Grove – Carbon Dioxide Sensor(MH-Z16) trial

In preparation for a student project to monitor the CO2 levels in a number of classrooms I purchased a Grove – Carbon Dioxide Sensor(MH-Z16) for evaluation.

I downloaded the seeedstudio wiki example code, compiled and uploaded it to one of my Arduino Uno R3 devices.

I increased delay between readings to 10sec and reduced the baud rate of the serial logging to 9600baud.

/*
  This test code is write for Arduino AVR Series(UNO, Leonardo, Mega)
  If you want to use with LinkIt ONE, please connect the module to D0/1 and modify:

  // #include <SoftwareSerial.h>
  // SoftwareSerial s_serial(2, 3);      // TX, RX

  #define sensor Serial1
*/


#include <SoftwareSerial.h>
SoftwareSerial s_serial(2, 3);      // TX, RX

#define sensor s_serial

const unsigned char cmd_get_sensor[] =
{
    0xff, 0x01, 0x86, 0x00, 0x00,
    0x00, 0x00, 0x00, 0x79
};

unsigned char dataRevice[9];
int temperature;
int CO2PPM;

void setup()
{
    sensor.begin(9600);
    Serial.begin(9600);
    Serial.println("get a 'g', begin to read from sensor!");
    Serial.println("********************************************************");
    Serial.println();
}

void loop()
{
    if(dataRecieve())
    {
        Serial.print("Temperature: ");
        Serial.print(temperature);
        Serial.print("  CO2: ");
        Serial.print(CO2PPM);
        Serial.println("");
    }
    delay(10000);
}

bool dataRecieve(void)
{
    byte data[9];
    int i = 0;

    //transmit command data
    for(i=0; i<sizeof(cmd_get_sensor); i++)
    {
        sensor.write(cmd_get_sensor[i]);
    }
    delay(10);
    //begin reveiceing data
    if(sensor.available())
    {
        while(sensor.available())
        {
            for(int i=0;i<9; i++)
            {
                data[i] = sensor.read();
            }
        }
    }

    for(int j=0; j<9; j++)
    {
        Serial.print(data[j]);
        Serial.print(" ");
    }
    Serial.println("");

    if((i != 9) || (1 + (0xFF ^ (byte)(data[1] + data[2] + data[3] + data[4] + data[5] + data[6] + data[7]))) != data[8])
    {
        return false;
    }

    CO2PPM = (int)data[2] * 256 + (int)data[3];
    temperature = (int)data[4] - 40;

    return true;
}

The debug output wasn’t too promising there weren’t any C02 parts per million (ppm) values and the response payloads looked wrong. So I downloaded the MH-Z16 NDIR CO2 Sensor datasheet for some background. The datasheet didn’t mention any temperature data in the message payloads so I removed that code.

The response payload validation code was all on one line and hard to figure out what it was doing.

    if((i != 9) || (1 + (0xFF ^ (byte)(data[1] + data[2] + data[3] + data[4] + data[5] + data[6] + data[7]))) != data[8])
    {
        return false;
    }

To make debugging easier I split the payload validation code into several steps so I could see what was failing.

/*
  This test code is write for Arduino AVR Series(UNO, Leonardo, Mega)
  If you want to use with LinkIt ONE, please connect the module to D0/1 and modify:

  // #include <SoftwareSerial.h>
  // SoftwareSerial s_serial(2, 3);      // TX, RX

  #define sensor Serial1
*/


#include <SoftwareSerial.h>
SoftwareSerial s_serial(2, 3);      // TX, RX

#define sensor s_serial

const unsigned char cmd_get_sensor[] =
{
    0xff, 0x01, 0x86, 0x00, 0x00,
    0x00, 0x00, 0x00, 0x79
};

unsigned char dataRevice[9];
int CO2PPM;

void setup()
{
    sensor.begin(9600);
    Serial.begin(9600);
    Serial.println("get a 'g', begin to read from sensor!");
    Serial.println("********************************************************");
    Serial.println();
}

void loop()
{
    if(dataRecieve())
    {
        Serial.print("  CO2: ");
        Serial.print(CO2PPM);
        Serial.println("");
    }
    delay(10000);
}

bool dataRecieve(void)
{
    byte data[9];
    int i = 0;

