Michael Stanley
Mike.Stanley@freescale.com
Freescale Sensor Solutions Division
TM

Michael Stanley
Employed at Motorola / Freescale
Semiconductor from June 1980 to the present, where I’ve had multiple careers.
Most recently:
–
SoC Integration / MCU Architecture
–
Sensors Architecture / Algorithms /
Product Definition
– basically, solving systems level problems
I blog on sensor related topics at http://www.freescale.com/blogs/mikestanley and http://memsblog.wordpress.com/
• Mike.Stanley@freescale.com

Introduction
Problem Definition
A Solution
Structural Details
Development Environment
Where Do I Get It?

Sensor fusion encompasses a variety of techniques which:
 Trade off strengths and weaknesses of the various sensors to compute something more than can be calculated using the individual sensors;
 Improve the quality and noise level of computed results by taking advantage of:
 Known data redundancies between sensors
 Knowledge of system transfer functions, dynamic behavior and/or expected motion

Sensor fusion means many things to many people. Our focus here is on modeling orientation of an object (in real time) relative to a global earth frame.
axis/angle
Euler Angle Illustration source: http://en.wikipedia.org/wiki/File:Euler2a.gif

Configure, Power State, Data Control
Pressure Trim
3-Axis Acc
3-Axis
Gyro
FoR mapping
FoR mapping
Trim
Trim
3-Axis
Mag
FoR mapping
Trim & Hard/Soft compensation
Raw data calibration parameters
FoR = Frame of Reference Mapping
Calculate hard/soft iron parameters hi/low/band pass filtering hi/low/band pass filtering hi/low/band pass filtering hi/low/band pass filtering
Sensor Hub Functions
Shake detection
Pressure shake event
Acc x,y,z
ω x,y,z
B x,y,z
Kalman
Filter or similar function
Geometric computations
Rotation matrix
Quaternion
Tilt-compensated mag heading
Orientation ( φ, Θ, Ψ)
Sensor Fusion

Developing from scratch is expensive and time consuming
The underlying mathematics are extremely complex
In the past, you could buy these functions (usually for six figures) from companies like Sensor
Platforms, Movea and Xsens.
A recent spree of purchases resulted in all of these being acquired by larger companies.

a complete solution, including: low-cost reference hardware free source code for a variety of sensor fusion options the ability to choose your own MCU and sensor supplier full datasheet with performance metrics user’s manual telling you how to customize the code free on-demand training materials free visualization tools to make sure you got it right free development tools free community support professional engineering consultation options


Optimized for the computation of orientation with respect to a global frame of reference as a function of sensor readings from:
– accelerometer
– and/or gyroscope
– and/or magnetometer
 Along with orientation, also computes:
– linear acceleration
– magnetic interference and correction factors for same
– magnetic inclination angle
– gyroscope zero-rate offset
– compass heading
– virtual gyro from accelerometer / magnetometer

Supplied under open source BSD license from Freescale
Implemented as pure C-code sitting on top of device driver and MQX

Lite implementations created via Freescale’s
Processor Expert
 tool
Shipped in the form of CodeWarrior
 and Kinetis Design
Studio projects compatible with the Freescale Sensor Fusion
Toolbox
Downloadable from http://www.freescale.com/sensorfusion
Community support available at https://community.freescale.com/community/sensors/sensorfusion
Contract support services offered by Freescale. Contact: sfsw@freescale.com
for details.

Now bundled into the sensor fusion library
–
4 and 7 and now 10 element solvers are available in source form
– the 10 element solver was previously available only under NDA
Freescale’s eCompass software received the
Electronic Products Magazine 2012 Product of the Year Award.

