[troydavis]

I bought an AirGradient indoor meter and ran into 3 significant undocumented problems, 2 of which affect CO2:

1. The VOC readings use a relative scale and it's relative to a floating value: the last 24 hours. If a VOC source is present for more than 24 hours, it will quietly become part of the baseline value (100). You can read more here: https://forum.airgradient.com/t/measurement-values-for-the-a... , https://forum.airgradient.com/t/measurement-values-for-the-a...

2. The CO2 measurements re-calibrate to the lowest value detected every week (as ~400 ppm clean air). If it's not exposed to clean air every week, the readings will drift. When users asked how this worked, calibration hadn't been thought about much, if at all: https://forum.airgradient.com/t/model-number-for-co2-calibra... , https://forum.airgradient.com/t/model-number-for-co2-calibra...

3. The graphs don't show reboots, even though reboots (always?) reset the baselines. Reboots happen relatively frequently due to software updates. So, even the relative scale can change without any entry on the graph.

I was pretty annoyed because it didn't seem like the company had thought through the user experience enough to even encounter these questions, let alone consider the best implementation (or at least document the flaws in whatever they're able to implement). This is the whole purpose of the product. The company means well and tries hard, but at some point good intentions matter less than the product.

[ahaucnx]

Achim, founder of AirGradient here.

These are all valid points that came up at the beginning of the year and we took this feedback very seriously.

I'm happy to let you know that we have implemented already point 1) and 2) in the latest firmware version.

You can now set the "learning gain offset" for TVOC and NOx from the default 24h to basically 1 year. As a result the monitor will behave much more like an absolute TVOC monitor as the baseline reset is much longer.

We have also implemented the automatic baseline calibration period, which you can now set to a much longer period than one week.

We are also looking into feeding the last vale back into the sensor to avoid the reboot issue but this is more tricky to implement.

By the way, if you compare the granularity that we offer now to fine tune the sensor modules, I'm not aware of any competitor offering the same feature set. But we won't stop here. We currently working on fine tuning compensation algorithms for PM.