noc18-ee28-Lecture 40-Semiconductor optical amplifier

IIT Kanpur July 201
20 followers
Dec 3, 2023
Hello and welcome to NPTEL, MOOC on Fiber Optic Communication Systems and Techniques.
In this module we will look at semiconductor optical amplifiers, which is one of the optical
amplifier that is used in fiber optic communications and fiber optics ah um signal processing ah
fields. Now, the reason I am discussing semiconductor
optical amplifiers is because you can convert a laser semiconductor laser into a semiconductor
optical amplifier by removing the optical feedback or by requiring the optical feedback
be very very low, ok. Of course, that is what differentiates between any optical amplifier
or in general any amplifier with a laser laser being the oscillator. So, when there is a
feedback, then this feedback will be sufficient for the device to generate self sustained
oscillations in the steady state. Whereas, when you remove the feedback or you adjust
a feedback such that the feedback is not important then you will not be able to you know build
on the you know oscillations and what you will actually have ah end up with having is
a optical amplifier. Much of the working that we ah look from a
semiconductor optical amplifier the principles of working is the same as that of the semiconductor
laser, you will have an inversion layer where there will be stimulated emission. But this
stimulated emission will be brought about by the externally applied signals the cavities
will still be there ah in some form of the other, ok but the cavity reflectivities are
usually kept very low. So, as to obtain a broadband lasing I mean broadband amplification
rather than lasing. And all the other principles of you know ah laser also remains. So, you
have to have a hetero structure laser or you know have a hetero structure material and
then you have to have an injection pumping, injection by ah way of applying an appropriate
current forward bias in the junction. For example, all that thing that we discussed
will still be there, ok. So, then what is the different features of semiconductor optical
amplifiers is what we want to look at, ok. Now, any amplifier the function of that any
amplifier would be very simple. So, you send in certain light here which has an electric
field E in then the output would be ideally square root G times E in. And using square
root G because G or gain is usually referred to the power quantity and not a electric field,
ok. So, G is usually what is called as the power gain and not to the electric field.
So, for electric field we simply square up the power gain and therefore, you will be
obtaining what is the field gain, ok. Of course, if it was an ideal amplifier this
is what it would have done but most physical amplifiers also add a certain amount of noise
to the output, ok. This is kind of a amplifier spontaneous emission noise that is added from
the semiconductor material itself. And the amount of noise that goes into the band depends
on the bandwidth of the input signal itself. Usually the gain is broadband and ah when
you send in light at a particular ah frequency or with a certain bandwidth around a particular
carrier frequency the in band noise is determined by what is the bandwidth of the input signal.
So, this is what one would expect from an ideal amplifier and what one would expect
from a non ideal amplifier, ok. But how exactly is this created? Well, if
you look at the structure of a semiconductor optical amplifiers most structure fall into
two types, one is called as the FP cavity type structure, ok wherein the lasing material
is actually kept or the active medium is actually kept in between two mirrors, ok. These mirrors
can be physical mirrors or depending may be in the form of a grating whatever way that
you make them. But they essentially are the properties instead of a mirror and then they
have reflection coefficients R, ok. If R is close to unity then what would happen? As
the mode start to propagate it would move and then it would get fed back here almost
all of the optical mode would be fed back which would then pass through the SOA material
in between or the gain material in between and induce more stimulated emission, and then
when it comes back to the first medium it will again be reflected if R is very close
to 1. And this process continues to happen thereby
you will essentially have a nice laser. But that is not what you want what you want is
not a layer what you want is an amplifier therefore, you keep this R values to be quite
small, ok. So, most R values are just about point one and one asked one tries to make
this R value even less than that, ok but most materials unless they are specifically made
then they will have reflectivity of about 0.1 1 or flight are less than 0.1, ok.
So, that most of the light is not getting reflected most of the light is actually coming
out of their cavity in the sense that when you launch a probe light at a certain wavelength
then what you would get at the output is an amplified version with with some amount of
noise but almost faithful replication as that of the input, ok. So, this is what you would
have. Of course, how do you inject here? What is
the pumping mechanism? This is the injection current that you need to supply and this is
a PN junction therefore, you take this SOA and then forward bias this junction and then
essentially send in some input and you obtain the output. But because this SOA material
is kept in the cavity there are differences in the way the gain actually comes out, ok.
So, what we are now going to do is not to go back to the semiconductor basics, we are
already looked at that theory enough. Whatever we have said about lasers most of that also
applies to the semiconductor optical amplifiers also. But what I am interested is to look
at the amplifier gain as a function of frequency omega.
