I’ve wanted to know for a long time whether DIY air purifiers are as effective as the Ferrari filters.
In an earlier post, I compared my data to the tests of Dr. Saint Cyr (whose excellent posts inspired me to look into filters in the first place). But I noted that the comparisons were far from perfect because:
- The rooms were different.
- The Cyr post did not specify how long the tests were (and that can make a big difference if you’re looking at times under an hour–see this time comparison).
- The Cyr post did not describe the particle counter or particle size.
How We Tested The Air Purifiers
But now I finally have directly comparable data! That’s because two kind souls donated a Blue Air 203/270E (3,600 RMB) and a Philips AC4072 (3, 000 RMB). That means I could finally test the DIY against expensive brands in the same room, for the same amount of time, with the same particle counter.
To do that, Anna ran 11 overnight tests with the Blue Air and 9 tests with the Philips. As always, I calculated effectiveness as percent reduction in particulates from the room air. Anna tested the air before she turned on the air filter, and then set a Dylos DC1700 laser particle counter to take hourly measurements of the air in her 15m2 room.
Anna used the highest setting on each purifier. (I’m putting the original data and more details about the methods for fellow nerds at the end of this post.)
And (drumroll!) here are the results:
The Cannon removed as much particulate as the expensive machines. Not bad for 450 RMB!
Yet all four filters were making the room air significantly cleaner. For particles 2.5 microns and above, all four removed over 90%. For particles 0.5 microns and bigger, all four removed over 80%. I’m not the first person to say: All you need to significantly reduce the particulate pollution in your home is a simple HEPA filter.
Based on the data, here’s how much you’re paying for each percentage reduction in 0.5 micron particles:
(And that’s not counting the cost of the exorbitantly priced replacement filters.)
Recently, a Chinese news article claimed air filter companies are making “falsely inflated profits.” That fits with this data showing that the Cannon removes as much particulate as the Blueair, yet costs about 10% per percentage of 0.5 micron reduction. Similarly, the Original DIY removes 4% less 2.5 micron particulate and 6% less 0.5 micron particulate than the Blueair on average, yet the Blue Air costs more than 16 times as much.
What About Gas Pollution?
Now, as I’ve said before, particulates are not everything. There are also gases like ozone and nitrogen dioxide (although I’m less concerned about those). Here’s how to know whether your home has serious gas pollution. If it does, tests show that activated carbon removes gases like formaldehyde and benzene.
Open Data and Methods
As always, I’m posting the original data and detailed methods for fellow nerds. Don’t believe me. Check out the data and decide for yourself.
Machines: The Original DIY and the Cannon are the same as the units we are shipping from Smart Air. The Blue Air was the 270E model. The Philips was the AC4072.
Calculating effectiveness: As always (1,2), to calculate effectiveness, I average the particle counts for the last four hours and divide that by the baseline number. As I’ve said before, I think this method of calculating effectiveness is more rigorous than calculating the reduction versus outside because indoor air is cleaner than outdoor air.
Sample timing: Anna took the baseline measurement before she went to bed. Then she set the particle counter to take hourly measurements until she woke up in the morning.
Room setup: The DIY filter was on one side of the room. The particle counter was on the opposite side. Anna made this schematic of the experimental setup:
Anna’s room is 15 meters squared. The windows were closed at all times, and the bedroom door was mostly closed before Anna went to bed, and it was always closed during sleeping hours.
Particle Counter: My particle counter gives counts for (a) particles 0.5 microns and above and (b) particle 2.5 microns and above. For more information on how laser particle counts compare to government machines and AQIs, check out my data comparing the two over 70 measurement occasions. (Spoiler: they correlate at r = .89.)
Filter Life: The owner of the Philips estimated it had been used previously for a few weeks. The owner of the Blueair estimated the current filters had been used for approximately 6 months, which means it was about time to replace the filters.
(I think we tend to overestimate how much the newness of filters matters, except for with very old filters. For example, the Blue Air was still doing an excellent job at 6 months. Keep in mind that, over time, HEPAs actually become better at catching particles because particles “fill up” the spaces in the filter. The downside is that air flow decreases. However, if anyone has a Blue Air with new filters, contact me and we’ll run the tests!)
Interpreting the speed of the drop: The time between (1) turning the filter on and (2) the nearest hourly measurement varies a lot based on when Anna took the baseline measurement and when she set the particle counter into hourly mode. Therefore, it’s difficult to compare how fast the particulates went down.
If you want to try and compare the speed of the drop in particulates in the first few hours, you’ll have to note the time that the filter was turned on (in the far right column). However, for speed comparisons, I think our controlled tests are much more useful (here’s an example). I’ll post more of those in the future.
The original data
Note that the outdoor air during these 11 tests was variable–more variable than I’d like. Ideally, the outdoor air would remain relatively stable. To analyze whether outside fluctuations affected the results, I re-calculated effectiveness by looking only at the 5 days where the outside concentration did not change by more than 50 micrograms from baseline. However, that gave nearly identical numbers for average effectiveness: 91% of 0.5 microns and 96% of 2.5 microns (versus 90% and 96% over all 11 tests). Therefore, I’m confident that outdoor fluctuations did not significantly influence the estimation of effectiveness.
Here’s the data for the Philips:
There was one outlier in the Philips data: the percent reduction was much lower on 11/30 (68% and 83%). This is probably because the outside concentration went from about 25 to 125 on that day–a fivefold increase. Thus, I removed this datapoint from my final calculation of effectiveness. However, you can see the calculation with that datapoint included above.