BerryGPS-IMU v4 "python-BerryIMU-gyro-accel-compass"

[bkfichter]

I am new at this so after reading and going through the tutorials and trying out the sample python code (/python-BerryIMU-gyro-accel-compass/berryIMU.py), I'm not seeing this as a usable compass. The values are off, even with calibration. And the calibration values change each time you run the calibration script anyway. The heading values even change for the same orientation if you continue to rotate the device. I can't even tell you which direction is "forward" on the unit because the values don't match any compass direction. And there is no tilt compensation even though there seems to be tilt compensation correction in the code.

So my question is, has anyone ever gotten this thing to behave like a real functioning compass like you have on a smart phone or is this unit not capable of doing it? Again, I am not an expert in this area.

[mwilliams03]

Have you tried to graph out your calibration results to confirm the calibration is correct?
Have a look at this guide

[bkfichter]

I have. I am now starting from scratch and going through the tutorial from the start using the C code. When using "compass_tutorial01", I am getting values from roughly 90-180. It occurs to me I don't now what UP or FORWARD is on the BerryGPS-IMU v4. I have it plugged into the headers on a Pi 4. The battery and blox chip are on top (see image). I see the X,Y, and Z indicator. Looking at the code, I am thinking Y is FORWARD. The values are so wrong in all directions I'm just not sure.

If Y is FORWARD using your default tutorial code, then the readings I'm getting for the heading are between 90 and 180 repeated 2x per rotation. Not 0-360 as expected.

From the code:

while(1){
		readMAG(magRaw);
		printf("magRaw X %i \tmagRaw Y %i \tMagRaw Z %i \n", magRaw[0],magRaw[1],magRaw[2]);

		//Compute heading
		float heading = 180 * atan2(magRaw[1],magRaw[0])/M_PI;

		//Convert heading to 0 - 360
		if(heading < 0)
			heading += 360;

		printf("heading %7.3f \t ", heading);

		//Sleep for 0.25ms
		usleep(25000);

[mwilliams03]

that is the correct direction.
try working with this code as it include calibration and tilt compensation

[bkfichter]

I think this is the exact same code that I started this thread with. It was producing wrong values even after calibration. That's why I started over with the 1st tutorial, to see if I missed something somewhere. And as you read, the 1st tutorial I was getting values between 90 and 180. This is with the device sitting flat on a table. I didn't want to introduce tilt error. My next step will be doing calibration again. Perhaps my values are so far off because it needs calibration. I'm not exactly sure how calibration will correct the clamped values I'm getting, but we shall see. But at least I have confirmed what forward is supposed to be and can test accordingly.

[bkfichter]

@mwilliams03
I have spent a bit of time on this. I have done the python. I have done the C. I have calibrated to a perfect circle dead center. I have applied those max/mins to the BerryIMU.py (not the BerryIMU-simple.py you linked me to. BerryIMU.py seems to have the same stuff plus the filtering). I have added the declination. I have done it all. The best I have gotten is comparing to my phone compass (which has been compared to a physical compass and is near accurate) is:

Compass	BerryGPS-IMU
0	250
90	358
180	80
270	165

So it is a rough 90 (or ~270) off of the FORWARD as indicated above. This can be adjusted in code and fine for a rough prototype, but not for anything real. I'm guessing at this point this is good as this unit gets using this code and for more accuracy and stability, I need either better calibration process and code and/or a better chip (?).

[mwilliams03]

Can show me the results of the below test;
Use an original version of this;
https://github.com/ozzmaker/BerryIMU/blob/master/python-BerryIMU-gyro-accel-compass/berryIMU-simple.py

Add your calibration values to the script

Right after these lines;

    MAGy = IMU.readMAGy()
    MAGz = IMU.readMAGz()

add;
print ("ACC %i %i %i GYR %i %i %i MAG %i %i %i " % (ACCx,ACCy,ACCz,GYRx,GYRy,GYRz,MAGx,MAGy,MAGz))
Disable all other print statements.

Start the script.
When it starts to output data, rotate the IMU(or Pi) 360 degrees and then stop the script.

please respond with the data and your calibration values.

[bkfichter]

Thanks @mwilliams03

From your instructions, I assumed rotate on the Z axis only. However I provided a file with rotation on all axes also.

magXmax = 933
magYmax = 2570
magZmax = 1567
magXmin = -2348
magYmin = -784
magZmin = -1604

output_AllRotation.txt

output_ZRotation.txt

[mwilliams03]

Your output looks strange.
This is the output from a test i just did. Vanilla code and just added calibration values.
Below is the compass heading graphed out. One turn on the Z axis and it goes between 0 and 360

Here is your output (including calibration)

Can you confirm that the changes you have made to the code are only to add the calibration values and the print statement to show the extra output.
Did you also only do one 360 degree turn in your test?

Can you do another test with the latest code from git, add the calibration values and print statement and then do another 360 rotation on the Z axis.
Can you also show the output of few hundred lines of the IMU not moving

[bkfichter]

@mwilliams03
Did you use "output_ZRotation.txt"?

[mwilliams03]

Yes, i used the file with Z rotation

[bkfichter]

@mwilliams03

I redid everything and made sure the GPS/IMU was perfectly level with one perfect rotation on the Z axis.
Here is still.txt, z_Rotate.txt, and the python file I used (renamed .txt in order to upload here.)
Also include a file with values with the unit facing north.

still.txt
z_Rotate.txt
berryIMU-simple.txt
north.txt

[bkfichter]

@mwilliams03
Any new thoughts? Thanks

[mwilliams03]

Sorry, i didnt noticed you responded.
I pumped your z_rotate.txt output into Excel which has all the formulas to calculate the headings.. and it looks correct to me.
It starts at around 75, goes 360 and circles back to 75

I have uploaded the xls for you to review.(https://github.com/ozzmaker/BerryIMU/files/6335555/tilt.compensation.formula.with.calibration.xlsx)

[bkfichter]

@mwilliams03
I graphed my north.txt file (#21 (comment)). The value for NORTH is around 265. I assume NORTH should be a value of 0. So the same general error as I have been getting, around 90-95 degrees off, as indicated in #21 (comment).

Is it common to have to adjust the value in code for the misaligned values coming off the unit? Could the magnetometer be mounted incorrectly on the pcb? Is there any other reason it is off? It is too late to return this unit to get another one to test.

I pointed the FORWARD as indicated in the picture in #21 (comment) to north to get the values.

[bkfichter]

@mwilliams03
Hey sorry to be a bother, but so close to determining if this is going to be a valid solution. Any thoughts on the previous post?