Magnetometer calibration and heading (yaw)

[pmavros]

Hi Kris,
Amazing work porting the MARG into a nice arduino sketch!

I have got it to work with the https://www.sparkfun.com/products/11486, but the yaw values don't range 360 degrees as I rotate the sensors on its axis (rotating around itself, flat on the table mag-z) and there is not good response, so I may turn it 90 degrees North and it only registers 10º change. The other two pitch and roll seem to work fine.

How good is the calibration? Is it worth implementing something like this? http://www.camelsoftware.com/firetail/blog/uavs/3-axis-magnetometer-calibration-a-simple-technique-for-hard-soft-errors/

Many thanks,
Panos

[kriswiner]

Panos,

The sketch has no calibration for the magnetometer except the manual calibration I might have done for my particular MPU9150. I suggest at least you record the min/max value on all three axes and subtract the average so as to re-center the mag response surface. This is all I did. If you suggest something more sophisticated and automatic and get it to work well I would love to incorporate it into my MPU9x50 and LSM9DS0 sketches.

Kris

From: ΠΜ
Sent: Monday, October 13, 2014 12:36 PM
To: kriswiner/MPU-9150
Subject: [MPU-9150] Magnetometer calibration and heading (yaw) (#3)

Hi Kris,
Amazing work porting the MARG into a nice arduino sketch!

I have got it to work with the https://www.sparkfun.com/products/11486, but the yaw values don't range 360 degrees as I rotate the sensors on its axis (rotating around itself, flat on the table mag-z) and there is not good response, so I may turn it 90 degrees North and it only registers 10º change. The other two pitch and roll seem to work fine.

How good is the calibration? Is it worth implementing something like this? http://www.camelsoftware.com/firetail/blog/uavs/3-axis-magnetometer-calibration-a-simple-technique-for-hard-soft-errors/

Many thanks,
Panos

—
Reply to this email directly or view it on GitHub.

[pmavros]

Kris,

will have a thought on that. The link I said has a good solution to remove both hard-iron and soft-iron bias. Also quite quick to implement.

My question holds. the yaw (headings) gives neither a 360º nor a -180º to 180º output but rather a almost random result. I am using the mpu9150 sparkfun SM11486. Also I am confused with which sensor alignment you finally choose as reference the acc or the mag? So which xyz should I use to find north etc?

Finally I had a look on the MPU6050_9Axis_MotionApps41.h and it seems it also has functions to calculate the euler quaternions etc (e.g. MPUgetyawptichroll) - but you are not using those, correct? If that's the case why?

Cheers,
Panos

[kriswiner]

Hi Panos,

I am using the x,y,z of the accel/gyro as reference. If you do not calibrate the magnetometer the Euler values of yaw, pitch, and roll can be quite non-sensical. I wrote my own AHRS routines based on open-source Madgwick and Mahoney sensor fusion algorithms. I don’t like the way Invensense programs are structured nor do I like their dependence on a binary blob for the DMP, which they also depend on.

Kris

From: ΠΜ
Sent: Tuesday, October 14, 2014 5:41 AM
To: kriswiner/MPU-9150
Cc: Kris Winer
Subject: Re: [MPU-9150] Magnetometer calibration and heading (yaw) (#3)

Kris,

will have a thought on that. The link I said has a good solution to remove both hard-iron and soft-iron bias. Also quite quick to implement.

My question holds. the yaw (headings) gives neither a 360º nor a -180º to 180º output but rather a almost random result. I am using the mpu9150 sparkfun SM11486. Also I am confused with which sensor alignment you finally choose as reference the acc or the mag? So which xyz should I use to find north etc?

Finally I had a look on the MPU6050_9Axis_MotionApps41.h and it seems it also has functions to calculate the euler quaternions etc (e.g. MPUgetyawptichroll) - but you are not using those, correct? If that's the case why?

Cheers,
Panos

—
Reply to this email directly or view it on GitHub.