FREESCALE END-USER SOFTWARE LICENSE AGREEMENT
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
 Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
 Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
 Neither the name of Freescale Semiconductor, Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL FREESCALE SEMICONDUCTOR, INC.
BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Feature
Filter Type
Roll / Pitch / Tilt in degrees
Yaw in degrees
Angular Rate 1 in degrees/second
Compass heading (magnetic north) in degrees
Quaternion and rotation vector
Rotation matrix
Linear acceleration separate from gravity
NED (North-East-Down Frame of Reference
ENU (Windows 8 variant) Frame of Reference
ENU (Android
 variant) Frame of Reference
Magnetic calibration included
Gyro offset calibration included
FRDM-KL25Z board support
FRDM-KL26Z board support
FRDM-KL46Z board support
FRDM-K20D50M board support
FRDM-K64F board support
Accel only Accel + gyro Accel + mag Accel + mag + gyro
Low Pass
Yes
No virtual 2 axis 2
No
Yes
Yes
No
Yes 3
Yes 3
Yes 3
No
N/A
Yes
Yes
Yes
Yes
Yes
Indirect
Kalman
Yes
No
Yes
No
Yes
Yes
Yes
Yes 3
Yes 3
Yes 3
No
Yes
Yes
Yes
Yes
Yes
Yes
Low Pass
Yes
Yes virtual 3 axis
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
N/A
Yes
Yes
Yes
Yes
Yes
Indirect Kalman
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
1. Angular rate for configurations with a gyro include corrections for gyro offset
2. Subject to well-known limitation of being blind to rotation about axes aligned with gravity
3. These solutions do not include a magnetometer, therefore there is no sense of compass heading

License
Feature
CPU selection
Development Version
Open Source BSD
The ANSI C99 source code was optimized on Freescale Kinetis MCUs based upon
ARM

Cortex

M0+, M4 and M4F processors, but should be portable to any CPU.
Board customizable Yes 1
Sensor sample rate
Fusion rate
Frame of Reference
Algorithms Executing
Sleep mode enabled between samples/calculations
RTOS
Programmable
Programmable, typically = sample rate/N
Programmable (NED, Android

, or Windows 8)
Any combination of those shown in the prior slide
Programmable
Code flexibility
Access to Processor Expert
Configuration
MQX Lite RTOS
All code is supplied in source form
Yes
Product Deliverables
* Datasheet, User guide, Application Notes
* Template CodeWarrior projects
* Pre-compiled s-record / binary files
[1] FRDM_KL25Z, KL26Z, KL46Z, K20D50M and K64F are supported “out of the box” and may be used as templates for other board/MCU combinations..

Access to raw fusion and magnetic calibration functions
Control over sampling and fusion rates
Ability to add custom Hardware Abstraction Layer
(HAL)
Access to MQX Lite RTOS customization via
Processor Expert Tool

MQXLITE include kernel psp
FLASH
<ProjName>.elf
(generated>
Library prepopulated with .s19 and
.bin files
<ProjName>
As shipped:
• FSFK_KL25Z; or
• FSFK_KL26Z; or
• FSFK_KL46Z; or
• FSFK_K20D50M; or
• FSFK_K64F
Sources approximations.c
approximations.h
build.h
drivers.h
Events.h
fusion.h
include_all.h
magnetic.h
drivers.c
Events.c
fusion.c
matrix.h
mqx_tasks.h
orientation.h
tasks.h
user_tasks.h
magnetic.c
main.c
matrix.c
mqx_tasks.c
orientation.h
tasks.c
user_tasks.c
ProcessorExpert.pe
.project
Files in bold red are most likely to be customized on a per project basis.

MQX Lite RTOS mqx_tasks magnetic fusion orientation, matrix, approximations tasks
I 2 drivers & Events
C Interface
UART
Interface
Hardware
GPIO
Interface generated by
Processor Expert Tool primary fusion & calibration functions
I 2 user tasks
C and UART communications to external devices are encapsulated by drivers.c and Events.c
Existing hardware platforms are discussed in later slides.
Main_task()
RdSensData_task()
Fusion_task()
MagCal_task()
RdSensData_Init()
Fusion_Init()
RdSensData_Run()
Fusion_Run()
MagCal_Run()
ApplyAccelHAL()
ApplyMagHAL()
ApplyGyroHAL()

FXOS8700
(Internal clock)
I 2 C
200 Hz MQX Lite Hardware Timer
FXAS21000
(Internal clock)
Sampling Task
25 Hz Software Event
Specific to hardware and sensors
Fusion Task
~1 per minute Software Event
Magnetic Calibration
Task sampling interval = 5 ms
Independent of hardware and sensors

Pre-built templates are targeted at specific
Freescale Freedom boards
User code easily added to a single .c file within any of the following functions:
– void UserStartup(void);
– void UserHighFrequencyTaskInit(void) ; // runs once, the first time through the 200Hz task
– void UserHighFrequencyTaskRun(void); // runs each time the 200Hz task runs
– void UserMediumFrequencyTaskInit(void); // runs once, the first time through the 25Hz task
– void UserMediumFrequencyTaskRun(void); // runs each time the 25Hz task runs
Sensor and fusion values are simply read from predefined global structures

#include "Cpu.h"
#include "Events.h"
#include "mqx_tasks.h"
#include "UART.h"
#include "include_all.h" void UserStartup(void) {
// The following UART function call initializes Bluetooth communications used by the
// Freescale Sensor Fusion Toolbox. If the developer is not using the toolbox,
// these can be removed.
//
// initialize BlueRadios Bluetooth module
BlueRadios_Init(UART2_DeviceData);
// put code here to be executed at the end of the RTOS startup sequence.
//
// PUT YOUR CODE HERE
// return;

void UserHighFrequencyTaskInit(void) {
// User code to be executed ONE TIME the first time the high frequency task is run.
//
// PUT YOUR CODE HERE
// return;
} void UserMediumFrequencyTaskInit(void) {
// User code to be executed ONE TIME the first time the medium frequency task is run
//
// PUT YOUR CODE HERE
// return;
} void UserHighFrequencyTaskRun(void) {
// The default frequency at which this code runs is 200Hz.
// This code runs after sensors are sampled.
// In general, try to keep "high intensity" code out of UserHighFrequencyTaskRun.
// The high frequency task also has highest priority.
//
// PUT YOUR CODE HERE
// return;
}

void UserMediumFrequencyTaskRun(void) {
// This code runs after the Kalman filter loop
// The default frequency at which this code runs is 25Hz.
// The following UART function constructs and sends Bluetooth packets used by the
// Freescale Sensor Fusion Toolbox. If the developer is not using the toolbox,
// it can be removed.
// transmit orientation over the radio link
CreateAndSendBluetoothPacketsViaUART(UART2_DeviceData);
//
// PUT YOUR CODE HERE
// return;
}
Steps to use:
1. Import project into CodeWarrior IDE or Kinetis Design Studio
2. Run Processor Expert Tool (you push a single button)
3. Add your code as shown above
4. Build
5. Download and run

Access Fusion Inputs & Outputs Via a Standard
Set of Global Data Structures
Input Global Data Structures defined in build.h
Pointer Function Structure Name Structure Type
Accelerometer
Magnetometer
Gyroscope thisAccel thisMag thisGyro
AccelSensor
MagSensor
GyroSensor
Output Global Data Structures defined in tasks.h
Pointer Function Structure Name Structure Type
Altimeter results
3-axis Accelerometer results
2D Magnetic-only eCompass results
Gyro-only orientation eCompass results accel+gyro results
9-axis results thisSV_1DOF_P_BASIC thisSV_3DOF_G_BASIC thisSV_3DOF_B_BASIC thisSV_3DOF_Y_BASIC thisSV_6DOF_GB_BASIC thisSV_6DOF_GY_KALMAN
SV_1DOF_P_BASIC
SV_3DOF_G_BASIC
SV_3DOF_B_BASIC
SV_3DOF_Y_BASIC
SV_6DOF_GB_BASIC
SV_6DOF_GY_KALMAN thisSV_9DOF_GBY_KALMAN SV_9DOF_GBY_KALMAN

