[bethekidyouwant]

I have these (inexpensive) sensors (except the barometer) hooked up to esp32’s in my home. After reviewing several months of the graph output I don’t have any confidence in the voc or co2 sensors accuracy. Additionally there is no good way to calibrate them outside of a lab. Imo the technology on the more interesting cheap sensors is not there yet.

[vosper]

For the AirGradient device I think the way to calibrate CO2 is to expose it to fresh air at some point, which has a known CO2 level. It’s supposed to auto-calibrate. But if you don’t do that regularly then the readings will drift (without anyway of telling they’re drifting). You also can’t tell when the sensor has calibrated.

IIUC VOC is way more complicated and hard to both calibrate and interpret. I’m not sure I have any faith in that value.

I do think a lot of people are going to be mislead by these monitors, the sensors and devices come with a bunch of caveats that aren’t clearly communicated.

For me what people should most care about is particulates, and at least as far as I know those sensors don’t come with the calibration issues of CO2 and VOC. That’s the sensor my AirGradient is set to alert on.

[formerly_proven]

Reasonable CO2 NDIR sensors use two chambers: one sampling ambient air for the measurement and a sealed chamber with a reference gas for autocal.

edit: It seems AirGradient uses one of the cheap sensors.

[Aspos]

I have a CO2 sensor mounted right next to the vent of my ERV which pulls in fresh air regularly so the sensor gets calibrated automatically.

[ajolly]

What erv are you using?

[Aspos]

I use Broan ERVS100S and Panasonic FV-04VE1

[hedora]

I was worried our sensors had that autocalibration mode. Our house has new concrete floors. They alternate between absorbing + outgassing CO2 (depending on room temperature), so sometimes our living room reads below 400 ppm.

From what I can tell, the sensors haven't drifted, so they're not using the "below 420 ppm => recalibrate" heuristic.

[GeoAtreides]

I have running for years both SCD30 and SCD41, and they're very much in-sync, although the technology is different. SCD41 is very sensitive to voltage, so you need to keep it as stable as possible. SCD30 is NDIR, so it doesn't need calibration (mostly...).

For VOC also run SEN55, BME680 with BSEC 2.0 and BME688 with BSEC 2.6, they mostly agree, with the SEN55 being more sensitive to 'chemicals' (think industrial chemicals vs living entities chemicals), and BME688 more 'agressive/sensitive' to changes than BME680.

[hedora]

Obviously, it's not as good as a lab, but I have an Ambient Weather indoor PM 2.5/10/CO2 meter, and the numbers it's produced over the last few years are extremely plausible. Someone might end up correlating the two.

Overall, their whole setup is a nice plug and play experience. My only complaint is that I want a 900MHz <-> WiFi / PoE bridge so that I can place temp + humidity sensors further from the house. e.g., in cellars, forested spots, etc.

Here's what would make me consider replacing / augmenting my system with Air Lab devices:

- Weather proofing - Integrated solar panel charging solution so the 7.5 day battery life turns into 7.5 days without sun. - Battery life estimates for WiFi mode that mean the solar panel would work, or a (weatherproof) BLE <-> WiFi/PoE bridge, assuming the BLE range (through a wood wall) is at least 100 feet. - PoE on the sensor board would be nice to have. - I like that these have VOC/NOX sensors, which my current setup is missing. I don't like that they're missing PM sensors. - I don't care about the screen; a blinking LED for debugging would too. - If it were small enough to go in my pocket, I'd care about the screen. I wonder if cell phones will eat that use case though.

[256dpi]

I'm curious, what's the range/variability you're seeing?

[radicality]

Oh yeah? That’s surprising, I’m doing something similar - with esp32s from M5stack, specifically their SCD41 sensor for co2 (same as in the device in the post), with an AtomS3 to drive it, and it’s super responsive and I think very accurate. Perhaps you have the scd40? It’s rated for high accuracy up to 2000ppm, while scd41 maintains accuracy up to 5000ppm.

Using with esphome and homeAssistant, I grab the data every 5 seconds (I did 1s too but seemed unnecessary), and push it to HomeAssistant. The moment I turn on the oven, or open a window, or just enter the room after the night from the bedroom, the graph moves very very visibly and very quickly. I like this setup much more than my uHoo Air sensor.

[Havoc]

Have a similar setup. Unfortunately the open a window thing isn't all that reliable as test. Even sensors cheaper than the SDC40/41 will happily do that.

The issue is more drift and that its a very localised measurement. The sharp spikes in the graphs are often more about the air is moving than the room's total CO2 changing. Like on mine (also sdc41) I can see when I wake up on the bedroom one. Being awake produces more CO2 sure but it's marginal. 80% sure the spike is simply more turbulence in room. Lying still for hours vs getting up and making bed etc

[ajolly]

Really? I've got a few cheap sensors and while they're not always accurate in an absolute sense they are in an directional sense. It's extremely evident for Windows opened if someone is cooking if I turn on an air purifier, etc