DIY cell culture CO₂ monitor

Parts

Wiring map

Build order

1. Flash the firmware

   Do this on the bare board, before any wiring.

   arduino-cli lib install "LovyanGFX" "OneWire" "DallasTemperature"
   cd firmware/CO2_CYD
   ./build.sh upload

2. Fit the pigtails

   Plug a JST pigtail into P1, P3 and CN1. Jumper pins do not grip these sockets.

   

3. Wire the MH-Z16

   Red to VIN, black to GND, yellow to IO35, through the lever connectors. Tie back the other leads.

   

4. Wire the DS18B20

   Clamp the probe into the adapter. Run VCC to 3V3, GND to GND, DAT to IO27.

5. Close the junction box

   Sensor leads in one side, board wires out the other.

   

6. Power up

   The start-up screen lists each part as it is found. Expect CO₂ SENSOR ONLINE and TEMP PROBE OK.

7. Calibrate the touch panel

   Open the serial monitor, type c, tap the four targets.

   cd firmware/CO2_CYD
   ./build.sh monitor

8. Place the sensor head

   Sensors inside the incubator, board and junction box outside.

   

   

Reading the screens

Tap anywhere to move to the next page. There are four: main, graph, diagnostics and connect.

Main page

1. LIVE while readings arrive
2. Gauge, 0 to 10 %, with the target band shaded
3. CO₂ in percent
4. CO₂ in ppm, raw sensor value in brackets, then the trend
5. Status
6. Last 15 minutes: CO₂ in teal, temperature in orange
7. Temperature

Graph page

1. CO₂ axis
2. Temperature axis
3. Now, minimum, maximum and average
4. Time span

How it measures

The sensor sends one pulse a second. The longer the pulse, the more CO₂.

flowchart TD
  A["Power on"] --> B["Start the display"]
  B --> C["Load touch calibration"]
  C --> D["Start CO2 capture on IO35"]
  D --> E["Find the probe on IO27"]
  E --> F["Look for SD card and Wi-Fi"]
  F --> G{"First CO2 reading in,<br/>or 12 s passed?"}
  G -- no --> G
  G -- yes --> H["Main page"]

flowchart TD
  A["Pulse edge on IO35"] --> B{"Glitch, under 200 µs?"}
  B -- yes --> X["Ignore"]
  B -- no --> C["Time the full cycle"]
  C --> D{"Cycle 900 to 1100 ms<br/>and timings agree?"}
  D -- no --> R["Reject, keep the last value"]
  D -- yes --> E["Convert pulse length to ppm"]
  E --> F["Average 8 cycles"]
  F --> K["Apply your calibration factor"]
  K --> G{"Sensor warm?"}
  G -- no --> H["Show only"]
  G -- yes --> I["Show, log, watch for door events"]

flowchart TD
  A{"Valid reading in<br/>the last 10 s?"}
  A -- no --> E["SENSOR ERROR"]
  A -- yes --> D{"Sensor warm?"}
  D -- no --> W["WARMING UP"]
  D -- yes --> O{"Raw reading at or<br/>above 99,000 ppm?"}
  O -- yes --> OR["OVER RANGE"]
  O -- no --> F{"Corrected CO2"}
  F -- "below 4.5 %" --> L["LOW CO2"]
  F -- "4.5 to 5.5 %" --> N["NORMAL"]
  F -- "above 5.5 %" --> H["HIGH CO2"]

An hour in the incubator

A log from the monitor itself: 384 readings, ten seconds apart.

• The door was opened in two bursts. Each time CO₂ fell from 5.1 % to about 1.8 %.
• After the last opening the incubator brought CO₂ back into the band in 4.7 minutes and settled in 7.7.
• Temperature stayed between 37.4 and 38.3 °C throughout.

Ten days in the incubator

One unbroken log from the SD card: 91,739 readings, summarised by the hour.

• The resting reading climbed from 4.84 % to 5.08 % over the first three days, then held.
• Every excursion was a dip, never a rise.
• The deepest dip reached 0.44 %.

Web dashboard

Open the Connect page on the monitor and scan its QR codes with a phone. One joins the monitor's own network and the other opens the page. The page shows the live reading and the history, sets the monitor's clock from the phone, and offers every log file on the SD card for download.

The monitor starts a new numbered log file at each power-on, such as co2_0007.csv, with one row every ten seconds.

Calibrate before trusting it

1. Let the monitor run in an incubator at a known setpoint for three days. The reading climbs while the sensor settles.
2. Read the raw value in brackets on the main page.
3. Divide the known value by the raw value. For 5.0 % known and 6.7 % raw that is 0.75.
4. Set CO2_CAL_GAIN to that number in Config.h and flash again.

The firmware ships uncorrected, with a gain of 1.0. The unit in these photos needed 0.72. Repeat the check from time to time.

Checks and faults

Shows about 5 % at power-on

Normal. The sensor reports 50,000 ppm while it preheats.

Temperature shows n/a

Check the adapter's screw terminals, then power-cycle.

Taps land in the wrong place

Run the touch calibration again.

Cannot join the lab network

After 60 seconds it starts its own network, CO2-monitor. Join it and open http://192.168.4.1/.