Description Data Type fLPPhi fLPThe fLPPsi fLPRho fLPChi
N/A
G
(accel) roll in degrees pitch in degrees yaw in degrees compass heading in degrees tilt angle in degrees magnetic inclination angle in degrees geomagnetic vector
(microTeslas, global frame) gyro offset in degrees/sec linear acceleration in the sensor frame in gravities linear acceleration in the global frame in gravities quaternion (unitless) float float float float float float float float float float
N/A
N/A
N/A
N/A angular velocity in dps orientation matrix (unitless) rotation vector time interval in seconds fquaternion fq float float fLPq fOmega[3] 1 fR[3][3] fLPR[3][3] float float
Footnotes:
1.
Yaw is not supported for 3-axis accelerometer
2.
Physical gyro angular rate corrected to subtract computed offset fLPRVec[3] fdeltat
Fusion Algorithm Options
GB
(eCompass) fLPPhi fLPThe fLPPsi fLPRho fLPChi fDelta fLPDelta
N/A
GY
(accel + gyro) fPhiPl fThePl fPsiPl fRhoPl fChiPl
N/A
N/A
GBY
9-axis fPhiPl fThePl fPsiPl fRhoPl fChiPl fDeltaPl fmGl[3]
N/A
N/A
N/A fq fLPq fOmega[3] fR[3][3] fLPR[3][3] fLPRvec[3] fdeltat fbPl[3] faSePl[3]
N/A fqPl fOmega[3] 2 fRPl[3][3] fRVecPl[3] fdeltat fbPl[3] faSePl[3] faGlPl[3] fqPl fOmega[3] 2 fRPl[3][3] fRVecPl[3] fdeltat

struct fquaternion fq; float q0, q1, q2, q3;
// quaternion
//fq = thisSV_3DOF_G_BASIC.fLPq; // OR
//fq = thisSV_6DOF_GB_BASIC.fLPq; // OR
//fq = thisSV_6DOF_GY_KALMAN.fqPl; // OR fq = thisSV_9DOF_GBY_KALMAN.fqPl; q0 = fq.q0; q1 = fq.q1; q2 = fq.q2; q3 = fq.q3;
// more details/examples are presented in the following section

Using 3-axis model: float roll = thisSV_3DOF_G_BASIC.fLPPhi; float pitch = thisSV_3DOF_G_BASIC.fLPThe; float yaw = thisSV_3DOF_G_BASIC.fLPPsi;
Using 6-axis accel + mag (eCompass) model: float roll = thisSV_6DOF_GB_BASIC.fLPPhi; float pitch = thisSV_6DOF_GB_BASIC.fLPThe; float yaw = thisSV_6DOF_GB_BASIC.fLPPsi;
Using 6-axis accel + gyro Kalman filter model: float roll = thisSV_6DOF_GY_KALMAN.fPhiPl; float pitch = thisSV_6DOF_GY_KALMAN.fThePl; float yaw = thisSV_6DOF_GY_KALMAN.fPsiPl;
Using 9-axis Kalman filter model: float roll = thisSV_9DOF_GBY_KALMAN.fPhiPl; float pitch = thisSV_9DOF_GBY_KALMAN.fThePl; float yaw = thisSV_9DOF_GBY_KALMAN.fPsiPl;

#ifndef BUILD_H
#define BUILD_H
// PCB HAL options
#define BOARD_WIN8_REV05 0
#define BOARD_FRDM_KL25Z 1
#define BOARD_FRDM_K20D50M 2
#define BOARD_FXLC95000CL 3
#define BOARD_FRDM_KL26Z 4
#define BOARD_FRDM_K64F 5
#define BOARD_FRDM_KL16Z 6
#define BOARD_FRDM_KL46Z 7
#define BOARD_FRDM_KL46Z_STANDALONE
// with sensor shield
// with sensor shield
// with sensor shield
// with sensor shield
// with sensor shield
// with sensor shield
// with sensor shield
// without sensor shield 8
// enter new PCBs here with incrementing values
// C Compiler Preprocessor define in the CodeWarrior project will choose which board to use
#ifdef REV05
#define THIS_BOARD_ID
#endif
#ifdef KL25Z
BOARD_WIN8_REV05
#define THIS_BOARD_ID BOARD_FRDM_KL25Z
#endif

