Bug in the calcQuaternion algorithm

[jtragtenberg]

Hi there,
I've been all this time learning C++ and figuring out how the Motioner and RazorIMU libraries work.
I make my own prototyping board with just an arduino UNO soldered with a HMC8553L magnetometer and MPU6050 gyro+accel. I'm doing all my tests connecting the board through USB.
I made a code out of the Motioner.ino where in the loop() I just added the lines:

float qx = motioner.mRazorIMU.getQuatX();
float qy = motioner.mRazorIMU.getQuatY();
float qz = motioner.mRazorIMU.getQuatZ();
float qw = motioner.mRazorIMU.getQuatW();

Serial.print(qw);
Serial.print('|');
Serial.print(qx);
Serial.print('|');
Serial.print(qy);
Serial.print('|');
Serial.println(qz);

where the getQuat* functions I added are just:

float getQuatX() const { return quat.x; }
float getQuatY() const { return quat.y; }
float getQuatZ() const { return quat.z; }
float getQuatW() const { return quat.w; }

and I made the mRazorIMU object public on the Motioner class.

Out of this I noticed a bug in the angles as I tilt them, where the quaternions change sign for no reason...
I have calibrated my board with the IMUDebug.ino firmware and it all worked well in the IMUvisualization Processing sketch.

here is what the serial monitor shows:
https://youtu.be/AYBm-mMuhkI

Does anyone know why this bug keeps bugging me?
Is this discontinuity and sign inversion expected?

thank you

[jtragtenberg]

Anyone?...

[jtragtenberg]

Has anyone an example of a Motioner IMU working just outputting the quaternion values ?
It really seems to me that the bug is in the calcQuaternion algorithm, the Gamasutra code that transforms the Matrix into quats....

[skistua]

you are right . It's wrong. Give me your email address . then I send your right one . it's project is long gone. Here is no one answer your question. @jtragtenberg

[jtragtenberg]

Hi there,

its joao@batebit.cc

Thank you