DIY Air Purifier - DIY Air Purifier

DIY Air Purifier

Last Update: April 2019

Table of Contents


As explained in our Introduction to Pai on, a sad but true modern reality all across Northern Thailand, Burma, and Lao is that, beginning around Feburary-March of every year and lasting until the first long rain in April/May, the hot season brings absurd levels of pollution caused by intentional burning of fields and hills:

The pollution includes jaw-dropping, China-levels of the most dangerous PM2.5 particles that are so high that they literally break the AQI pollution scale defined by the US EPA, which only goes to 500. Here are a few of the many days in April 2019 that were above 500 AQI (for the current pollution levels in Pai and Chiang Mai, see

PM2.5 US AQI levels above 500

During this period, PM2.5 concentrations (which are different from AQI but 500 concentration is about equal to 500 AQI) are regularly measured above 600ug/m3 and sometimes as high as 858 ug/m3 (4:22am March 30, 2019):

PM2.5 US AQI levels above 500

Vast numbers of especially old patients crowd the hospitals with lung-related conditions.

Without the use of masks or air purifiers, levels inside even well-sealed rooms will still regularly rise to 200-300 ug/m3, which is 4 times the Thai government safety standard of 50 ug/m3 and 8 times the WHO safety standard of 25 ug/m3.

While this insane burning obviously decimates the tourism industry, at least the foreigners and some more wealthy Thais have the option to flee Northern Thailand for cleaner air in the islands and beaches of Southern Thailand.

However, the vast majority of the thousands of citizens and immigrants living in Pai District cannot simply leave. In some cases, their work or family tie them down, in some cases due to their immigrant status they are actually forbidden from leaving the district, and in some cases money is the main factor.

Those people must stay and, ideally, use masks and room air purifiers.

Proper PM2.5-blocking N95 masks are somewhat affordable (when they are not sold out!) but of course are a nuisance for indoor use.

But there is a big problem with room air purifiers.

There is currently a price gouging going on in the air purifier industry that puts all the units out of reach to most people in Pai valley (cost of a unit is minimuim 2500 THB but more often 6000-10000 THB, plus 1000 THB per 2-3 months for an overpriced replacement filter in a proprietary size/shape that prevents substitution).

In reality, an air purifier is just a box with a fan, a HEPA filter, and maybe some (largely unnecessary) extras like a PM2.5 meter or a bluetooth interface to a privacy-intruding mobile app.

This has given risen to creative methods of building lower-cost DIY air purifiers.

On this page we present one such design that uses the typical floor/desk fan that almost everyone already owns, along with some of the design process and performance measurements using a PM2.5 measuring device.

Then we present some of the other cool ideas we've found along the same lines.

We hope that this site will help people share, adapt, and improve DIY air purifier designs in order to give the most people in the world as possible access to clean air.

If you have comments or more ideas, please contact us at:

How To Build It

DIY Air Purifier 1/8

DIY Air Purifier 2/8

DIY Air Purifier 3/8

DIY Air Purifier 4/8

See: Where to Buy HEPA Filters in Thailand

DIY Air Purifier 5/8

DIY Air Purifier 6/8

DIY Air Purifier 7/8

DIY Air Purifier 8/8

How To Build It: ภาษาไทย

เครื่องฟอกอากาศทำเอง 1/7

เครื่องฟอกอากาศทำเอง 2/7

เครื่องฟอกอากาศทำเอง 3/7

เครื่องฟอกอากาศทำเอง 4/7

See: Where to Buy HEPA Filters in Thailand

เครื่องฟอกอากาศทำเอง 5/7

เครื่องฟอกอากาศทำเอง 6/7

เครื่องฟอกอากาศทำเอง 7/7

Where to Buy HEPA Filters in Thailand

HEPA filters are always in short supply in the burning season. Be sure to stock up ahead of time.

Make sure the filter you are buying has dimensions of at least 30cm x 30 cm (you can buy smaller filters and tape them together if necessary) and is accordion-folded, not just a flat sheet. Also take note of how thick the accordion-folded filter is: thicker means more surface to catch particles and thus better filtering and longer life.