#ifdef K20D50M
#define THIS_BOARD_ID
#endif
#ifdef FXLC95000CL
#define THIS_BOARD_ID
#endif
#ifdef KL26Z
#define THIS_BOARD_ID
#endif
#ifdef K64F
#define THIS_BOARD_ID
#endif
#ifdef KL16Z
#define THIS_BOARD_ID
#endif
#ifdef KL46Z
#define THIS_BOARD_ID
#endif
#ifdef KL46Z_STANDALONE
#define THIS_BOARD_ID
#endif
// coordinate system for the build
#define NED 0
#define ANDROID 1
#define WIN8 2
#define THISCOORDSYSTEM ANDROID
BOARD_FRDM_K20D50M
BOARD_FRDM_FXLC95000CL
BOARD_FRDM_KL26Z
BOARD_FRDM_K64F
BOARD_FRDM_KL16Z
BOARD_FRDM_KL46Z
BOARD_FRDM_KL46Z_STANDALONE
// identifier for NED angle output
// identifier for Android angle output
// identifier for Windows 8 angle output
// the coordinate system to be used

// sensors to be enabled
. . .
#define USE_MPL3115
#define USE_FXOS8700
#define USE_FXAS21000
//#define USE_FXAS21002
//#define USE_MMA8652
//#define USE_MAG3110
. . .
// normally all enabled: degrees of freedom algorithms to be executed
#define COMPUTE_1DOF_P_BASIC // 1DOF pressure (altitude) and temperature: (1x pressure)
#define COMPUTE_3DOF_G_BASIC // 3DOF accel tilt: (1x accel)
#define COMPUTE_3DOF_B_BASIC // 3DOF mag eCompass (vehicle): (1x mag)
#define COMPUTE_3DOF_Y_BASIC // 3DOF gyro integration: (1x gyro)
#define COMPUTE_6DOF_GB_BASIC // 6DOF accel and mag eCompass: (1x accel + 1x mag)
#define COMPUTE_6DOF_GY_KALMAN // 6DOF accel and gyro (Kalman): (1x accel + 1x gyro)
#define COMPUTE_9DOF_GBY_KALMAN // 9DOF accel, mag and gyro (Kalman):(1x accel+1x mag+1x gyro)

// int16 build number sent in Bluetooth debug packet
#define THISBUILD 420
// sampling rate and kalman filter timing
#define FTM_INCLK_HZ 1000000 // int32: 1MHz FTM timer frequency set in PE: do not change
#define SENSORFS 200 // int32: 200Hz: frequency (Hz) of sensor sampling process
#define OVERSAMPLE_RATIO 8
// int32: 8x: 3DOF, 6DOF, 9DOF run at SENSORFS / OVERSAMPLE_RATIO Hz
// power saving deep sleep
//#define DEEPSLEEP // define to enable deep sleep power saving
// UART (Bluetooth) serial port control
//#define UART_OFF // define to measure MCU+algorithm current only

Make any desired changes to the template boardspecific template run
Processor
Expert Tool updated project with MQX
Lite RTOS
Build using
Code
Warrior or KDS
For K64, there is one intermediate (and temporary) step here. Manually edit
CPU_Config.h, change the value for
NV_FSEC to 0xFE. This works around a bug in the MBED bootloader firmware.
Test via Freescale
Sensor Fusion Toolbox for Windows or
Android


You must have either Kinetis Design Studio 1.1.1 or CodeWarrior IDE 10.6 and
Processor Expert tool to build sensor fusion applications using the Freescale project templates.
–
KDS is FREE, with no code size restrictions.
In order to experiment with the demo program, you will need an Android 3.0 or higher device running the Freescale Sensor Fusion Toolbox OR the PC-based variant of the toolbox.
Development board(s) 1 . We (of course) recommend Freescale components:
–
Kinetis ARM

Cortex

-M0+, M4 or M4F MCU
–
Freescale FXOS8700CQ 3-axis magnetometer + 3 axis accelerometer
–
Freescale FXAS21000 3-axis gyroscope
1 See details on the next slide. Additional sensor combinations are supported in build.h. And of course, you can add your own!
Future expansion boards may replace the FXAS21000 with the FXAS21002, which is also supported.

