Table des matières

INTRODUCTION ....................................................................................................................................... 2
THEORETICAL ANALYSIS .......................................................................................................................... 3
ANALYTICAL FINDINGS ........................................................................................................................ 3
STABILITY OF THE SYSTEM............................................................................................................... 3
NUMERICAL RESULTS .......................................................................................................................... 4
EXPERIMENTAL OBSERVATIONS USING FIELD PROGRAMMABLE GATE ARRAYS (FPGA) ....................... 6
CONCLUSION ......................................................................................................................................... 10

KOUEKAM SEDRIC JOEL 1

 INTRODUCTION
In recent years, there has been a growing interest in theoretical, numerical and practical
implementation research of chaotic system. This great interest is explained by the multiple
applications of these systems in the fields of science, technology and engineering. The nonlinear
dynamics present in a system can lead to chaos and is characterized by sensitivity to initial
conditions. Chaos theory primarily concerns nonlinear systems in nature, such as turbulence,
weather, stock market, biological systems and many more. These systems are characterized by
the inability to predict their behavior.
OPTOELECTRONIC OSCILLATORS (OEOs) are autonomous nonlinear systems
whose feedback loop is constituted with an optical and an electrical branch. In the optical part,
OEOs are seeded with a continuous-wave (cw) semiconductor laser (typically in the near-
infrared range). The output laser beam is generally modulated using a nonlinear electrooptic
modulator and subsequently launched into an optical fiber line which is long enough to induce
a significant time delay. The optical beam is then converted into an electrical signal using a fast
photodiode (PD). This radio-frequency (RF) signal is eventually bandpass or lowpass filtered,
amplified, and then used as a driving signal of the modulator, thereby closing the feedback loop.
Following this mechanism, OEOs can output RF signal with a frequency range from 1 kHz to
100 GHz, and they
display a very wide range of complex dynamical behaviors. These oscillators were initially
investigated by Neyer and Voges, and they were rapidly identified as ideal experimental
benchmarks to investigate some of the dynamical features of the paradigmatic Ikeda equation.
Many applications are related to OEOs, such as optical chaos cryptography and neuromorphic
computing, for example (see review article and references therein). However, the most well-
known application is ultra-pure microwave generation for aerospace engineering applications,
and this area of research was pioneered by Yao and Maleki in 1994, who are also those who
coined the term OEO. In most architectures of OEOs, the nonlinear conversion between the
electrical and the optical signal is performed by a phase or an intensity modulator with
sinusoidal transfer function. Models to investigate the dynamical properties of such
architectures have been proposed for both the narrow-band and wideband configurations of
OEOs. However, it has been demonstrated that other electro-optic or optoelectronic devices can
be used to perform this nonlinear conversion. The simplest example is to consider the seeding
laser-diode (LD) itself as an electrical-to-optical converter through its power-intensity or
“PI”—transfer function. It is noteworthy that in the adiabatic limit (modulation frequency much
slower than the relaxation frequency of the laser), the PI transfer function of a semiconductor
laser is a “piece-wise linear” function equal to zero below a given threshold current Ith, and
linearly increasing above. We therefore consider that this architecture corresponds to the
simplest OEO possible, and it is relevant to perform a detailed theoretical and experimental
study in order to understand its dynamical properties. Moreover, since this OEO does not
require an electro-optic modulator (which is definitely one of the most expansive element of
conventional OEOs), it appears to be a cost-effective alternative for frequency-versatile
microwave generation. The aim of our research is to analyze this new OEO architecture where
the LD nonlinearity is used to perform the conversion from the electrical to the optical domain.
We show that the system displays a complex dynamic which depends on slow and fast-scale
limit-cycles that can arise in the system depending on the gain of the delayed feedback. We also
successfully compare our analytical findings with numerical simulations and experimental
measurements.

KOUEKAM SEDRIC JOEL 2

 THEORETICAL ANALYSIS

ANALYTICAL FINDINGS
y=x (1.1)

1
 x = −x − y +  D XT −  (1.2)

To study the dynamics of a system it is always interesting to study the stability of its fix points.
Given the dynamic system below
dX (t)
= F (X(t)) (1.3)
d (t)

With F(X(t)) being a nonlinear function that depends on the state variable X(t). The fix points
of the system given above are obtained by resolving the following equation
dX (0)
=0 (1.4)
d (t)

STABILITY OF THE SYSTEM

The linear stability analysis is based on analyzing the time dependent trajectory of the system
when slightly perturbed from the steady state ( xst ; yst ) . The flow of Equations. (1.1) and (1.2)
has a single stationary point which obeys:

( xst ; yst ) = (0;  − )

(0; − ), for   0
= (1.5)
(0;0), for   0.
The solutions x(t) and y(t) can be represented as a steady state and a small perturbation
following

x(t) = xst +  x(t) (1.6)

y(t) = yst +  y(t) (1.7)

and inserting the preceding equations into Equations. (1.1) and (1.2) leads to

 y =x (1.8)
1
 x = − x −  y +  H  −  xT (1.9)

From equations (11) and (12) the stability of the system can be investigated after setting
 x,  y  et through the eigenvalue’s equation:

KOUEKAM SEDRIC JOEL 3

 2 +  1 −  e− t H  −  +
1
=0 (1.10)

and it appears that the stability of the system critically depends on the value of the dimensionless
offset voltage α, and more specifically, on its sign.