[akhilmohan4487]

Hello Kris,

After implementing Madwick filter, we have the quaternion representing orientation of MPU9150 in global co-ordinate frame. Now if I want to represent the quaternion in the body co-ordinate frame what should I do ?
I have tried to implement (qSB = qGB x q*GS where x is a quaternion multiplication) where s -sensor, b- body and G - global coordinate frame, am I right in doing this ?. After this orientation of MPU in body coordinate frame and expressed in the sensor frame (qSB) is converted to euler angles and I plotted the same. But now if am rotating MPU by 180 I should be able to see a corresponding change in the plot, am I right ? (I have some confusion in using quaternions and am trying to figure it out).

Thanks in advance,
Akhil

[kriswiner]

Hi Akhil,

I am not sure what you want to do. The quaternion represents the sensor's
absolute orientation with respect to an external (yes, global) standard.
This is usually True North. What do you mean by the body coordinate frame in
this case?

Kris

-----Original Message-----
From: Akhil Mohan [mailto:notifications@github.com]
Sent: March 12, 2015 7:13 AM
To: kriswiner/MPU-9150
Cc: Kris Winer
Subject: Re: [MPU-9150] Magnetometer calibration and heading (yaw) (#3)

Hello Kris,

After implementing Madwick filter, we have the quaternion representing
orientation of MPU9150 in global co-ordinate frame. Now if I want to
represent the quaternion in the body co-ordinate frame what should I do ?
I have tried to implement (qSB = qGB x q*GS where x is a quaternion
multiplication) where s -sensor, b- body and G - global coordinate frame, am
I right in doing this ?. After this orientation of MPU in body coordinate
frame and expressed in the sensor frame (qSB) is converted to euler angles
and I plotted the same. But now if am rotating MPU by 180 I should be able
to see a corresponding change in the plot, am I right ? (I have some
confusion in using quaternions and am trying to figure it out).

Thanks in advance,
Akhil

Reply to this email directly or view it on GitHub
#3 (comment) .
<https://github.com/notifications/beacon/AGY1qsn3HDq27TfiuRGSIbtV28fnTfhaks5
n0ZZngaJpZM4CuKSq.gif>

[akhilmohan4487]

Hello,
I have a wireless board ready where multiple MPU's can be plugged in and the data can be logged on to a PC. Now on the PC i wanted to animate the motion that humans do in real-time. Say, I am attaching the MPU on elbow and forearm and wanted to measure the joint angle, I have to represent the information in the body coordinate frame (the location where MPU is mounted). I am bit uncertain how to do this ? What is your opinion in this regard ?

Thanks for the quick reply.
Akhil

[kriswiner]

If you have absolute orientation for each sensor, can you not rather simply calculate the relative angle between, say, the two IMU's x-axes? You have two orientation vectors wrt the same global absloute reference; hwy can you not simply compare these two orientations directly? Why do you need a transform?

[akhilmohan4487]

Yes, we will get a relative orientation of one segment with other in this way. But the problem is i don't know the value of a at time = 0. This can be known only when you keep your body in a fixed pose. For that, initially the segment under measurement is calibrated.

For calibration - (R)GB = (R)SB x (R)GS
(R)GB - orientation info of body wrt global (known for a fixed pose)
(R)SB - orientation info of body wrt sensor
(R)GS - orientation info of sensor wrt global (Madgwick out data converted in rotation matrix form)

We can do all these algebra in quaterion form and is mathematically and computationally useful.

For me, I am nor sure how do we represent (R)GB (Is it a (I)3x3 matrix at rest) ? I don't know.

[kriswiner]

I am not sure what you are asking. If you know each sensors absolute orientation (i.e., quaternion) wrt a fixed global reference like True North, don't you know everything you need to know?

[akhilmohan4487]

Hello,
I would try my best to explain what I wanted to do. After the Madgwick filter update, I have a quaternion which represents the absolute orientation of sensor wrt a global co-ordinate frame(NED). Now I am attaching my Imu's to forearm(segment1) and upper-arm(segment2). Now I want to express segment kinematics in global frame. For that, I have to perform an initial calibration where the orientation of the sensor module wrt segment and the relative distances are measured.

