Hi Achim, do you think my graph exhibits the issue? I’ve only had it since yesterday, but I can see there are couple of large periods where the sensor is reading 0 for extended periods.
It seems to be reporting higher concentrations well, a couple of spikes from outside air and traffic that seem to be about expected levels, and burning candles inside last night. I also run a HEPA air purifier. Many Thanks
Thanks! This seems to fix the PM2,5 reading on the dashboard by calibrating my sensor with batchnumber PMS5003-20231218.
Will this also fix the PM10 and PM1 readings / measurements which are suspiciously flat, i.e. same as the PM2,5 reading was previously? What I can read from your report (part of it as depicted below) this seems not to be the case?
” PM1 and PM10 are currently not the focus, so no correction algorithms have been developed”*
Our focus was to fix 2.5 first and we are starting to look now into the PM1 and 10.
It’s not easy to spot in your case as you are running an air purifier and have no outdoor monitor to compare with.
Best would be to open the enclosure and have a look on the label of the PM sensor.
Thanks for your efforts and glad to have installed both the indoor- and the outdoor monitor. Very cool and helpful instruments…. !
Thanks for getting back to me so fast. My PM sensor is one of the affected batch numbers - PMS5003-20231030, but I wasn’t sure if every sensor in the batch was definitely affected though. The only other monitor I have to compare it with is a Qingping Air Monitor Lite, which is reporting PM 2.5 values around 2-3 μg when the AG is reporting zero, and maybe 2 μg higher up until around the 10μg range.
Not sure how accurate the QP is however, but its CO2 sensor is usually within 10ppm, and often 5ppm of the AG which is reassuring (both calibrated at the same time). I believe it uses the same sensors as the AG.
So far we have seen the same behavior for every single unit within a batch.
So I am reasonably confident that applying the algorithm based on the batch number should work reliably.
Thanks again for your expertise Achim. I’ll apply the correction. Very happy with my monitor overall. 
I had to do a ctrl + F5 refresh before it showed up in my dashboard
Cleared cache, local storage, session storage, cache/db storage, logged out and back in. Still not showing up
Quick question, will it be possible to push these corrections (both the EPA and the specific Plantower batch correction) to the display on monitor itself via a firmware update in the future?
On a side note, the two corrections seem to working well, and correspond very closely to another portable monitor I use, both at low and higher PM2.5 concentrations. I think it uses a different Grandway PM sensor.
Yes. It will come to the firmware in the next few days. Then the value on the display will also be corrected.
We are in the final test of the new firmware and it’s looking good.
Monitor Maker: AirGradient
Firmware Version: 3.1.9
Model: I-1DIY
Try and switch to 9PSL model and reload the web app. Let me know if that hekos,
That worked. Thank you
I’m looking at the blog post, and I’m not sure whether I’ve found a tiny typo, or I’m not understanding something about how the fix is meant to work.
Under the heading “Update #2 (29 October 2024)”, there’s a written version of an algorithm. The value PM2.5_calibrated_low is calculated using two constants and the PM0.3_count. In the case that the result of the calculation is below 31, it will be is carried forward, replacing the PM2.5 value.
Is it intentional that the sensor value for PM2.5 is never used in this case, only the PM0.3? If so, why does this work?
Yes, this is intentional. The PM2.5 raw readings of some batches are lower than expected. Often they report 0 ug/m3. Unfortunately, this happens when the actual PM2.5 is indeed 0 ug/m3, as well as when the PM2.5 is supposed to be higher. This means that based on the raw reading of 0 we are unable to tell whether it was indeed 0, or whether this value should be corrected. That’s why we decided to do the workaround via the PM0.3 count instead.
Ok, that makes sense, thanks!
