Semiconductor Optical Amplifiers (Soas, Pp. 232-243 in Agrawal)

Semiconductor Optical Amplifiers
(SOAs, pp. 232-243 in Agrawal)

• Amplifier Design
–requirements on the mirror reflectivities
• Amplifier Characteristics
–Saturation power
–Gain clamping
–Noise figure
–Polarization sensitivity
–Pattern effect
• Pulse Amplification (next lecture)
SOA Design
We already discussed basic SOA design in the general lecture about optical amplifiers
where we emphasized that a traveling wave (TW) amplifier, which relies on a single-
pass gain, ideally has no mirror reflections and is the most desirable layout. This is
typically achieved by either AR coatings on the cleaved mirror facets and/or a tilted
waveguide with respect to the facets. Remember that any reflections from the mirror
facets are usually undesirable because they can severely modulate the amplifier gain
and also considerably narrow the amplifier bandwidth. The effect of these reflections
on a TW amplifier is called “gain ripple.” Any design that deliberately uses this feature
to enhance the gain due to cavity resonance is called a Fabry-Perot (FP) amplifier.
The SOA schematics are repeated below from the previous lecture.
Pump current
Signal in Signal out

Active region
AR = Antireflection AR Partial mirror Partial mirror

coating
(a) Traveling wave amplifier (a) Fabry-Perot amplifier
Simplified schematic illustrations of two types of laser amplifiers

© 1999 S.O. Kasap, Optoelectronics (Prentice Hall)
Equations for SOA Design
Assuming the facets of the SOA have reflectivities, R1 and R2, a single-pass gain
of G, Fabry-Perot cavity resonance frequencies νm (eqn. 3.3.5 in the text), and
frequency spacings ∆νL (p. 96), the amplification factor of the SOA is:
GFP (ν ) =
(1− R1 )(1− R2 )G(ν )
(1− G R1R2 ) + 4G R1R2 sin 2 [π (ν − ν m ) /∆ν L ]
2
Essentially this is the equation for the residual gain ripple in a traveling-wave (TW)
SOA. In other words, what we ideally want in a TW SOA is GFP(ν) = G(ν). In FP
SOA’s we use the GFP(ν) expression to calculate the amplifier gain enhancement
around one of the cavity resonance frequencies, νm.
Note: what Agrawal calls the “amplification factor”, G, (unit-less) is also

frequently called the “amplifier gain”. Remember to distinguish this from
the “gain coefficient”, g, which has units of cm-1.
SOA Design continued
I don’t have a good figure showing lots of gain ripple, but if you look closely,
you can see the residual gain ripple in the TW amplifier data shown below.
Note: FP amplifiers can achieve a very large cavity enhancement of the
amplifier gain when ν = νm and G(R1R2)1/2 is close to 1, but the tradeoff is a
very, very narrow amplifier bandwidth. Here’s the equation for the 3-dB
bandwidth of an FP SOA:
⎛ ⎞
2∆ν L −1⎜ 1− G R1R2 ⎟
∆ν A = sin
⎜
( )
12⎟
π 4G R R
⎝ 1 2 ⎠
Plugging in typical numbers for an FP
SOA, we get a ∆νA of only 10 GHz, less
than one channel spacing in a WDM
system. Thus, FP SOA’s are undesirable
for most lightwave system applications
SOA Design continued
The clear way to reduce gain ripple in a TW SOA is to reduce the reflectivities,
R1 and R2. But what do those values need to be? At worst, the gain ripple, ∆G,
cannot exceed 3 dB. Otherwise the amplifier bandwidth is determined by cavity
resonances. It is easy to show from our previous equation that the gain ripple is:
⎛ 1+ G R R ⎞
2
max
G
∆G = FP
min
= ⎜⎜ 1 2
⎟⎟
GFP ⎝ 1− G R1R2 ⎠
A typical TW SOA gain is 30 dB, so to keep ∆G < 3 dB, (R1R2)1/2 < 0.017%.
Since R1 is usually chosen to equal R2 in the design of the SOA, that means
each facet must be AR-coated to around 0.017% reflectivity. That is very
difficult to achieve over let’s say a 50 nm bandwidth (C-band). Hence, the
need for a tilted waveguide geometry, which can easily achieve 0.1%
reflectivity with a 7° tilt angle. Combining the AR coating and the tilted
waveguide gives some added cushion to achieve about 0.01% reflectivity.
SOA Characteristics-
The Saturation Power, Ps
We talked generally about power saturation in optical amplifiers in the last
lecture, but let’s investigate the physical origins of Ps in SOA’s specifically.
Following Agrawal’s derivation on pp. 234-235, it is now more convenient to

use the peak gain coefficient, g(N) (cm-1), the optical confinement factor, Γ
(unitless), the differential gain, σg (cm2), the cross-sectional area of the
waveguide mode, σm (cm2), the active region volume, V (cm3), the carrier
number, N (unitless), the transparency carrier number, N0 (unitless),the carrier
lifetime, τc, the injection current, I (mA), and the signal power, P (mW). I also
will remind you that the amplifier has a length L.
One important concept here is that the optical mode volume (where the photons
are) does not equal the active region volume (where the gain is). In other words,
σ mL ≠ V also Γ=
V
σ mL
SOA Power Saturation continued
Here are the relevant equations for the peak gain coefficent and the time rate
of change of the carrier number (or population):
g(N) = (Γσ g /V )(N − N 0 ) (Compare this to eqn. (3.5.3) in which the units are s-1)
dN I N σ g (N − N 0 ) Compare this to eqn. (3.5.2). It’s in the same units, but