Use any one of the FRDM-FXS boards with any one of the Freedom development boards shown.
FRDM-FXS-
Part Number Description
FRDM-FXS-MULTI* Freescale Sensor
Expansion board
MPL3115A2
KL25Z or
KL26Z or
KL46Z or
K20D50M or
K64F
MMA8652
FXAS21000
FXOS8700
FXLS8471
MMA955x
MAG3110
FRDM-FXS-MULTI-B* Freescale Sensor
Expansion board with
Bluetooth and Battery
MPL3115A2
MMA8652
FXAS21000
FXOS8700
FXLS8471
MMA955x
MAG3110
FRDM-FXS-9AXIS* Freescale Sensor
Expansion board with only
2 sensors
FXAS21000
FXOS8700
Pricing
(USD)
$50
Availability
Now
$125
$30
Now
Now

The Freescale sensor boards on the prior page are
Arduino

R3 pinout compatible
They are also compatible with the following
Freescale Freedom boards :
–
The FRDM-KL25Z, KL26Z and KL46Z utilize ARM

Cortex M0+
–
The FRDM-K20D50M utilizes ARM Cortex M4
–
The FRDM-K64F utilizes ARM M4F (F = floating point unit)
Prebuilt programs for each of the Freedom boards above are available in CodeWarrior and Kinetis
Design Studio formats.

Function
Cpu
LED_RED
LED_GREEN
LED_BLUE
FTM
FSFK_KL25Z FSFK_KL26Z
MKL25Z128VLK4 MKL26Z128VLH4
FSFK_KL46Z
MKL46Z256VMC4
FSFK_
K20D50M
MK20DX128VLH5
FSFK_K64F
MK64FN1M0VLL12
Description
PTB18 PTE29 PTE29 PTC3 PTB22
PTB19
PTD1
LPTMR0
PTE31
PTD5
LPTMR0
PTD5
PTE31
LPTMR0
PTD4
PTA2
LPTMR0
PTE26
PTB21
LPTMR0
UART
I2C
TestPin_KF_Time
UART0 on
PTA2:1
I2C1on PTC2:1
UART0 on
PTA2:1
I2C1 on PTC2:1
UART0 on
PTA2:1
I2C1 on PTC2:1
UART1 on
PTE1:0
I2C0 on PTB1:0
PTC10 PTC10 PTC10 PTC10
UART3 on
PTC17:16
I2C1 on
PTC11:10
PTC7
TestPin_MagCal_
Time
PTC11 PTC11 PTC11 PTC1 PTC5
Illuminated when a magnetic calibration is in progress
Flickers when fusion algorithms are running
Currently unused
Low frequency timer drives the
200 Hz sensor read process
Used for Bluetooth communications
Communicates to sensors
Output lines used for debug purposes

Accel Gyro Mag
=
Kinetis MCU
Bluetooth Module Embedded board running the Freescale Sensor
Fusion Library for Kinetis
Sensor output data is “fused” using
Freescale-developed code running on
Kinetis, and then “beamed” to a PC or
Android device, where it drives the GUI
Phone/tablet running the Freescale
Sensor Fusion Toolbox for Android

Provides visualization functions for the fusion library
Allows you to experiment with different sensor/algorithm choices
Gives you access to raw sensor data
Allows you to log sensor and fusion data for later use
Works with demo and development versions of the Freescale
Sensor Fusion Library
Platforms
• Android
• Windows PC

The Freescale Sensor Fusion Toolbox Features by Platform
Feature
Bluetooth wireless link
Ethernet wireless link
UART over USB
OS requirements
Support for native sensors
Device View
Panorama View
Statistics View
Canvas View
Orientation XY Plots
Inertial XY Plots
Magnetics
Kalman
Altimeter XY Plots
Data Logging Capability
Integrated documentation
Availability
Price
Android

On WiGo board only
-
>= Android 3.0





-
-
-
-
-


Google Play
Free
PC
Requires BT on PC (built-in or dongle)
-
 1
>= Windows 7.0
-

-
-






-

Freescale website
Free
1 FRDM_K64F and FRDM_K20D50M projects require a Processor Expert configuration change to run in wired mode.

