Tip:
Highlight text to annotate it
X
Good afternoon.
Today I'm going to be talking about using
off-the-shelf multimode fibre couplers
in combination with
a spatial light modulator (SLM)
to perform all optical Mode Division Multiplexing (MDM)
and we'll be doing that over
2km of standard graded-index OM2 grade fibre.
Although I'm going to be doing optical MDM,
I think the wider point to take away from this talk is actually
about how you can use a Spatial Light Modulator to modally
characterise components.
So the system starts with two
Gaussian beams, each modulated with its own channel of information
one of those beams passes straight through the beamsplitter and goes on
to excite the fundamental mode of the fibre.
The other beam
reflects off the SLM
which imprints a new phasefront onto that beam
which will become the desired higher-order mode
at the waist
of the refocusing lens.
You couple both of those into the same fibre
and at the receiving end
use a
standard off-the-shelf
multimode fiber coupler
to demultiplex them
but first I want to
talk about the phase masks
used on the Spatial Light Modulator
In most MDM experiments thus far people have used
few-mode fibre, and I've taken the case here
of a 6 mode fibre which supports
LP(0,1), LP(1,1), LP(0,2) and LP(2,1)
and if you wish to excite
a particular mode in the fibre. Typically people have
used phase masks which take the form
of the desired mode
and they illuminate that with a Gaussian
and in the case of an LP(2,1) mode we can see the far-field that produces
in the middle right there
and you can see it's not exactly an LP(2,1) mode but it's fairly close
and when you couple that into
the few-mode fibre, the spatial filtering properties of the fibre itself
clean that up for you and you get a nice
pure LP(2,1) mode excitation.
As seen on the right there.
Problem is,
as you move to
fibres that support more and more modes
the fibre becomes far less forgiving, so if you try that same technique
on say a 45 mode fibre
then you can see on the right there, that although you are still coupling mostly
the LP(2,1)
there's other modes being excited as well.
And that's because as you increase the number of modes
the fibre
performs less spatial filtering
and the onus for generating a really high quality mode is shifted
back into the plane of the phase mask. You can't rely on the fibre itself
to clean it up for you
the phase masks I use are a little bit different and they use techniques from
computational holography
to generate (fields) which more accurately
(match) the
desired modal profile.
So as you can see in the middle there. The phase mask still has the same base 4 lobe
structure to it
but it's a bit more complicated
but generates a mode which is
much more accurate
and when you couple that into the fibre you get a nice pure LP(2,1) mode.
and the trick is
this more complicated mask
still has noise power. There's still
light that it can't convert into the LP(2,1) mode just as there is for the
simple mask
but whereas the simple mask tends to dump
all that noise power directly onto the core of the fibre, exciting
unwanted modes
this algorithm scatters
the noise power away from the core of the fibre
so that it's not going to
excite any undesired modes.
and this is what the
phase masks look like
on the left there
that consists of the
binary phase pixels of the SLM
illuminated with a Gaussian beam
which then produces the
far-field
on the right observed on a camera.
As I said, I'm going to be using a multimode fibre coupler to demux
the modes
and to start with, I'm justing going to plug the coupler directly into the
SLM based mode launch system. So now transmission fibre,
just launching modes directly into the coupler. And doing that you can
characterise the modal properties
of the fibre coupler.
So here for instance we have
launched LP(0,1) mode into the coupler
and we see that almost all the power comes out of the primary port and you can't see anything
coming out of the secondary port
and when you launch the LP(1,1) mode you can see that most of the power still exits
via the primary port
but there's a rotational
variation to it
and as you head to higher and higher-order modes
that coupling ratio between the port becomes more 50/50
but there's still a rotational
variation.
So what I'm going to do is
connect a single-mode fibre to the primary port such that the LP(0,1) mode
is still free to leave via that port
and all other modes will be extinguished
and of course the higher-order modes
are still able to leave by the second port
So you get a very simple demux
whereby LP(0,1) leaves via the primary port and all higher-order modes leave
via the secondary port.
So to put some actual number on that,
I've taken,
three different couplers from the same manufacturer
rated at 50/50, 70/30 and 90/10
nominal splitting ratios
but obviously that splitting ratio actually depends on the
modes propagating in the fibre
and what I've done is, launched each mode one at a time at its optimal
orientation
along x-axis
and measured the coupling out of the two ports
along the y-axis
for the primary port with the single-mode fibre attached
we can see that all three couplers perform
pretty similarly and that's because the modal properties are dominated
by the single-mode fibre attached to the coupler rather than the coupler itself.
