I would need some good CO2 sensor. I have a MHZ-19B and have been watching it drift. I sealed it in an airtight box. Here is the result, it drifts 250ppm per day, which is really a lot.
As far as I know, there are no sensors that can do absolute CO2 measurements for extended periods without requiring occasional calibration.
So CO2 sensors for indoor air quality calibrate themselves on the knowledge that outside air is approx. 400ppm and that the sensor should periodically be exposed to fresh air.
So putting it in an airtight box with high CO2 (which by the way is also headache causing quantities) and seeing the calibration drop is not unexpected. I imagine all consumer sensors would do something similar.
See point 4 and 5 in this explanation:
The quality of the sensor shows from the wideness/denseness of the band of dots. The angle of the average line is a calibration effect of the driver.
First of all, if those are real numbers, that is extreme danger territory. If you are in that space and not feeling sick, headaches, fatigue, that sensor is not working…
This is my scd30 since I last purged my DB. The high points are when we have people awake in the room, the dead part we were out of town…
We had a time in Jan when were out of power and using alternate heating, and we were feeling health affects. Numbers were around 2-3k…
Using espHome integration on an esp32.
@Edwin_D I know the sensor drifts, but this is really a lot. For the SCD30, they say it should drift by 50 ppm over its lifetime, but I don’t know how I can trust that. What do you mean by “The angle of the average line is a calibration effect of the driver.”?
Read the linked post about calibration, calibration type 4 and 5 (mostly 5).
The sensor is not drifting, the calibration algorithm is causing the measurements to seemingly drop, because it never gets exposed to outside air inside the box.
So what you call drift (the fact that measurements seem to be getting lower) is the algorithm changing calibration to make the lowest measurements closer to 400 over time. Place the sensor in a place exposed to fresh outside air, over time you’ll see the values change to the 400ppm mark and staying there. After that you’ll get a good reading, and the fresh air should happen periodically to keep it good.
The sensor calibration (and thus your measurements) won’t work properly if not exposed to fresh air from time to time. It is all there in the article I linked. A quote:
The advantage of Automatic Baseline Calibration is that the CO2 sensor is virtually self-calibrating over the life of the sensor. The disadvantage of the ABC algorithm is that it will not work properly if the sensor never “reads” 400 ppm fresh air. For example, ABC calibration would not work in an indoor livestock environment or an office building that was staffed 24/7.
The CO2 in your closed box is worse than having livestock placed in an office building that was also staffed 24/7.
Also check the documentation on the sensor. You can turn ABC calibration on or off, but if you set if to off then you will have to do your own calibration:
A similar setting is present on the SCD30 sensor, but called ASC:
My advice: keep the setting on, and trust the CO2 value. (If used under normal conditions where you open the window occasionally. As you should if you care about CO2 levels.)
I need that sensor for closed ecosystem research. Now I have turned off ABC and am taking measurements, then I will send an update. Anyway, thanks for the responses.
In that case you definitely want to turn off any form of autocalibration, because it would invalidate any measurement.
I tested it without ABC and the drift is the same. Doesn’t it work more like it takes the lowest value in the last 24 hours and then suddenly adjusts the data?