2
3
1
6
4
7
Figures are from 28 August 2014 build of the application.
Appearance may vary for other versions.
5
5.
6.
7.
1.
2.
3.
4.
Rotating 3D PCB display
Image align function
Navigation Tabs for:
• Sensors Data Tab
•
Dynamics Tab
•
Magnetics
• Kalman
• Altimeter
• Help
Packet information choice of PC comm port packet activity indicator
# of packet errors
Roll/Pitch/Yaw & MagCal status
Choice of sensor set & algorithm
Sensor board run time and build parameters, Data logging on/off
This is the most intuitive way to confirm that your sensor fusion is working properly.

1
2
3
1.
Raw Accelerometer
Values
2.
Calibrated
Magnetometer Values
3.
Raw Gyroscope Values
The PC is used for display only. All values are computed on the embedded board.

2
3
1 1.
Roll, pitch & compass heading
2.
Current quaternion
3.
Angular velocity
4.
Linear Acceleration
4
The PC is used for display only. All values are computed on the embedded board.

2
1
3
1
4
1
6
6
1.
2D representation of the data point “cloud” used for hard/soft iron compensation
2.
Computed hard iron vector
3.
Soft iron matrix
4.
Statistics
5.
Calibration status light
6.
Save to text file
You can use this display to view how the magnetic constellation evolves over time in response to changing magnetic environments.

1
2
3
Use this tab to view how well your sensor fusion “digests” changes in its environment.
1. Error in orientation estimate (X,Y,Z)
2. Computed gyro offset
3. Error in gyro offset estimate (X,Y,Z)

output selection sensor selection algorithm selection
Sensors
Native to your
Android
Device
Remote
Sensor
Interface accel only mag (2d) gyro only accel/mag accel/gyro
9-axis
Panorama View
Device View
Log Window
Statistics View
Canvas View
Not Shown

Android Action Bar
Fusion Settings Bar
Sensor & Algorithm
Selection Here
Navigation
Control The Options
Menu is a button on some devices
If present, click to enable low pass filter

For mag / accel / gyro and rotation, the “Statistics” Views displays:
• sensor description
• current sensor value
• min / mean / max values
• standard deviation
• noise / √Hz
When used with the “local” sensor sources, this is a great way to gain insight into devices from the competition!

Component
Sensor Fusion Library for Kinetis
MCUs
Freescale Freedom Development
Platform
Freescale Freedom Development
Platform for Multiple Freescale
Sensors
Freescale Sensor Fusion Toolbox
For Windows
Freescale Sensor Fusion Toolbox
Android
Freescale Sensors
Kinetis Design Studio
CodeWarrior Development Studio for
MCUs
Community Support
Professional Support Services
Price (USD)
Free
KL25Z = $12.95
KL26Z = $15.00
KL46Z = $15.00
K20D50M = $18.00
K64F = $29.00
$30
$50
$125
Free
Free varies
Free varies
Free varies
Prices are current as of 6 Sept, 2014. They may vary in the future.
Location http://www.freescale.com/sensorfusion http://www.freescale.com/freedom http:www.freescale.com/FRDM-FXS-9AXIS http:www.freescale.com/FRDM-FXS-MULTI http:www.freescale.com/FRDM-FXS-MULTI-B http://www.freescale.com/sensorfusion https://play.google.com/store/apps/details?id=com.freesc
ale.sensors.sfusion
http://www.freescale.com/sensors http://www.freescale.com/kds http://www.freescale.com/codewarrior https://community.freescale.com/community/sensors/sens orfusion sfsw@freescale.com

