We are planning to use the BME 688 purely for sensing eCO2. Does Bosch have any data from training that supports certain specific and common applications like ours? It would make sense if they provided training result data to the community for specific applications like dedicated eC02 rather than have multiple clients individually train the 688 for the same functionality.
Solved! Go to Solution.
Yes, eCO2 is not absolute value, but comparative value.
BME688 doesn't directly measure CO2, but derives it from average between VOCs and CO2 in human's exhaled breath.
It could be different depends on test their environment.
Therefore, yes, users need to test CO2 from their side.
After digging into what VOCs would be appropriate and expected levels of those VOCs within human breath, I have come to the conclusion that really only Ethanol and Methanol are useful. Other VOC concentrations are simply too low and not repeatable.
To use Ethanol and Methanol I will have to teach the AI with quite low concetrations. I was expecting approx 10-20 ppb for each to indicate low values and 100-200 ppb to indicate high values. Note that is parts per billion and not parts per million.
The critical question - is the BME 688 able to measure concentrations of Ethanol and Methanol in such low concentrations? Even if this is yes, testing against some accurate measurment will be difficult as a complicated and expensive device like a mass spectrometer may be requried.