Lesson E1 made the case for the hot-water loop; this lesson designs one. Four decisions shape every hot-water system — the ΔT, the pumps, the balancing, and the storage — and each one made lazily becomes a permanent operating cost. The good news: hot-water design is honest arithmetic all the way down, and the arithmetic is here.

Decision one: the ΔT — the master variable

Choose the temperature drop across the users and you have chosen almost everything else. The duty equation (lesson A10 with water on both sides):

flow (m³/h) ≈ duty (kcal/h) ÷ (1000 × ΔT)

A 500,000 kcal/h load at ΔT = 10 °C needs 50 m³/h of circulation; at ΔT = 25 °C, 20 m³/h. Wide ΔT means smaller pipes, smaller pumps, and pumping power falling faster than linearly (it scales roughly with the cube of flow through a given circuit). The price of wide ΔT: users at the return end see cooler water, so heat-transfer surfaces must grow (lesson A11 — LMTD shrinks). The working compromise in process systems is 20–30 °C; tight-uniformity duties argue it down, pumping economics argue it up, and the argument should happen on paper, once, at design.

Decision two: pumps and the loop's shape

The pump is the loop's boiler-equivalent: sized for flow (from the ΔT decision) against the circuit's resistance, on the pump curve where efficiency lives — not at its edge. Two disciplines: NPSH respect — hot water is always near flashing; the pump wants the coolest, highest-static-head point of the loop (pump into the heater on the return side, classically), or it cavitates every summer afternoon (the E1 site check). Variable speed where loads swing — a VFD trimming pump speed to hold loop ΔP converts the cube law from enemy to ally; on systems whose demand halves at night, it is routinely the fastest-payback item on the drawing.

Decision three: balancing — reverse return beats valves

Multiple users in parallel all want their share of flow; left alone, the nearest user gorges and the farthest starves — the universal disease of piped networks. Two cures combine: reverse return layout, where the return line is routed so every user sees the same total pipe length (first fed, last returned), making the network self-balancing by geometry; and balancing valves (or per-user flow controllers) to trim what geometry cannot. Reverse return costs some metres of pipe once; permanent throttling at every branch costs pump head forever — buy the geometry when the layout allows it.

direct return — near user wins U1 — gorges U2 U3 — starves short circuit for U1, long haul for U3 reverse return — equal paths U1 U2 U3 first fed = last returned → every path the same length geometry balances forever, free; throttling balances today, at pump-head prices
The reverse-return idea in one look. Route the return the "long way" and the network hands every user the same resistance.

Decision four: storage that actually stratifies

Where demand is peaky (lesson D9's charts), a buffer tank rides the peaks — but only a stratified tank does it well. Hot water floats on cold in stable layers if you let it: tall tank, gentle flows in and out through diffusers, hot connections at the top, cold at the bottom. Preserved stratification means the process draws true-hot water until the tank is genuinely spent, and the generator sees true-cold return (maximising its recovery — condensing sources and heat pumps, lesson F6, live or die on cold return). One crude mixing inlet destroys the whole effect and turns the asset into a lukewarm barrel. Tank connections are the cheapest place in the system to be careful.

stratified — layers preserved destroyed — the lukewarm barrel diffuser diffuser hot — floats cold — sits charge in hot, at the top true-hot draw → process cold return — bottom, gentle true-cold → generator thermocline — keep it thin one crude jetting inlet ≈ one temperature everywhere draws lukewarm same steel, no layers — buffer value gone tall beats wide · hot top, cold bottom · gentle in and out through diffusers site check: thermometers top, middle, bottom — three similar readings = the barrel
Stratification in one look. Keep the layers and the process draws true-hot water until the tank is genuinely spent while the generator sees true-cold return; one jetting connection turns the same steel into a lukewarm barrel.
At site
  • Measure the real ΔT across your loop at working load. If design said 25 and the pipes say 8, flow is triple the need — the pump is burning the difference (cube law). Rebalance before resizing anything.
  • Walk parallel users feeling return legs: uniform warm = balanced; one hot and one cold = the network disease. Check for reverse-return possibility before buying balancing valves.
  • Buffer tanks: thermometers top, middle, bottom. Three similar readings = destroyed stratification — inspect the diffusers, not the insulation.
  • Summer cavitation on the loop pump = NPSH margin gone; check cushion pressure (E1) and suction layout before condemning the pump.
Pin this
  • Flow = duty ÷ (1000 × ΔT). The ΔT decision sizes pipes, pumps and surfaces — make it once, on paper, at 20–30 °C unless the product argues.
  • Pumping power follows the cube of flow: half the flow, an eighth the power. VFDs collect this.
  • Reverse return balances by geometry, forever, free. Throttle only what geometry can't reach.
  • Storage works only stratified: tall, gentle, hot-top cold-bottom, diffused.
Steam stories

Stratification was the household engineer's secret first: the classic Indian overhead "geyser and drum" bathroom, like the English airing-cupboard cylinder, worked because nobody stirred it — hot stayed up, cold stayed down, and a modest heater served a big family from a small tank. Solar water heating rediscovered the same physics at panel scale, and district heating industrialised it into thousand-cubic-metre stratified accumulators. The principle never changed: heat floats — design so it may.

FAQ

What ΔT should I design a hot-water loop for?

Start the argument at 20–30 °C and settle it on paper, once. Wider ΔT means smaller pipes, smaller pumps and much smaller pumping power — but cooler water at the return-end users, so heat-transfer surfaces must grow. Tight-uniformity products argue it down; pumping economics argue it up.

How do I work out the circulation flow for a heating duty?

Flow (m³/h) ≈ duty (kcal/h) ÷ (1000 × ΔT). A 500,000 kcal/h load at ΔT = 25 °C needs 20 m³/h; let the ΔT collapse to 10 °C and the same duty demands 50 m³/h — same heat, two-and-a-half times the flow, and the pump pays for the difference by the cube law.

Why is the farthest user on my loop always cold?

Because the nearest user is gorging: in a plain parallel network the short path wins and the long path starves. The lasting cure is reverse-return routing — first fed, last returned, so every user sees the same total pipe length — with balancing valves only to trim what geometry cannot reach. Throttling everything instead costs pump head forever.

Is a VFD on the circulating pump worth it?

On systems whose demand swings, usually yes — and often the fastest-payback item on the drawing. Pumping power scales roughly with the cube of flow, so a drive that trims speed to hold loop ΔP collects an eighth of the power at half the flow. Steady, flat-load loops have less to collect.

My buffer tank reads the same temperature top, middle and bottom. What went wrong?

Stratification is destroyed — the tank has become a lukewarm barrel. Hot water floats on cold only if the tank is tall, the flows are gentle, and the connections diffuse rather than jet: hot in and out at the top, cold at the bottom. Inspect the diffusers and connection velocities, not the insulation.

Check yourself

Five quick questions on this lesson. No marks, no records — the score is for you.