No matter what, do not cheap out on the filter. Only a true HEPA filter blocks PM2.5 particles: cloth, foam, and car air filters do not work. You cannot make a HEPA filter yourself.

Other non-HEPA filtering features like "carbon" filtering are OK (they filter stuff other than PM2.5 particles) but might actually reduce the amount of useful PM2.5 filtering you get, because they make it harder to push air through the filter. Probably better to get a basic HEPA filter.

Retail shops:

Global Home, Home Pro, Lotus/Tesco, Baan and Beyond, ...


When searching online, try searches like:

Obviously, Lazada and Taobao are going to have a lot of fakes (filters that are not really HEPA and don't really block PM2.5) so be wary and check reviews. If you buy a sketchy filter, be sure to get/borrow a PM2.5 meter to make sure it's really working.

Lazada ships HEPA filters from Bangkok (1 week delivery) and China (3 weeks delivery), with all the Bangkok ones running out of stock once the burning starts.

Amazon in the US, and Taobao in China, both ship some items to Thailand. It's more expensive of course, but it may be your only choice if Lazada is totally sold out.

Some example links (NOT checked for whether they are fake or real):


General "HEPA Filter" Search 190B and up

General "HEPA กรอง" search 190B and up

30cm x 30cm DIY one for 329B

Two of above for 203B each

36cm x 28cm one for 202B

Lazada smaller ones (can tape together as needed):

Two-Pack of 25.7cm x 16.5cm for 209B

24.8cm x 20cm one for 194B

26cm x 16.5cm one for 197B

Taobao (Chinese site; SmartAIR-branded filters; ask a Chinese person for help):

Sample Filter 1

Sample Filter 2 US (many choices and some will ship to Thailand):

Search for "HEPA replacement filter" without the quotes


IAQ Virus Keeper Nano-Silver Filter (they claim it blocks PM2.5; we could not corroborate this from other sites)

DIY Design and Performance

Testing Standards

Testing location: all tests done in the same 32 square meter space (twice as big as that recommended in the images above), all 12 windows taped on outside, door weatherstripped and holes in door taped, door to bathroom (which is exposed to outside air) held shut with wooden shim, aircon off (ouch), other fans off (ouch). Unfortunately the space has a wood floor exposed to the air below and I was not able to get carpet or linoleum to seal this obvious dirty air entrance.

PM2.5 Testing device: A roughly 600-700 THB portable PM2.5 Chinese measuring device similar to this one on Lazada. The accuracy seems reasonably good because the unit was always within about 5% of another PM2.5 meter in two different Xiaomi air purifiers.

"Healthy" Levels: the WHO healthy level of PM2.5 particulates is 25 ug/m3, with the Thai government standard of course being double that at 50 (because, you know, physics is just different in Thailand).

Testing standard: all tests are done with the air purifier (whether commercial or DIY) on the corner of my bed "pointed" at my desk, with PM2.5 sensor about 2.5 meters away, at about head level, not in the airstream from the purifier device. The PM2.5 sensor's intake grille is pointed towards the nearest window and away from the purifier device's airstream. All "outside air" measurements below are from just inches outside the front door of the room where testing took place (they are not numbers from AirVisual or any other public service).

Note: it's very important not to test with the PM2.5 sensor right in the airstream of the purifier device (right next to the filter material, as seen in so many "magic" filtering videos and ads) as that doesn't tell us about the overall ability of the filter to clean the air in a whole room, which is what we really care about. We already know that the air coming out of a HEPA filter is 97% free of PM2.5 or more. What we really care about is the ability of the purifier to suck in and clean air from all over the room, and "keep up" with new dirty air being added into the room.

Comparison Commercial Purifier: A Hatari HT-AP12. Cost about 6000 THB, 5000 THB on sale. Filter had been used for about 2 weeks but was not visibly darkened by dust yet. Tests always performed with fan on maximum setting.

Testing filter A: The 36cm x 37cm accordion-fold HEPA filter from a Hatari HT-AP12. Filter had been used for about 2 weeks but was not visibly darkened by dust yet.

Testing filter B: A brand new 30cm x 40cm Gruenluft HEPA filter from Lazada.

