VHDL-AMS Modeling of VCSEL including Noise

Mohamed KARRAY , Patricia DESGREYS and Jean-Jacques CHARLOT

École Nationale Supérieure des Télécommunications

Département COMELEC
46, rue Barrault, 75013 Paris, France

karray@[Link]

Keywords Output beam 840nm

Anode Ti/Au
Behavioral modeling, VHDL-AMS, Opto-electronics, VCSEL. 19 pairs (n dope)
Al 0.12Ga0.88As/AlAs

Abstract Proton implant or lateral 2.466µm Upper mirror

oxydation
3 Quantum well AsGa 0.006µm Active area
This paper presents a VCSEL diode model including electrical, +2 potentiel barrer Al0.3Ga 0.7As
625µm
thermal and optical behaviors where noise is present in the rate 3.894µm Lower mirror
30 pairs (p dope)
equations. VHDL-AMS language is used to write this model. Al 0.12Ga0.88As/AlAs
N dope
Langevin forces are introduced in equations and generated AsGa substrat
in Mentor Graphics ADVanceMS© by white noise signal
generators, using the UNIFORM statement. The comparison Linear variation of the Al Cathode AuGe/Au
concentration
of transient results with Spice ones published in literature AlAs intermediate
shows a good agreement. 3 Quantum well AsGa
+2 potentiel barrer Al0.3Ga0.7As

Al 0.12Ga0.88As
Introduction
In the last decade, the Vertical Cavity Surface Emitting Laser
(VCSEL) appeared as a reliable low-cost high-speed solution Figure 1 : VCSEL structure.
for data communication applications and interconnects. It
started to challenge the well-established edge-emitting lasers VCSEL modeling is based on the resolution of semiconductor
in telecom and data storage applications. laser rate equations [1], expressed for single mode VCSEL
This paper recalls the mathematical modeling described and operation as function of photon, carrier numbers and phase
presents the VHDL-AMS modeling of the proposed VCSEL modulation: S, N and ψ , respectively.
and comparison with Spice models results published in
literature. dS S N S
=− +β + GN ( N − N0 ) (1)
VCSEL Model development
dt τp τn 1+ε S

The fundamental difference between an edge-emitting laser dN I N S

and VCSEL is the fact that the laser oscillation as well as the = ηi − − GN (N − N 0 ) (2)
out-coupling of the laser beam occur in a direction dt q τn 1+ ε S
perpendicular to the gain region and the surface of the laser
chip. dψ α  1 1 
The advantages of VCSEL is that the surface emission and =− h G N ( N − N 0 ) −  (3)
dt 2  1 + ε S τ p 
the small size make it possible to fabricate very dense two-
dimensional arrays of VCSELs, suitable for multi-channels
parallel transmission modules. Parameters defined in these well known equations are
Figure 1 shows the schematic of an GaAs processed at TRT physical internal parameters in general not available. They are
(France) which was used to compare some results. defined on Table 1 with their derivation from system
parameters [2]. With η i is the current injection efficiency and
I is the injected current.
The noise is an important factor when laser diode are used for
signal transmission in equipment and in systems such as
optical fiber communication. These equations can be
rewritten as following:
 Tableau 1: VCSEL System-Physical Parameters Conversion [2].

Physical Unit Conversion towards physical Needed system Needed

Quantity
parameter parameters parameter physical parameter

GN Slope ηLI
Differential GN ≈
s−1 Power P0
gain [(e.ηLI)/2P0]. (2π.f3dB) Bandpass f−3dB
τp Photon τp ≈ Slope ηLI
s Optical frequency ν Cavity loss αm
lifetime ηLI.2q/(hν.vg.αm)
τn Carrier τn−1 ≈ Threshold current It Transparence number
s
lifetime (q/I).[N0 + (τ[Link])−1] Those of GN and τp N0
β Spontaneous
β −1≈ Thresh current It Transparence number
emission -
fraction (τp/e).(It – e.N0.τn−1) Those of τp and τn N0
ε normalized Bandpass f3dB
gain ε≈ Damping ξ in P
compression
- GN. [ξ. .(2π.f3dB)−1−τs] function of mod.
freq. fm
factor Those of τp
αΗ Linewid. in the chirp band Chirp function of
enhancement αΗ≈2.[(δν/δI)/ε )] mod. freq. fm
- out of the band Slope ηLI
factor
(Henry) αΗ≈2.[(δν/δI).fm]/ (ηLI.P) Damping ξ