Using rotation matrix I know that (R)GB = (R)SB x (R)GS.
Now how can I do the same using quaternion where I wanted to know qSB given qGB abd qGS ?

I have tried doing qSB = qGB x q*GS (where * corresponds to Quaternion conjugate and x quaternion multiplication). Here qGB = [0.707, 0, 0, -0.707] and which results in singularities and I don't know how to solve this.

Hope this explanation is fine. I am really sorry if I am not able to convey what I want.

Thanks for your patience Kris.

[Ecollot]

Hi Kris,
Do you now have a calibration for the magnetometer? Just to know if I need to figure one out myself or not =)
Great work by the way, thanks so much!
Emilie

[kriswiner]

Here is a mag calibration routine for the MPU9250 which can be adapted for the MPU9150:

void magcalMPU9250(float * dest1)
{
uint16_t ii = 0, sample_count = 0;
int32_t mag_bias[3] = {0, 0, 0};
int16_t mag_max[3] = {0, 0, 0}, mag_min[3] = {0, 0, 0}, mag_temp[3] = {0, 0, 0};
Serial.println("Mag Calibration: Wave device in a figure eight until done!");
delay(4000);
sample_count = 64;
for(ii = 0; ii < sample_count; ii++) {
readMagData(mag_temp); // Read the mag data
for (int jj = 0; jj < 3; jj++) {
if(mag_temp[jj] > mag_max[jj]) mag_max[jj] = mag_temp[jj];
if(mag_temp[jj] < mag_min[jj]) mag_min[jj] = mag_temp[jj];
}
delay(135); // at 8 Hz ODR, new mag data is available every 125 ms
}
// Serial.println("mag x min/max:"); Serial.println(mag_max[0]); Serial.println(mag_min[0]);
// Serial.println("mag y min/max:"); Serial.println(mag_max[1]); Serial.println(mag_min[1]);
// Serial.println("mag z min/max:"); Serial.println(mag_max[2]); Serial.println(mag_min[2]);
mag_bias[0] = (mag_max[0] + mag_min[0])/2; // get average x mag bias in counts
mag_bias[1] = (mag_max[1] + mag_min[1])/2; // get average y mag bias in counts
mag_bias[2] = (mag_max[2] + mag_min[2])/2; // get average z mag bias in counts
dest1[0] = (float) mag_bias[0]_mRes_magCalibration[0]; // save mag biases in G for main program
dest1[1] = (float) mag_bias[1]_mRes_magCalibration[1];
dest1[2] = (float) mag_bias[2]_mRes_magCalibration[2];
Serial.println("Mag Calibration done!");
}

[Ecollot]

Thanks so much for answering so fast !

Still working with the code, probably got another few questions popping up,
but the main one is : you calculate the biases for the accelerometer and
gyroscope, but you never use these when considering the raw values in any
way it seems. If you do, where do you use these corrections ?
And for now, the magnetometer biases are just put in by hand right ?

On 10 April 2015 at 11:13, Kris Winer notifications@github.com wrote:

Here is a mag calibration routine for the MPU9250 which can be adapted for
the MPU9150:

