8bit（两个值）数组到16bit数组（一个值）

Question

I'm reading data from a sensor, that sends them as two 8bit values and user have to convert them into one signed 16bit value. I save these values into an array of 5400 elements (the sensor is accelerometer+gyro+magnetometer, so 9 axes, so in the end I'll have 300 values of each). So lets say I need reading from accelerometer's X axis. In my array I save these in first two places, so my accX readings are on indexes 0,1 then 18,19 then 36,37 and so on..... Now I need to extract these into new 16bit array with 300 elements that will represent only the X axis from accelerometer. What I came up is following:

uint8_t rawData[5400]={0};

int16_t *AccX = malloc(300*sizeof(int16_t));

for(int i=0;i<300;i++){
    AccX[i]=(rawData[i+1+i*17]<<8 | rawData[i+0+i*17]);
}
//do some calculations
free(AccX);

Yet in my AccX array are values I do not expect to be there. Any idea what might have gone wrong? The sensor is LSM9DS1.

Edit: sorry I didn't give more info about the device. I know for sure this works:

x_accel = (rawData[1]<<8 | rawData[0]);

also I inspected the rawData array and it contains correct data. It goes wrong when I try to fill that data to AccX array

Answer 1

Look at the datasheet . Essentially, there are two separate devices in a single package.

You need the accelerometer/gyroscope device registers:

• 21 and 22 (0x15 and 0x16) for temperature

• 24 and 25 (0x18 and 0x19) for gyroscope X axis

• 26 and 27 (0x1A and 0x1B) for gyroscope Y axis

• 28 and 29 (0x1C and 0x1D) for gyroscope Z axis

• 40 and 41 (0x28 and 0x29) for accelerometer X axis

• 42 and 43 (0x2A and 0x2B) for accelerometer X axis

• 44 and 45 (0x2C and 0x2D) for accelerometer Z axis

For the magnetic field, you need the magnetic sensor registers:

• 40 and 41 (0x28 and 0x29) for X axis field strength

• 42 and 43 (0x2A and 0x2B) for Y axis field strength

• 44 and 45 (0x2C and 0x2D) for Z axis field strength

These are all in two's complement format, with least significant byte first.

Since practically all microcontrollers use two's complement format for signed integers (STM32 definitely does, as do ARMs and AVRs, including all Arduinos), you can write five simple helper functions. In pseudo-C:

static int16_t  temperature;
static int16_t  gyro_x, gyro_y, gyro_z;
static int16_t  accel_x, accel_y, accel_z;
static int16_t  mag_x, mag_y, mag_z;
static uint8_t  buf[6];


static void update_temperature(void)
{
    /* Read 2 bytes from registers 0x15 and 0x16 (21 and 22)
       from the accelerometer/gyroscope device, to buf[] */
    temperature = (int16_t)((((unsigned int)buf[1]) << 8) | (unsigned int)(buf[0]));
}

static void update_gyro(void)
{
    /* Read 6 bytes from registers 0x18..0x1D (24 through 29)
       from the accelerometer/gyroscope device, to buf[] */
    gyro_x = (int16_t)((((unsigned int)buf[1]) << 8) | (unsigned int)(buf[0]));
    gyro_y = (int16_t)((((unsigned int)buf[3]) << 8) | (unsigned int)(buf[2]));
    gyro_z = (int16_t)((((unsigned int)buf[5]) << 8) | (unsigned int)(buf[4]));
}

static void update_accel(void)
{
    /* Read 6 bytes from registers 0x28..0x2D (40 through 45)
       from the accelerometer/gyroscope device, to buf[] */
    accel_x = (int16_t)((((unsigned int)buf[1]) << 8) | (unsigned int)(buf[0]));
    accel_y = (int16_t)((((unsigned int)buf[3]) << 8) | (unsigned int)(buf[2]));
    accel_z = (int16_t)((((unsigned int)buf[5]) << 8) | (unsigned int)(buf[4]));
}

