eWizzard wrote^ Particulate matter (PM) is one category of pollutant and does not account for gases.
Then according to your results, we should've all been dead a long time ago. The "cleanest" reading is about 6% pollutants. If that 6% was CO2, everyone would've died within minutes. If it was methane, the highway would've probably exploded. Even with baseline correction (using early morning hours as "clean" baseline) the peak amount of pollutants during heavy traffic hours is around 9% which is impossible.
Sorry, but totally wrong, because you are assuming percentage is relative to specific gases, while you don't know anything about circuit, composition of sensor, linearity of sensor, sensitivity to specific chemical compounds (it is different to each).
While in reality it means just ADC value scale. It wont be <1-2% unless sensor is failed, for example. In future i might adjust it by opamp circuit with shifting zero level to improve sensitivity.
eWizzard wrote
Anyway, for serious chemical monitoring and analysis of air pollutants, you need expensive instruments, standards, and a dedicated work environment. Cost can easily be in the hundreds of thousands of dollars. Cheap semiconductor sensors are not recommended for such delicate quantitative studies, but they're good and very cost-efficient for other applications such as alarm systems, which don't require high accuracy, dynamic range, and specificity, and they still have to be periodically calibrated. I wouldn't waste too much time with them in this experiment.
It's way not hundreds of thousands, but still cost too significant for hobby.
And again almost totally wrong. This sensor is not semiconductors based at all (sure except amplification, ADC and etc), and this is why they are not precise, it is typical "catalytic" sensors, that measure changes in materials sensitive to specific gases. Same way is used by oxygen sensor in exhaust system of cars, but sure composition is quite different. And as each application has its own use - car lambda sensor is useful specifically there, even it has all drawbacks specific to this type of sensors. As all such sensors - they are "drifting" over time and need frequent recalibration (or they will be expensive as car lambda sensor), additionally i need to do several corrections based on environmental parameters, which i am not doing entirely correct, but it affects data very little.
For alarm systems totally different sensors used - mostly photoelectric or α-ray particles detectors. They are just detecting particles in air, and they are not precise at all about density and etc, they just react to their presence above quite big threshold.
While real semiconductor sensors is much _more precise_ on short/longterm and usually calibrated (but expensive as well, while i might still get them), their principle are based on gas adsorption spectrum, and effect that laser change wavelength with applied voltage. But unfortunately they are sensitive only to very specific gases.
And even if i get precise sensors, calibrate them - in most of cases it is totally useless, because you can detect many chemicals and miss one, even more harmful.
This is why in decent sewage systems such sensors are used and they are the best detectors for contaminants harmful for humans and marine life: