Water-Powered Fan for Radiators and Fan Coil Units: Forced Convection Without a Mains Cable

The water in a heating or cooling circuit carries more than heat. It also carries pressure, and in every hydraulically balanced system the balancing valve deliberately wastes part of that pressure by throttling. This concept takes that energy away from the valve and gives it to a fan that pushes air through the heat exchanger. No socket, no cable, no power supply.

This is my own innovation, built on over a decade of hands-on experience across several hundred buildings. I described a first version in an earlier article, Wireless fan in a radiator and fan coil unit (in Polish). Here I go further: I work through the power balance, correct the mistakes of that first version, and show where the idea makes the most sense.

Why this idea

When retrofitting buildings for heat pumps, the same two problems come up again and again:

• Existing radiators are too small for low flow temperatures. Drop the supply from 55–70°C to 35–45°C and their output falls sharply. The options are bigger radiators or chasing walls for underfloor heating.
• Fan coil units need power, controls and condensate drainage in every room. In an existing building each cable and each drain pipe means a chase in the wall, cost and downtime.

Both cases lack the same thing: air movement at the heat exchanger. The energy for that movement is already in the pipes. The circulation pump creates pressure, and the valves destroy part of it. A turbine in place of the throttle recovers what is today converted into hiss at the valve seat.

How it works

A turbine instead of the presetting orifice

In a hydraulically balanced installation, valves with a preset (kv value) throttle the excess available pressure. On radiators close to the pump that is typically a few to a few dozen kPa. All of it ends up as noise and frictional heat.

If an adjustable inlet guide vane with a micro-turbine takes over the throttling role, the branch resistance stays exactly as designed. The only thing that changes is where the pressure drop goes: into shaft rotation instead of noise.

Note: this works only where there is surplus pressure to throttle. At the farthest radiator (the index circuit) there is usually no margin, and a turbine would simply add resistance. Those locations must be identified at the design stage.

Magnetic coupling and a dielectric partition

The turbine shaft never passes through a wall. A magnetic coupling transmits torque through a sealed partition, so there are no glands or dynamic seals that start to leak after a few years.

This is where I correct the first version. There I wrote about aluminium parts near the magnets, but aluminium is out. It is not ferromagnetic, yet it conducts electricity well, so the rotating field induces eddy currents in it (the principle of an electrodynamic brake, the Arago disc). With a power budget of a fraction of a watt, such a partition would brake the rotor almost to a standstill. The same applies to stainless steel. The partition must be a dielectric: a polymer composite (PEEK, PPS, glass-fibre-reinforced PA) or technical ceramic.

A magnetite filter upstream of the turbine

Older systems carry magnetite (Fe₃O₄) in the water. Magnets on the wet side of the coupling would act as a dirt separator and soon be buried in sludge. For that reason a compact mesh-and-magnet strainer with a drain plug goes upstream of the turbine. In a heat pump retrofit such a filter should be in the system anyway, but here it is a hard requirement.

The physics: how much power can you get from a pipe?

A 1,000 W radiator at 40/35°C

With a heat pump and a low flow temperature (ΔT = 5 K), a 1,000 W radiator needs:

• a flow rate of 1,000 W / (4,186 J/(kg·K) × 5 K) ≈ 0.048 kg/s ≈ 172 L/h (4.8 × 10⁻⁵ m³/s),
• hydraulic power, at a 15 kPa drop across the turbine, of 4.8 × 10⁻⁵ m³/s × 15,000 Pa ≈ 0.72 W.

A micro-turbine in a DN15/DN20 pipe is not efficient: low Reynolds numbers and clearance leakage hurt. I assume roughly 30–40% for the turbine and about 85% for the coupling with its bearings. That leaves on the order of 0.2–0.25 W at the shaft.

It is little, but enough for a radiator. A light cross-flow fan does not have to push air through a filter. It only has to break up the boundary layer at the fins and raise the air-side heat transfer coefficient. For comparison, a 120 mm PC fan delivers power to the air of the same order, a few tenths of a watt. Fan boosters for radiators are reported to add tens of percent to output at low flow temperatures. How much this design will add can only be shown by measuring a prototype.

At a conventional ΔT of 10 K the flow is half as large (about 86 L/h), so at the same pressure drop the shaft power also halves. The concept suits low temperature differences and high flow rates best, which is exactly what heat pumps deliver.

