crossing is the operating point, and it is the number your pond turns over at, your sprinkler zone runs at, and your sump actually moves. Enter three points off your pump's curve and this tool finds it.

Your system

Describe the run
What you know
Units
Fitting method

 

 

The flow

If you already know the flow, this returns the head your pump has to produce to sustain it.

Fittings and valves

Each fitting costs some multiple of one velocity head. Under Crane's method that multiple is not a fixed number — it depends on the pipe size the fitting sits in, which is why the K beside each row moves when you change the pipe.

Velocity
--
Enter a flow
Reynolds number
--
 
Friction factor f
--
 
Pipe friction
--
 
Fitting losses
--
 
Total dynamic head
--
 
Hazen-Williams check
--
 
Turnover
--
Enter a system volume

The run, and where the curves cross

Above: the pipe in elevation, with the hydraulic grade line falling across it. The steeper that line, the more head the pipe is eating. Below: the system curve — the head your pipe demands at every flow — against the pump curve fitted through your three points. Where they cross is what you actually get.

Hydraulic grade line System curve Pump curve Fitting

Where the head goes

The whole point of a total dynamic head figure is that it decomposes. If one line here dominates, that is the line worth attacking.

Component Basis Head Share

Enter a flow to see the cross-check.

Why the answer is a crossing, not a number

A pipe does not have a flow rate. It has a relationship between flow and the head needed to sustain it, and that relationship rises steeply — friction goes roughly as the square of velocity, so doubling the flow roughly quadruples the loss. A pump does not have a flow rate either. It has a curve that starts at its shutoff head with the discharge closed and falls away as you open it up. Neither one determines the answer alone.

Put both on the same axes and there is exactly one flow where the head the pipe demands equals the head the pump supplies. Everywhere left of it the pump is producing more head than the pipe needs, so the flow accelerates. Everywhere right of it the pipe demands more than the pump has, so the flow falls back. The system settles at the crossing whether you calculate it or not.

Darcy-Weisbach — NRCS NEH-634 Ch.4 eq. 4-13

hf = f × (L/D) × V² / 2g

Reynolds number — 634.0403.C

Re = V D / ν

Total dynamic head — NRCS NEH-15 Ch.8

TDH = static lift + hf + ΣK·V²/2g + V²/2g + pressure head

Why Darcy-Weisbach and not Hazen-Williams

Hazen-Williams is easier and more familiar, and this page computes it too — as a cross-check, never as the answer. The reason is that it carries no viscosity term at all. Its constant was calibrated for water at about 60 °F and the temperature dependence is baked into the number, so it cannot know that the same pond loop is a different hydraulic problem in April than in August. Darcy-Weisbach carries viscosity explicitly, through the Reynolds number, which is why it works for any fluid at any temperature.

It also has published limits that are narrower than most people realise. Diskin's 1960 analysis found Hazen-Williams applicable only in part of the turbulent transition zone, only for C between 100 and 160, and only inside a Reynolds window that is bounded above as well as below and that shifts with relative roughness. The familiar "valid above Re 105" rule of thumb is not what the source says. This page checks your case against Diskin's own table and tells you when the cross-check does not apply rather than printing a number that looks authoritative.

The critical zone, where nobody has an answer

Between Reynolds 2,000 and 4,000 the flow is neither reliably laminar nor reliably turbulent, and the friction factor genuinely is not a well-defined function of Reynolds number and roughness. It depends on the disturbance history of the flow. Reynolds himself found the transition could be pushed from about 2,000 up to 13,000 by keeping the inlet quiet enough.

Calculators handle this in one of three ways: silently extrapolating the turbulent equation (wrong, and wrong by a lot — at Re 2,000 it runs 60% high), refusing to answer, or interpolating. This one interpolates using the published cubic from EPANET, the EPA's water distribution model, which is the closest thing to a reference implementation in this field. It joins continuously to 64/Re at 2,000 and to the turbulent equation at 4,000. It is also not monotonic — the interpolated factor dips before it rises — and that is a real property of the published algorithm, not a bug here. Whatever the method, a result in this band deserves the warning it gets.

What Crane actually says about fittings

Nearly every fitting table on the internet gives one number per fitting: a 90° elbow is 0.75, a gate valve is 0.17, and so on. Crane's Technical Paper 410, which is where most of those numbers ultimately come from, does not say that. Crane publishes an equivalent length ratio L/D and defines the resistance coefficient as:

Crane TP-410 — the two-friction-factor method

K = fT × (L/D)

fT = 0.25 / [ log10(3.7 D / ε) ]²   with ε = 0.0018 in

fT is the friction factor for that pipe size in fully turbulent flow, evaluated for clean commercial steel. Because the roughness is fixed by that assumption, fT becomes a function of pipe size alone — and it falls as the pipe gets bigger. The consequence is that the same 90° elbow is K ≈ 0.78 in half-inch pipe and K ≈ 0.49 in four-inch, a spread of more than 1.5 to 1. A single-number K table is a small-pipe approximation, and this page shows the size dependence rather than hiding it.

Note also what Crane does not scale this way. Entrances, exits, sudden contractions and enlargements are geometric losses with bare K values — an exit is one velocity head whatever the pipe size, always. Multiplying those by fT is a common and silent error.