    //transmit command data
    for(i=0; i<sizeof(cmd_get_sensor); i++)
    {
        sensor.write(cmd_get_sensor[i]);
    }
    delay(10);
    //begin reveiceing data
    if(sensor.available())
    {
        while(sensor.available())
        {
            for(int i=0;i<9; i++)
            {
                data[i] = sensor.read();
            }
        }
    }

    for(int j=0; j<9; j++)
    {
        Serial.print(data[j]);
        Serial.print(" ");
    }
    Serial.println("");

    // First calculate then validate the check sum as there is no point in proceeding if the packet is corrupted. (code inspired by datasheet algorithm)
    byte checksum = 0 ;
    for(int j=1; j<8; j++)
    {
      checksum += data[j];
    }
    checksum=0xff-checksum; 
    checksum+=1;
       
    if  (checksum != data[8])
    {
      Serial.println("Error checksum");
      return false;
    }

    // Then check the start byte to make sure response is what we were expecting
    if ( data[0] != 0xFF )
    {
        Serial.println("Error start byte");
        return false;
    }

    // Then check the command byte to make sure response is what we were expecting
    if ( data[1] != 0x86 )
    {
        Serial.println("Error command");
        return false;
    }


    CO2PPM = (int)data[2] * 256 + (int)data[3];

    return true;
}

From these modifications I could see the payload was messed up and based on the datasheet message descriptions it looked like it was offset by a byte or two.

15:58:32.509 -> get a 'g', begin to read from sensor!
15:58:32.578 -> ********************************************************
15:58:32.612 -> 
15:58:32.612 -> 255 134 6 238 76 0 0 1 255 
15:58:32.647 -> Error checksum
15:58:42.631 -> 57 255 134 6 246 76 0 0 1 
15:58:42.666 -> Error checksum
15:58:52.667 -> 49 255 134 5 125 76 0 0 1 
15:58:52.702 -> Error checksum
15:59:02.704 -> 171 255 134 4 86 76 0 0 1 
15:59:02.750 -> Error checksum

I had a look at the code and the delay(10) after sending the sensor reading request message caught my attention. I have found that often delay(x) commands are used to “tweak” the code to get it to work.

These “tweaks” often break when code is run on a different device or sensor firmware is updated changing the timing of individual bytes, or request-response processes.

I removed the delay(10) replaced it with a serial.flush() and changed the code to display the payload bytes in hexadecimal.

/*
  This test code is write for Arduino AVR Series(UNO, Leonardo, Mega)
  If you want to use with LinkIt ONE, please connect the module to D0/1 and modify:

  // #include <SoftwareSerial.h>
  // SoftwareSerial s_serial(2, 3);      // TX, RX

  #define sensor Serial1
*/


#include <SoftwareSerial.h>
SoftwareSerial s_serial(2, 3);      // TX, RX

#define sensor s_serial

const unsigned char cmd_get_sensor[] =
{
    0xff, 0x01, 0x86, 0x00, 0x00,
    0x00, 0x00, 0x00, 0x79
};

unsigned char dataRevice[9];
int CO2PPM;

void setup()
{
    sensor.begin(9600);
    Serial.begin(9600);
    Serial.println("get a 'g', begin to read from sensor!");
    Serial.println("********************************************************");
    Serial.println();
}

void loop()
{
    if(dataRecieve())
    {
        Serial.print("  CO2: ");
        Serial.print(CO2PPM);
        Serial.println("");
    }
    delay(10000);
}

bool dataRecieve(void)
{
    byte data[9];
    int i = 0;

    //transmit command data
    for(i=0; i<sizeof(cmd_get_sensor); i++)
    {
        sensor.write(cmd_get_sensor[i]);
    }
    Serial.flush();
    
    //begin reveiceing data
    if(sensor.available())
    {
        while(sensor.available())
        {
            for(int i=0;i<9; i++)
            {
                data[i] = sensor.read();
            }
        }
    }

    for(int j=0; j<9; j++)
    {
        Serial.print(data[j],HEX);
        Serial.print(" ");
    }
    Serial.println("");

