Ways of sending a Float array as a value of a characteristic

Question

want to send a float as a value of a characteristic

I have successfully achieved this using the simple way

Below is the example of just two float variables in an array.

float temperature[2]=  {1.1,2.2}; 

attr_char_value.init_len  = 8;
attr_char_value.init_offs = 0;
attr_char_value.max_len   = 8;
attr_char_value.p_value   = (uint8_t*)&temperature;

And this works fine, in the future I want to be able to send180 variables so i will need to change

attr_char_value.max_len   = 180;

While reading up upon Heart monitor and Blood pressure monitoring services, they use this method:

static uint8_t hts_measurement_encode(ble_hts_t      * p_hts,
                                      ble_hts_meas_t * p_hts_meas,
                                      uint8_t        * p_encoded_buffer)
{
    uint8_t  flags = 0;
    uint8_t  len   = 1;
    uint32_t encoded_temp;

    // Flags field
    if (p_hts_meas->temp_in_fahr_units)
    {
        flags |= HTS_MEAS_FLAG_TEMP_UNITS_BIT;
    }
    if (p_hts_meas->time_stamp_present)
    {
        flags |= HTS_MEAS_FLAG_TIME_STAMP_BIT;
    }

    // Temperature Measurement Value field
    if (p_hts_meas->temp_in_fahr_units)
    {
        flags |= HTS_MEAS_FLAG_TEMP_UNITS_BIT;

        encoded_temp = ((p_hts_meas->temp_in_fahr.exponent << 24) & 0xFF000000) |
                       ((p_hts_meas->temp_in_fahr.mantissa <<  0) & 0x00FFFFFF);
    }
    else
    {
        encoded_temp = ((p_hts_meas->temp_in_celcius.exponent << 24) & 0xFF000000) |
                       ((p_hts_meas->temp_in_celcius.mantissa <<  0) & 0x00FFFFFF);
    }
    len += uint32_encode(encoded_temp, &p_encoded_buffer[len]);

    // Time Stamp field
    if (p_hts_meas->time_stamp_present)
    {
        flags |= HTS_MEAS_FLAG_TIME_STAMP_BIT;
        len   += ble_date_time_encode(&p_hts_meas->time_stamp, &p_encoded_buffer[len]);
    }

    // Temperature Type field
    if (p_hts_meas->temp_type_present)
    {
        flags                  |= HTS_MEAS_FLAG_TEMP_TYPE_BIT;
        p_encoded_buffer[len++] = p_hts_meas->temp_type;
    }

    // Flags field
    p_encoded_buffer[0] = flags;

    return len;
}

This code uses a struct called

    typedef struct
{
  int8_t  exponent;                                                         /**< Base 10 exponent */
  int32_t mantissa;                                                         /**< Mantissa, should be using only 24 bits */
} ieee_float32_t;

So my question is, is this another way or better way of sending a float variable?? I don't understand what they are doing here as almost every NRF ble example uses this method.

Hung Bui · Accepted Answer

Hi Nicko, 
 In our example we have to do with IEEE-11073 float (with 24 bit mantissa and 8 bit exponent) because it's defined in the spec as in here . 
 If you are fine with C float, you can simply convert it to 4 bytes like what you are doing now. Note that each float requires 4 bytes, if you have 180 values, you would need 720 bytes. This may exceed the max size of characteristic we support now. I would suggest you to send them in 20 bytes trunk, so you keep the characteristic size at 20 bytes, and send the result with 36 notification packets.

Ways of sending a Float array as a value of a characteristic

Top Replies