A Note on Fan Types

The Hatari commercial air purifier, like many air-blowing devices, employs a centrifugal fan, which produces much more air pressure against an obstruction (like the filter) than the normal everyday axial fan like the desk/floor fans we all have already. Cross-flow fans also have a similar strength.

If money and access were not an obstacle, and especially if electrical efficiency or noise reduction were the main goal, we would clearly use centrifugal or cross-flow fans.

But money and access are the primary issue. Centrifugal/cross-flow fans are expensive, hard to find, and hard to work with, whereas nearly everyone already has an axial desk/floor fan. Even consumer bathroom/ceiling exhaust fans are always axial fans here in Thailand.

A typical floor/table fan here costs about 200-300THB.

Our goal here is to increase access by seeing how good of an air purifier we can make with the type of fan that people already have.

Interestingly, the popular Xiaomi 2S commercial air purifier uses a custom-made axial fan with two blades in a row (one blade on either side of the motor, on the same shaft) where the custom blades nearly touch the edge of the air pipe they are mounted inside.

No box fans available: the square "box fans" so popular in America, with the nice flat front and back grille, are almost impossible to find in Thailand. Nobody has them (though a few small models are available on Lazada). So we did not create a design with those.


First I had to establish a baseline for a test.

Using the Hatari HT-AP12 with the fan on the maximum speed (speed #4):

    Before test outside air: 244 ug/m3
    Test measurements:
    10:58 97
    11:00 99
    fan on
    11:02 108
    11:05 86
    11:06 86
    11:08 81
    11:09 73
    11:10 73
    11:12 74
    fan off
    11:14 74
    11:15 76
    11:16 81
    11:18 82
    After test outside air: 275 ug/m3
So the Hatari brought the PM2.5 from about 100 to about 75 in 20 minutes (-25 ug/m3 in 20 minutes).

NOTE: As I later discovered in this cool analysis, doing the test for only 20 minutes doesn't mean that much because even the commercial devices took the PM2.5 down about 66% (in their 15 square meter room) during the first 1 hour and to 97% after about 2 hours. So eventually I need to re-run the tests for at least 2 hours. But this is the data we have for now.

Prototype #1: Filter in Front

I used cardboard and a bunch of tape to build an air guide that forces all the air coming out of my trusty 48cm diameter desk fan through the filter. There was no air guide behind the fan as shown in the final design above.

Prototype #1: Filter in Front

With the fan on full power, using testing filter A (the filter from the Hatari), there was barely any airflow at all coming out of the filter and a lot of air being ejected backwards out the back of the fan. The airflow was not even enough to move a piece of that Thai cellophane package-wrapping ribbon.

    Before test outside air: 194 ug/m3
    Test measurements:
    9:25 52
    fan on
    9:27 54
    9:29 50
    9:31 43
    9:33 35
    9:35 41
    9:37 39
    9:39 46
    9:41 41
    9:43 40
    stopped fan
    9:45 50
    9:47 51
    9:49 47
    After test outside air: 198 ug/m3
So prototype #1 brought the PM2.5 from about 52 to about 40 in 20 minutes (-12 ug/m3 in 20 minutes).

Prototype #2: Filter in Back

I rejiggered it so the filter was before the fan, which was a pain because the neck of the fan now hits the air guide and needed a cutout, as seen in the final design above.

Prototype #2: Filter in Back

An advantage of configuration #2 is that now the fan can stand on a table without falling over due to imbalance.

But a huge potential problem with #2 is that since there is little airflow going on around the fan motor, the motor is free to heat up more and more and more without any airflow to carry the heat away. And the motor is under even more severe stress than normal because of the filter. I wasn't able to actually reach in and see how hot the motor was while it was running (since the air guide blocks me from every angle) but after I took it apart once, the motor was quite hot.

Now, with the fan on full power, again using testing filter A (the filter from the Hatari) there was a tiny bit of airflow going into the filter (not enough to feel with your hand, but enough to suck a little strip of Filtrete so it sticks). On the output side, there was actually air being sucked INTO the fan near the center of the fan, and there was a lot of airflow coming out of the fan along its circumference, but the air was actually flowing back towards the back of the fan where the filter was!

It seems certain that most of the air being shot out the circumference came in at the center and did not pass through the filter.