= − − P
dt q τ c σ m hν expressed just a little differently
The differential gain, σg, is basically the active region cross-section. If σg is

large, then the gain medium is easy to stimulate with photons. It’s very
responsive. We will see, however, that this also makes the gain medium
easier to saturate.
SOA Saturation Power continued
To find the steady-state value for the carrier number, N, we set dN/dt = 0
and solve for N. Then we put the solution for N into the peak gain
coefficient expression and we find that the optical gain saturates as:
g0
g= Note: the SOA saturates in the same way as the simple 2-level
1+ P Ps system described in the general optical amplifier lecture
Where the small-signal gain, g0, is:
g0 = (Γσ g V )(Iτ c q − N 0 )
And, most importantly, the saturation power of the SOA is:
Ps = hνσ m (σ g τ c )
Desirable features for high Ps in an SOA
Ps = hνσ m (σ g τ c )
• Large cross-section for the optical mode, i.e., for a given power, the photons are
spread out.
• Small differential gain, which means that the gain medium should have a small
cross-section. It is more difficult to stimulate for a given gain value.
• A fast carrier lifetime, the resupply of carriers to the gain medium should be fast.
Remember from last lecture that the saturation output power of the 2-level system is:
s
Pout ≈ 0.69Ps
Typical saturation output powers nowadays for SOA’s are 10-30 mW (better than
what Agrawal quotes)
Gain-clamped (GC) SOA’s
We want to eliminate the gain saturation effect in SOA’s because in multichannel applications the gain
saturation leads to severe crosstalk. The channels are trying to rob gain from each other. Why don’t we just
operate in the linear regime of the amplifier? Answer: then you don’t get enough power output.
Gain clamping extends the linear regime of the amplifier and thus the useful output power. Within the same
waveguide, gain clamping is accomplished by introducing laser action at a wavelength remote from the signal
wavelength. Once lasing begins the carrier population is clamped at a fixed value. This has the effect of
making the amplifier gain relatively insensitive to the input signal power. Wavelength selective feedback is
accomplished with either a DFB grating throughout the optical cavity or at the ends in which case it is called a
distributed Bragg reflector or DBR mirror. Notice the very hard saturation of the GC-SOA in the figure on the
right. Compare that to Fig. 6.2 in Agrawal for a typical 2-level optical amplifier.
SOA Noise Figure
The noise figure expression for an SOA is pretty simple. From the last lecture,
remember the general expression for the optical amplifier noise figure, Fn:
N2 N
Fn ≈ 2n sp = 2 ≈2
N 2 − N1 N − N0
Looking at this equation, we need to minimize N0, which takes into account the
carrier population that generates those noisy spontaneous emission photons that
get amplified by the gain.
For the SOA, there is another term that we must introduce in the noise figure
equations to treat the internal loss of the optical waveguide. This term reflects the
fact that not all the gain generated in the optical cavity is coupled to the external
world, some of it is scattered and lost. So the final expression is
⎛ N ⎞⎛ g ⎞
Fn = 2⎜ ⎟⎜ ⎟
⎝ N − N 0 ⎠⎝ g − α int ⎠
Polarization Sensitivity in SOA’s
A typical SOA is dominated by the TE mode as opposed to TM. Please accept this as a fact for right
now. This strong TE behavior is undesirable for lightwave systems because the polarization coming out
of the optical fiber is random unless we use polarization maintaining fiber (which is relatively
expensive). The brute force (and more expensive) methods of getting around this problem are shown
below. The more usual solution nowadays is to introduce tensile strain into the SOA bulk active region.
This has the effect of balancing the TE and TM gain and only a single amplifier chip is required.
The Pattern Effect in SOA’s
Now the pattern effect is a big deal in SOA’s and our text appears to be silent on this matter.
Because their carrier lifetime, τc, is roughly the same as the bit rate in modern lightwave systems,
SOA’s when pushed towards gain saturation have the nasty habit of being sensitive to the pattern of
the bit signal. This is because as the digital signal rises and falls, the SOA is getting pushed back
and forth between the linear and gain saturation regimes. Gain-clamping helps, but not entirely.
See the next slide for some real data showing the pattern effect in typical InGaAsP SOA’s and its
elimination in quantum dot based SOA’s. EDFA’s don’t suffer from this pattern effect problem at
all because the fluorescence time (analogous to carrier lifetime) in EDFA’s is so long, milliseconds,
that the amplifier responds to average signal power, not instantaneous power.
Quantum Dot 1310 nm SOA - Removes Pattern Effect
• Operation in Saturation Regime is desirable. Typical SOA suffers bit distortion.
• Quantum dot has very fast τc < 3 ps (10X faster than typical InGaAsP SOA) solves the problem.

Semiconductor Optical Amplifiers (Soas, Pp. 232-243 in Agrawal)

Uploaded by

Copyright:

Available Formats

Semiconductor Optical Amplifiers (Soas, Pp. 232-243 in Agrawal)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Semiconductor Optical Amplifiers (Soas, Pp. 232-243 in Agrawal)

Uploaded by

Copyright:

Available Formats

Semiconductor Optical Amplifiers

(SOAs, pp. 232-243 in Agrawal)

Signal in Signal out

AR = Antireflection AR Partial mirror Partial mirror

Simplified schematic illustrations of two types of laser amplifiers

Note: what Agrawal calls the “amplification factor”, G, (unit-less) is also

Following Agrawal’s derivation on pp. 234-235, it is now more convenient to

dN I N σ g (N − N 0 ) Compare this to eqn. (3.5.2). It’s in the same units, but

The differential gain, σg, is basically the active region cross-section. If σg is

Where the small-signal gain, g0, is:

And, most importantly, the saturation power of the SOA is:

You might also like