Autocalibration od CO2 sensors

Measuring principle

Room and duct COsensors for building control systems are based mostly on the NDIR (Nondispersive Infrared) measuring elements. They measure the diffusion of infrared light caused by the presence of COmolecules in the measuring chamber. The main parts of the measuring element are:

  • infrared diode
  • measuring chamber
  • filter to remove all wavelenghts except that one absorbed by the CO2molecules
  • IR radiation detector
  • electronic for signal amplifying and conditioning.

The output is usually an analogue signal (mostly as PWM) or a simple communication protocol which provides the measured value in ppm (parts per million) CO2. Remember that the outside air usually contains about 400 ppm CO2, which means that the measured value should never drop below this level. During normal room occupancy the concentration rises up to 1000 or 1500 ppm, a very bad air contains 4000 ppm and more. Commonly used sensors provide measuring range of 0…2000 ppm or 0…5000 ppm, the range may also be configurable. The output signal from the measuring element is further processed and brought to the sensor output, usually as a 0…10 V or 4…20 mA signal or on a communication bus.

Because of mechanical stress (especially at transport and installation), temperature changes, and ageing of components, the output signal may change in time and the sensor accuracy decreases, i.e. the sensor is „drifting“. This is why regular calibration is necessary. The following image shows two non-calibrated sensors of different manufactures (with different measuring elements) installed in the same room. In a year’s time the absolute difference of measured values rose from 80 to about 300 ppm, as seen in the plot. The black curve drifts down on a long-term basis, the red curve drifts up (which is not visible at this image as it only shows a short part at the end of the testing period). The lowest measured values of both curves should approach the 400 pm line from above.


Two CO2sensors of different manufacturers, non-calibrated


Part of the calibration process is exposition of the sensor to air with defined CO2concentration. Using calibration gases with fixed gas composition is difficult or impossible in the daily practice, so the expected COconcentration of about 400 ppm in the outside air is used as a standard. Manual calibration at sensors with analogue output is usually started by a calibration button. The steps then may be as follows:

  • the sensor is exposed to outside air (with known CO2concentration)
  • the calibration button is pressed to activate calibration
  • the sensor indicates the end of the calibration process by a LED or another indicator.

The potential point of failure of this process is e.g. that manipulating with the sensor may lead to immediate decalibration because of mechanical shocks, and the sensor may be influenced by the presence of the operator. Therefore, automatic calibration is recommended.


The autocalibration process is programmed in the sensor’s control processor. The sensor remembers the lowest measured value to which it has been exposed over a certain time period, and it is supposed that this air was the outside air with known CO2concentration. See more detailed description e.g. in [1]. A typical calibration time is 8 days to cover also one weekend when the building is supposed to be unoccupied and the interior COconcentration drops down to the outside air values after 4-8 hours, be it due to ventilation or natural infiltration by leakages. The lowest measured level is assigned the 400 ppm value. This algorithm is called ABC (Automatic Background Calibration) and its implementation may slightly differ among manufacturers.

The Senseair company uses 7.5 days as the calibration interval. The lowest measured leved is assigned the value of 400 ppm. Maximum correction is limited to 30 ppm weekly to prevent steep changes. See also [2].

The S+S Regeltechnik sensors require uninterrupted operation (sensor must be powered) for at least 24 hodin to start the autocalibration process at all. The COconcentration should drop to the outside air level at least once during this period. The calibration is complete if the sensor has been exposed to the outside air at least 4 times during  21 days of operation. It is recommended to execute manual calibration at the sensor commissioning time to speed up the adaptation, which, of course, requires outside air brought into the room where the sensor is installed.

The Domat UT090 sensor (output 0…10 V) keeps recording the minimum value for 8 days. The minimum measured value is updated continually. The algorithm is thus very similar to this above.

Communicative sensors Domat UI90x use the T-100C measuring element by ELT Sensor Corp., which is able to run the autocalibration [3]after the first 2 days after power-up, and then repeats every 7 days. Both autocalibration and manual calibration are implemented in the supervised communication processor: after power-up, the minimum signal level is being recorded for 200 hours, and this is assigned the 400 ppm value. The correction is saved in EEPROM and the whole process repeats. In case of manual calibration, the correction value can be written in a Modbus register so that the correction added to the output of the measuring element provide the correct measured value on the bus. This makes possible to execute the correction using a reference (portable) COsensor, even without fresh air supplied.

The autocalibration algorithm called ABCLogic by Thermokon records the lowest measured value during 24 hours, and calibrates the sensor to 400 ppm after 2 weeks of operation.

When the autocalibration does not work

The process described above is meaningless if the sensor can not be exposed to the outside air for any reason. This is the case of greenhouses, rooms with special requirements, rooms supplied by AHUs with mixing dampers (unless they are possess a purge function), but also air quality sensors installed in the extract air duct. In these situations, the autocalibration must be switched off and manual calibration must be used.

Service and maintenance

Sometimes, sensors are being deployed for years without any calibration. What are the consequences for the controlled air handling unit? It very depends on the controlled technology and the application software. A two-point (on-off) controlled AHU may even not switch off if the measured value drifts, or will be switched off only on a time scheduler event (if implemented and set up correctly). In systems with analogue control (VAV units), a drifted sensor results either in unnecessarily high energy consumption, or in bad air in the room. This is why not only measured values but also the functionality of the technologies must be evaluated on a long-term basis. Service engineers should not only check the trend values, but also the current value and measuring range of the sensor.

In the control system settings, the room COsetpoints and hystereses in the on-off controllers should be revised. It is also advisable to verify that the controller setting corresponds with the required functionality: There is CO2setpoint of 800 ppm at both images below. However, the first one switches on when the setpoint is reached („the critical pollution level was achieved“), while the other switches off at the same setpoint („the air is now clean enough“). The hysteresis (Xp = 300 ppm) is set up so that the unit does not start too frequently.


The COsetpoint is set so as to switch the ventilation on or off?

If the system is set according to the first sequence, the time to achieve that low carbon dioxide level (500 ppm) in the room may be too long as the gas concentration decrease is exponential; in the worst case when the sensor had drifted too much the measured value will never be able to achieve the 500 ppm level and the unit will never stop.

Properly calibrated and maintained COsensors are thus a must for efficient and healthy operation of the air handling unit.

[1] [2] [3]

Taken from the original: