Accelerometer and Gyroscope are switched

[Cacciuc]

When using just the simplest implementation, it seems that the accelerometer values and the gyroscope values are switched. My implementation in main ` . . . //initialisations from CubeMx SD_MPU6050 mpu; if (SD_MPU6050_Init(&hi2c1, &mpu, SD_MPU6050_Device_1, SD_MPU6050_Accelerometer_4G, SD_MPU6050_Gyroscope_1000s) != SD_MPU6050_Result_Ok) { LedRed(1); asm("nop"); return -1; } / USER CODE END 2 /

/* Infinite loop */
/* USER CODE BEGIN WHILE */
while (1) {
    for (int i = 0; i < 5; i++) {
        if (SD_MPU6050_ReadAll(&hi2c1, &mpu)
                != SD_MPU6050_Result_Ok) {
            LedRed(1);
        }
        HAL_Delay(100);
    }

}` In the STMStudio the values are shown in a bar graph. This is a screenshot of the MPU6050 sitting still in a weird angle. All values are stable. I will switch the accelerometer and gyroscope variables in my Code, but since no one else has the same error, i will not make a pull request.

[git-oliver]

Hello Cacciuc,

the answer ist no, because of multiple reasons:

1) If you look insideSD_MPU6050_READ_ALL() the register to read out is MPU6050_ACCEL_XOUT_H (0x3B). And from this reg we will read 14 bytes (0xE) If you now do 0x3B + 0xE = 0x49.
And if you look inside the datasheet [1] (page. 29 to 31) all reg between 0x3B and0x49 are the

• Accelerometer (0x3B to 0x40)
• Temperature (0x41 to 0x42)
• Gyroscope (0x43 to 0x48)

So the function does everything correct

2) If you mount the MPU6050 sensor on your table and fix them with simple tape, then only the z-Acceleration should be unequal to zero. Set the following properties:

SD_MPU6050_Init(    hi2c1,
                    mpu1,
                    SD_MPU6050_Device_0,
                    SD_MPU6050_Accelerometer_2G,
                    SD_MPU6050_Gyroscope_250s
                     );

After setting these properties you need to calculate the Offset. For that read out 1000 datapoints. Then calculate the Offset with the code. Note I don't use the first 400 values. Thats only for safety, because I found out that the the first values are garbage. To be sure to not get garbage values, I dont care about the first 400 values.

%% Offset für acc_x
offset_acc_x = raw_acc_x(400:end);
offset_acc_x = (mean(offset_acc_x)); 

%% Offset für acc_y
offset_acc_y = raw_acc_y(400:end);
offset_acc_y = (mean(offset_acc_y));

%% Offset für acc_z
offset_acc_z = raw_acc_z(400:end);
offset_acc_z = (mean(offset_acc_z));
offset_acc_z = offset_acc_z - 16384;    % z =! 1g (@range +-2g)

%% Offset für gyro_x
offset_gyro_x = raw_gyro_x(400:end);
offset_gyro_x =(mean(offset_gyro_x));

%% Offset für gyro_y
offset_gyro_y = raw_gyro_y(400:end);
offset_gyro_y =(mean(offset_gyro_y));

%% Offset für gyro_z
offset_gyro_z = raw_gyro_z(400:end);
offset_gyro_z =(mean(offset_gyro_z));

After that you can read out new datapoints and subtract them with you offset values. You should get nearly zeros on x and y acceleration axis and 16384 on the z-axis. This is because the z-axis measures the gravity.

So long story short:

The function works as expected and you your assumption is wrong

In a few weeks I will fork this repo and add a complementary filter to calculate the angle. And some others improvements of this repo/code.

[1] https://invensense.tdk.com/wp-content/uploads/2015/02/MPU-6000-Register-Map1.pdf