One problem in air purifier design is to try to prevent the already-filtered air from getting sucked back into the filter, which wastes energy and makes the filter device work more slowly. I think configuration #2 actually encourages this, because most of the airflow coming out of the front was actually pointed back to the back again so more likely to get sucked up.

Notice that the outside pollution was MUCH worse for this test, meaning that the air inside the room was slowly getting worse and worse even if I did nothing. I had been running my aircon (with a Filtrete filter in it) before the test while building the air guide, and I stopped the aircon for 10 painful minutes before running this test.

    Before test outside air: 366 ug/m3 ARGH!
    Test measurements:
    12:02 85
    12:04 96
    12:06 90
    fan on, with fan cage sitting right on the bed as with the previous test
    12:08 81
    12:10 81
    12:12 83
    12:13 80
    12:15 77
    12:17 77
    12:19 70
    12:21 69
    12:22 62
    12:23 70
    12:25 68
    12:27 66
    12:29 64
    12:31 70
    12:32 62
    12:33 67
    put fan upright, still pointing at same desk corner
    12:35 66
    12:37 69
    12:39 74
    12:41 74
    12:43 73
    12:45 75
    put fan back down again
    12:47 75
    12:49 75
    12:51 75
    12:55 72
    After test outside air: 415 ug/m3 KILL ME NOW
So prototype #2 brought the PM2.5 level from about 90 to about 66 after 20 minutes (-24 ug/m3 in 20 minutes) or about the same as the Hatari. But then it seems outside pollution was so bad that prototype #2 could not maintain the same PM2.5 level and it crept up to 75.

This is promising in that the numbers are similar to the Hatari, but I need to re-run the tests for at least 2 hours to have a real meaningful result.

Prototype #2 Overnight Test

I left prototype #2 running overnight (but on fan speed 2/3 instead of 3/3 because I was really worried about the fan motor overheating).

When I woke up in the morning, PM2.5 levels were above 500 ug/m3 outside (!) and the air inside was at 120 ug/m3.

So prototype #2 was not able to bring PM2.5 levels below 120 even with a long time running.

I put the filter material back into the Hatari and it brought PM2.5 levels down to about 60 within 20 minutes and maybe 40 after 1 hour, albeit with the Hatari and the sensor in different positions in the room than the tests above.

So in this case the Hatari was vastly superior.

Prototype #3: Cannon

My next idea is to put an air guide at both the front and the back, and put the filter in the front (reducing the motor overheat problem), so the air guide forms a "Cannon" with the fan blades in the middle. This is essentially the final design shown above.

Prototype #3: Cannon

Based on this very cool analysis it seems important to have an air guide on both front and back to prevent air from "bouncing" out of the same side of the fan, as I was seeing so much. The guy who made that post actually has a business selling DIY purifiers in China, Philipines, and India (more on that later), and his current favored design also puts the fan in the middle of a big tube that extends both ways.

Also I finally did longer tests.

Using a strip of Filtrete as a wind vane showed that the air entering and exiting the purifier was much more sane. There was still some air bouncing back out the intake, but nowhere near as much as when there was no back guide.

And more importantly there was noticeably more air coming out of the filter (which was downstream of the fan), but still nowhere near the amount of air that comes out of the Hatari. Since this test with prototype #3 used the exact same filter A as the Hatari, I think it is because the Hatari unit is using a centrifugal fan as mentioned above.

Still, at least there is some air.

I ran the test for 2 hours.

    Before test outside air: 153 ug/m3
    Test measurements:
    08:51 52
    fan on
    08:55 53
    09:00 53
    09:05 37
    09:14 36
    09:21 36
    09:26 36
    09:30 33
    09:44 31
    09:49 29
    09:55 26
    noticed that about 8" of tape at filter had breached, fixed it
    forgot to measure difference in flow after fix
    10:05 31
    10:12 32
    10:18 35
    10:25 32
    10:33 34
    10:40 34
    10:49 34
    10:51 40
    After test outside air: 174 ug/m3
So prototype #3 brought the PM2.5 level from about 52 to about 36 after 20 minutes (-16 ug/m3 in 20 minutes) and to 26 after 1 hour (-26 ug/m3 after 1 hour). After 1 hour, I think the slightly rising pollution outside might have been affecting the result because we crept back up to 40 after 2 hours (-12 ug/m3 in 2 hours).