You first of all agree that this will be frequency dependent when it will be frequency dependent
because you have to agree it because the SOA material itself is a frequency dependent material.
Remember all lasers have a certain line shape, this line shape will be either in the form
of a or in the form of a Gaussian line shape. This is of course, a function of frequency
omega. You do not have any lasers which generate or any gain medium which has a delta like
spectrum. So, most of them actually have a certain variation and you would expect that
the gain of this FPA also to be something like this with respect to omega with a maximum
gain which we will call as this is FPA max and a minimum gain that would occur at that
tail which we would call as G min FPA this is what you would expect. So, this is what
I would actually expect from this one. But what in reality you get is a different
thing. So, you would actually get something like this. So, you notice that you have an
overall Gaussian type function, I did not write the other difference currently but essentially
this is a Gaussian type functions, ok. But there are these regions where the gain is
at a minimum level. So, there are these blanks in between and there are these peaks, multiple
peaks within the gain spectrum of the amplifier material this is a characteristic of the cavity.
So, if you assume that the cavity has a certain length L, then the spacing between these two
peaks is about c by nL or maybe c by 2 nL, ok. We can fit a half length or something.
And the idea is essentially the same. So, you have a cavity it would serve to have standing
modes rather than the traveling modes they are actually the standing modes and only certain
standing modes that satisfy by the boundary conditions will be sustained by the cavity.
And because you have now have some sort of a periodicity, right periodicity in the form
of wave going here coming back and then moving again in the same direction.
So, one round trip propagation can be considered to be a spatial thing, ok and because periodicity
is there in one domain there actually discreetness in the other domain and this is preferably
the discreteness that you are actually seeing. And this spacing between the gain medium is
called as free spectral range, ok, to increase the free spectral range, to increase the spacing
between the peaks you want to make n to be as large as possible. Of course, there are
practical limitations as to how long you can make these devices therefore, you have to
live with the fact that the gain spectrum is discrete and actually shows up multiple
peaks, ok. Of course, within these peaks the peak themselves are reasonably broad so that
most of the wavelengths can be fit into it. But not every wavelength in the operating
fiber optic communication system can be fit into one of those two peaks.
If you ah are interested in the expression for what would be that gain ah as a function
of frequency. You may actually see that this will be given by G s of omega, where G s of
omega is the line shape function of the basic lasing material whatever the ah active gain
material that you have used times 1 minus R G s cos omega square plus 4 R G s of omega
sin squared omega minus omega 0 into f by v, ok, where v of course, is given by c by
n is the phase velocity, ok. Because sin square is an oscillatory function the overall gain
will be an oscillatory function. Omega 0 is called as the design frequency it is the frequency
at which the gain will be maximum, ok. So, if you are making a fabricate in a semiconductor
optical amplifier and you want to maximize the gain at a particular wavelength you choose
that wavelength to be omega 0 or corresponding to the frequency omega 0 and you get a maximum
gain. What is the maximum gain? When when omega is equal to omega naught you have G
s omega naught 1 minus R whole square divided by 1 minus R G s omega naught squared because
you will get a maximum gain when the sin square term goes to 0 of their sin does not. This
term could have gone to negative but because it is a sin square term it will not go to
negative, and G s of course, is always a positive quantity because this is a gain of the basic
material or sometimes also called as single pass gain, ok. That is the gain that the optical
medium will undergo when it passes through the semiconductor optical amplifier of length
L, ok. So, this is what you would get. And if you now look at different values of
R, ok you can then change the different I mean you will get different values of gain,
ok. That is something that you can easily observe from this equation which I do not
bother you too much but notice that the gain can actually go all the way to infinity, right.
So, it I can go to infinity when R G f omega naught at that particular frequency approaches
1, right. So, when the denominator goes off to 0 when the gain goes to infinity and essentially
you would have obtained instead of an amplifier, you would have obtained a laser therefore,
this condition is not suitable for us and R G s would always be kept less than one inside
it should be kept much much less than 1, that is the first observation.
The second observation is if you look at the ratio of the maximum gain to the minimum gain,
right. I leave you to find out what is the minimum value of the gain or at least show
that this minimum value of the gain is given by G s 1 minus R square divided by 1 minus
R G s whole square plus 4 R G s and it will happen at those frequencies where this sin
square term goes of to 1, ok. So, when the sin square term goes to 1, you can find out
what frequencies that that condition hold and for those conditions you will actually
have a minimum value of G or the gain. The maximum value anyway we have written it
is G s 1 minus R square in the numerator divided by 1 minus R G s in the denominator. So, what
you can see that you can cancel off these two and you get 1 minus R G s whole square
plus 4 R G s divided by 1 minus R G s whole square, ok. You can of course, take this 1
minus R G s whole square as a common thing and then call this one as delta G, delta G
stands for the gain rippled, ok. Gain ripple means ah certain frequencies or certain bands
will have a larger gain and certain bands will have a minimum gain.