D ss = R sp * S (7)
dS S N S
=− +β + GN ( N − N 0 ) + Fs (4) N
dt τp τn 1+ε S Dnn = Rsp * S + (8)
τn
R sp
dN I N S Dψψ = (9)
= ηi − − GN (N − N 0 ) + Fn (5) 4S
dt q τn 1+ ε S and Dij = 0 for i ≠ j

dψ α  1 1  where R sp is the rate of spontaneous emission and given by

=− h G N ( N − N 0 ) −  + Fψ (6)
dt 2  1 + ε S τ p  (10) :
n sp
Where Fn , Fs and Fψ are the Langevin Forces. They R sp = , (10)
τp
assumed to be Gaussian random process with zero as average and n sp is known as the spontaneous-emission factor and
value and have a correlation function of the form:
Fi (t ) F j (t ′) = 2 Dij δ (t − t ′) ,
N
given by (11) : n sp = (11)
N − N0
where i , j = S , N and Dij , the diffusion coefficient. The
dominant contribution to laser noise is supported by only two To generate the Langevin forces we use random signal
coefficients Dij and Dij ; others can be assumed be nearly generators. Only white noise sources have been considered.
To generate a white noise source, we use the function
zero [3].
UNIFORM provided by the library math_real of Mentor
These different diffusion coefficients are given by (7), (8),
Graphics ADVanceMS©. This function returns a pseudo-
(9):
random number x with uniform distribution. Next figure
shows an example of a random signal generator process
written with VHDL-AMS and using the UNIFORM
statement.
 Detection:process
I th (T ) = a0 + a1 .T + a 2 .T 2 + a 3 .T 3 + a 4 .T 4 (13)
variable seed1:integer:= 3456 ;
variable seed2:integer:= 4563 ;
variable unf : rea l;
where a1 … a 4 are temperature dependent parameters and can
begin be deduced from experimental VCSEL curves.
wait for tau; Another parameter depends on temperature, it’s the differential gain
uniform(seed1,seed2,unf); and it written as (14):[5]
If unf>0.0 Then out1<= unf;
ag 0 + ag 1 * T + ag 2 * T 2
ELSE null;
G = G0 (14)
End If; bg 0 + bg 1 * T + bg 2 * T 2
uniform(seed1,seed2,unf);
If unf>0.0 Then out2<= unf; VHDL-AMS Modeling
ELSE null;
End If; In this section we present different model written with
end process; VHDL-AMS language.
We use pseudo random bit sequence to modulate the VCSEL,
Figure 2 : VHDL-AMS Random Signal Generator. we use a CMOS driver between the VCSEL and a pseudo
random generator (Figure 3).
To obtain a white noise with Gaussian distribution, we use Figure 4 shown the VHDL-AMS model of the generator.
Box-Muller transformation, it looks like:
Periode Dimensions
y 1 = − 2 ln( x1 ) cos(2πx 2 )

y 2 = − 2 ln(x1 ) sin(2πx 2 )
Voltage Current
Generator Driver

We start with two independent random numbers, x1 and x2,

which come from a uniform distribution (in the range from 0 VCSEL
Optical
to 1). Above transformations applied to get two new Power

independent random numbers which have a Gaussian Gnd

distribution with zero average value and a standard deviation

of one. Figure 3 : Optical transmitters.
The Langevin Forces in equations 4,5 and 6 are written using
x1 and x 2 by:[4] clock <= not clock after ( periode);
PROCESS
 Fn = Dnn .x1n BEGIN

