Section 13 · Deep Dives

Advanced hydraulics

The rules of thumb on this site will get a working ram built. This page is for the cases where they will not — large duties, unusual geometry, or a number you have to defend — and it starts by being honest about what the simple formulas actually say.

Where the first-order model stops

Everything practical on this site rests on one relation and one constant:

q = η · Q · H / h, η ≈ 0.6

The formula is not wrong — it is an accounting identity. It says how much water you get if you have already decided how efficiently the machine converts power. It does not tell you what η is, and it does not know why η falls.

Treating η as a constant is the useful lie, and it survives because most sites sit in a forgiving band. It stops being adequate when any of the following matters:

Efficiency is not constant. It depends on the lift ratio, the drive pipe's proportions, the valve setting and the beat rate. The empirical "sweet spot" of 2–6 : 1 is not a physical law — it is the region where the machine's internal timings happen to line up. Well beyond it, the spike can still clear the delivery head but only for a fraction of the beat, so less of each cycle pumps and η collapses.
The drive pipe has friction. During spin-up, friction caps the velocity the column reaches before the valve closes, and the spike is proportional to that velocity. The terminal-velocity relation v_max = √(2gHD/(f·L)) is the ceiling — which is why a scaled or undersized drive pipe delivers a feeble hammer no matter how the valve is tuned (the friction budget).
Air is a cushion, not a working fluid. A small pocket changes the effective wave speed, softens the closure and shifts the timing. The model assumes a single-phase column of known wave speed.
The air chamber has dynamics. Its volume, its precharge and its orifice all shape the pressure seen at the check valve. A first-order calculation treats the chamber as a perfect integrator.
Closing time is not zero. The Joukowsky rise is the maximum, valid only when the valve closes faster than the wave can return. Everything slower earns less — the subject of the last section.

📏

Use the simple model the way a chart is used

Run it to size the site, then derate for what it leaves out — the field expectation of 70–80% of theory is exactly that correction, applied in advance (measuring delivery). When a design has to be defended rather than built, the tools below are the defence.

Dimensionless scaling

Two rams of different sizes behave identically if the same dimensionless groups match. That is what makes small-scale testing mean anything, and why manufacturer tables are families of curves rather than single numbers.

The groups that govern a ram, in rough order of importance. Everything else — absolute size, fluid, units — drops out.
Group	What it fixes
h / H — lift ratio	How much of the spike's energy is useful. The dominant group, and the one the front-of-page rules of thumb are really about.
L / D — drive-pipe slenderness	The column's inertia against its friction, and therefore the velocity it can reach and the timing of the beat.
L / H — length against fall	The 3–7 rule, expressed dimensionlessly: the drive pipe must be long enough for the valve to close before the wave returns, and short enough to accelerate promptly.
τ = T_close / (2L / c) — closure ratio	Whether the closure counts as sudden (τ ≤ 1, full Joukowsky rise) or slow (τ > 1, a reduced rise).
v / √(2gH) — velocity index	How near the column runs to its frictionless free-acceleration limit; small values mean friction is ruling the machine.

Note what is absent: Reynolds number and pipe roughness appear only inside the friction term, so they matter far less to overall performance than the geometry groups above. Two geometrically similar rams of different diameters will therefore deliver the same q/Q at the same lift ratio — which is the justification for testing a half-scale model and trusting the result.

Wave timing & the method of characteristics

Fast and slow closure

The Joukowsky rise is the sudden-closure case. Whether a real closure is "sudden" depends on comparing the time it takes with the time the wave needs for its round trip.

τ = T_close / (2L / c)

For L = 13 m of steel (c = 1300 m/s) the wave returns in 2 × 13 / 1300 = 20 ms; any closure faster than that is a sudden closure and earns the full Joukowsky rise, while a closure in 100 ms earns only about a fifth of it. This is exactly why a ram's valve weight matters so much: it is not only setting how fast the water is moving, it is setting how much of the spike survives.

The characteristic equations

To predict a beat rather than estimate it, the pipe is divided into reaches of length Δx and the transient is marched forward in time steps of Δt = Δx/c. Within one time step, disturbances travel only along the two characteristic lines — C⁺ carrying information up from the upstream node, C⁻ carrying it back from the downstream one — and the two equations solve simultaneously for the head and flow at the new point.

C⁺: H_P = H_A − B(Q_P − Q_A) − R·Q_A|Q_A|
C⁻: H_P = H_B + B(Q_P − Q_B) + R·Q_B|Q_B|

B = c / (g·A) is the characteristic impedance and R = f·Δx / (2g·D·A²) the friction term. A and B are the neighbouring nodes at the previous time step; P is the point being solved. This is the standard form used for water-hammer analysis generally; for the texts it comes from, and for the ram-specific treatments that apply it, see further reading.

Eliminating H_P gives the flow at the new point directly:

Q_P = [ H_A − H_B + B(Q_A + Q_B) − R(Q_A|Q_A| + Q_B|Q_B|) ] / 2B

and H_P follows from either characteristic. The absolute-value terms are what make the friction always oppose the flow, whichever way it is going — which is essential here, because a ram's drive column reverses direction twice a beat.

Closing the loop: the boundaries

The interior solution is only half the problem. The character of a ram comes from its boundaries, and each of them replaces one of the two characteristic equations with a physical law:

The source is a constant-head reservoir, so H_P is fixed and only the C⁻ characteristic applies.
The waste valve is an orifice whose area varies with time as it swings shut — a valve law, usually written as a discharge coefficient against percentage open. This is the boundary that generates the beat.
The delivery check valve opens and closes on the sign of the pressure difference across it, which makes the problem a switching one: the machine changes its own boundary conditions.
The air chamber is a gas law — P·Vⁿ = constant — so its boundary condition couples the pressure to the volume of water that has entered it, using the same polytropic relation as the sizing tool.

Run that for enough cycles and something notable falls out: the beat rate is not an input. It is a consequence of the valve's weight, its stroke, the column's inertia and the wave speed, and it settles to whatever value the system can sustain. That is the precise sense in which a ram tunes itself — and why the field method (adjust the weight, listen, measure at the tank) is not a crude approximation of the modelling so much as the modelling run in hardware.

⚙️

When it is worth the effort

For a single household or village ram, model nothing: build it, tune the weight, and keep the log. Modelling earns its keep when the duty is large, the geometry unusual (long or steeply stepped drive pipes, cascades, very high lift ratios), the cost of being wrong is a failed community scheme, or someone is asking you to sign your name to a performance figure. In those cases the first-order model is a starting point and the method of characteristics is the answer.

If you are here for practical reasons rather than theoretical ones, the pages that will actually build you a pump are Design & Sizing, Pipe & Materials and Build Your Own.