void magcalMPU9250(float * dest1)
{
uint16_t ii = 0, sample_count = 0;
int32_t mag_bias[3] = {0, 0, 0};
int16_t mag_max[3] = {0, 0, 0}, mag_min[3] = {0, 0, 0}, mag_temp[3] = {0,
0, 0};
Serial.println("Mag Calibration: Wave device in a figure eight until
done!");
delay(4000);
sample_count = 64;
for(ii = 0; ii < sample_count; ii++) {
readMagData(mag_temp); // Read the mag data
for (int jj = 0; jj < 3; jj++) {
if(mag_temp[jj] > mag_max[jj]) mag_max[jj] = mag_temp[jj];
if(mag_temp[jj] < mag_min[jj]) mag_min[jj] = mag_temp[jj];
}
delay(135); // at 8 Hz ODR, new mag data is available every 125 ms
}
// Serial.println("mag x min/max:"); Serial.println(mag_max[0]);
Serial.println(mag_min[0]);
// Serial.println("mag y min/max:"); Serial.println(mag_max[1]);
Serial.println(mag_min[1]);
// Serial.println("mag z min/max:"); Serial.println(mag_max[2]);
Serial.println(mag_min[2]);
mag_bias[0] = (mag_max[0] + mag_min[0])/2; // get average x mag bias in
counts
mag_bias[1] = (mag_max[1] + mag_min[1])/2; // get average y mag bias in
counts
mag_bias[2] = (mag_max[2] + mag_min[2])/2; // get average z mag bias in
counts
dest1[0] = (float) mag_bias[0]_mRes_magCalibration[0]; // save mag biases
in G for main program
dest1[1] = (float) mag_bias[1]_mRes_magCalibration[1];
dest1[2] = (float) mag_bias[2]_mRes_magCalibration[2];
Serial.println("Mag Calibration done!");
}

—
Reply to this email directly or view it on GitHub
#3 (comment).

[kriswiner]

I write the accel and gyro biases to user bias registers. These bias values
are automatically subtracted from the raw sensor data. You can also choose
to not write any values there and manually subtract the accel and gyro
biases like the mag biases are done. I sometimes do this because there is a
temperature bit in the accel bias register that has to be preserved and I
find sometimes the biases applied automatically by using thebias register
value are not quite right for the accelerometer, so I often still do this
one by hand.

Kris

-----Original Message-----
From: Ecollot [mailto:notifications@github.com]
Sent: April 9, 2015 7:19 PM
To: kriswiner/MPU-9150
Cc: Kris Winer
Subject: Re: [MPU-9150] Magnetometer calibration and heading (yaw) (#3)

Thanks so much for answering so fast !

Still working with the code, probably got another few questions popping up,
but the main one is : you calculate the biases for the accelerometer and
gyroscope, but you never use these when considering the raw values in any
way it seems. If you do, where do you use these corrections ?
And for now, the magnetometer biases are just put in by hand right ?

On 10 April 2015 at 11:13, Kris Winer notifications@github.com wrote:

Here is a mag calibration routine for the MPU9250 which can be adapted for
the MPU9150:

void magcalMPU9250(float * dest1)
{
uint16_t ii = 0, sample_count = 0;
int32_t mag_bias[3] = {0, 0, 0};
int16_t mag_max[3] = {0, 0, 0}, mag_min[3] = {0, 0, 0}, mag_temp[3] = {0,
0, 0};
Serial.println("Mag Calibration: Wave device in a figure eight until
done!");
delay(4000);
sample_count = 64;
for(ii = 0; ii < sample_count; ii++) {
readMagData(mag_temp); // Read the mag data
for (int jj = 0; jj < 3; jj++) {
if(mag_temp[jj] > mag_max[jj]) mag_max[jj] = mag_temp[jj];
if(mag_temp[jj] < mag_min[jj]) mag_min[jj] = mag_temp[jj];
}
delay(135); // at 8 Hz ODR, new mag data is available every 125 ms
}
// Serial.println("mag x min/max:"); Serial.println(mag_max[0]);
Serial.println(mag_min[0]);
// Serial.println("mag y min/max:"); Serial.println(mag_max[1]);
Serial.println(mag_min[1]);
// Serial.println("mag z min/max:"); Serial.println(mag_max[2]);
Serial.println(mag_min[2]);
mag_bias[0] = (mag_max[0] + mag_min[0])/2; // get average x mag bias in
counts
mag_bias[1] = (mag_max[1] + mag_min[1])/2; // get average y mag bias in
counts
mag_bias[2] = (mag_max[2] + mag_min[2])/2; // get average z mag bias in
counts
dest1[0] = (float) mag_bias[0]_mRes_magCalibration[0]; // save mag biases
in G for main program
dest1[1] = (float) mag_bias[1]_mRes_magCalibration[1];
dest1[2] = (float) mag_bias[2]_mRes_magCalibration[2];
Serial.println("Mag Calibration done!");
}

Reply to this email directly or view it on GitHub
#3 (comment).