 wait until clock ='1';
 F p = Dnn .x1 p
 FOR i IN 1 TO n -1 LOOP
 Fψ = Dψψ .x1ψ SR(i+1) <= SR(i);
IF SR(i) ='1' THEN Vout_int <=V_high;
there different xij are modeled with the same process only ELSE Vout_int <= V_ low;
timeout-clause of wait statement is modified. END IF;
END LOOP;
Thermal dependence
CASE n IS
Different parameters depend on temperature in VCSEL rate when 7 => SR(1) <= SR(3) xor SR(7);
equation. That is the case of the threshold current and the
when 10 => SR(1) <= SR(3) xor SR(10);
output power.
The output power can written now by (12): when 15 => SR(1) <= SR(1) xor SR(15);
END CASE;
Popt = η (T ) [I − I th (T )] (12) END PROCESS;
Vout == Vout_int'ramp;
where the threshold current is given by (13):[5] END;

Figure 4 : Architecture of pseudo random generator.

Next Figure present a comparison between the Probability
Density Function of Matlab© Random noise and VHDL-
AMS one.

Figure 7 : Transient Results.

Figure 5 : PDF of two Gaussian Noise.

The VCSEL is modeled with VHDL-AMS (Figure 6)and

simulated with ADVanceMS©. Figure 7 presents some
transient results, clock, injection current and optical power
respectively. The comparison with other works [6] reported in
figure 8 shows a good agreement in the output optical power Figure 8 : Response to Square Current pulse. [6]
signal.
Conclusion
We have shown that VHDL-AMS is able to model
ARCHIT ECTURE be_laser OF laser IS advantageously VCSEL where different type of multi-
technological quantities (here optical, thermal and electrical)
--Laser rate equations are coexisting. A methodology for modeling the noise in the
N'dot == etai * I/physical_Q - N/TauN - Go * (N - No) * S / (1.0 + eps * S ) + Fn ; VCSEL in transient domain is presented, simulations are run
S'dot == - S/TauP + Beta * N / Tau N + Go * (N - No) * S / (1.0 + eps * S ) + Fs; with ADVanceMS©. Simulation results are positively
Psi'dot ==AlphaH / 2.0 * ( Go * (N - No)/(1.0 + eps * S) - 1.0/TauP ) + Fpsi; compared with Spice models results published in literature.
--Threshold current
Ith == a 0 + a 1 * T i + a 2 * T i**2.0 + a 3 * Ti**3.0 + a 4 * Ti**4.0 ;
--Noise Source References
Fn == Vnoise * sqrt( Dnn ) ;
Fs == Vnoise2 * sqrt( Dss ) ; [1] Z. Toffano and all. “Multilevel Behavioral Simulation of VCSEL based
Fpsi == Vnoise3 * sqrt( Dpsipsi ) ; Optoelectronic Modules” , IEEE Journal of Selected Topics in
--Diffusion coefficient
Quantum Electronics Accepted Juin 2003.
[2] Z. Toffano, Optoélectronique, composants photoniques et fibres
Dnn == Rsp * S+N / Tau N;
optiques, Paris, Editions Ellipses, 2001, chap.V, pp. 178-197.
Dss == Rsp * S ; [3] G.P Agrawal, Fiber-optic communication systems, Third Edition, 2002
Dpsipsi == Rsp / ( 4.0 * S ) ; Chapter3, pp. 77-127.
--Spontaneous emission
[4] S.F. Yu,” Nonlinear Dynamics of Vertical-Cavity Surface-Emitting
Lasers”, IEEE Journal of Quantum Electronics. Vol.35, No.3, March
Rsp == nsp / TauP ;
1999.
nsp == N / ( N - No ) ; [5] P. V. Mena, “A Simple Rate-Equation-Based Thermal VCSEL Model”,
--Optical power IEEE Journal of Lightwave Technology. Vol.17, No.5, May 1999.
Popt == (physical_H * nu * Alpham * Vg / 2.0 ) * S ;
[6] A. Przekwas and all, “ Multi-level simulation of VCSEL devices and
packages in CFD-ACE+ environment”, International Workshop on
Numerical Simulation of Semiconductor optoelectronic Devices Santa
END ARCHIT ECTURE be_laser; Barbara, March 25-27, 2001.

Figure 6 : VCSEL Architecture.