Air Quality Sensor

  • Hero Member

    Air Quality Index is a combination of several sensors values:

    • sulfur dioxide (SO2) MQ136 / (avoid:2SH12 no datasheet)
    • nitrogen dioxide (NO2) MiCS-2710 MiCS-2714 MICS-6814
    • carbon monoxide (CO) MQ-7 MiCS-5525 TGS2442 MICS-6814
    • ozone (O3) MQ131 MICS-2610 MICS-2614
    • Particles concentration: PM2.5 and PM10 (SamYoung DSM501, Shinyei_PPD42NS, avoid: Sharp_GP2Y1010AU)

    Additional gases can be interesting to track:

    Current work being done:

    • manage default Ro once
    • take into account humidity
    • manage to get proper values from TGS2600
    • possibility to have a push if a value goes up too quick (for deadly gases even without calibration)

    What is done:

    • remove powerdown to keep heaters on
    • add a five minute pre-heat on each startup
    • removed the powerdown
    • 2SH12 datasheet doesn't exist

    WARNING: all sensors differs even within the same reference, a calibration has to be done each time. Do not rely upon a sketch result with deadly gases.

    Working sketch in 1.4/1.5:

  • Mod

    @epierre The MQ-135 sensor, for example, has an internal heater inside, which consumes 800mW max.
    This disqualifies the sensor for battery powered applications as the batteries would run out very fast.
    A sensor module like the MQ-135 has an analog output voltage, which you can feed to an analog input of your Arduino. Most modules also have a digital output, which can be tuned using the onboard potmeter, which triggers when a certain level has been reached.
    These sensors come calibrated (to a certain amount) but calibrating them yourself will be complicated as you need to replicate an environment with a defined relative amount of the gas the ensor is sensitive to...

  • Hero Member


    sadly the discussions were lost...

    So let's start again on the subject:

    Here is the list of my current experimentations:

    • MQ135.ino : for CO2/COV ... validated
    • MQ2.ino : for ethanol... needs pre-heat : ongoing
    • MQ6 sensor : just received
    • MQ131 sensor : just received
    • TGS2600 sensor : just received

    I'm getting serious a with a mega and a sensor board, 4 of them are connected together !

    The MQ135 is not on the list, I use it to measure CO2 in children's room...

    • MQ2: 85 - Combustible gas and smoke
    • MQ6: 25 - Isobutane, Butane, LPG
    • MQ131: 57 - O3, CL2
    • TGS2600: 943 - ethanol, hydrogen, iso-butane, CO, Methane

    As discussed before, I would appreciate someone helping me reading the logarithmic curves in the datasheet (link on each sensor name above) to get the linear value based on analog reading. I need to learn reading it... thanks!

    as an example here in sandboxelectronics we saw that several values could be read from the datasheet:

    float           LPGCurve[3]  =  {2.3,0.21,-0.47};   //two points are taken from the curve.
                                                        //with these two points, a line is formed which is "approximately equivalent"
                                                     //to the original curve.
                                                    //data format:{ x, y, slope}; point1: (lg200, 0.21), point2: (lg10000, -0.59)
    float           COCurve[3]  =  {2.3,0.72,-0.34};    //two points are taken from the curve.
                                                    //with these two points, a line is formed which is "approximately equivalent"
                                                    //to the original curve.
                                                    //data format:{ x, y, slope}; point1: (lg200, 0.72), point2: (lg10000,  0.15)
    float           SmokeCurve[3] ={2.3,0.53,-0.44};    //two points are taken from the curve.
                                                    //with these two points, a line is formed which is "approximately equivalent"
                                                    //to the original curve.
                                                    //data format:{ x, y, slope}; point1: (lg200, 0.53), point2: (lg10000,  -0.22)                                                    
    float           Ro           =  10;                 //Ro is initialized to 10 kilo ohms

  • Mod

    @epierre To decorate your codeblock use 4 spaces or one tab character first on each line.

    This forum uses markdown.
    For help: press the little questionmark-icon in the compose window.

  • Mod

    @epierre you say you need some help reading the logarithmic curves, right?
    I copied one curve from the MQ2 datasheet :

    Along the horizontal axis you see the values 100, 1000 & 10000.
    Between 100 & 1000 each vertical line adds 100, between 1000 & 10000 each vertical line adds 1000.
    On the vertical axis it works just the same: each horizontal line between 0.1 & 1 adds 0.1, between 1 & 10 it adds 1.
    So the first H2 lies at (200,2)

  • Hero Member

    @Yveaux Thanks, but can you confirm the measure that I quoted above, for I don't find them... (LLPG, CO, what they call smoke ???)

    @marceltrapman thanks for the hint... I spent some time trying to find it... I'm more a wikiforumman

  • Hero Member

    another very interresting post on the theory behind:


  • Hero Member

    Ok mesurement understood...

    values must be in logarithmic... sometime they display lg for log, or directly the log calculated value which does not help understand it...

    so for the slope from the values read : (LOG(0,4)-LOG(1))/(LOG(10000)-LOG(1000)) and I agree to the curves.

    Now let's go to the next point: resistive protection.
    My boards (to few exception) are this :

    There are 1K and 10K resistors, plus a variable one.

    From the article I base myself, it says:

    The protection resistor (4.7Kohms) and the adjustable (0-50Kohms) are in serial which forms a load resistor RL (4.7-54.7Kohms). The sensor’s resistance RS and RL forms a voltage divider. The output voltage on the signal pin could be read by Arduino or MCU via ADC.  Given a value of RL , Power Supply Voltage, and output voltage, RS could be derived.  Based on the chart provided in the MQ-2 datasheet, RS in clean air under given temperature and humidity is a constant,which is the “initial” resistance of the sensor named RO. RO of the resistor could be derived from RS. 