Reply to this email directly or view it on GitHub
#3 (comment) .
<https://github.com/notifications/beacon/AGY1qj7yBPwEtmTuSvDNUBHBT4Sqsnoxks5
n9yqsgaJpZM4CuKSq.gif>

[Ecollot]

Ok, got that. One more thing, how do define the user-environmental x,y,z
axis corrections for the magnetometer? Where do these values come from?

On 10 April 2015 at 14:24, Kris Winer notifications@github.com wrote:

I write the accel and gyro biases to user bias registers. These bias values
are automatically subtracted from the raw sensor data. You can also choose
to not write any values there and manually subtract the accel and gyro
biases like the mag biases are done. I sometimes do this because there is a
temperature bit in the accel bias register that has to be preserved and I
find sometimes the biases applied automatically by using thebias register
value are not quite right for the accelerometer, so I often still do this
one by hand.

Kris

-----Original Message-----
From: Ecollot [mailto:notifications@github.com]
Sent: April 9, 2015 7:19 PM
To: kriswiner/MPU-9150
Cc: Kris Winer
Subject: Re: [MPU-9150] Magnetometer calibration and heading (yaw) (#3)

Thanks so much for answering so fast !

Still working with the code, probably got another few questions popping up,
but the main one is : you calculate the biases for the accelerometer and
gyroscope, but you never use these when considering the raw values in any
way it seems. If you do, where do you use these corrections ?
And for now, the magnetometer biases are just put in by hand right ?

On 10 April 2015 at 11:13, Kris Winer notifications@github.com wrote:

Here is a mag calibration routine for the MPU9250 which can be adapted
for
the MPU9150:

void magcalMPU9250(float * dest1)
{
uint16_t ii = 0, sample_count = 0;
int32_t mag_bias[3] = {0, 0, 0};
int16_t mag_max[3] = {0, 0, 0}, mag_min[3] = {0, 0, 0}, mag_temp[3] = {0,
0, 0};
Serial.println("Mag Calibration: Wave device in a figure eight until
done!");
delay(4000);
sample_count = 64;
for(ii = 0; ii < sample_count; ii++) {
readMagData(mag_temp); // Read the mag data
for (int jj = 0; jj < 3; jj++) {
if(mag_temp[jj] > mag_max[jj]) mag_max[jj] = mag_temp[jj];
if(mag_temp[jj] < mag_min[jj]) mag_min[jj] = mag_temp[jj];
}
delay(135); // at 8 Hz ODR, new mag data is available every 125 ms
}
// Serial.println("mag x min/max:"); Serial.println(mag_max[0]);
Serial.println(mag_min[0]);
// Serial.println("mag y min/max:"); Serial.println(mag_max[1]);
Serial.println(mag_min[1]);
// Serial.println("mag z min/max:"); Serial.println(mag_max[2]);
Serial.println(mag_min[2]);
mag_bias[0] = (mag_max[0] + mag_min[0])/2; // get average x mag bias in
counts
mag_bias[1] = (mag_max[1] + mag_min[1])/2; // get average y mag bias in
counts
mag_bias[2] = (mag_max[2] + mag_min[2])/2; // get average z mag bias in
counts
dest1[0] = (float) mag_bias[0]_mRes_magCalibration[0]; // save mag biases
in G for main program
dest1[1] = (float) mag_bias[1]_mRes_magCalibration[1];
dest1[2] = (float) mag_bias[2]_mRes_magCalibration[2];
Serial.println("Mag Calibration done!");
}

Reply to this email directly or view it on GitHub
#3 (comment).

Reply to this email directly or view it on GitHub
#3 (comment) .
<
https://github.com/notifications/beacon/AGY1qj7yBPwEtmTuSvDNUBHBT4Sqsnoxks5
n9yqsgaJpZM4CuKSq.gif>

—
Reply to this email directly or view it on GitHub
#3 (comment).