Two other published methods are offered above, and they disagree with Crane by enough to matter: for a 2-inch flanged 90° elbow at Re 105, Crane gives K = 0.57, Hooper's 2-K gives 0.38 and Darby's 3-K gives 0.39 — Crane runs about 45% higher. Peer-reviewed comparison work judges the 3-K method the most accurate of the three for elbows. None of them is "the" answer; what would be wrong is silently blending them, so switching methods here switches the whole calculation and says so.

Where these numbers come from

The friction equations, the flow-regime thresholds and the roughness table are from the USDA-NRCS National Engineering Handbook, Title 210 Part 634 Chapter 4 (2022) — a US government engineering handbook, chosen over textbook reproductions because it is fetchable, citable and prints its own worked examples. Fitting resistance is Crane TP-410. Inside diameters are computed from the published outside diameter and wall thickness in the ASTM and ASME dimensional standards rather than copied from a supplier's friction chart, so the arithmetic is visible.

Two things this page will not tell you

An accuracy figure for the Swamee-Jain equation. The explicit friction-factor approximation used above turbulent Re 4,000 is widely quoted as being "within 1%" of the implicit Colebrook equation. That figure appears only on aggregator sites, and direct computation contradicts it — peer-reviewed comparison over a million-node grid finds a worst case of −3.36%, and an independent check agrees at about 3%, worst at low Reynolds and high roughness. Rather than print a tolerance, this page solves Colebrook iteratively as well and shows you both numbers and the gap between them for your actual case.

Crane's tabulated fT. Crane prints an fT-by-size table on appendix page A-26. That page could not be verified, and the three transcriptions of it in circulation disagree — precisely in the small sizes a homeowner uses. So fT is computed here from the fully-rough equation Crane itself prints, and labelled as computed rather than quoted. Against the widely-copied two-figure table it is exact at 2 inches and runs up to about 4% low elsewhere, so the two are close but not interchangeable. Where a fitting dominates your total — many valves on short pipe — that difference is worth knowing about.

What this tool is for, and what it is not

Water only, and specifically: irrigation mains, shop and yard supply lines, rainwater and greywater transfer, pond and water-feature loops, pool circulation plumbing, and aquarium sump returns. Every equation on this page carries the properties of liquid water and nothing else.

It is not for gas or compressed-air piping, steam, fire-protection sprinkler hydraulics, plumbing-code fixture-unit sizing, backflow prevention, or any determination about pool or spa suction outlets. That last one is deliberate and not a technicality: suction-outlet velocity limits exist to prevent bodily entrapment, they are governed by their own standard, and a general-purpose friction calculator is the wrong instrument for a life-safety decision. If that is the question you have, it belongs with a pool professional working to the applicable code.

Reference tables

Every figure this calculator uses, with the source and the arithmetic for each group.

 

Frequently Asked Questions

What flow will my pump actually deliver?

Whatever flow makes your pipe's head demand equal your pump's head output — no more, and the pump has no say in it beyond its curve. Switch this page to "The pump", enter your pipe run and fittings, then read three points off the curve on your pump's datasheet: the shutoff head at zero flow, something near the middle, and something out near the right-hand end. The tool fits the curve through them and finds the crossing. A common surprise is how much of the head goes to fittings on a short run and how completely the pipe dominates on a long one.

How much pressure do I lose in 200 feet of 3/4 inch pipe?

It depends on flow, and steeply. At 5 GPM through 3/4 inch Schedule 40 PVC you lose roughly a foot and a half of head over 200 feet; at 15 GPM the same pipe costs you around eleven feet, because friction goes as roughly the square of velocity. It also depends on which 3/4 inch pipe: 3/4 inch PEX has a 0.671 inch bore against Schedule 40 PVC's 0.810, and since loss goes as roughly the fifth power of diameter, that difference alone nearly doubles the friction. Pick the actual material and size above rather than trusting a generic table.

Why does the same elbow have a different K in different pipe sizes?

Because Crane's method says it does. Crane publishes an equivalent length ratio L/D per fitting and computes the resistance coefficient as K = fT × (L/D), where fT is the fully-turbulent friction factor for that pipe size. Since fT falls as pipe gets larger, so does K: a standard 90° elbow is about 0.78 in half-inch pipe and about 0.49 in four-inch. The single-number tables everywhere on the internet are that relationship collapsed to one small-pipe value. It matters most where you have many fittings on small pipe, which is exactly the aquarium and pond-loop case.

Is my flow laminar or turbulent, and does it matter?

Almost certainly turbulent, and yes it matters enormously. Below Reynolds 2,000 the friction factor is exactly 64/Re and roughness is irrelevant; above 4,000 roughness dominates and viscosity matters less. Water in household pipe at anything above a trickle is turbulent — you need very low flow or very cold, viscous fluid to get laminar. The band between 2,000 and 4,000 is the awkward one: the friction factor there is genuinely indeterminate, this page interpolates using EPANET's published method, and it flags the result because it can be off by tens of percent.

Should I use Hazen-Williams or Darcy-Weisbach?

Darcy-Weisbach, for anything you care about. Hazen-Williams is an empirical fit with the viscosity of 60 °F water baked into its constant, so it cannot represent temperature at all, and its published validity is narrower than its popularity suggests — Diskin's 1960 analysis limits it to C values between 100 and 160 and to a Reynolds window bounded above as well as below. It survives because it is easy to do by hand and because C values for aged pipe are well catalogued. This page computes it alongside and tells you when your case falls outside its range.