Đắk Song is a 6 × 4 × 3 m insulated drying chamber on a coffee farm in Vietnam's Central Highlands, loading four tonnes of parchment coffee at a time. A batch gives up about 480 kg of water over forty hours, and parchment is dried at 40–60 °C — hotter cracks the bean — so the chamber is held at 55 °C under a 30 % RH ceiling, 0.030 kg of water per kg of air, the wet bulb 36 °C. The batch is drawn as what it is: 8.2 kW of moisture into the air and the same 8.2 kW of heat taken out of it to evaporate it — the product dries into the room, it does not heat it. Batches run back to back; the chamber's thermal mass is the coffee's.

The machine is a closed-loop heat-pump dryer: the chamber's own air goes round through a DX coil that takes the water out and a condenser that puts the heat back, and nothing is vented — a real dryer bleeds hot air only when the chamber runs over, drawn here as the room's own exhaust damper. Inside is a Copeland ZWD81KBC digital scroll on R134a, one of the 676 maker's maps with an air-conditioning publication code that the catalog admitted for this case: its envelope runs to 85 °C condensing, which is why the family is sold for dryers, and the case study that prompted the example — a Vietnamese dryer maker's, for a coffee exporter in these highlands — puts its modulation at 10–100 %. At its design point — the chamber's 55 °C air at a 36 °C wet bulb — the scroll sits at 24 °C evaporating and 85 °C condensing, 22 kW of cooling for 7.3: on its ceiling, with a fifth of a kelvin to spare. (This page used to say six kelvin under. Since 2 Oct an air stream carries one mass from where it enters, the condenser lost the air it had been credited beyond that, and the machine was given 0.9 m³/s where it had 0.8 to stay under its ceiling at all.)

Two things had to be right. The coil sees 45 % of the air through a face-and-bypass: on the whole stream it would cool all of the 55 °C air to 33 °C and take 7.5 litres of water an hour out of it, while on less than half it takes its share to 27 °C and the mix leaves four grams a kilogram drier — 14 litres an hour, nearly twice the water — and the condenser still gets every kilogram of the stream to reheat. And the unit is run humidity-led, judged, modulated and cycled on the chamber's ceiling alone: a machine whose condenser is in its own supply is always warmer out than in, so no temperature test can steer it. The pool hall's dehumidifier had this rule already — but its over-delivery was still read off the dry bulb, which an in-stream condenser never overshoots, so it ran flat out on every hour and only cycled. The dryer, with a ceiling under what its flat-out supply already makes, found it; since 5 September a humidity-led inverter unloads to its ceiling.

The alternative is the dryer it replaces: the same chamber on a 30 kW heater and outside air, the moist air vented — the simplest hot-air dryer, once-through; a coal or LPG dryer is the same air path on another fuel. On Buôn Ma Thuột's year the heat pump takes 301 kWh a batch — 0.62 kWh per kilogram of water, a specific moisture extraction of 1.60 kg/kWh, inside the 1–4 published for heat-pump dryers — and the hot-air dryer 985 kWh, 2.04 kWh/kg. The model's saving is 69 % of the energy; the case study's customer counted 30 % of the cost against a baseline it does not describe, and the two are not the same quantity: a recirculating hot-air dryer vents less than this once-through one, and against a coal dryer at a third of electricity's price per kilowatt-hour a 69 % energy saving shows only 8 % on the bill — 30 % would take 77 %. Between the resistance baseline and a cheap-fuel one the example brackets the customer; nothing in it is fitted to the case study, which states no temperature, batch, duration or airflow — all four are declared here.

What it reads on Buôn Ma Thuột

Heat pump	301 kWh per 40 h batch (290–306 over the year's 219), 0.62 kWh per kg of water, SMER 1.60 kg/kWh; 187 MWh at the coil on 66 MWh of electricity a year
Hot air	985 kWh per batch (815–1 143), 2.04 kWh/kg, SMER 0.49; 216 MWh a year
Chamber	mean 55.0 °C, 47.7 at its coldest hour — the exhaust damper bleeds to the setpoint and no further — RH mean 0.24, 1 hour over the 0.030 ceiling, 0 failed hours
The scroll	24.4 °C evaporating / 84.8 °C condensing at the design point, 22.0 kW for 7.3 — 0.2 K under the map's 85 °C ceiling, where the solver warns that the map's published mass flow and its capacity part by 31 %

What it taught

• A closed-loop dryer's energy is the compressor's alone — the latent heat the coil recovers is what the condenser puts back — and the chamber runs over on it, not short: the exhaust damper is the machine's thermostat.
• A dehumidifying coil on the whole stream of hot air spends most of its duty cooling; on less than half the stream it condenses nearly twice the water — face-and-bypass is how a dryer's coil is made to dry.
• A machine that reheats with its own condenser is judged on the humidity ratio for everything — meeting, overshooting, cycling — or it never unloads.
• A customer's 30 % of cost and a model's 69 % of energy are different quantities against different baselines; say which one you are quoting.

The machines

• Heat-pump dryer, 20 kW class — open in the builder

Sources

• Copeland, DeAir JSC — ZWD digital scroll heat-pump dryers (case study): a coffee exporter in Vietnam's Central Highlands, drying cost per batch down about 30 %, 2–4 year payback, the digital scroll's 10–100 % range. No temperature, batch size, duration or airflow stated
• Copeland ZWD81KBC-TFP (R134a, 50 Hz) — the published map and envelope, in the catalog
• PVGIS TMY, Buôn Ma Thuột

Readings of the shipped model on the building's own PVGIS typical meteorological year; engineering estimates from catalog data and published correlations. The building runs on every plan, including Free, a few times a day.