CO2-Ampel

Yet another CO₂-Ampel based on an ESP8266 (Wemos D1 mini) to continuously measure CO₂ concentration indoors with a SCD30 CO₂ sensor accompanied by a BME280 for temperature, humidity and air pressure readings. Current air condition (good, medium, critical, bad) is shown using a NeoPixel ring with WS2812 LEDs illuminating the device in either green, yellow or red.

Unlike other devices this one runs on two 18650 Li-Ion batteries, offers a sophisticated web interface for configuration, live sensor readings and OTA firmware updates. Sensor data can be logged to local flash (LittleFS), retrieved RESTful, published using MQTT or even transmitted with LoRaWAN if a RFM95 shield is installed.

Hardware components

• SCD30 CO2 sensor
• BME280 break-out board
• Wemos D1 mini board
• Wemos Battery Shield
• Wemos ProtoBoard Shield or LoRa Shield
• DS3231 AT24C32 RTC Module
• NeoPixel ring with 12 WS2812 LEDs
• 3-pin slide switch
• 2 x 18650 rechargeable Li-Ion cells
• 2 x single-slot 18650 battery holders
• 5 x Dupont cables with female connector (3 x 4-pin, 1 x 2-pin, 1 x 1-pin)
• 2pin JST PH2.0 female plug
• 2.54mm single row pin headers for Dupont connectors

Looking around for a suitable housing for the CO₂-Ampel we (my son and I) spontaneously choose this because it was almost empty, pretty robust and seemed perfect in respect to shape and size. 😉

To help stack all components inside the translucent plastic tin I designed a simple 3D printed inner frame which has a place for all components and is glued into the orange lid. The SCD30 and BME280 sensors are placed into a seperate 3D printed sensor-mount which is glued on top of the frame after the NeoPixel ring has been screwed onto the frame. Air ventilation is achieved by a 3D printed grille pressed into a 40 mm hole cut into the rather thick bottom of the plastic tin.

Putting it all together

All Wemos components are simply stacked together. The battery shield is placed at the bottom, the Wemos D1 mini in the middle and the (optional) LoRa shield on top. The I2C connector of the LoRa shield was extended with two more 4-pin headers to connect the three I2C modules (SCD30, BME280, DS3231). Alternatively you can replace the LoRa shield with a Wemos ProtoBoard shield and solder the three 4-pin headers in parallel to D1 (SCL), D2 (SDA), 3.3V and GND on top of the board. If you're using a ProtoBoard you also have to add two 4.7k pull-up resistors from D1 and D2 to 3.3V or I2C communication will proably fail.

The SCD30 CO₂ sensor, BME280 break-out board and DS3231 RTC are each soldered to 4-pin Dupont cables which in turn connect to the 4-pin I2C header on either the LoRa or a Wemos ProtoBoard shield. Double check that all SCL pins are connected to D1 and all SDA pins to D2. The Vcc input of all I2C modules must be connected to 3.3V of the Wemos stack and not 5V or you'll destroy all I2C components in no time!

The Vin pin of the NeoPixel ring must be connected to 5V, its data input port to GPIO0 and GND to GND of the Wemos shield stack. I used a 2-pin Dupont cable for 5V and GPIO0 and a 1-pin cable for GND to connect the ring to the 2.54mm pin headers soldered onto the LoRaWAN shield. If you're using a Wemos ProtoBoard you probably want to merge GPIO0, 5V and GND into a single 3-pin header.

The two 18650 battery holders are wired in parallel and are connected to the Wemos battery shield using a JST PH2 female plug. A rubber band will prevent the Li-Ion cells from becoming loose. The Vcc line is looped through the slide switch mounted into the orange lid serving as the external power switch.

Two 3400mAh Li-Ion cells will last for about two 9-to-5 working days. If the voltage falls belows 3.55V the CO₂-Ampel will enter deep sleep to protect the cells from deep discharge. It will then wake up once every hour, flash red six times before returning to deep sleep again. You can configure a simple daily scheduler to adjust operating times to your needs.

The Li-Ion batteries can either be changed if empty or charged using the micro usb port connected on the Wemos battery shield. On the bottom of the battery shield you can choose to close a solder bridge (J1) to increase charging current from 0.5A to 1A. Since only recent Wemos battery shields (v1.3) offer a solder bridge (J2) to connect the battery Vcc to A0 of the Wemos board (using a 120k resistor) you need to add a resistor (100-220k) yourself from the Vcc pin of the JST male connector on the shield to the A0 pin to enable voltage monitoring.

Compiling the firmware

First adjust settings in config.h according to your needs (language) and hardware setup. For an initial setup you should probably leave the default setttings untouched since most of them can be changed anyway using the extensive web interface and are then stored in the EEPROM of the DS3231 RTC module. Only support for the optional LoRa Shield shield must be explicitly enabled by uncommenting HAS_LORAWAN_SHIELD. Same holds true for the parameter VBAT_ADJUST which sets the multiplier for the voltage divider used on the A0 input of the ESP8266 to measure the battery voltage on the Wemos battery shield's JST battery connector.

Before trying to compile and flash the sketch to your Wemos D1 mini make sure that you added support for ESP8266 based boards and the following libraries have been installed. Don't be too intimidated by the long list, the Arduino IDE library manager will give you a hand. Make sure that you set the flash size to 4MB (FS:2MB OTA:~1019k) to spare flash for the logging. To save battery set the CPU-Frequency to 80 MHz.

