Window AC Sizer

Use when buying or troubleshooting a room air conditioner — Manual-J-style BTU sizing from room geometry, sun exposure, insulation and occupancy, with the oversizing trap (short-cycling => no dehumidification), 110V vs 230V circuit checks, and noise/heat-pump guidance.

Overview

The most common window air conditioner mistake is not undersizing — it's oversizing. A too-big unit hits setpoint in minutes, shuts off, and never runs long enough to pull water off the coil: the room ends up cold, clammy, and mold-prone, with the compressor thrashing itself to death. Yet store shelves and "just buy 10,000 BTU" folklore push people exactly there. Correct sizing is a 5-minute load estimate: area, ceiling height, sun exposure, insulation quality, occupants, heat-producing appliances, and kitchen load — plus honest adjustment factors instead of fear-driven padding.

This skill computes a Manual-J-flavored room load (with every assumption printed), converts it to a recommended BTU band with the classic +600 BTU/person-over-two and +4,000 kitchen corrections, checks the oversize ceiling (never more than ~15% over calculated load), flags the short-cycle signature for existing units, and covers the practical purchase gates: 110V/15A vs 230V circuits, window type compatibility, CEER efficiency math ($/season between two models), noise placement, and the heat-pump-vs-AC question for shoulder seasons.

When to Use

• Buying a window/portable AC for a bedroom, living room, office, or rental
• Existing unit "cools but the room feels damp" — short-cycling diagnosis
• Comparing two models' CEER and runtime cost at your electricity price
• Deciding between a second small unit vs one big one (spoiler: zone them)
• Landlord/rental constraints: no permanent install, window types, noise rules

Don't use for: whole-house central AC replacement (needs real Manual J from a contractor), multi-split design, or garage/server-room cooling with continuous high internal loads — those need different load assumptions, not a bigger fudge factor.

Quick Start

bash

# Bedroom: 12x14 ft, 8 ft ceiling, shaded, average insulation, 2 people
python3 scripts/window_ac_sizer.py size --length 12 --width 14 --ceiling 8 \
    --sun shaded --insulation average --occupants 2

# West-facing living room with big windows, poor insulation, kitchen adjacent
python3 scripts/window_ac_sizer.py size --length 18 --width 14 --ceiling 9 \
    --sun west --insulation poor --occupants 3 --kitchen yes

# Is my existing 12,000 BTU unit the reason the room is clammy?
python3 scripts/window_ac_sizer.py check --length 12 --width 14 --ceiling 8 \
    --sun shaded --insulation average --occupants 2 --unit-btu 12000

# Season cost compare: 8,000 BTU CEER 12 vs 10,000 BTU CEER 15
python3 scripts/window_ac_sizer.py cost --btu 8000 --ceer 12 --price 0.17 \
    --hours 8 --days 120

The Load Model (transparent, not magic)

text

base = area_sqft × 20 BTU/h            (the classic starting point)
ceiling_adj = base × (ceiling_ft − 8) × 0.05      per extra foot (5%/ft)
sun: heavily shaded ×0.90 | average ×1.00 | sunny ×1.10 | west-facing ×1.15
insulation: good ×0.90 | average ×1.00 | poor ×1.15 | very poor (drafty old) ×1.25
occupants: +600 BTU/h per person beyond the first two
kitchen: +4,000 BTU/h if used for cooking
recommended band = load × 1.00 … × 1.15        (NEVER round "up for safety" beyond this)

Every factor exists because heat gain is real: west glass loads in the afternoon, a drafty 1900s wall assembly leaks both ways, a kitchen exhausts a meal's worth of heat into the adjacent room. The model prints the full stack so you can argue with any single line.

The Oversizing Trap (the core teaching)

An AC does two jobs: drop temperature, and condense humidity. The second needs sustained coil time. Size a 6,000 BTU/h room at 12,000 and you get: 5-minute blasts, cold-but-wet air, compressor short-cycling (wear, efficiency collapse), and 60%+ RH that feels warmer than the thermostat claims — so you lower the setpoint and make it worse. The check subcommand quantifies oversizing and names the clamminess mechanism.

The one legitimate oversize case: if the room's load estimate sits within 10% below a standard size (5k/6k/8k/10k/12k/15k...), buy the size up — but not two sizes up.

Purchase Gates (before you click buy)

1. Circuit: ≤8,000 BTU @110/120V on a standard 15A outlet (dedicated preferred); 10–15k often 230V (check for the right receptacle — a 230V unit on an adapter is a fire meme, not a plan); 12k+ @110V pulls ~11–12A — don't share the circuit with a gaming rig.
2. Window: double-hung fits most bracket-mounted units; casement/sliding windows need specifically designed units or a plexi panel kit. Measure width AND the free sash opening height.
3. Weight & install: 55–90 lb; sleeve vs bracket; landlord rules for ground-floor street-facing units.
4. Efficiency: CEER (combined, includes standby) — compare $/season with the cost command at YOUR price and hours, not the Energy Guide national average.
5. Noise: bedroom target ≤ 55 dB max / low-50s dB on low fan; inverter-driven units (variable speed) run quieter and dehumidify better because they modulate instead of cycling.
6. Heat pump variant: if you'd also use it for spring/fall heating, a window heat pump handles the shoulder seasons and dodges resistive-heat cost — but check its heating rating at your design low.

Common Pitfalls

1. "Bigger is safer" — the #1 error; cold + clammy + dead compressor in 4 seasons.
2. Sizing by room type folklore ("bedrooms need 8k") instead of the actual envelope.
3. Ignoring west glass — a 12×14 west room can need 40% more than the same room facing north.
4. Two rooms, one big unit — an open doorway halves delivered cooling to the far room; two right-sized units beat one oversized.
5. Portable "14,000 BTU" single-hose units — the hose dumps conditioned air outside, pulling hot air IN; effective capacity is often 40–50% of the label. Dual-hose only, if portable at all.
6. Skipping the tilt/sleeve drainage check — modern units often route condensate to the condenser fan slinger; mounting dead-level when the manual wants a slight tilt floods the room side.

Verification Checklist

• Load stack printed with every factor (argue line by line, not the total)
• Recommended band vs nearest standard sizes: within +15%, never 2 sizes up
• Circuit voltage/amperage confirmed for the chosen model
• Window type + sash opening measured
• CEER season-cost comparison run at local kWh price
• Bedroom units: dB on low fan checked against sleep needs

Example Session

text

$ python3 scripts/window_ac_sizer.py size --length 12 --width 14 --ceiling 8 \
    --sun shaded --insulation average --occupants 2
Base load: 168 sqft x 20 = 3,360 BTU/h
Sun (shaded): x0.90 -> 3,024 | Insulation (average): x1.00 -> 3,024
Occupants: 2 (no add) | Kitchen: no
CALCULATED LOAD: ~3,024 BTU/h
Recommended band: 3,024 - 3,478 BTU/h -> buy 6,000 BTU (nearest sane size,
within +15% doctrine NOT met: explain 5k/6k gap; 6k = oversize +98% -> prefer
inverter unit that modulates down, or accept slightly clammy corners on design days)

Note: the script names the honest gap between tiny room loads and the smallest market sizes — inverter modulation is the modern fix, and the output says so instead of pretending.