More Data!

A while back, I posted data comparing how effective the DIY purifier was compared to a 6,000 RMB Blue Air and 11,000 RMB IQ Air. In that test, I calculated effectiveness as percent difference from outside air. At the time, I said I didn’t like that way of calculating effectiveness, but I did it that way so I could compare my results with the results of Dr. Saint Cyr.

Now I’m posting data from new tests that improve on the old tests in three ways:

  1. The new tests calculate effectiveness as the percentage reduction from indoor air. This gives smaller percentages because outside air is dirtier than inside air. In my opinion, this is a more rigorous way to test air purifier effectiveness.
  2. The new tests were conducted by my Smart Air collaborator Anna Guo in her apartment. This helps test whether the results are generalizable. In other words, this way we can know that my original results were not a fluke just based on the layout of my room or how well my windows are sealed. Anna’s room (15 m2) is also slightly larger than mine (13.5 m2).
  3. Finally, the new tests are with the Meidi fan I’m now shipping to people from my Smart Air site. Thus, these results are more relevant for people using our DIY filter kits. (However, the Meidi fans are very similar to the blue Kadeer fans I was using before. I haven’t seen a noticeable difference in the results with the two fans.)

Methods: Anna used my particle counter to take a baseline 2.5 micron reading before turning on the filter. Then she turned on the filter, put the particle counter on the other side of the room from the DIY purifier, set it to take hourly measurements, and went to bed.

To calculate effectiveness, we took the average 2.5 micron reading of the four hours prior to waking up and calculated the percent reduction from the baseline reading before turning the filter on.

Filter: We used the same DIY kit with the Meidi fan that we’re shipping from Smart Air:


Results: Here’s what the data looks like. This data is from September 9th:


In these real-world tests, we need to be careful to check that outdoor particulate didn’t go down during the test. If it did, we would think the decrease is due to the purifier when really it’s due to outdoor air changes. To test that, we can overlay the outdoor PM 2.5 readings from the US Embassy:


Outdoor air actually got slightly worse during the test, so the reduction inside is not because outdoor air got cleaner.

And here’s the average reduction in 2.5 micron particles over the three test nights:


Conclusion: The DIY filter removed an average of 92%. That’s similar to the earlier result of 94%. I suspect the new result is 2% lower than before because this new calculation uses reduction from inside air rather than outside air.

Again, the effectiveness ratings are similar to Dr. Saint Cyr’s tests with the Blue Air and IQ Air, although remember that his room is larger (20 m2), and the post does not specify the length of testing time or the particle counter used. Ideally we’d have tests with the same particle counter in the same room.

As always, I’m posting more details on the methods for fellow nerds here:

The raw data is here:


Timing: 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 up for hourly mode. Therefore, I caution against interpreting the exact time it takes the counts to drop too closely.

To understand how quickly the counts go down, minute-by-minute tests like these are much more useful. What is clear is that it’s best to turn on the filter before you plan to stay in the room, if possible.

Outside AQI: Note that the outside AQI fluctuated a lot on the second and third tests. The first test is probably the most reliable because outdoor AQI was very stable that night.

Placement: Per usual, the particle counter was placed on the other side of the room from the fan. Anna made a schematic of the experiment layout:




Leave a Reply

Your email address will not be published. Required fields are marked *