What's more interesting is the
seconadary port,
which acts as a kind of modal high-pass filter
where the modal roll-off is defined by the nominal splitting ratio
So you can see the 50/50 coupler
couples anything into the secondary port that's higher than about the second mode-group.
The 70/30 will couple anything higher than about the fourth mode-group
and the 90/10 will couple anything higher than about the
6 or 7th mode-group
So the two ports together: our primary port acts like a modal low-pass filter,
the secondary port acts like a modal high-pass filter. Taken together,
you've got a kind of modal diplexer
with a modal guard band between the two ports which consists of modes too lower-order
to couple to the secondary port
but too higher-order to couple to the primary port.
So if you send modal channels either side of that modal guard band,
that modal guard band will serve to keep them isolated
even in the presence of some modal mixing.
Subtract the coupling between the two ports and you can get the isolation for a two-channel system.
You can see that all the higher-order modes perform
fairly similarly, quite well
and that's because the single-mode fibre is very effective at stripping off the higher-order modes
the constraint on performance is really
the fundamental mode
so like how well you can pass
the fundamental mode through the coupler
without coupling it to the secondary port
and as you can see with the 50/50 coupler, there's about 12dB of isolation, for the
70/30 coupler there's about 27dB
and for the
90/10 there's about 31dB
now as I said this is all in a back-to-back configuration, where the modes are
being launched directly
into the multimode
fibre coupler.
Now if you attach an actual transmission fibre in between
everything's going to get a lot more statistical and it's going to start to vary with time.
But luckily, using a Spatial Light Modulator it's possible to track
that modal evoluation in time.
For instance,
here we have a 50/50 coupler at the end of a 2km length of OM2 fibre
and you can see that by repeatedly launching the mode at different orientations
anti the by repeatedly launching by the motive different orientations
you can track
the optimal orientation
with time
It actually varies fairly slowly in a 2km length of fibre,
this video is playing back at about 15 minutes per second
So a more rigorous investigation is this one here.
Where we've got each of the three fibre couplers
and across the top
it's measuring the modal isolation between the two ports for each of the modes
at 16 different orientations
so it's repeatedly launching about 325 phase masks
which correspond with the 25 LP modes
at 16 different orientations.
At the bottom, we're building up a histogram
for each of the couplers, for each mode at its optimal orientation
at any point in time
now there's a lot
of information that gets collected in sweeps like these
and I think that's really the wider point to take away from this.
I'll be performing all-optical MDM
in a few slides
but even if for a real MDM system you want to use
fixed phase masks or spots
or integrated devices, or whatever you choose
it's hard to beat a Spatial Light Modulator in the context of R&D for experimenting and
characterising and
measuring modal properties.
So, I'm going to perform optical MDM
we've got 3 couplers to choose from here,
the 50/50 coupler
only has about 7dB of isolation for the fundamental mode
and that value's quite variable
so that's not particularly useful to us.
The 70/30
looks pretty good, it's got about 19dB of isolation.
The 90/10 looks even better it's got about 27dB of isolation.
Unfortunately,
it has that isolation due to the very large modal guard band and because of that very
large modal guard band
you can only use the very highest-order modes
for transmitting a channel.
Unfortunately,
those higher-order modes have a lot of modal mixing and a lot of
modal dispersion.
Which makes them unusable in the experiment that follows as I'm using
direct detection.
So we can see here, what I've done is I've launched each mode of the fibre in
one at a time at it's optimal orientation
and measured its impulse response
Which is yet another handy thing you can do with an SLM
And you can see that as you head to higher and higher-order modes
the impulse response becomes more diffuse
and in the frequency domain you can see that it starts to roll-off as well.
So that's not much good.
So basically it comes down to the 70/30 coupler which is a good
compromise nice isolation of the fundamental mode
between
and a wide choice of
modes to use as a second channel
and as it turns out, the best mode to use for the second channel is something in
the 4th mode-group, so the LP(1,2) or the LP(3,1)
which has good impulse response
and isolation of about 23dB
So that's what I'm going to do.
Here I've got the input 12.5Gbps NRZ signal
and if you just do an overfill launch (OFL)
which is like a standard way of characterising
the bandwidth of a multimode fibre,
you get something like this
Obviously a complete disaster, because the fibre's only got
about 940MHz worth of modal bandwidth.
However if you do MDM and look at the signal out of the fundamental port
you get something like that
and if you look out the secondary port where we're propagating on the fourth
mode-group
you get the signal like this.
and that's it. Thank you very much.