http://play.google.com/store/apps/details?id=com.freescale.sensors.sfusion

at http://www.freescale.com/sensorfusion

We believe this package offers the lowest cost, most complete, sensor fusion solution available anywhere, with:
BSD style open source license (it’s FREE!)
3, 6 and 9-axis sensor fusion options
Source code for all functions
Working template programs
Low cost hardware options
Extensive documentation (data sheet, user manual and multiple app notes, training slides and videos)
Free development Kinetis Design Studio development environment
Free Windows and Android applications to visualize fusion results
Freescale community support at https://community.freescale.com/community/sensors/sensorfusion
Paid support available from Freescale’s Software Services team
( sfsw@freescale.com
)
For more details, please visit http://www.freescale.com/sensorfusion

Freescale, the Freescale logo, CodeWarrior, Kinetis,
MQX and Processor Expert are trademarks of Freescale
Semiconductor, Inc., Reg. U.S. Pat. & Tm. Off. All other product or service names are the property of their respective owners.
ARM
 and Cortex
 are trademark(s) or registered trademarks of ARM Ltd or its subsidiaries in the EU and/or elsewhere. Windows is a registered trademark of
Microsoft Corporation in the United States and other countries.
All rights reserved. © 2014 Freescale Semiconductor,
Inc

Files approximations.c
approximations.h
build.h
drivers.c
drivers.h
Events.c
Events.h
fusion.c
fusion.h
include_all.h
magnetic.c
magnetic.h
Description
Reduced accuracy/power trig functions
Build options consolidated into a single file
Initialization of hardware timers and I 2 C drivers for inertial and magnetic sensors. Contains
CreateAndSendBluetoothPacketsViaUART ().
Callback functions for hardware events. Contains
UART_OnBlockReceived ()
This is where the primary sensor fusion routines reside. All 3, 6 and 9-axis fusion routines are here.
A catchall for all the other .h files
Magnetic calibration functions

Files Description main.c
matrix.c
matrix.h
mqx_tasks.c
mqx_tasks.h
Initializes and executes MQX Lite RTOS
Optimized matrix manipulation functions orientation.c
orientation.h
tasks.c
tasks.h
Creates and runs the Sampling, Fusion and Calibration tasks which in turn call functions in tasks.c
This file contains functions designed to operate on, or compute, orientations. These may be in rotation matrix form, quaternion form, or
Euler angles. It also includes functions designed to operate with specific reference frames (Android, Windows 8, NED).
tasks.c provides the high level fusion library interface. It also includes the option to apply a Hardware Abstraction Layer (HAL).
With proper attention to sensor orientations during PCB design, tasks.c may never need modification.
user_tasks.c
user_tasks.h
Placeholder functions for // Put your code here

Soft Iron in fixed spatial relationship to
the sensor distorts the measured field.
The sphere is distorted into an ellipsoid.
Hard Iron (permanent magnet) in fixed
spatial relationship to the sensor adds an offset.
Ideal Measured
Both are linear effects 1 , and can be reversed – if you know what you are doing!
1 Assuming there is no magnetic hysteresis present

B c
= W -1 (B p
– V) where:
B c
W -1
B p
V
Calibrated magnetic vector
Inverse Soft Iron Matrix
Physical magnetic measurement
Hard Iron Offset Vector
B cx
B cy
B cz
= s
1 s
2 s
3 s
2 s
4 s
5 s
3 s
5 s
6
B px
B py
B pz
V x
V y
V z
The 4-element calibration computes V x
, V y vector . W -1 = identity matrix and V z hard iron offsets plus magnitude of the geomagnetic
The 7-element calibration also computes s
1
, s
4 and s
6
. Off diagonal components of W -1 are 0.
W -1 = s
1
0 0
0
0 s
0
4 s
0
6
The 10-element calibration computes all elements of W -1 , including s
2
, s
3
, and s
5
W -1 = s
1 s
2 s
3 s
2 s
3 s
4 s
5 s
5 s
6
Everyone uses the same equation.
The magic is in how you compute the coefficients.