    // First calculate then validate the check sum as there is no point in proceeding if the packet is corrupted. (code inspired by datasheet algorithm)
    byte checksum = 0 ;
    for(int j=1; j<8; j++)
    {
      checksum += data[j];
    }
    checksum=0xff-checksum; 
    checksum+=1;
       
    if  (checksum != data[8])
    {
      Serial.println("Error checksum");
      return false;
    }

    // Then check the start byte to make sure response is what we were expecting
    if ( data[0] != 0xFF )
    {
        Serial.println("Error start byte");
        return false;
    }

    // Then check the command byte to make sure response is what we were expecting
    if ( data[1] != 0x86 )
    {
        Serial.println("Error command");
        return false;
    }


    CO2PPM = (int)data[2] * 256 + (int)data[3];

    return true;
}

The initial values from the sensor were a bit high, but after leaving the device running for 3 minutes (Preheat time in the documentation) they settled down into a reasonable range

16:14:31.686 -> get a 'g', begin to read from sensor!
16:14:31.721 -> ********************************************************
16:14:31.789 -> 
16:14:31.789 -> 255 134 6 224 75 0 0 1 72 
16:14:31.823 ->   CO2: 1760
16:14:41.824 -> 255 134 6 224 75 0 0 1 72 
16:14:41.824 ->   CO2: 1760
16:14:51.824 -> 255 134 5 189 75 0 0 1 108 
16:14:51.858 ->   CO2: 1469
16:15:01.868 -> 255 134 3 157 75 0 0 1 142 
16:15:01.868 ->   CO2: 925
16:15:11.857 -> 255 134 3 223 75 0 0 1 76 
16:15:11.892 ->   CO2: 991
16:15:21.882 -> 255 134 6 56 75 0 0 1 240 
16:15:21.917 ->   CO2: 1592
16:15:31.911 -> 255 134 4 186 75 0 0 1 112 
16:15:31.945 ->   CO2: 1210
16:15:41.927 -> 255 134 3 131 75 0 0 1 168 
16:15:41.962 ->   CO2: 899
16:15:51.940 -> 255 134 3 30 75 0 0 1 13 
16:15:51.975 ->   CO2: 798
16:16:01.986 -> 255 134 2 201 75 0 0 1 99 
16:16:01.986 ->   CO2: 713
16:16:11.985 -> 255 134 4 133 75 0 0 1 165 
16:16:12.019 ->   CO2: 1157
16:16:22.020 -> 255 134 6 62 75 0 0 1 234 
16:16:22.053 ->   CO2: 1598
16:16:32.041 -> 255 134 5 80 75 0 0 1 217 
16:16:32.041 ->   CO2: 1360
16:16:42.057 -> 255 134 3 204 75 0 0 1 95 
16:16:42.092 ->   CO2: 972
16:16:52.084 -> 255 134 3 191 75 0 0 1 108 
16:16:52.084 ->   CO2: 959
16:17:02.102 -> 255 134 2 230 75 0 0 1 70 
16:17:02.102 ->   CO2: 742
16:17:12.094 -> 255 134 3 106 75 0 0 1 193 
16:17:12.129 ->   CO2: 874
16:17:22.111 -> 255 134 2 227 75 0 0 1 73 
16:17:22.145 ->   CO2: 739
16:17:32.139 -> 255 134 3 225 75 0 0 1 74 
16:17:32.172 ->   CO2: 993
16:17:42.170 -> 255 134 3 109 75 0 0 1 190 
16:17:42.204 ->   CO2: 877
16:17:52.174 -> 255 134 2 188 75 0 0 1 112 
16:17:52.207 ->   CO2: 700
16:18:02.218 -> 255 134 2 70 75 0 0 1 230 
16:18:02.253 ->   CO2: 582
16:18:12.239 -> 255 134 2 163 75 0 0 1 137 
16:18:12.239 ->   CO2: 675
16:18:22.251 -> 255 134 2 110 75 0 0 1 190 
16:18:22.285 ->   CO2: 622
16:18:32.246 -> 255 134 2 83 75 0 0 1 217 
16:18:32.280 ->   CO2: 595
16:18:42.277 -> 255 134 2 48 75 0 0 1 252 
16:18:42.312 ->   CO2: 560
16:18:52.305 -> 255 134 2 62 75 0 0 1 238 
16:18:52.339 ->   CO2: 574