So, kind of disappointing. Prototype #3 worked same or a little worse than prototype #2 and crept up after like prototype #2 did.

Prototype #4: Cannon with new Gruenluft filter B

Air quality is truly hideously awful right now (US AQI 423 in Vieng Neua and 623 ug/m3 outside my door, which is WAY over AQI 500) so it provides good test conditions...

And there is some good news...

I kept the same cannon configuration and switched from using Filter A (from the Hatari) to Filter B (30cm x 40cm Gruenluft filter from Lazada).

Filter B cost more than I wanted (770B) due to massive over-demand all over Northern Thailand. It was the ONLY HEPA filter available in less than 2 weeks. All shops across Chiang Mai were totally out of stock of all filter material, as were all Bangkok-based Lazada stores. The only Lazada shops with stock were in China with no express shipping option.

Lazada has quite a few comparable (usually larger) filters available from China at a cost around 200-400B (see examples).

Next year I hope to be able to test those out.

Surprisingly, when I put Filter B into the Cannon, suddenly there was a LOT more airflow (still not as much as the Hatari but getting in the same ballpark).

Not sure if this is only because the filter is brand new (remember the Hatari filter A only had 2 weeks of use on it), or because there is something inherent about the Gruenluft filter B that it lets more air through compared to the Hatari filter.

And the numbers were much more impressive...

I ran the test for 1 hour, after which the number seemed to bottom out:

    Before test outside air: 420 ug/m3 OUCH
    Test measurements:
    10:42 150
    fan on
    10:43 162
    10:44 133
    10:45 118
    10:47 103
    10:49  95
    10:51  89
    10:55  69
    10:59  74
    11:01  71
    11:06  75
    11:08  75
    11:11  65
    11:15  57
    11:17  52
    11:18  58
    11:23  61
    11:25  64
    11:29  60
    11:31  61
    11:33  61
    11:34  66
    11:42  61
    After test outside air: did not measure; continued onto prototype #5 below
So prototype #4 brought the PM2.5 level from about 150 to about 71 after 20 minutes (-79 ug/m3 in 20 minutes) and to 61 after 1 hour (-89 ug/m3 after 1 hour) which is way way more than the commercial Hatari device (remembering that the Hatari has 2 weeks of dust stuck in its filter; the Hatari device still kicks Prototype #3/#4's ass when using the same clogged Hatari filter A).

So this is actually the first DIY filter device in my attempts that seems like it might be somewhat useful.

It's not clear whether the Greunluft filter will quickly become clogged like the Hatari one did, leading to the disappointing performance of Prototype #3, or whether it is something else about the Greunluft filter that allows more airflow. Only way to test that is to keep using the Gruenluft filter.

But since the Hatari Filter A is not visibly dirty, the much better performance seems more likely to be connected with some other aspect about the design of the Greunluft Filter B. Perhaps Filter B lacks some other layers of filtering (not PM2.5) that Filter A has, and those extra layers were adding unnecessary drag.

Prototype #5: HumanCentiFan with Filter B

I had noticed that the Xiaomi 2S filter device has a single fan motor but two sets of axial fan blades in a row (mounted on either side of the motor shaft), and that is how it is able to get enough pressure despite being an axial fan. I was thinking about a way to hack a regular fan to have two blades in a row.

Then a friend had the idea...why not use two fans?

So was born the idea of the HumanCentiFan (seen in the SUPER POWER image above).

Prototype #5: HumanCentiFan

Having a second fan increases the overall cost of the project from:

300 (fan) + 300 (filter, bought ahead of time) + 100 (tape and cardboard) = 700B


300 (fan) + 300 (fan) + 300 (filter, bought ahead of time) + 100 (tape and cardboard) = 1000B

but in most cases Thai people will already have 1 fan (-300B) or maybe even 2 fans (-600B).

I continued the test above, but shoved another fan up the butt of Prototype #4 and cranked both fans to the max setting.

When turning on the second fan, the small piece of filtrete I have taped to the outlet shows me there was an increase in pressure. I tried temporarily adding a third air guide upstream of the second fan, and that increased pressure a bit more, but I did not leave this third air guide on (too lazy to cut a hole for the neck of the second fan).

