Real-World Tests of Auto Modes
Smart Air co-founder Anna tested two of the most popular auto modes in China—the Xiaomi 2 and the Philips AC4072.
These tests are simple and realistic. Over 13 days in an ordinary 15m2 Beijing apartment, Anna turned the Xiaomi or the Philips on auto mode before bed.
She used a Dylos Pro laser particle counter to track 0.5 micron and 2.5 micron particles every hour until she woke up in the morning. The Dylos has shown high accuracy (r = .89) in several comparison tests with official PM2.5 numbers (1, 2). I calculated percentage reductions from before turning on the purifier versus the average over the last four hours before waking up.
The pictures below show the Xiaomi at the start and end of a test. Despite running in a closed room all night, the Xiaomi left indoor air in the unsafe range.
The Xiaomi averaged 62% reductions, and the Philips averaged 59% reductions.
For comparison, this is more than 20% worse than what I found by just strapping a filter to a small fan with a Velcro strap.
But wait, maybe the Philips and Xiaomi are just bad purifiers. Maybe they use low-grade filters, or maybe their fans aren’t strong enough.
Is there something with the purifiers?
Our other tests say no. In tests of the high setting, the Philips scored near the top of my earlier comparison tests.
And in tests of the Xiaomi’s high setting, it also performed relatively well but cannot be kept on high. The machine switches to auto mode whether users want it or not at the third hour. Really? You really can’t set the Xiaomi to stay on high?
But is it actually unsafe?
OK, OK, skeptical readers might say, “well a 60% reduction isn’t bad. Isn’t that good enough?”
It’s certainly better than nothing, but based on my reading of the research on the health effects of air pollution, this is meaningfully bad air. On average, indoor air was double the WHO 24-hour PM2.5 limit (and that’s the looser of their limits!).
Why Are Auto Modes So Bad?
The tests on the high setting above showed that it isn’t a problem of the fan or the filter. These machines work well on high. So what explains it?
Cause 1: The Built-In Air Quality Monitor is Wildly Inaccurate
The auto mode is governed by the built-in air quality monitor. The problem is, I systematically tested three different Xiaomi’s against three independent particle counters, and found the Xiaomi monitor was wildly inaccurate. At its worst, it was underestimating actual PM2.5 by 218 micrograms!
So sometimes these machines might think the air issafe when “toxic” would be a more accurate description.
Cause 2: The Xiaomi Has a Loose Definition of “Safe Air”
It could also be that they just have a loose definition of “safe air.” For example, China’s PM 2.5 limit is 35 micrograms, whereas the World Health Organization’s annual limit is 10 micrograms. Is 10 micrograms too low? I think not, because studies have found important health effects even below 20 micrograms.
As always, I’m publishing the original data and more details on the test method for fellow nerds. They are available here.