Inverter capacity range: why the minimum matters most

A published cooling capacity is three numbers, not one. Most specification sheets print the nominal figure between a low and a high value, in kW or in Btu/h. The nominal is the one that reaches the quote and the showroom sticker. The other two mark the ends of what the machine can hold.

The nominal figure is a measured result from one fixed condition. ISO 5151:2017 fixes that condition for a single split. The chamber holds 27 °C dry bulb indoors against 35 °C outdoors, with wet bulbs of 19 °C and 24 °C. NEA adopted the same standard and class for regulated single-split units from 1 April 2021.

The maximum is the ceiling at that same condition. Compressor top speed sets part of it. Coil area and the air the fan can move set the rest. It describes the machine asked for everything at once.

The minimum is the lowest output the machine can hold without stopping. ISO 16358-1 defines the test for it in almost those words. The minimum capacity test runs at the lowest capacity control setting which allows steady-state operation. Below that setting there is no steady state left to measure.

All three figures move together when conditions change. ISO 16358-1 carries correction factors for exactly that. At 29 °C outdoor air, cooling output is taken as 1.077 times the 35 °C value, drawn on 0.914 times the power. A rated airflow figure carries the same kind of small print.

The rating condition is hotter than most Singapore afternoons. ISO files 35 °C under class T1 and describes that class as moderate. Its hot-climate class sits at 46 °C. The Meteorological Service describes the local daily range with a maximum not rising above 31 to 33 °C during the day.

Operating hours therefore sit below the nominal figure almost all of the time. A bedroom at two in the morning carries a fraction of the load it carried at three in the afternoon. People leave, the sun comes off the wall, and a modulating machine follows the load down to its own floor.

What happens at that floor is decided by the published minimum. If the room asks for less cooling than the figure, the compressor cannot balance against it. It stops, waits and restarts, which is short cycling and carries its own consequences.

Moisture is what raises the stakes locally. Mean relative humidity runs around 82 percent, and above 90 in the hour before sunrise, by the Meteorological Service's own account. Indoors the rating pairs that same 27 °C reading with a wet bulb of only 19 °C, which is drier air than a room here usually holds. The test understates the drying half of the work.

Low output does not guarantee good drying, and the evidence is more conditional than the marketing. A Department of Energy contract report on part-load dehumidification found that reducing refrigerant flow lets the coil surface run warmer. That raises its dew point and shifts the work toward cooling and away from removing water. The controlling variable is the ratio of air volume to capacity.

Where airflow falls at least as fast as output, the coil stays below the entering dew point and the drying holds up. Read that as a mechanism to ask about, and not as a measured result for a wall unit. The systems studied were ducted American installations.

The span between minimum and nominal is the comparison worth having, and no manufacturer publishes it. Divide nominal by minimum and the result is a turndown ratio. It says how far a machine can back off before it runs out of adjustment. That figure appears in no specification sheet, databook or service manual checked for this article.

Makers plainly think in these terms and stop short of printing the number. Mitsubishi Electric's databook arranges its performance tables in blocks headed maximum frequency, rated frequency and minimum frequency, so the concept is labelled in their own pages. The Hz values behind those headings are left out. Samsung, Midea and Gree publish no frequency range either, and Gree gives a frequency only at the rated point.

The arithmetic is simple once both figures are in hand. Take the Mitsubishi Electric product information sheet covering the MSZ-AY and MSZ-AP wall-mounted range, effective February 2026. It prints cooling capacity as a nominal figure with the band in brackets. The figures below belong to those models and that document, and the third column is arithmetic instead of a published value.

Every maker publishes the nominal figure. What happens to the other two varies by company, by market, and sometimes between two documents from the same company.

One Daikin product sheet gives a ceiling and no floor. The document covering the FTKM and RKM PVMK cooling-only range carries a row headed nominal and maximum cooling capacity, and lists the 18,000 Btu/h model as 18,000 and 20,500 Btu/h. No minimum row appears anywhere on that page. The same page adds a separate actual maximum of 18,000 Btu/h taken at the hot-climate condition, so the sheet is detailed. The floor sits outside what it discloses.

Toshiba publishes the minimum somewhere a buyer would not look. Engineering data for one residential inverter range gives capacity in the specifications section as a single nominal figure. The minimum turns up later, in the part-load section, under a heading of operation range, and in watts while the specification page works in kilowatts. Comparing specification pages side by side would suggest no floor is published. One is.

Gree splits its own practice by market. Service manuals for North America carry a cooling capacity row written as minimum to maximum. Manuals for the same class of machine sold into Europe, Australia and other export markets give one figure and no minimum at all. Same maker, same machine, different disclosure.

Units and document type shift as well. Samsung keeps a minimum, standard and maximum row as a standing feature of its residential technical data books, in kW and Btu/h together. Midea's ranges appear almost entirely in Btu/h and in market catalogues, and its service manuals from 2019 onward dropped the full specification sheet. Hitachi names the method on the page, stating that its nominal capacities are based on ISO 5151.

Start by getting the model number off the outdoor unit itself. A published range belongs to one pairing of indoor and outdoor units, and it is filed under that pairing. Without it there is no row to look up.

Then ask for the minimum in the same units as the nominal. An answer in watts against a nominal in kW is normal and needs converting before any comparison. An answer that no minimum is published is also a real answer, and it means the comparison has to rest on something else.

Work out the turndown and write it beside each model on the list. Nominal divided by minimum, to one decimal place. The machine with the lower floor holds a small load without stopping, and two units sharing a nominal figure can sit far apart at that end.

Treat the maximum as information about pull-down and nothing more. It describes the opening stretch of a run, when the room is furthest from the setpoint. Sizing to it means sizing to a condition the machine leaves behind within the hour.

The range cannot settle the sizing question by itself. It describes what the machine can do. The heat load describes what the room asks for. Both are needed. Only the first reaches the carton.

A quote that leads with the maximum has chosen the least representative number available. A seller who can produce the minimum, the nominal, and the conditions all three were measured under has handed over something worth working from. That beats a conversation held over a size bracket alone.