So, what is the difference between the max and the min gain? In terms of usually in terms
of the ratio would be about 1 minus R G s whole square 1 plus 4 R G s divided by on1e
minus R G s whole square, ok. So, you can if you wish to simplify this further and you
can calculate certain numbers which are actually going to be interesting.
Suppose I take G s equal to 3, and I take R equals 0.32, ok. So, what would be the product
R G s? This would definitely be less than one because the product is actually about
point. So, this is 3 times, right. So, 3 2 is a 6, if it is the 9, so you are very close
but if you are not exactly close. But if you look at what is a gain in you know
the maximum gain for this particular case the maximum gain will actually be about 867
in the linear scale or about 30 dB approximately in the degree scale, ok this is a huge value.
Whereas, the minimum gain for the same condition would actually be even less than 1, in fact
it would be a loss if you choose this particular value and, in fact it would be less than about
0.36, right. If you look at the difference between these
two which would of course, be less than 0 dB the gain ripple is about 30 dB. So, if
you have signals of two different wavelengths. So, or two different frequencies omega 1 and
omega 2, if omega 1 and omega 2 happen to fall into this unfortunate situation that
this is your omega 1 and this one is or rather this one is omega 2, ok. So, this is omega
2 then you can see that omega 1 channel will experience an extremely large gain omega 2
channel will experience almost no gain or, in fact if you experience for loss and when
you had initially same power but after passing through the SOA now you have a large power
here and almost no power in omega 2. So, therefore, this is not a very good mechanism
or this is not very good value to choose because you are then essentially creating unequal
gain distribution in the multi channel system which is not at all good, because you expect
all the channels to be amplified with the same level not with different levels. How
do I you know ensure that that is happening? So, the way to ensure that one would be to
further reduce R G s consider taking the single path gain to be G s equal to 2 while keeping
R to be the same 0.32. In this case the product is about 0.64 which is less than 1, ok.
The difference is not too much you would think you have 0.96 in one case, of 0.64 in the
other case but it turns out that the maximum gain again you have to verify that these numbers
are correct, ok. Please do that one the maximum gain is about 8.5 dB, ok. So, this is 8.5
dB the minimum gain is still quite less than 1. So, the minimum gain is just about 0.32,
but now you see that the gain ripple is down to just about 8 dB. So, this is the ripple
that you actually get which is much much smaller than this one. So, this is actually very important.
But if you can reduce R further which is what normally we do you can keep the single pulse
gain high but reduce R further which is entirely possible by the technology that we have today,
ok. So, what you would essentially convert or what you would essentially do is to minimize
this gain ripple, ok. So, but still the gain ripple would be present if you were to use
an FP cavity type filters, ok. We are also interested in the bandwidth of
these ah um devices. So, if you actually look at the bandwidth you would you can actually
find out what would be the bandwidth of this one. But the bandwidth actually depends on
how the wave shape or the line shape itself goes or the single pass gain itself goes,
and the single pass gain itself is given by E to the power or exponential ah of gamma,
g of omega comma P minus alpha times L. Where, alpha corresponds to the resonators or the
reflection losses that would happen in the cavity. Gamma corresponds to confinement which
is kind of an overlap between the optical mode and the cavity design itself and most
importantly this is the gain medium frequency dependence. So, this is where you are actually
getting all of your gain differences. Of course, for this to actually be positive
gamma g of omega comma P at every frequency and for every power must be greater than alpha
so that is overall we will have gain and not absorption or loss, ok. So, these conditions
are to be satisfied for any type of op-amplify for a gain. But this is what you would actually
have. And further this g of omega comma P itself
is given by a frequency independent gain function, and divided by this term in the denominator.
And this term in the denominator where we have written P sat we called the saturation
power. At the saturation power what will happen? For P much much less than P sat the denominator
can be approximated to 1 and the gain would be essentially independent of the power, ok.
