Blog & White

The last issue of Make: magazine had an article about building a CO₂ monitor. The concentration of CO₂ is apparently a good estimator for the concentration of SARS-CoV-2 in the air, which is correlated to the risk of transmission. The article suggests keeping the concentration of CO₂ under 800 ppm when people are wearing masks.

The Make: article proposes a green-yellow-red light indicator, where the light is green when the concentration is below 1000 ppm, yellow above that level and red when it goes above 2000 ppm. These values are rather about indoors air quality and not directly related to limiting the transmission of SARS-CoV-2.

An article from NIST however indicates that the 1000 ppm limit has no basis whatsoever so I decided to use the 800 ppm limit instead and a rather arbitrary limit of 1300 ppm based on looking at Figure 2 in this article which seems to indicate that some cognitive abilities drop around that concentration.

Description

The device itself is quite simple: it has an on/off switch and a single two-color LED indicator, red and green. The yellow color is obtained by turning on both green and red colors at the same time. It also has a small hole where a paperclip can be inserted for triggering the calibration procedure. The holes next to the power switch expose the CO₂ sensor and its temperature/humidity sensor.

Inside the box there is an Arduino-like Jeenode I had lying around with an AA Power board (I have no idea if these are still sold, I've had them around for over ten years). The AA Power board is meant for a single AA battery, but it cannot provide enough current, so I removed the battery clips and connected it to a 2-AA battery holder, via the switch. It seems to be working well with two NiMH rechargeable batteries.

The CO₂ sensor is a Sensirion SCD30. It is quite expensive (about 50 EUR), but has an easy to use I2C interface and is the most accurate of the sensors presented in the Make: article. The software is quite trivial (if you except the calibration procedure, see below), and available here. The device automatically makes a measurement ever two seconds, the program reads it and updates the LED accordingly. That's it.

Calibration

There is an automatic calibration procedure that requires to keep the sensor powered for at least 7 days and put it in fresh (outdoors) air at least an hour per day. This is not very practical given that I've estimated that the batteries would last about 25 hours (20 mA for the ATmega, 17 mA for the SCD30, 10 mA for the LED at 3.3 V, with maybe a 80% efficiency for the power board with two 1900 mAh batteries at 2.4 V). There is however also the possibility to expose the sensor to air with a known CO₂ concentration and tell it the actual value it is measuring. It then uses this value as a reference point for subsequent measurements. The device is apparently sensitive to changes in shape (e.g. when subjected to mechanical stress during transport) so as a portable device it probably needs to be regularly re-calibrated.

The calibration procedure is simple: place the device outdoors, wait about 2 minutes for it to settle, then introduce a paperclip into the hole and press the button underneath it; the LED will be flashing red. The device will take repeated measurements every 2 seconds, and when a succession of 10 measurements is considered stable enough (i.e. the absolute value of the difference between the first and last of those is at most 1, and the sum of the absolute values of each measurement with its previous one is at most 10), it sets the calibration value to 416 ppm (which seems to be about the average value in 2021). As the device is accurate to +/- 30 ppm, the exact value does not matter so much. The device then returns to its normal operation mode, showing a green, yellow or red light.

What next?

Put a sticker on it with some indications about the LED colors and the reset button hole.