I ran the test for 1 MORE hour after the test above:

    Test measurements:
    11:50  63
    11:53  70
    11:56  67
    11:59  54
    12:03  53
    12:06  42
    12:07  47
    12:11  50
    12:12  49
    12:15  55
    12:21  47
    12:27  47
    12:34  53
    (second fan fell off and I kind of retaped it; may have been intake air blowing towards the sensor at this point)
    12:37  40
    12:40  46
    12:42  45
    12:43  44
    12:53  50
    12:55  55
     1:06  60
     1:20  52
     1:27  51
     1:35  45
    After test outside air: 623 ug/m3 <-- WTF?
Prototype #5 (with the first hour being Prototype #4) brought the PM2.5 level from about 150 to around 50 after 2 hours (-100ug/m3 after 2 hours including time as prototype #4 and more time putzing with the fans).

A bit silly-looking and obviously a huge waste of energy, but the cleaning performance is not bad.

The extra fan really only brought 10 ug/m3 extra cleaning according to the readings, though keep in mind the already-awful air (420ug/m3) was getting ludicrously more awful (623ug/m3) during this 2-3 hour period. So the advantage of the second fan might be a lot more actually.

While more exhaustive testing is needed to be sure, I think it is safe to say the HumanCentiFan outperforms the Hatari compared with its performance in the 20-minute and overnight test, especially since Prototype #4 already outperformed the Hatari.

Final Design

The final design seen in the images above essentially shows Prototype #4 (cannon) with Prototype #5 (HumanCentiFan) as an option for those with two fans.

The main difference between the final design and my testing situation is that my final design specifies a room of 15 square meters or less, half the size of my (one and only available testing) 32 square meter room. And it specifies to cover up wood floors with carpet or linoleum.

I believe these changes in spec give a safer margin that might actually allow any of the devices (including the commercial ones) to reach the WHO "safe" PM2.5 level of 25 ug/m3.

Future Tests Needed

At the conclusion of the tests and data-gathering above, I could no longer handle the horrific pollution, so I fled Northern Thailand. However, the pollution will be back next year, and I plan to improve the tests to account for many confounding factors you probably noticed above:

These changes would bring my tests closer to the gold standard established by Smart Air Filters in their open-data trials.

Inspirations and Other DIY Designs


The DIY purifier idea has come up everywhere there has been pollution, including China and India and Mongolia and during the California wildfires.

My first inspiration was actually from 6:47 of this BBC report about toxic pollution in Mongolia (PM2.5 greater than 999 ug/m3!) where they talk about Smart Air bringing in their low-cost/DIY air purifiers:

There are a ton of identical YouTube videos showing people lashing HEPA filters to box fans (and again box fans are almost totally unavailable in Thailand), but this one is more popular because he is wearing a lab coat:

However, in both of the two videos above they make the mistake of putting the PM2.5 sensor right up against the filter material and in the airflow of the air purifier. We already know that HEPA filters filter 99%...what we want to know is how does the unit work overall for a room?

Smart Air Filters

Smart Air Filters seems to have gotten further along on DIY air purifiers than anyone else. They sell very cheap DIY air purifiers in China, India, Mongolia, Nepal, and the Philippines.

But much more importantly, they are very cool in that they release their design, raw data, and testing methods in detail:

Here is a source that is not selling filters, comparing DIY filters against commercial ones.

Chiang Mai University Design

Here is a fun Thai DIY air purifier design from engineers at Chiang Mai University that uses an (axial) exhaust fan and one of the cylindrical HEPA filters sold for Xiaomi air purifiers:

They provide build instructions here.

They didn't provide any performance information (they committed the same sin as the other videos above by holding the PM2.5 meter right in front of the filter, which doesn't tell us anything useful about the device's ability to clean an actual room).

The Xiaomi filters cost at least 600 THB (typically more like 1000 THB) and the exhaust fans also cost around 500-600 THB so this design is much more expensive.

I am not sure if the axial exhaust fan will pull that much air through. In their video they talk about room sizes of 20-30 square meters, which seems too much for the small fan.