So, and it would have some sort of a variation ah that is dependent on the gain shape function
G of omega itself. But if you were to plot the gain as a function
overall gain let us say as a function of the input power or the probe power or or the input
channel power then for P is less than saturation landed a gain is almost 1, but ah at this
point actually starts to drop and then eventually drops off for that. So, the value of the gain
at the saturation value then P equals P sat is just about G by 2, ok. So that would essentially
correspond to the bandwidth in terms of the saturation power. It is kind of quite difficult
to obtain you know proper equation or an understanding between gain and P because it is a complicated
ah process that takes place. But approximate equations have been derived,
so approximately G s depends on the input power in this particular manner. It is a transcendental
equation, so you will have to solve this equation, ah um numerically most of the times otherwise
you will not be able to find because there is G s on the, right hand side G s on the
left hand side as well, ok. Again please note that there are many competing
processes that are happening in the semiconductor optical amplifier, the actual gain depends
on cavity designed, reflectivity if you know ah um the the tray light that the enters into
it, the packaging all those things have to be optimized if you want to create a good
amplifier. And of course, you have to also do some sort of a optimization if you want
to create good lasers, ok. So, this was about ah semiconductor optical amplifiers.
The last type of semiconductor optical amplifier that I am going to look at is what is called
as traveling wave amplifier, ok. The principles of traveling wave amplifier are slightly difficult
and different and we do not want to go into that, but the basic idea is that you do away
with the reflections you know R will not be there R, in fact will be approximately 0.
In fact, you cannot get R approximately 0, so what you would actually have is a you know
a range of values of R say from ideal 0 value to about 0.1 let us say ah wherein you would
treat the commitment or optical amplifier as traveling wave amplifier and beyond that
we would treat them as Fabry-Perot amplifier, ok. So, beyond the value of R of say 0.1,
0.12, 0.15 you would build them as a Fabry-Perot amplifier.
The gain of the traveling wave amplifier is actually much smaller compared to the Fabry-Perot
design because there are no cavities here. Cavities serve to have ah um standing modes
and therefore, the gain can actually shoot up. So, you are trading off a large gain,
ok in ah your ah Fabry-Perot cavity, so you will have a larger gain here at the appropriate
frequencies, and if this is the only frequency of interest that you are working with then
there is no problem. You can easily have multiple channels here between that band and then you
almost all of them will have equal amplification and you are all very good and very nice, ok.
But if your number of channels actually go around this region then you are in trouble,
because some of the channels will see lesser gain. The traveling wave amplifier overcomes
this problem it gives you almost kind of equal gain frequency dependence. So, this is the
traveling wave amplifier, but the gain peak value is actually quite small, ok. So, you
have to of course, improve the single pass gain because that is exactly equal to the
gain of a traveling wave amplifier by continuously changing g increasing lamb gamma increasing
g and then reducing this alpha and increasing L. So, this has to be increasing, this has
to be increasing, loss has to be reducing and L has to be increasing all with certain
constraints and approximations, ok. Even the bandwidth of a traveling wave amplifier
if you were to look at it will actually be much wider compared to that of the ah Fabry-Perot
cavity, ok. So, this is the traveling wave amplifier and this is the overall gain that
we talk about to the saturation value for the traveling wave amplifier is usually much
larger than that of the Fabry-Perot cavity. The big advantage of course is that traveling
wave amplifier will never lays because there is very very little feedback that the amplifier
is sustaining. On the other hand Fabry-Perot amplifiers can be made to lays you know even
externally by actually putting the entire SOA in the middle of a certain cavity and
a bigger cavity and then increasing the feedback of that external cavity. So, SOAs are also
used for lasers not a very good way but it is also used for lasers. And most importantly
SOAs offer high non-linearity much higher than the non-linear fiber non-linearity and
therefore, they are actually excellent for many non-linear optical signal processing.
For example, you can send in two different ah frequencies into the SOA and under, right
conditions you actually end up having 4 different frequencies, ok. So, you can actually create
multiple frequencies by employing the non-linearity. You can also clean up your input signal by
what is called as a phase conjugation, you can improve the data rate by actually putting
in a SOA based which conjugating the middle of the fiber optic communication link in the
form of what is called as a phase conjugation. So, all these advantages that you would get
are from the ah semiconductor optical amplifier. The down side of semiconductor optical amplifier
is that there is a large gain fluctuation, because the lifetime of ah electrons in the
conduction band is about 1 nanosecond any bandwidth which is greater than about 1 or
any bandwidth which is less than 1 Giga Hertz will experience severe losses, right. Because
if you have lesser than 1 nanosecond then you will ah or if you have a larger this one
then the smaller lifetime will actually cause the gain pulsations to occur.
So, compared to an erbium doped fiber amplifier the gain pulsations or the gain is very noisy
from a semiconductor optical amplifier. And therefore, for this reason it is not used
in optical amplification but it is widely used in other non-linear signal processing
applications that I just mentioned, ok. There are other differences between semiconductor
optical amplifier and erbium doped fiber amplifier. We will not get into these differences at
this point and we close this by saying; Thank you.
