AHU Airflow Arrangements — Stacked, Side-by-Side and Cross-Flow Bridges

Before a single fan or coil is selected, one decision fixes the footprint, the plant-room height, which side every duct leaves from, and how hard the heat recovery section is to service: how the supply and extract airstreams are arranged relative to each other. Get it wrong and the unit is technically correct and physically un-installable.

Two decks, three ways to stack them

A double-airstream AHU has a supply deck and an extract deck. There are only three practical ways to put them in the same plant room.

Stacked (supply over extract, or the reverse). Halves the footprint, doubles the height. The standard choice when floor area is the binding constraint — rooftop plinths, tight technical floors, retrofits into an existing plant room. The penalty is service access: the upper deck needs a platform or ladder, and filter changes on a unit whose top deck sits at 2.4 m are a two-person job for the rest of the building's life.

Side-by-side. Both decks at floor level, full footprint, low profile. Preferred wherever headroom is limited or maintenance frequency is high (hospitals, food processing, anywhere with ePM1 filters on a short change interval — see filter classes after ISO 16890). Every access door is reachable without a platform.

Single-deck (supply only, or 100% extract). No arrangement question, but also no heat recovery — which in most EU applications now means an ErP/NRVU justification is required rather than assumed.

Arrangement	Footprint	Height	Service access	Typical driver
Stacked	~50%	~2× per-deck height	Platform needed for upper deck	Floor area constrained
Side-by-side	100%	1× per-deck height	All doors at floor level	Headroom / maintenance constrained
Single deck	~50%	1×	Full	No HRS required

Note that casing height in a stacked unit is specified per deck, not for the assembly. A 1600 mm entry on a stacked unit's data sheet means 1600 mm per deck and roughly 3.2 m plus base frame overall — a recurringly misread figure when a data sheet is skimmed rather than read against the drawing.

The bridge: where the two airstreams meet

Heat recovery is the only place the decks physically interact, and the component type dictates the geometry.

A rotary wheel sits vertically in the plane shared by both decks — the airstreams pass through the same disc, so the deck-to-deck relationship stays straightforward: supply enters one side, leaves the same side.

A counter-flow or cross-flow plate exchanger is different. The plate block routes the two airstreams across each other on a diagonal, which means the chain that entered as "supply on the left" continues downstream of the exchanger on the other side of the unit. That diagonal swap is the single most common source of routing errors in AHU documentation, because it breaks the assumption that a connection's physical side tells you which airstream it carries.

Frost strategy interacts here too: a preheater or bypass damper serving the plate block has to be positioned before the swap, on the correct airstream — a detail that gets mislabelled for exactly the same reason.

Deck-scoped keys, and why they must be translated

Connection identifiers such as left_supply or right_extract are scoped to the deck, not to the airflow direction: left_supply means the left-hand end of the supply deck. On a side-by-side unit with a rotary wheel, deck side and airflow side coincide and nobody notices the distinction. Insert a cross-flow plate exchanger and they no longer coincide — the supply chain crosses the bridge diagonal and exits where a naive reading expects extract.

The consequences are not cosmetic. Every downstream number keyed off that mapping inherits the error:

• Connection velocity and dP get computed against the wrong airflow (the lost 50 Pa).
• The DXF and IFC export label duct stubs for the wrong airstream, and the BIM coordinator routes ductwork to a connection that carries the opposite air.
• The technical data sheet shows the right total airflows against the wrong physical sides — a discrepancy consultants do catch, and one that costs a resubmission.

The fix is structural, not procedural: derive the aperture-to-airflow mapping from the unit's actual internal chain every time, and never let a human, a template or an export script hardcode "left = supply".

What to lock down before selection

1. Footprint vs height budget — decide stacked or side-by-side from the plant room, before component selection, because it changes casing height per deck and therefore face velocity.
2. Access clearances on the constrained side — a stacked unit still needs its full withdrawal length for coils and its door swing on both decks.
3. HRS type before layout is frozen — a plate block imposes the diagonal swap; a wheel does not.
4. Every connection re-derived from the chain, not from where it appears to sit on the drawing.

In AirSelect3D, the arrangement is a first-class property of the unit: stacked and side-by-side layouts, bridge modules and cross-flow swaps all resolve in the 3D model, and every duct aperture's airstream is derived from the actual chain — so the label on the drawing, the velocity in the pressure budget and the stub in the DXF all agree by construction rather than by review.

Build a stacked or side-by-side unit and watch the routing resolve in 3D →