Sensor Fusion FXS-9AXIS + KL25Z don´t work well in Gyro Stabilized Compass Mode.

cancel
Showing results for 
Search instead for 
Did you mean: 

Sensor Fusion FXS-9AXIS + KL25Z don´t work well in Gyro Stabilized Compass Mode.

682 Views
amadeunasciment
Contributor I

Hi,

 

I trying to use the Fusion Toolbox in Windows 7 (Build date 20 Nov 2014 Rev B) with the board KL25Z + FXS-9AXIS connected to computer via USB but I don´t get the same results as in the demostrations videos (http://www.freescale.com/webapp/video_vault/videoSummary.sp?code=XTRINSIC_SFTANDROID_VID).

 

For example, if a put a magnet close to the board the orientation varies even if the board is using the Gyro Stabilized Compass algorithm.  Another example is if I rotate the board 90 degrees the software show a rotation of 70 or 100 degrees.

 

I have followed the instructions on the Toolbox User Guide release 422 (Section 2.3 - Windows Quick Start Guide), and I have the OpenSda driver P&E Micro version 1.09

 

 

What I can possibily doing wrong?

 

Thanks

Labels (1)
0 Kudos
3 Replies

158 Views
michaelestanley
NXP Employee
NXP Employee

There are several things to consider.  My first question is have you rotated the board sufficient for the magnetic buffer to fill enough for a clean magnetic calibration?  You can tell this by clicking on the "Magnetics" tab in the GUI.

MagTab.png

The figure above shows a full constellation of points used for magnetic calibration, with a fit error of 3.8%.  You don't normally need a full constellation for the math to short things out.  A portion of the "magnetic sphere" should do the job.    I created this one specifically to show you a full constellation.

My next question is: can you confirm that the sensor selection matches the radio button selection as I've outlined below?

DeviceView.png

You WILL see some variation, even on the gyro stabilized setting, when you introduce a magnet into the vicinity of the sensor.  But it should have much less effect than when using the "tilt compensated" eCompass setting.

Another thing to watch for... are you doing this on a table or desk top?  Did you look underneath that surface to see if any ferrous materials are in a drawer, or used to construct the surface?  Or do you have a PC or speakers within a foot or so of your sensor?  If so, the magnetic constellation may shift as you move the board.  Generally the algorithm is much more immune to fast changes in magnetic field.  It cannot differentiate between very slow changes and the normal ambient field.

On wireless systems, I've seen these get out of sync when the battery is low.  Given that you are using a wired 9-axis board, this should not occur.  But it doesn't hurt to check.

If you move your board faster than the maximum rate of the gyro, it may take a short period for the algorithm to reconverge.  This maximum rate is  +/-1600 dps for boards with the FXAS21000, +/-2000 dps for the FXAS21002.  99.9% of the 9-axis boards in existence today use the FXAS21000.

BTW, if you are communicating to the Sensor Fusion Toolbox (and you are if you can get the displays above), then you are OK with regard to the OpenSDA communications.  Problems with that link are usually of the "all or nothing" category.

I hope that helps.

Mike

158 Views
amadeunasciment
Contributor I

My first question is have you rotated the board sufficient for the magnetic buffer to fill enough for a clean magnetic calibration?

I have waited for magnetic calibration until that completes 240 measurements on buffer. But my fit was never less than 5%.

My next question is: can you confirm that the sensor selection matches the radio button selection as I've outlined below?

The selection and the sensors also match.

And I have tried in many different indoor places, but I always can't get the right rotation. I think I'm very electromagnetic polluted environment. =/. And for my application I need to move between rooms. So I think any sensor fusion using magnetometer will be distorted.

Thanks for the help

0 Kudos

158 Views
michaelestanley
NXP Employee
NXP Employee

You're probably right.  We have one location in our office area where an old fashioned magnetic compass indicates north towards our west wall!  No amount of computation can correct for that in the steady state case.

FYI, +/- 5 degrees is roughly the best you'll see without a lot of fine tuning for the eCompass.

0 Kudos