• EPS8266Wifi
• ESP8266WebServer
• ESP8266HTTPUpdateServer
• ESP8266mDNS
• uEEPROMLib
• RTClib
• FastLED
• PubSubClient
• ArduinoJson
• BME280I2C
• MCCI LoRaWAN LMIC Library
• LittleFS
• NTPClient
• RunningMedian
• Time
• Timezone

You only need to upload the firmware from the Arduino IDE to the Wemos D1 mini using its micro USB connector once. For further updates just use the Export compiled Binary option under the Sketch menu in the Arduino IDE and upload the binary using the firmware update option in the web interface.

Initial startup

After flashing the firmware onto a hopefully working hardware setup you should enable the serial monitor in the Arduino IDE to closely monitor the startup process and initial operation of the CO₂-Ampel.

Right after power up you'll see two white flashes inidicating a successful inialization of all components connected to the I2C bus, followed by two green flashes showing a successful initialization of the RTC module. An error at this stage will result in continous red flashing...time to check all connections.

If you compiled in support for the optional LoRa shield and also enabled OTAA by defining LORAWAN_USEOTAA in config.h the CO₂-Ampel will now try to connect to a LoRaWAN gateway with a timeout of 10 seconds. During OTAA it will flash magenta followed be two green flashes if the network join succeeded. LoRaWAN session keys and settings are stored in EEPROM and reused on next power up. An OTAA failure will trigger two red flashes.

Next WiFi is turned on (two blue flashes), a local access point is fired up (two green flashes) and a webserver is started to further configure the devices and show current sensor readings. If you enabled the optional WiFi uplink in config.h a few LEDs will now turn blue while trying to connect to the WiFi network specified with WIFI_STA_SSID and WIFI_STA_PASSWORD. On success you'll see two green flashes otherwise red. You need to connect the CO₂-Ampel to a local WiFi (with internet access) at least once to initially set its RTC.

If a WiFi uplink is available and MQTT is enabled the CO₂-Ampel will initially connect to the configured MQTT broker. A successlful initialization of the MQTT uplink is indicated by two orange flashes.

The initial setup ends with a few more green flashes indicating the successful initialization of the SCD30 and BME280 sensor modules. At last the CO₂-Ampel will switch to white for about a minute giving the SCD30 time to warm up. Eventually monitoring of the CO₂ concentration will start indicated by a continous green (good), yellow (medium) or red (bad) illumination. If you start to see two repeating red flashes the air condition is critical...time to open the windows!

Continous four red flashes indicate repeated errors while trying to get valid CO₂ readings from the SCD30 sensor. Turning the CO₂-Ampel off for a few seconds will usually fix this.

If the reset button on the Wemos D1 mini is pressed all settings are reset to their default values as set in config.h at compile time.

Adjust device settings

During the initial setup phase the CO₂-Ampel will start a WiFi access point with the SSID CO2-Ampel followed by six alphanumeric characters derived from the Wemos D1 mini's MAC address. You can connect to the access point within a timeout initilially set to 10 minutes using the password specified by WIFI_AP_PASSWORD (default is __mysecret__). Then point your favorite browser to http://192.168.4.1 and you'll be presented an extensive web interface to further configure your CO₂-Ampel. Its language is preset at compile time by either defining LANG_DE or LAND_EN in config.h. All settings are saved in the EEPROM of the RTC module and will thus survive a power off.

CO2 sensor calibration

Since the SCD30 CO₂ sensor has to be exposed to fresh air for at least one hour every day to make its automatic self-calibration (ASC) work as expected, it has been disabled in this setup. The SCD30 is pre-calibrated at the factory, but the accuracy (approx. +- 60 ppm, drift approx. 80 ppm/year) may change e.g. due to mechanical effects during transport.

However the CO₂ sensor can be manually (re)calibrated in fresh air (outdoor air is actually between 350 ppm and 450 ppm in normal geographic locations) using its forced recalibration (FRC) command which can be triggered in the web interface. Just place the CO₂-Ampel into (windless) fresh air, start the calibration process (CO₂-Ampel will switch to cyan color) and keep an eye on the web interface. The fully automated calibration might take a few minutes since the CO₂-Ampel now continously monitors the standard deviation of the CO₂ readings. It will only set a new reference value (410 ppm) if the readings are more or less stable. This value can be changed in sensor.h with SCD30_CO2_CALIBRATION_VALUE.

If CO₂ readings still keep floating within a timeout of 5 minutes the calibration is eventually aborted, keeping the previous reference value. The CO₂-Ampel will flash red for 15 seconds to indicate calibration failure.

The temperature offset between the onboard RH/T sensor and the BME280 is continously corrected automatically based on their standard deviation. Intial offset is set to 1.9 in sensors.h with SCD30_TEMP_OFFSET.

Disclaimer

Even though this device consists more or less of the same components as professional systems it does not have any certification. It's a home made product for educational purposes and thus cannot be used as medical device or serve as personal protective equipment in the sense of any legislation.

Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change.

License

Copyright (c) 2020-2021 Lars Wessels
This software was published under the Apache License 2.0.
Please check the license file.