What happens when the thermostatic head throttles the valve

An early estimate claimed that at 50% flow the fan stops, because turbine power drops eightfold. That is wrong.

• With fixed turbine geometry the pressure drop scales with the square of the flow, so available power falls roughly with the cube of the flow.
• But the fan’s power demand also falls with the cube of its speed (the fan affinity laws).
• Both sides scale the same way, so fan speed falls roughly in proportion to flow. At half the flow the fan turns at about half speed: quieter and gentler.

The fan stops only when the turbine torque cannot overcome static friction: bearing breakaway torque, the coupling’s cogging torque and clearance leakage at very low flow. That threshold depends on the design and has to be measured on a prototype. Below it the radiator simply carries on as an ordinary convector, so the worst case is a return to today’s situation.

The result is natural self-regulation with no electronics:

• Room too cold: the head opens the valve, flow rises, the fan speeds up and the room warms up quickly.
• Setpoint reached: the head closes the valve, fan speed falls in proportion, and in the end only quiet natural convection remains.

Fan coil unit: the balance closes, but only just

Fitting a turbine to a conventional fan coil unit as it is makes no sense. A dense coil (fins about 1.8 mm apart) with a non-woven filter presents 25–40 Pa of air-side resistance and needs several to a dozen watts at the fan. A fraction of a watt from the water will not move it.

The coil has to be redesigned towards a chilled beam: fins 4–5 mm apart, more copper, air-side resistance of 2–4 Pa.

The balance gives a simple rule of thumb. At typical temperature differences (about 3 K on the water side, about 6 K on the air side) the volumetric air flow is roughly 1,700 times the water flow. With an overall turbine–coupling–fan chain efficiency of 15–20%, this gives:

about 10 kPa of pressure drop across the turbine for every 1 Pa of air-side resistance.

At 3 Pa that is about 30 kPa, the upper limit of what is typically throttled across valves. The balance closes, but with no margin. Every extra pascal (a finer mesh, dust, condensate between the fins) has to be paid for with more kilopascals on the water side.

A worked module: 150 m³/h of air (0.042 m³/s) at 3 Pa and 50% rotor efficiency needs about 0.25 W at the shaft. 50% is optimistic for a small cross-flow fan and the real figure may be lower. Cooling that airflow by about 6 K gives on the order of 300 W of sensible cooling. More capacity comes from a longer unit or several modules, not from a stronger fan.

What we gain: forced air velocity of 0.8–1.0 m/s instead of about 0.15 m/s from natural convection raises the air-side heat transfer coefficient several times over (from about 10 to a few tens of W/(m²·K)). The same copper therefore delivers noticeably more output than a passive beam.

A low-resistance 4-pipe fan coil unit

Two cooling modes

• Dry cooling at 16/19°C (most of the summer): operation above the dew point, so the coil stays dry. The chiller runs at a higher water temperature, and its EER is clearly better than at 7/12°C (roughly 20–30%, depending on the machine). Free cooling from the ground is also an option. Air movement in the occupied zone lowers the perceived temperature further (about 1 K at 0.3 m/s).
• Dehumidification at 12/16°C (humid days): when the dew point exceeds about 15°C, the water temperature is lowered and moisture is collected in a drip tray with gravity drainage. The power budget has no room for a condensate pump.

A wide fin pitch helps with condensate too. The capillary length of water is about 2.7 mm. At a 1.8 mm fin pitch, droplets bridge into films between the fins and air-side resistance roughly doubles. At 4–5 mm with a hydrophilic coating, condensate runs off as a thin film.

Instead of a dense non-woven filter (+10–15 Pa, far outside the budget), use a coarse mesh that catches lint and fibres, with resistance below 1 Pa. The wide fin pitch also makes the coil easy to vacuum.

One coil, one six-way valve

Two coils would double the air-side resistance and sit as dead ballast for half the year. Instead, the 4-pipe system is combined on one oversized coil through a six-way valve. A single turbine on the common branch is then enough.