But they do provide a design for a nice-looking plastic case :)

If anyone has actually tried this design and measured with a PM2.5 meter, please let us know your testing methods and results.

Other Fan on Cylinder Designs

Lazada and Taobao offer a number of options similar to the CMU design in that they place a fan unit on top of a cylindrical Xiaomi filter:

Fan on Cylinder Designs

Here is a design most often seen on Lazada, with a variable-speed fan:

Fan on Cylinder Designs

I am even more skeptical of this design than the CMU design above because the fan is so small. It might take 12 hours to bring a room to the same level that a more powerful fan (either DIY or commercial) could do in 2 hours, or maybe the room would never reach a safe level because the fan cannot keep up with new dirty air entering the room.

If anyone has actually tried this design and measured with a PM2.5 meter, please let us know your testing methods and results.

3M Filtrete for Aircons and Fans?

3M Filtrete is a thin sheet material that blocks PM2.5 pollution. It is available in rolls 15" by various sizes (24", 48", 96").

3M Filtrete

A lot of people who are lucky enough to have aircons here tape sheets of Filtrete to their existing aircon filters to turn their aircons into air purifiers:

This works ok but some non-scientific experimentation with my PM2.5 meter indicates that it filters a lot less PM2.5 out than either Prototype #4/#5 above or the Hatari commercial filter.

It might be possible to build a DIY fan-based filter using Filtrete.

But we then run into the big problem: because Filtrete is a single sheet and not an accordion-folded thick pad, it has a much lower surface area, which means you need to replace it much, much more often.

Here is a filter that was in my aircon for no more than 1 month (probably 3 weeks) of use, along with some new Filtrete for comparison:

Filtrete fills quickly

At this point it was completely saturated and not filtering any PM2.5, according to my meter.

Tower Fan Design

Another type of fan available in Thailand is the stand-up tower, which presumably uses the much more powerful/efficient/quiet Cross-flow fan type inside.

A Pai resident tried taping some PM2.5-blocking filter material on the output of this tower fan:

Tower-style Fan

The good news: the picture above shows the huge amount of black deposit after just 5 days of use. So the tower fan is pulling PM2.5 out of the air even more quickly than an aircon with a sheet of Filtrete.

Bad News: After 5 days, the unit started to smell like burnt electronics, indicating that the additional strain imposed by the filter on the fan motor was too much.

In this experiment, the Pai resident was using a filter material that is about 5 times thicker than Filtrete. So, it is possible that with thinner Filtrete, this combination may work without over-stressing the fan motor.

Or, it is also possible that the fan overheated only because the filter material got too clogged up. It could be that replacing the filter more often avoids the problem.

General Northern Thailand Smog Info

Here is a presentation about the causes and patterns of the Northern Thailand smog by Craig Houston, an expat who was a pioneer in getting more PM2.5 measuring stations installed and publicly visible all over Northern Thailand:


This documentary provides more information on causes and has interviews with some experts:

This Bangkok Post article goes over some other causes.

When checking PM2.5 pollution levels, it is very important to understand the difference between AQI (Air Quality Index) and raw concentration measurements like ug/m3. There's more than one kind of AQI (e.g. US AQI, China AQI, and I'll bet you can guess which one is more "optimistic"), AQI is not always derived solely from PM2.5 (it may include amounts of other particles or substances too, or there might be a PM2.5-based AQI, a PM10-based AQI, etc.), and even when it is derived solely from PM2.5, the numbers do not match up: 400 AQI is not the same as a PM2.5 concentration of 400 ug/m3, and the mapping from one to the other is not even linear, so it's not just as simple as multiplying by a constant. AQI is designed to have certain ranges (and colors) that tell you how dangerous the air is, regardless of whether the AQI is derived from PM2.5 or other toxic air pollutants.

When using apps/websites that give pollution data, it is crucial to understand the units and meaning of the numbers they are showing you, and only compare apples to apples. On some apps/websites you really have to hunt to get this critical info.

There are several websites/agencies which measure smog in Northern Thailand.

AirVisual Warning: Don't be Fooled by Green-Washed Stats

It is tempting to assume that all pollution stats can be compared with each other.

