Table of contents
6.The magnetic fields sensors
7.The inclination measurement system
9.The data acquisition system
11.The power supply
12.Realisation of the PCB
13.The embedded system
14.Static Library Util.a
17.Sensors controller commands
XCompass are two programs used to manage the Smart Compass card throw the
serial port. To do so, they share the same static library: util.a
This library contains the classes and
functions used by the programs
To control of the serial port (class Cport)
To display the notifications and the errors and
log them in a file (class Clog)
To construct and interpret the data received and
send to the Smart Compass (class CNavigation).
Others common functions are in the file util.cpp
Cport class contains the functions used
to handle the serial port.
14.1.1. Access port on Linux
The way to access the port on Linux is
the same way to access a file. It can be open, read, write and close by the
same function. Linux names its serial ports in the UNIX tradition. The first
serial port has the file name /dev/ttyS0, the second serial port has the file
name /dev/ttyS1, and so on. Once open, a termios structure can alter the
characteristic of the port.
14.1.2. Termios structure
This structure gathers the characteristic
to access on a serial device (including the standard output: cout, cin, cerr)
under Linux. Many option can be setup there. The most common are the baudrate,
the frame type (parity, stop bits, data bits) and the flow control.
A Cport’s subclass is Cbuffer. It emulate
a circular buffer that store the last data received from the port.
Excluding the methods used to receive and
send the data (char, string, structure), the Cport class can execute a shell command throw a pipe (ex: setserial, sz …).
Clog class saves and displays the data
from the programs. A member function searches for keyword and highlights them
with one of the following colours (red, green, yellow, blue, purple, grey and
cyan) or form (bold, underlined, blink, inverse). Once this class called, a
pointer to file is created and a list of keyword is loaded from the file
“HIGHLIGHT”. Each message send to the static member write() is saved in the
opened file and sent to the std::cout output with the appropriate colour.
This class also deals with
the errors thanks to the static member function error_detected(). The error
message is display and the errno error if exist, is commented.
If the class is launch by an
Xwindow program, the method m_ouput allow adding some specific function.
Therefore, pointers to three functions are passed:
To output in a multi-line text edit
To open a messagebox after each notificationalert
To advance a curser while progressing a task
class is an abstract class that defines the structure to the different modules
of the Card (magneto, accelero, gyro, power and compensation). Each of these
modules shares the same behaviour.
14.3.1. Middle-level methods
The module can
request value from the card. They have to be calibrating first, then they can
be execute (1). The instruction is then send to the command method that create
the request frame and insert the headers (2). Then, the corresponding structure
is send to the micro controller where the frame is interpreted and executed.
The reply is also a structure that is received (3) and transmits back to the
command that save the value into the module (4). The process is similar for the
calibrate method exec that except that the value saved is the average value of
high number of value (5). The initialised data can be saved or loaded from / to
a file (6).
exec() and calibrate() do not return directly the value but register them in
private structure inside the class. Therefore, to access the data, the virtual
function getInit() and getVal() have to be called (7).
The calibration is essential for the rest
of the measurement. The initial values are used many times in the calculation
of the azimuth. Therefore, we should have the most accurate values as possible.
The calibration’s program has been realized in this purpose. The process of the
calibration is different regarding the module:
For the magneto sensors, the initial values are
the averages voltage read after a negative and a positive pulse
For magnetic field compensation, we have to make
a first series a measurement in an initial direction, and then a second with
the card turn of 180º. The values of the magnetic field interference are the
For all the other modules, the calibration of
the module is done in two steps:
The calibration is done according to the
When the calibration is finished, the
initialized values are saved on a calibration file chosen by the user. If the
file does not exist, a new one is created. This file could be loaded at
anytime. In addition, the initial values are load each time ‘Xcompass’ or
‘viewport’ start. Nevertheless for optimized result a calibration have to be
done every day.
14.3.3. Request the value
method provides an easy way to send and receive any type of data. In addition,
this method deals with the header. If errors occur during the communication or
in the microcontroller, the error code is interpreted there and forwarded to
the Clog class to be registered, commented and notified.
14.3.4. Direct control:
value from the sensors can be access through the method getValue() from the
corresponding module: magneto, gyro, accelero, power that returns respectively
the magnetic fields, the rotation velocity, the acceleration and the voltage of
methods are Usage(), resetSystem(), pingSystem()
14.3.5. Matter of inheritance
language allows the multiple inheritances in contrast of other object oriented
language like the Java. However, the method is a little tricky when a great
child (like Cnavigation) inherits from two children (ex: CMagneto and Cgyro)
that have the same parent (CModule). In that case, the parent is duplicated and
two distinct instance of this object co-exist. It means that if a variable is
change by a child, the modification does not occur on the second. This problem
does not really affect the methods, especially virtual methods that have no
To by-pass this
problem, the solution implemented have been create a private inner class
(Common) which the only instance have a static linkage. In that way, all the
shared data that should have a space dedicated, are not duplicated and can be
access by any child.
This class is
specific to the Smart Compass card. It holds the object of the different module
and implements the methods that have to receive information from more than one
module. Consequently, CNavigation gathers the method to calibrate the different
modules, to calculate the angle between the card and magnetic north.
getNorth() calculate from the different parameters, the angle between the heading
direction and the magnetic north.
calculation of azimuth tries to compensate the different source of errors.
Many kinds of
errors may occur during the measurement:
field may be distorted by other magnetic fields or by nearby ferrous materials.
These errors can be classified into two types:
Deterministic interference: The interference
source is at a fixed position relative to the compass and its magnitude is
constant over the time. The effect will be to shift circle’s centre of the
interference fields. This error can be reduced by a calibration: the shift of
the centre can be compensating by electro-magnetic feedback.
Non-Deterministic interference: the interference
source happens from time to time with unknown magnitude: for example the robot
passes close to a speaker or a computer. This kind of error cannot be
compensating. However, once detected, the lecture of the angle can be switch on
the integration of the gyro value over a short period.
important error from gyro is a slow drift of the average value during the time.
This error is non-deterministic and its value can be some m°/s per minute. A
way to compensate this error can be to add a numerical low-pass filter with Fc
lead to an important deviation as the gyro velocity is integrated to obtain the
The inclination leads to an error due to
the measurement of the component of the field.
Saturation can appear from:
The ADCs Max386 and Atmega32, which cannot read a value outside the range [0 V– 4,096 V]
The amplifier INA2126 has an output range: [0,65 V – 4,35 V]
appear in the magneto-sensor if the magnitude of the field detected is twice
the earth field.
numerisation of the analogical values by the ADC produce a noise from measure.
This noise can be modelize by the following function where Va is the tension
from the sensor and Vn the value read.
The noise is directly dependent of the read frequency
Shannon theorem: Fe ³ 2 * Fmax
=> Problem for the gyro.
Gaussian white noise:
Voltage and movement noise are measured.
Refer to the section.
In regards to these perturbations, the
following algorithm has been implemented to determine the corresponding
This file contains some useful functions:
Conversion method (ASCII to hexadecimal or binary, int to string)
A non blocking keyboard input reader.