Heating 28/25°C ──► [SIX-WAY VALVE] ◄── Cooling 16/19°C or 12/16°C
                           │
                 [Magnetite filter]
                           │
                 [Guide vane (kv) + micro-turbine]
                           │
              [Thermosensitive air-deflector element]
                           │
              [One Cu/Al coil, 4–5 mm fin pitch]

The large surface area designed for cooling lets you heat in winter with a very low flow temperature, about 28/25°C. At that supply temperature the heat pump runs at a high COP. With favourable conditions (a ground source, for example) it can reach 5–6.

Temperature symmetry: the difference between the heating return (25°C) and the cooling return (18–19°C) is only a few kelvin. In a classic system at 45/35°C and 7/12°C it is over 20 K, and at 55/45°C over 30 K. Add a dead band in the controls (hysteresis and a 15–30 minute pause before changeover): the standing water in the coil drifts towards room temperature, so warm water does not reach the chilled water return. Mixing losses are negligible.

A practical note: classic ball-type six-way valves have high seal friction and their actuators draw several watts. With water-powered operation a pressure-balanced valve (spool or poppet type) with an interlock against opening both circuits at once is the better choice.

An air deflector without a motor

At low supply velocity (about 0.8–1.2 m/s) buoyancy plays a major role. A thermosensitive element on the pipe downstream of the six-way valve (a wax capsule or a Nitinol spring) moves the louvre through a linkage:

• Cooling (water below about 19°C): the deflector sits horizontal. Thanks to the Coandă effect the cool air clings to the ceiling, travels deep into the room and sinks gently, with no draught on the neck.
• Heating (water above about 25°C): the wax expands and tilts the deflector down by 45–60°. Warm air reaches the occupied zone instead of rising straight to the ceiling.

It works without power, sensors or a stepper motor.

The variant with an energy buffer (described below) can have the deflector optionally battery-powered: a small actuator or micro-servo instead of, or alongside, the capsule. Wax responds to water temperature with a lag of several minutes and knows only two positions. An actuator sets the deflector the moment the mode changes, at any angle, following a schedule, occupancy or a command from the BMS. The energy cost is negligible: a move at about 0.5 W for 2 s is roughly 0.3 mWh, so a few moves a day take only a fraction of the daily standby budget. The wax capsule stays as a fallback: with a flat battery, the deflector still sets itself.

Option: an energy buffer and LoRaWAN to the BMS

To move the valve after weeks without flow and to stay connected to the BMS, a small energy buffer can be added:

• The coupling as a generator: a flat stator coil outside the partition picks up on the order of 20–40 mW from the rotating magnets while water flows. That is about 10% of the shaft budget, so it has to be counted in the balance, not treated as free.
• A LiFePO4 14500 cell (about 600 mAh, 3.2 V, roughly 1.9 Wh) or a lithium-ion capacitor (LIC). Both are much safer thermally and last longer under shallow cycling than ordinary Li-ion.
• Optionally, a deflector actuator running from the same buffer (about 0.3 mWh per move), with the wax capsule as a fallback.
• A local PI loop plus Class A LoRaWAN (868 MHz): the device wakes every 5–10 minutes, sends temperature and humidity and receives any new setpoint. Average consumption of about 0.15 mW (roughly 3.6 mWh per day) means, in theory, more than a year of standby with no flow, before self-discharge. One valve move (1.5 W for 20 s) costs about 8 mWh.

Charge only during normal heating or cooling operation. This is another correction to the first version, where I wrote about charging the battery “when the fan coil is idle but water is flowing”. Running water just to charge a battery overheats or overcools the room and spoils the return temperature across the whole building.

Comparison with conventional solutions

Parameter	Classic 4-pipe FCU	Passive chilled beam	Active (induction) beam	Hydro-active FCU (this concept)
Cooling	7/12°C	15/18°C	15/18°C	16/19°C dry / 12/16°C dehumidification
Heating	45/35 or 55/45°C	none or 35/30°C	35/30°C	about 28/25°C, same coil
Air-side resistance	25–40 Pa	below 1 Pa	2–5 Pa	2–4 Pa (4–5 mm fins)
230 V in the room	yes, a dozen to a few dozen W	no	no (pressure from the air handling unit)	no (a fraction of a watt from the water flow)
BMS communication	wired	wired to the actuator	wired to the actuator	optional LoRaWAN with a buffer
Air deflector	manual or motor	none	fixed nozzles	thermomechanical (wax/Nitinol)
Condensate	tray plus often a pump	not permitted	not permitted	gravity tray in dehumidification mode