Bill of materials (prices as at Jan 2019)

Arduino Uno R3 USD24.95 NZD54
Grove Base shield V2 USD4.45 NZD14
Grove – Carbon Dioxide Sensor(MH-Z16) USD89

After these tentative fixes for the MH-Z16 sensor I think going to see if there are any other libraries written by someone smarter than me available.

5 thoughts on “Grove – Carbon Dioxide Sensor(MH-Z16) trial”

hackybaby says:

December 7, 2019 at 1:05 am

thank you so much for writing down your process !
I am going to use your code as a starting point for my master thesis. ❤

- devmobilenz says:
  
  December 7, 2019 at 2:32 pm
  
  Hi
  Are you using the MH-Z16 or SCD30 sensor?
  
  I found the SCD30 worked well (plus the temperature & humidity values were useful) and was a bit cheaper though the multi-day calibration process was a pain.
  
  Keep us posted on your progress…
  
  @KiwiBryn
  
  - hackybaby says:
    
    December 30, 2019 at 6:53 am
    
    I am using MH-Z16.
    
    Currently trying to combine the CO2 Sensor (MH-Z16) with a Dust-Sensor (Seeed HM330X) with the Seeedstudio SD-Card Shield. Already running into multiple problems:
    -MH-Z16 wants Baud 9600
    -HM300x wants Baud 115200
    -SD-Shield data logging didn’t work (solved)
    -dynamic Memory usage ~95%
    -program storage usage ~90%
    -GPS Shield still not tested
    
    I now ordered the Seeeduino Cortex-M0+ because it offers more RAM.
    
    And these are all stationary problems; I need the setup to fit into an airtight case with an air pump and battery.
    
    I will post a link below to open up all my code with pictures on GitHub as soon as I feel confident. ❤
Mike Stone says:

March 11, 2020 at 1:34 am

I’m just getting started on a Co2 probe project and had a question regarding connections to the MH-Z16 probe. Noticed you had a 4 wire connection from the probe to the Grove Base Shield. probe has 7 pins . Mine will ship with a 7 pin cable , please explain what jumper you used . I also have a Sandbox MH-Z16 module that allows 12c/UART connections, maybe i don’t even need this ?

Thanks

devmobilenz says:

March 11, 2020 at 6:17 pm

Hi

On the manufacturer’s site (https://www.winsen-sensor.com/sensors/co2-sensor/mh-z16.html) their datasheet details the pins on the end of the sensor.

Pad1 HD
Pad2 Vout (0.4～2V,custom made)
Pad3 GND
Pad4 Vin(input voltage 4.5V～5.5V）
Pad5 UART（RXD） 0～3.3V input digital
Pad6 UART（TXD） 0～3.3V output digital
Pad7 PWM

The Grove cable has pad 3 GND (black wire), pad 4 Vin (red wire), pad 5 UART RXD (white wire) and pad 6 UART TXD (yellow wire).

You should be able to connect your sensor up using the same wiring and use one of the serial libraries discussed in my blog.

@KiwiBryn

devMobile's blog

Random wanderings through Microsoft Azure, Windows 10 Iot Core, .Net Micro Framework, .Net Nano Framework and GHI Electronics TinyCLR

Grove – Carbon Dioxide Sensor(MH-Z16) trial

5 thoughts on “Grove – Carbon Dioxide Sensor(MH-Z16) trial”

Leave a Reply Cancel reply

devMobile's blog

Random wanderings through Microsoft Azure, Windows 10 Iot Core, .Net Micro Framework, .Net Nano Framework and GHI Electronics TinyCLR

Share this:

Like this:

5 thoughts on “Grove – Carbon Dioxide Sensor(MH-Z16) trial”

Leave a Reply Cancel reply