April 2012 — sixteen months, a lot of concrete dust, and one system that works

The previous article covered the design. This one covers what it actually takes to build it in an Amsterdam apartment block with a flat concrete roof, existing floors that need to come up, and a flue that needs to pass through multiple reinforced concrete slabs to reach the sky.

Starting at the roof

The project started at the top because the flue is the critical path. Everything else can be sequenced flexibly. The flue cannot — it determines where the stove can go, which determines where the buffer tank goes, which determines the pipe routing for everything else.

A double-wall insulated stainless steel flue running through a building has non-negotiable requirements: the penetration through each concrete floor slab must be fire-rated, sealed against flue gas leakage, and properly supported at each level. The roof penetration must be waterproofed to the roofing membrane — in this case EPDM — without compromising either the membrane or the flue’s ability to expand thermally.

Getting through a Dutch concrete flat roof requires a diamond core drill. Not an SDS drill with a large bit — an actual core drill rig, clamped to the concrete, water-cooled, running a 200mm+ diamond core bit. The water keeps the bit cool and evacuates the slurry. It takes longer than you’d expect and makes less noise than you’d fear. The core that comes out is a solid concrete cylinder that lands on the roof membrane with a satisfying thud.

We needed three penetrations: flue, solar collector pipes, drain. Three core operations. Three concrete cylinders recovered from the gravel ballast.

The chimney in winter

The flue installation happened in January. The flat roof in January in Amsterdam is not a pleasant working environment: EPDM membrane covered in frost, pliers getting too cold to grip properly, and a stainless chimney section that conducts cold efficiently through work gloves.

The roof penetration is built up with a lead flashing collar bonded to the EPDM, sealed with fire-retardant sealant at the flue interface. The collar needs to accommodate the differential thermal movement between the stainless flue (which expands and contracts with temperature) and the concrete roof (which doesn’t care). Too tight and the flue binds; too loose and the seal fails. The design uses a telescoping upper section with a sliding seal — the flue can move vertically without breaking the waterproof joint.

The chimney cap is polished stainless, sized to prevent rain ingress at the outlet velocity the stove produces at normal firing rate. It is currently the most reflective object on the roof and serves as a useful landmark when finding the building from above on Google Maps.

The solar collector frame

The vacuum tube collector can’t sit flat on a gravel ballast roof. It needs to be at an angle — roughly 45° for Dutch latitude, trading off summer collection efficiency for winter gain, since summer is when we have surplus and winter is when we need it.

The mounting frame is welded steel section, fabricated in the basement and brought up in pieces. The design distributes its load across the flat roof without penetrating the membrane — it sits on rubber pads and relies on its own weight and the weight of the filled collector for stability against wind uplift. Wind loading calculations for a flat roof in an urban environment are not trivial: the collector is a large surface at an angle, and the aerodynamics of the gap between it and the roof are not intuitive.

The frame was overbuilt. This was intentional. A solar collector failure on a flat roof in an occupied building is not a small problem. The extra steel is cheap compared to the consequences of getting it wrong.

Opening the floors

Floor heating requires the pipes to be embedded in screed at the correct depth. The existing floor was concrete slab — structural — with a screed layer on top. The screed came up. All of it.

An SDS hammer drill with a wide chisel attachment removes screed efficiently and loudly. The neighbours were informed in advance. Not all of them were happy. This is the unavoidable cost of doing the job properly rather than laying pipes on top of an existing floor and building the floor level up — which loses ceiling height and creates transitions between rooms.

The floor heating pipes are laid on a layer of thermal insulation — without it, you heat the concrete slab and the ground below, not the room above. Steel mesh reinforcement goes over the pipes. New screed, mixed on site and poured to the correct depth for the pipe diameter, then levelled.

The screed must cure before it can be loaded. For a heated screed, the correct curing procedure is also the commissioning procedure — you increase the flow temperature gradually over several days to drive out residual moisture without cracking. Getting this wrong means cracked screed and an insurance conversation you do not want to have.

The copper pipework

The buffer tank in the plant room is the hub. From it, pipes run to the floor heating manifolds, the domestic hot water coil, and the solar collector loop on the roof. Every connection is hard copper, soldered.

The solar collector loop contains glycol — antifreeze — because the vacuum tubes can freeze in Dutch winters and because the loop operates at elevated temperatures during stagnation in summer. Push-fit fittings are not rated for glycol service at elevated temperatures over multi-year periods. Solder is. The choice was not difficult.

The lathe proved useful here. Several connection points required custom fittings — adapters between metric and imperial pipe standards, compression-to-solder transitions in locations too tight for standard fittings, sensor pockets machined to the correct diameter for the temperature probes. Buying custom fittings takes weeks and costs more than the material. Machining them takes hours and costs the material plus time.

The buffer tank

The buffer tank is a large insulated cylinder in the plant room. It contains the domestic hot water coil (a coiled tube through which tap water flows and is heated by the buffer water surrounding it), connection ports for the solar loop and wood stove circuit, temperature sensors at multiple heights to track stratification, and a pressure relief valve sized to the combined output of both heat sources.

Thermal stratification is a feature, not a bug. Hot water rises to the top of the tank; cooler return water sinks to the bottom. If you connect the domestic hot water coil at the top (hottest), the solar collector return at the bottom (where it’s coolest, improving collector efficiency), and the floor heating return partway up (at intermediate temperature), you can extract different temperature grades from the same tank without mixing them. This requires careful pipe entry geometry and avoiding turbulence at the inlets.

Getting this right is the difference between a buffer that behaves as designed and one that homogenises its temperature within an hour of any heat addition and loses most of the stratification benefit.

Commissioning

A hydronic system this complex doesn’t work correctly from day one. Commissioning is its own phase.

Air purging: every circuit needs air removed. Air pockets cause hot spots, flow noise, and pump cavitation. The system has automatic air vents at high points and manual bleed valves at low points. Initial commissioning involves running the pumps while systematically bleeding every circuit until the flow is bubble-free and the pump current matches its rated load.

Balancing: floor heating zones don’t naturally receive equal flow. The circuit furthest from the manifold has higher resistance than the nearest one. Without balancing valves set correctly, close zones overheat and far zones receive insufficient flow. Balancing requires measuring actual flow through each circuit with a flow meter and adjusting the lockshield valves on each manifold circuit until distribution is even.

Control calibration: the differential temperature controller for the solar loop needs to be calibrated so it starts the pump when the collector is genuinely hotter than the tank bottom (not just on a noise spike in the sensor reading) and stops it before the collector temperature drops so low that the pump is moving cooled fluid back into the tank.

Each of these takes time. Each reveals something the design didn’t anticipate. The design absorbs the learnings and gets updated.

Two winters in

The system works. In a typical Dutch October through March, the solar thermal covers a useful fraction of the heating load — not the majority, but enough to be meaningful on the gas bill that no longer exists. The wood stove covers the winter deficit and does so with fuel that is locally sourced, carbon-cycle neutral, and stockpilable.

The floor heating makes the apartment warm in the way that only thermal mass can — not hot air that stratifies and disappears when you open a window, but a floor that is consistently 22°C and radiates heat upward through the room uniformly.

The gas boiler is gone. The gas connection is gone. The meter was removed. There is no monthly gas standing charge.

That last point is the one that, in the long run, matters most.