NUMERICAL RESULTS

(a) (c)

(b) (d)

Fig 1.1 Numerical time traces of the OEO for, a)  =1.33,b)  =1.36, c)  =1.6, and d)  =1.9

KOUEKAM SEDRIC JOEL 4

Tab1.1 Variation of the period and amplitude according to the control parameter (  )

Control parameter (  ) period Amplitude

1.2 0.22 0.2241
1.21 0.22 0.2351
1.23 0.22 0.2547
1.25 0.22 0.273
1.27 0.22 0.2921
1.29 0.22 0.3113
1.3 0.22 0.3206
1.31 0.22 0.3311
1.33 0.22 0.3518
1.35 1.79 0.2316
1.37 1.78 0.2407
1.4 1.79 0.2544
1.5 1.82 0.3138
1.6 1.855 0.38
1.7 1.9 0.4569
1.8 1.95 0.5443
1.9 2.01 0.6514

(a) (b)
Fig1.2 Evolution of the Amplitude (a) and Period (b) of the asymmetric wave oscillation for a
constant delay and different gain parameter (  ) values.

KOUEKAM SEDRIC JOEL 5

EXPERIMENTAL OBSERVATIONS USING FIELD PROGRAMMABLE GATE
ARRAYS (FPGA)
The physical realization of time-delay systems pose challenges due to the implementation
of the time delay. Although physical realization of delays using analog electronic circuits such
as tunable delay and RC filter have been studied, an attractive solution is to implement the
differential equation on an FPGA. These devices are massively parallel architectures whose
hardware functionality can be appropriately configured. The attractiveness of the FPGA stems
from the fact that these devices can be configured with any nonlinearity. One can also
implement numerical algorithms (such as Euler’s method) and specify time delays quite easily
(without the burden of a processor) using a hardware description language (HDL). Hence these
platforms can serve as engineering solutions for conducting scientific experiments. Moreover,
100MHz clock frequencies are quite common on modern (21st century) FPGAs. Such high
frequencies, coupled with a variable data-path (32-bit, 64-bit, 128-bit etc.) specification,
provide us with a platform that is capable of handling large parallel blocks of data transfer,
unlike a traditional processor. So, in this work, we utilized a soft-processor on the FPGA to
sequentially emulate chaotic DDEs. The HDL block diagram of our system is setup using the
HDL coder library in Simulink as shown below.

Fig 1.2. We are using a fixed-step ode4 (Runge Kutta) solver in Simulink, with a step size of
0.001. We run the simulation for 50 seconds.

KOUEKAM SEDRIC JOEL 6

 (a) (b)

(c) (d)
Fig1.3 Result from hardware co-simulation, plotted using the scope in Simulink, For
a)  =1.33,b)  =1.36, c)  =1.6, and d)  =1.9
After following the different steps of the work flow advisor, we generate the VHDL
project which is stored in a folder of our choice. We open the project with vivado, synthesize,
implement and Run.

Fig1.4. Results from VHDL code simulation in Vivado, For  =1.33

KOUEKAM SEDRIC JOEL 7

Fig1.5. Results from VHDL code simulation in Vivado, For  =1.36

Fig1.6. Results from VHDL code simulation in Vivado, For  =1.6

KOUEKAM SEDRIC JOEL 8

Fig1.7. Results from VHDL code simulation in Vivado, For  =1.9

KOUEKAM SEDRIC JOEL 9

 CONCLUSION
As conclusion we can say that we have investigated the simplest architecture of OEO,
which is characterized by a piece-wise linear LD nonlinearity and a delayed feedback loop.
After the analytic and numerical studies, we have built the experimental system in sumulink,
generated the VHDL code and established the time-domain equations describing its dynamics.
We showed that the time traces exhibits, slow-scale limit-cycles which appears when the
feedback gain increased. The theoretical study of the dynamics of our system critically depends
on how the LD is polarized. We have compared the numerical simulations of our model with
experimental measurements and we have found an excellent agreement between both sets of
data.

KOUEKAM SEDRIC JOEL 10