Pros and cons

Pros

• A heat pump without replacing the radiators. Forced air movement helps hold output at a lowered heating curve, with no wall chasing and no larger radiators.
• No wiring in the room. No 230 V, power supply or control cable to every unit.
• Recovery of energy that is wasted anyway. The turbine replaces the valve’s throttle instead of adding to it.
• Self-regulation. Fan speed follows flow, and flow is set by an ordinary thermostatic head.
• Leak-tight. Magnetic coupling and no dynamic seals.
• Fail-safe. If the fan stops, you are left with an ordinary radiator or beam.
• Ground-source and air-source heat pumps. Lower supply temperature in winter, warmer chilled water in summer. With ground source, a much longer free-cooling season and borehole regeneration.
• Upgrading existing beams. A passive beam becomes an active one with no ducts and no cable, with an estimated 2–3 times the output and the ability to heat.
• One coil for heating and cooling at very low and very high temperatures, which gives the heat source good operating conditions.

Cons and risks

• A very small power budget. A fraction of a watt rules out non-woven filters, condensate pumps and dense coils. The design has to be low-resistance from the start.
• A tight balance in the fan coil. About 10 kPa per 1 Pa of air resistance; at 3 Pa that is already the upper limit of typical throttling.
• Surplus pressure is required. On the index circuit (the farthest terminal) the turbine would be an extra resistance.
• Start-up threshold. At very low flow static friction will stop the fan. The threshold must be measured.
• Water quality. A magnetite filter is mandatory and needs periodic cleaning.
• Materials. The partition must be a dielectric. Aluminium and stainless steel near the magnets are ruled out.
• Spreading the drive across the radiator width. The pipe enters at one point, while a radiator is 60–140 cm wide. A long cross-flow fan along the bottom edge works best. The alternative is a micro-generator with several small BLDC fans, which brings the electronics back.
• Concept stage. The figures above are order-of-magnitude calculations. Turbine and rotor efficiencies at this scale need to be confirmed on a prototype.

Applications

Heat pump retrofits

The largest and simplest market: a radiator variant with no electronics, fitted with a thermostatic head. It allows a lower supply temperature (for example from 55°C to about 40–42°C) and lets you keep existing radiators in many rooms. It works best where valves throttle the most today, that is, closer to the pump. It complements the usual retrofit path: hydraulic balancing, a lower heating curve, and only then decisions about replacing terminals.

The concept suits both common heat sources, ground-source and air-source, because the same mechanism works in both directions: forced air movement lets a terminal deliver the same output with water closer to room temperature. In winter that means a lower supply temperature; in summer, warmer chilled water. Both raise the efficiency of the heat source.

Older single-family homes: a heat pump without underfloor heating

In an older house with radiators sized for 70/55°C, a heat pump usually runs into one wall: at 40–45°C the radiators fall short, and underfloor heating means breaking up floors and renovating the whole ground level. Here the simplest mechanical variant is enough: a radiator plus a plain cross-flow fan along its bottom edge, driven by the turbine at the valve. No electronics, no cable, no socket under the window. Installation comes down to swapping the valve insert and adding a fan strip, an afternoon’s work for a plumber rather than a job for an electrician and a flooring crew.

An order-of-magnitude check: radiator output scales roughly with the water-to-room temperature difference to the power of 1.3. If forced air raises output by about 30% at the same water temperature, the same output is reached with a temperature difference about 18% smaller. For a radiator running today at a mean of 60°C in a 20°C room, that is roughly 7 K off the supply temperature. A common rule of thumb gives 2–3% better heat pump SCOP for every kelvin lower, so in total something like a mid-teens percentage less electricity over the season, without touching the floors. How much the fan adds on a specific radiator has to be measured on a prototype. The order of work stays the same: first a proper heat-loss calculation and the weakest room, which sets the heating curve. The fan goes where the radiator is undersized, not everywhere.

This opens an easy and economical route from a boiler to a heat pump in homes where, until now, the only honest answer was “underfloor heating first, or replace every radiator”.

Ground-source heat pumps: a much longer free-cooling season

A ground-source heat pump has free cooling built into its boreholes: in summer, brine from the ground loop passes through a heat exchanger and cools the building with the compressor off, at the cost of the circulation pumps alone. A conventional fan coil at 7/12°C cannot use it, because the ground does not deliver 7°C. A low-resistance fan coil or beam at 16/19°C needs water at about 16°C, so about 13–14°C on the borehole side, allowing for the heat exchanger. The ground at borehole depth is typically around 8–12°C. Rejecting heat over the summer warms it gradually, but at high chilled-water temperatures free cooling can run much longer into the season before the compressor is needed, and in buildings with moderate heat gains, for the whole summer. On humid days, when dehumidifying at 12/16°C, the compressor comes in anyway.

A bonus: heat rejected to the ground in summer regenerates the borehole field before winter, so the heat pump starts the heating season with a warmer source. How many free-cooling days you get depends on the borehole field, the ground and the load, and has to be calculated per building, but the direction is clear.

An air-source heat pump has no free cooling, but the rule is the same: cooling at 16/19°C instead of 7/12°C raises EER, and heating at a lower supply temperature raises COP exactly when it matters most, in frosty weather.

Upgrading passive chilled beams

A passive beam has the lowest air-side resistance of any terminal (below 1 Pa) but also the lowest output: air moves by buoyancy alone, at about 0.15 m/s. An active (induction) beam is more powerful but needs primary air from an air handling unit, which means ductwork and fan pressure. Adding the turbine and a cross-flow fan turns a passive beam into an active beam with no ducts and no cable: forced flow at 0.8–1.0 m/s raises the air-side heat transfer coefficient several times, and the output of the same beam rises by an estimated 2–3 times (to be confirmed by measurement).

In energy terms this is the most comfortable case in the whole concept. At around 1 Pa of air resistance the turbine needs about 10 kPa, well inside the typical throttling margin rather than at its limit as in a fan coil. Heating comes as a bonus. A passive beam heats poorly because warm air stays under the ceiling; a fan with a thermomechanical deflector pointing down in winter lets the same coil heat the room. The usual beam limitation remains: cooling must stay dry, above the dew point, because a beam has no condensate tray.

In existing offices this is a retrofit with no ceilings torn out and no new ductwork: more cooling from the same beams, and with a ground-source heat pump, a longer free-cooling season.

High-temperature cooling

A low-resistance fan coil at 16/19°C combines the comfort and efficiency of a chilled beam with the option of dehumidification and heating on the same coil. It suits offices and renovated buildings, especially where free cooling from the ground is available or where running cables to every unit is expensive.

Explosive atmospheres (ATEX)

In a hazardous zone the purely hydraulic variant brings no electrical components into the zone: only water pipes enter the room, while the pump and electrics stay outside. Electrical ignition sources disappear, and with them the costly explosion-proof design of motors and cabling. This does not mean certification disappears altogether. A spinning fan and bearings are potential non-electrical ignition sources (friction, overheating, static electricity) and are assessed under the standards for non-electrical equipment. The variant with a buffer and LoRaWAN brings electronics back, so it suits Ex zones only once the requirements for electrical equipment are met.

This is also the one place where a dedicated circulation pump just to drive the fans makes sense. Outside an Ex zone that chain (electricity → pump → pipe → turbine → coupling → fan) has an efficiency of only about ten-odd percent. If 230 V is available at the unit anyway for the pump, it is better to power the fan directly.

Wet rooms

No voltage at the unit also simplifies installation in wet rooms and wherever bringing in electrical supply is awkward.

Summary

The water in a hydronic system carries more than we use today. In a radiator on a heat pump, a turbine in place of the throttle gives enough power to assist convection. The balance holds up there, and the system regulates itself. In a fan coil the balance also closes, but only with a coil redesigned for a few pascals of resistance and full use of the typical pressure margin. That is why every pascal counts.

It pays off most where it changes the investment decision: in an older single-family home that, thanks to simple fan assistance at the radiators, can move to a heat pump without underfloor heating, and in an office with chilled beams that gain output, heating and a longer free-cooling season from the ground, with no new ducts or cables.

The next step is a prototype and measurements: micro-turbine efficiency, start-up threshold, and the output gain of a radiator at 35–40°C. I will write up the results in a follow-up article.

The concept is my own innovation, built on over a decade of hands-on experience across several hundred buildings. The first version (in Polish): Wireless fan in a radiator and fan coil unit.