static void update_mag(void)
{
    /* Read 6 bytes from registers 0x28..0x2D (40 through 45)
       from the magnetic sensor device, to buf[] */
    mag_x = (int16_t)((((unsigned int)buf[1]) << 8) | (unsigned int)(buf[0]));
    mag_y = (int16_t)((((unsigned int)buf[3]) << 8) | (unsigned int)(buf[2]));
    mag_z = (int16_t)((((unsigned int)buf[5]) << 8) | (unsigned int)(buf[4]));
}

static void update_all(void)
{
    update_temperature();
    update_gyro();
    update_accel();
    update_mag();
}

The assignments look funky, but the compiler should optimize them to sane machine code.

You'll probably want to write a helper function to do the actual reads (or perhaps three functions; one for reading 2 bytes from the accelerometer/gyroscope device to buf[], one for reading 6 bytes from the accelerometer/gyroscope device to buf[], and one for reading 6 bytes from the magnetic sensor device to buf[], all from consecutive registers given as a parameter to the function).

This approach uses 26 bytes of RAM, and should be suitable for all microcontrollers, both 8-bit (AVRs) and 32-bit (STM32, ARMs). The above code is not sensitive to the byte order (endianness) of the microcontroller.

---

If your device sends the gyroscope, accelerometer, and magnetic sensor values via USB or serial to your computer, you need to know the byte order (endianness) of the output values is.

The sensor itself provides them in least significant byte first. STM32 is, as far as I know, little-endian too, so it should also provide them least significant byte first.

So, lets assume you receive 18*N bytes from the device, with gyroscope axes first, then accelerometer axes, and finally magnetic sensor axis readings. The following should convert those values to signed integers for easy manipulation:

#include <stdlib.h>
#include <inttypes.h>

static inline void extract16toint(int *const                 to,
                                  const unsigned char *const from,
                                  const size_t               n,
                                  const size_t               stride)
{
    const unsigned char *const end = from + n * stride;
    const unsigned char       *src = from;
    int                       *dst;

    while (src < end) {
        *(dst++) = (int16_t)(  (unsigned int)(src[0])
                            | ((unsigned int)(src[1]) << 8) );
        src += stride;
    }
}

static void extract_gyroX(int *const                  to,
                          const unsigned char *const  data,
                          const size_t                n)
{
    extract16toint(to, data + 0, n, 18);
}

static void extract_gyroY(int *const                  to,
                          const unsigned char *const  data,
                          const size_t                n)
{
    extract16toint(to, data + 2, n, 18);
}

static void extract_gyroZ(int *const                  to,
                          const unsigned char *const  data,
                          const size_t                n)
{
    extract16toint(to, data + 4, n, 18);
}

static void extract_accelX(int *const                  to,
                           const unsigned char *const  data,
                           const size_t                n)
{
    extract16toint(to, data + 6, n, 18);
}

static void extract_accelY(int *const                  to,
                           const unsigned char *const  data,
                           const size_t                n)
{
    extract16toint(to, data + 8, n, 18);
}

static void extract_accelZ(int *const                  to,
                           const unsigned char *const  data,
                           const size_t                n)
{
    extract16toint(to, data + 10, n, 18);
}

static void extract_magX(int *const                  to,
                         const unsigned char *const  data,
                         const size_t                n)
{
    extract16toint(to, data + 12, n, 18);
}

static void extract_magY(int *const                  to,
                         const unsigned char *const  data,
                         const size_t                n)
{
    extract16toint(to, data + 14, n, 18);
}

static void extract_magZ(int *const                  to,
                         const unsigned char *const  data,
                         const size_t                n)
{
    extract16toint(to, data + 16, n, 18);
}

In OP's case, calling something like

int *accelX = malloc(300 * sizeof (int));
extract_accelX(accelX, rawData, 300);

should work.