In the previous section we learned one reason why that is not always the case: pollution can be measured in raw concentration units like ug/m3, or it can be "cooked" into AQI figures that are not the same as concentration numerically (and there is even more than one type of AQI). See above.

But there are even more sinister pitfalls that can trip you up when you try to interpret pollution data you see online...

The popular website/app AirVisual is full of them. The more you look into AirVisual, the more you realize that their website/app design seems more motivated by marketing (and perhaps government relations) than by helping people be healthy.

Here are two of the most egregious examples:

EXAMPLE 1: pretty much all the stats you can find on the AirVisual site are "smoothed" (to use a euphemism) in time, seemingly over a window of about 12-24 hours (that is, a reading at a given time is actually a moving average of the real raw data point from that instant of time with other raw data points from up to 6-12 hours before and (if available) after that time). So the net effect of this "smoothing" is that the pollution peak of each day is lowered (sometimes significantly) compared to the true value. For example, if PM2.5 really peaks at 500 AQI at some instant on a given day, you will likely see a peak value of more like 300 AQI on AirVisual. For a blatant example of this, notice how the AirVisual's Mae Hong Son readings from the smoky season of 2019 never went above AQI 325 when obvously the AQI was well above 500, as shown by legitimate (non-smoothed) measurements all over Northern Thailand, including in Pai. There are some possible (but weak) excuses for why one might want to do this smoothing, but given the preponderance of other evidence of design decisions on their site, I suspect AirVisual is doing this to curry favor with governments who might otherwise create problems for AirVisual (such as by blocking access to the AirVisual website and servers). The same "smoothing" process that basically green-washes the historical data from the true AQI to a lower, fake AQI also applies to the "current" reading you get if you just pop over to AirVisual (app or website) and check out the big number. It's not really the current reading—it's the current reading, averaged with some number of previous readings, and therefore peaks are again "shaved off" or "flattened" like an eroded mountain. It's insane!

The charts on our sister site, where possible, use the raw, un-smoothed data direct from the AirVisual hardware so do not suffer from this problem. But it is not always possible to get access to this raw data. Where doesn't have access to that raw data, uses the smoothed data from AirVisual and indicates that the data has been "peak-flattened" so you know the chart is lying to you.

Look at the differences in February 2020. The top chart is raw data from CCDC and the bottom chart is from AirVisual. The two stations are located about 100m (300 feet) away from each other:

AirVisual misleading numbers

EXAMPLE 2: On AirVisual, they way they show that historical bar chart of PM2.5/AQI data from a given station is extremely misleading. Suppose on a given day there were actually only 2 measurements (say, noon and 1pm) and the data from all other hours is missing (this is extremely common for all AirVisual data sources because many AirVisual hardware owners have flaky Wi-Fi). Well, AirVisual is not going to let you know there are huge gaps in the data. Instead, AirVisual will put the 2 bars for today right next to the (say for example) 24 bars from yesterday with NO visual indication that there is in fact a 12-20 hour gap in the data. If you are using their website from a mobile device, there is basically NO way to know this is happening—the bar charts are essentially worthless. If you are on desktop, you can mouse over the bars and sometimes (depending on which page you are on) AirVisual might give you a day of the week (e.g. "Tue") but it doesn't give you a date or time, so you can't ever be sure whether there might be a 1-week gap (which also happens pretty often). Why do you suppose AirVisual is doing this incredibly misleading thing? Well, I'll bet it's because it makes their site look better (because you won't know there are often huge gaps in the data, which you might then blame on AirVisual; while any gaps in data caused by AirVisual hardware owner Wi-Fi problems are not the fault of AirVisual, it is clear that AirVisual has a basic ethical responsibility to make those gaps clear to users).

The charts on never play this deceptive game. If there are gaps in the data longer than 1.5 hr (normally data comes once per hour), they show as visible gaps in the chart so you know when you have good data and when you don't, and you can immediately and easily determine the date and time for every data point.

Give Credit

Hope you have found this all useful.

If you use this data, please provide some credit and a link to this website:
If you have comments or more ideas, please contact us at:

CopyrightEntire website copyright 1999-2023 Chris Pirazzi unless otherwise indicated.

License for use:


Entire website copyright 1999-2023 Chris Pirazzi unless otherwise indicated.

License for use: