icm20948 incorrect mangetometer sensitivity

[ydvo]

Answers checklist.

• I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
• I have searched the issue tracker for a similar issue and not found a similar issue.

What is the expected behavior?

The icm20498 magnetometer should return values in a +/- 4900 uT range as indicated by the datasheet.

What is the actual behavior?

The magnetometer returns absurdly high values.
ex: -188374.5781 x 18856.1875 y 190086.9531 z

Steps to reproduce.

Call read_magnetometer(ec) from icm20948.cpp component

Build or installation Logs.

More Information.

It seems like the scaling for the magnetometer in the icm20948 component is applied incorrectly at line 699 in icm20948.cpp:

template <espp::icm20948::Interface I>
Icm20948<I>::Value Icm20948<I>::read_magnetometer(std::error_code &ec) {
  std::lock_guard<std::recursive_mutex> lock(base_mutex_);
  RawValue raw = get_magnetometer_raw(ec);
  if (ec) {
    return {0.0f, 0.0f, 0.0f};
  }
  float sensitivity = get_magnetometer_sensitivity();
  Value value = {
      static_cast<float>(raw.x) / sensitivity,
      static_cast<float>(raw.y) / sensitivity,
      static_cast<float>(raw.z) / sensitivity,
  };
  // update values
  mag_values_ = value;
  return value;
}

get_magnetometer_sensitivity() returns the scaling factor in uT/LSB so should be applied by multiplying not dividing.

Changing this locally returns values in the expected range.

[ydvo]

Was still getting weird values even after making this change locally, so I looked into it further:

template <espp::icm20948::Interface I>
Icm20948<I>::RawValue Icm20948<I>::get_raw(Icm20948<I>::RegisterBank0 reg, std::error_code &ec) {
  std::lock_guard<std::recursive_mutex> lock(base_mutex_);
  // select bank 0
  if (!select_bank(Bank::_0, ec)) {
    return {0, 0, 0};
  }
  uint8_t data[6];
  read_many_from_register(static_cast<uint8_t>(reg), data, 6, ec);
  if (ec) {
    return {0, 0, 0};
  }
  return {
      static_cast<int16_t>((data[0] << 8) | data[1]),
      static_cast<int16_t>((data[2] << 8) | data[3]),
      static_cast<int16_t>((data[4] << 8) | data[5]),
  };
}

Looks like the get_raw() function expects the data to be in big endian, but from the the AK09916 (magnetometer on the icm20948) datasheet, the data is stored in Little Endian. Switching this around finally gives reasonable values for the magnetometer but breaks the accelerometer and gyroscope

[finger563]

Ah good catch! I'll make a PR to fix it soon, thanks for the investigation! If you'd rather do it that's fine too, I should get to it in the next few days.