Hacker News Reader

Comment by Havoc | Hacker News Reader

Have you actually gotten any reliable co2 data out of the bme680?

Even with the Bosch library and their calibration the values seem rather odd at times

It's not capable of CO2 measurement, it only estimates CO2 levels based on measurements of VOCs. These estimated readings are garbage and have close to zero correlation with CO2 levels.

Unless I've got another source of VOCs around there should be at least some correlation with actual CO2 levels though? I don't have a reliable way of checking this, but the bottom end seemed well calibrated...outside air leads to a reading consistent with ambient co2

We tested a lot of VOC sensor modules and the estimated CO2 values they give is -in our experience- 80% of the time totally wrong. Also the absolute ppb values of the TVOC are most often not to be trusted.

What they are good at is detecting spikes, e.g. from cleaning liquids, desinfection etc. but I would not trust the absolute value at all.

The big sensor manufacturers recognized that and newer generations of the Sensirion sensors are now just displaying an "index" instead of absolute measurements.

Thank you for asking this question because all of my experience so far with "environmental sensors" that aren't laboratory grade has been that they are absolute garbage and only barely correlated to anything.

How are companies like Bosch allowed to pass this crap off as functional? They don't meet even basic fitness for purpose criteria.

When developing the book, I didn't have any additional meters or sensors to cross reference measurements with so I can't comment on the accuracy unfortunately. But I did notice that the changes in measurements made sense when the environment changed. I.e when I would exhale on the sensor, the measurement would rise and fall.

Another example was when I opened my window in my office. CO2 immediately dropped off (15:42 timestamp): https://twitter.com/akoutmos/status/1443233937015418891/phot...

Yeah same - there is definitely a practical cause & effect practical correlation, but also a lot of effect without obvious cause in an environment that should be relatively stable. (one person, small apt, no co2 sources)

> additional meters or sensors

I've got a CCS811 as well but no luck yet. Acquired 2nd hand so unsure if broken from shipping or I'm being stupid with code or I broke it while soldering

This reminds me how my DIY build of Luftdaten kit (PM sensor) started reporting huge particulate spikes indoors, with no obvious cause. Took me ~2 days to figure out they're being caused by our ultrasonic humidifier, and another 2 days of looking through blogs and scientific papers to finally learn that one shouldn't be putting tap water into them (as they efficiently aerosolise the minerals and contaminants dissolved in it).

Both the SGP30 and the BME680 have a big weakness that I experienced some years ago when I trying putting an air quality sensor together. They correlated the TVOC readings with the temperature. And sometimes it was bad, like a difference of just 1 or 2 degrees Celsius could double or triple the TVOC readings at the lower end of the scale (~50ppb), and add maybe ~50% at the upper end (~500ppb).

Are you seeing the same behavior?

Yeah, it's only a propietary approximation.

Reply on news.ycombinator.com