    I guess it is a different board here, so values must differ. Have you an idea how to mesure this with my multimeter ? else each time the sensor would start, the value would be calculated, clean or not :

    The main job of the calibration is to calculate the RO by sampling and averaging the readings when the module is placed in the clean air. Once the RO is derived, the concentration of target gas could be calculated by using the RS/RO ratio as the input. To achieve more accuracy, a segmented look-up table should be used. However, in the demonstration, a linear formula is used as an approximation to the original curve.

  • Hero Member

    So tonight with these new formulae, I updated my mega sketch to get :

     MQ2    :LPG   :0ppm    CO    :0ppm    SMOKE :0ppm
     MQ6    :LPG   :0ppm    CH4   :0ppm
     MQ131  :CL2   :1ppm    O3    :1ppm
     TGS2600:H2    :0ppm    C2H5OH:0ppm    C4H10 :0ppm

    Going through the "clean air calibation" I have:

    float MQResistanceCalculation(int raw_adc)    {
      return ( ((float)RL_VALUE*(1023-raw_adc)/raw_adc));

    Here are the readings without:


    and with all is 0... BTW...

    What do you think of that ? nothing indicated that MQ6 should be so big, and tgs so low... same for MQ131 that should be 100k-200k per datasheet....

  • Mod

    @epierre are you totally sure of the calculation in MQResistanceCalculation? Maybe the evaluation is correct, but I always add some braces to make the evaluation order explicit: first do the substraction, then the float multiplication and the integer division, cast to float, as last.

  • Hero Member

    @Yveaux you mean something like this ?

    return ( (((float)RL_VALUE*(1023.0 - (float) raw_adc)) / (float) raw_adc));

    I somehow forked the discussion on and sent an email in parallel do Davide Gironi for I would like to understand his formula as opposed to the curve approach.

  • Hero Member

    This post is deleted!

  • Mod

    @epierre something like that, yes. You could also feed it some artificial values in a debug build and see if the calculated values match your expectations.
    It wouldn't be the first time that some calculations (divisions) are unexpectedly calculated as ints and always return either 0 or 1 instead of the expected fraction...

  • Hero Member


    Having sorted out how to make a power regression, I now have the sames values as Davide Gironi.

    I applied this to MQ2, MQ6, MQ131, MQ135 and TGS2600 with resistance calibration at start.

    MQ2    :LPG   :12ppm    CO    :0ppm    SMOKE :21ppm
    MQ6    :LPG   :3ppm    CH4   :39ppm
    MQ131  :CL2   :2ppm    O3    :3ppm
    TGS2600:H2    :5191ppm    C2H5OH:8825ppm    C4H10 :16954ppm
    MQ135  :CO2    :0ppm    CO    :0ppm    CH3    :0ppm    NH4    :0ppm
    GP2Y1010AU0F Raw Signal Value (0-1023): 19 - Voltage: 0.09 - Dust Density: -84.23

    the MQ135 is still doing its first heating period, do not care about its values.

    Clearly the TGS shows abnormal values, I'm quite puzzled by this, using the previous method it was always 0...

  • Mod

    @epierre I'm still a bit wondered by the way in which you present the values.
    I, for instance, have an MQ135. Is "sensitive to noxious gas concentration. MQ135 sensor also sensitive to sulfide, benzene steam, smoke, etc". The logarithmic charts in the datasheet show the sensitivity to the various gases.
    This sensor has a single analog output from which IMHO you cannot determine the fraction of individual gases. Yet you present them as individual values in your measurement data... How can you distinguish?

  • Hero Member

    @Yveaux in fact it is a correlation, if you have a value from the analog, then from the curve you can guess it is one of the group, not one individually... it is the way they are designed...

  • Mod

    @epierre That's exactly my point!

  • Hero Member

    @Yveaux and there's worse : if you don't have the environment to calibrate it with real gas, it is just an indication... that is maybe why some projects that may have taken such sensor may hae a production process issue at the moment... In fact despite my researches, I've never found someone that said it made it to mesure correctly with gas sensors, mostly theory or just analog reading only...

    An example: my MQ135 was silencious, I put its nose in the Aceton bottle:

    Q2 :LPG :11ppm CO :0ppm SMOKE :18ppm
    MQ6 :LPG :3ppm CH4 :36ppm
    MQ131 :CL2 :2ppm O3 :3ppm
    TGS2600:H2 :7602ppm C2H5OH:13706ppm C4H10 :26350ppm
    MQ135 :CO2 :0ppm CO :0ppm CH3 :0ppm NH4 :0ppm
    Dust :raw : 18 Voltage: 0.09 - Dust Density: -85.06

    MQ2 :LPG :208ppm CO :0ppm SMOKE :832ppm
    MQ6 :LPG :44ppm CH4 :172ppm
    MQ131 :CL2 :16ppm O3 :17ppm
    TGS2600:H2 :7774ppm C2H5OH:13818ppm C4H10 :27274ppm
    MQ135 :CO2 :999ppm CO :0ppm CH3 :484ppm NH4 :1904ppm
    Dust :raw : 58 Voltage: 0.28 - Dust Density: -51.86

    The MQ135 has reacted, but all others around...

  • Hero Member

    doing the same on my basic sketch here what I have:

    2014-08-03 11:29:57 3 0 1 24 -28393

    looks like my old sketch needs some update...

  • Hero Member

    Doing the same on the basic sketch :

     Vrl / Rs / ratio:4.21 3731.17 0
     CO ppm :inf


    Vrl / Rs / ratio:4.21 3731.17 0
    CO ppm :inf