MPU6050 accelerometer offset routines don't set the low byte properly - DFRobot SEN0142 library outdated

[acrylic-origami]

I'm running into an issue with MPU6050 calibration where the registers for setting offsets don't seem aligned to my chip. The high byte of each accelerometer axis does work (i.e. the MPU6050_RA_*A_OFFS_H registers) but I've discovered that the low byte actually sets the low byte of the next axis over (e.g. setting the X low byte actually sets the Y low byte), that is to say:

+--------+-------+--------+-------+--------+-------+
| 0x06   | 0x07  | 0x08   | 0x09  | 0x0A   | 0x0B  |
+--------+-------+--------+-------+--------+-------+
| X HIGH | Y LOW | Y HIGH | Z LOW | Z HIGH | X LOW |
+--------+-------+--------+-------+--------+-------+

Also, there are very specific conditions needed to set the low byte: it only persists if the high byte of the next axis is written after that, and for the first time. This is the only way I've discovered to set offsets, by twisting the bytes and registers:

int16_t xoff = -392, yoff = -213, zoff = 1301;

mpu.setZAccelOffset((zoff & 0xFF00) | (xoff & 0xFF));
mpu.setXAccelOffset((xoff & 0xFF00) | (yoff & 0xFF));
mpu.setYAccelOffset((yoff & 0xFF00) | (zoff & 0xFF));
// mpu.setZAccelOffset((zoff & 0xFF00) | (xoff & 0xFF)); // <- this call won't persist the Z low byte because it's not the first call, so the Z low byte is already locked

And because of this I also can't set the low byte of the first axis (Z in the example above) because it's already locked. The obvious alternatives don't work:

• Setting the high byte as a byte after setting all the low bytes first won't persist them;
• Shifting all the register values down by a byte won't work: it seems mandatory to set the high byte of one and the low byte of the next as a single word;
• I can't set all the bytes in the bank at once like:

  uint8_t bank[6] = { xoff >> 8, yoff & 0xFF, yoff >> 8, zoff & 0xFF, zoff >> 8, xoff & 0xFF };
  I2Cdev::writeBytes(MPU6050_DEFAULT_ADDRESS, MPU6050_RA_XA_OFFS_H, 6, bank); // nope. doesn't work.

More specifically, I have an MPU6050 04F974L81 EL 1532 E, on a breakout board from DFRobot. Does this issue sound familiar? The hoops to jump through seems to my eye as something of an anti-tampering measure, although I don't know if my chip is necessarily that new.

[ZHomeSlice]

Per the document
MPU-6500 Register Map and Descriptions
page 8
see Table 2 MPU-6050 compatible mode register map

0x06 6 XA_OFFSET_H R/W XA_OFFSET [14:7]
0x07 7 XA_OFFSET_L R/W XA_OFFSET [6:0] 
0x08 8 YA_OFFSET_H R/W YA_OFFSET [14:7]
0x09 9 YA_OFFSET_L R/W YA_OFFSET [6:0] 
0x0A 10 ZA_OFFSET_H R/W ZA_OFFSET [14:7]
0x0B 11 ZA_OFFSET_L R/W ZA_OFFSET [6:0]

While calibrating the gyro
The calibration routine starts at address 0x06
It cycles through each access one at a time using the
I2Cdev::writeWords(devAddr, SaveAddress + (i * shift), 1, (uint16_t *)&Data);
to store the calibration on each pass.
in the function writeWords the most significant bit is sent first then the least significant bit. so XA_OFFSET_H then XA_OFFSET_L get written in that order note that the pointer advances automatically to the next memory location.

            Wire.write((uint8_t)(data[i] >> 8));    // send MSB
            Wire.write((uint8_t)data[i]);         // send LSB

I don't see a problem with this...

this should read:

+--------+-------+--------+-------+--------+-------+
| 0x06   | 0x07  | 0x08   | 0x09  | 0x0A   | 0x0B  |
+--------+-------+--------+-------+--------+-------+
| X HIGH | X LOW | Y HIGH | Y LOW | Z HIGH | Z LOW |
+--------+-------+--------+-------+--------+-------+

I also checked the code for
The offsets will work for the mpu6050

void MPU6050::setXAccelOffset(int16_t offset) {
	uint8_t SaveAddress = ((getDeviceID() < 0x38 )? MPU6050_RA_XA_OFFS_H:0x77);
	I2Cdev::writeWord(devAddr, SaveAddress, offset);
}

Z

[acrylic-origami]

Turns out the problem was that I2Cdev provided by the manufacturer through DigiKey was a couple versions too old, and a change in the intervening time fixed the issue. To hopefully hit some searches, this is SEN0142 by DFRobot. I've brought it up with DFRobot.