Set each profile by PWM % or by a noise target, and read off the PWM % to send from Home Assistant for each fan. RPM is there to check against the fan's tach reading. Home Assistant sets duty, not RPM.
Independent measurements from HouseFresh and the CADR tests it cites. The fan make and model aren't published: it uses four PC-style fans on top, spinning up to 2,900 RPM, pulling through four H11 HEPA panels about 11.4×11.4 in each.
| Setting (adapter volts) | Clean air (CADR, CFM) | Noise at 3 ft (dBA) | Power (W) |
|---|---|---|---|
| 3.6 V (lowest) | — | 36.1 | 2.2 |
| ~6 V | 213 | 43.3 | 9.0 |
| ~9 V | 328 | 49.1 (8.6 V) | 17.5 (8.6 V) |
| 12 V (top) | 413 | 56.3 (12.6 V) | 43.2 |
Four fans have to push 400+ CFM through dense HEPA, so each one runs near 2,900 RPM. DIY boxes do the same job with 6–13 fans at 35–50% speed through larger MERV 13 filters. Noise climbs steeply with fan speed, so many slow fans win.
The AirFanta figures are clean air, at one meter-ish in a room. The fan figures above are raw airflow per fan, in a lab chamber. For a rough match, a DIY box's clean air is about half to three quarters of its airflow with MERV 13. Its noise in a real room will read higher than the lab numbers. Put one phone meter at 3 ft from each and you'll have a fair head-to-head.
Sources: HouseFresh AirFanta 3Pro review · AirPurifierFirst review (measured 40–62.5 dBA across speeds 1–6, a different method) · Jeff Kaufman comparison
You won't run these at 100% in a bathroom. Every fan gets throttled to whatever noise you'll tolerate, so the fair question is how much air each moves at that noise. Full-speed spec sheets hide this.
The filter pushes back harder the more air you move. For each speed step the page finds where the fan's measured pressure curve meets the filter's resistance line. It uses each filter's verified manufacturer resistance and the fan's full measured curve instead of a straight line. Higher-pressure fans lose less here.
Five fans are fully lab-measured. The NF-P14s redux-1500 uses the lab curve of its 1200 RPM sibling, which has the same blade and frame, and only its top end is extrapolated. The TL-C14C was never lab-tested. It's modeled from Thermalright's spec with a +5.5 dB correction, because spec sheets read 3.7–7.6 dB quieter than Cybenetics. Hatched bars and dashed lines mark modeled data.
Noise was measured in free air. Behind a restrictive filter, fans usually get a bit louder at the same speed, more so with a hard-working fan. Box leaks and bends aren't modeled. The ARCTIC Pros only stop below 5% PWM (ARCTIC's spec). Up to 20% they read at the chamber's 6 dBA floor, so they're effectively silent. Their pressure stops rising above 80%.