April 10, 2024 • 55 mins
Unlock the mysteries behind the cables and connections that power our digital lives with Nokia Bell Labs' Xi (Vivian) Chen and Fatima Gunning, from Tyndall National Institute. This episode promises a fascinating exploration into the world of high-speed optical fiber communications—a vital yet often overlooked cornerstone of modern technology. As we journey from the switch from copper to the latest in fiber technology, you'll discover the incredible engineering feats that allow you to stream, browse, and connect at the speed of light.
Join us as the experts also discuss topics like the transition from materials like lithium niobate to the realms of silicon photonics and indium phosphide—all pivotal in our quest for faster data transmission. Further conversations are had on the challenges of adapting these technologies to new fiber types and how innovations like multicore and few-mode fibers could redefine the landscape of global connectivity.
However, the chat isn't just about bits and bytes; reflect on the values that underpin successful leadership in science and the fresh perspectives brought in by emerging talent. By the end of this episode, you'll be looking at the tangle of cords behind your desk with a newfound reverence for the photons racing through them and the people behind the innovation!
Host:
Thierry Lapinte-Leclerc
PhD Student
Boston University, USA
Moderator:
Fatima Garcia-Gunning
Senior Staff Researcher
Tyndall National Institute, Ireland
Expert:
Xi (Vivian) Chen
Department Head, Optoelectronic Sub-systems
Nokia Bell Labs, USA
Have a topic you're interested in hearing about? Let us know!
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:02):
Illuminated by IEEE.
Photonics is a podcast seriesthat shines light on the hot
topics in Photonics and thesubject matter experts advancing
technology forward.
Speaker 2 (00:14):
Hello everyone, welcome to today's episode of
Illuminated, an IEEE Photonicspodcast.
My name is Thierry LaPointe-Leclère and, as the IEEE
Photonics Society EmergingTechnology Task Force
representative for the YoungProfessionals Program, it's my
pleasure to be your host today,I guess.
Before getting any further, Iwould like to introduce myself,
(00:34):
because all of you will beunfamiliar with who I am not
that I blame you so I basicallyam a first year PhD student at
Boston University, working inProfessor Siddharth's or
Mashantran group, who specializeat the moment in optical
orbital angular momentum beams.
I have a master's degree inphotonics from Laval University
(00:58):
and a bachelor's degree inengineering physics from the
same university.
I think it's fair at this pointto say that I enjoy the field
of optics in Photonics quite abit, because I've spent already
at least like seven yearsworking in various groups and on
various projects in Photonicsand I'm absolutely delighted to
be hosting this episode ofIlluminated.
(01:18):
So on to today's podcast.
This one is particularlyinteresting because we're
looking at how we keep findingnew ways of sending more
information and at faster ratesin optical systems.
In particular, we're going tobe discussing high speed optical
system and high throughputfibers, and our expert yesterday
is Viviane Cheng from Bell Labsand our moderator is Dr Fatima
(01:43):
Gunning from Tyndall NationalInstitute.
We will go into their technicalwork, career journeys and
advice for young professionals.
Well, thank you for listeningto this episode.
I think it's going to be a goodone.
First of all, I'd like tointroduce Fatima.
I'll give a brief bio.
So Fatima received a PhD inOpto-electronics in 2000 from
(02:07):
Pontifica University at Catalicade Rio de Janeiro I hope I said
that correctly Investigatingfundamental optical properties
of glasses, chemical fibers andplanar waveguides transformed by
a technique called thermalpolling.
As a postgraduate researcherstudent, or as a postgraduate
research student, she visitedBritish Telecom Research
(02:29):
Laboratories in the UK andjoined Corning Research Centre
at Adelstral Park, also in theUK, in 2001 to work on advanced
electroabsorption based opticalmodulators for applications in
high speed optical transmissionnetworks.
She co-founded the PhotonicsSystem Group at Tyndall National
Institute in Cork, ireland, in2003.
(02:50):
There are research focuses onspectrally efficient techniques
for high speed opticaltransmission, including all
optical OFDM, coremwdm andfrequency comb generation,
expanding communications to newwavelength windows, namely to
micron, and with fibers such asOOCOR fibers.
She actively works closely withPhotonics Devices integration
(03:15):
and packaging teams to enablenext generation components for
communications.
She was at IEEE, so her rolewithin IEEE Photonics Society
over the years have involvedbeing a VP of Multicultural
Outreach from 2018 to 2019 andVP of Membership Outreach for
the IEEE Photonics Society from2020 to 2022.
(03:38):
And even though right now Idon't see a position listed, an
official position, she's stillvery active within the society,
being a volunteer throughmentorship, conference
committees and co-hosting events, so I think she's probably
quite busy.
Speaker 3 (03:56):
So thanks, Therese, for the introduction.
It's a delight to be here andit's in delight to introduce our
guest speaker today, which isVivian Cheng.
Vivian Cheng is a distinguishedmember of technical staff and
department head ofOptoelectronic subsystems group
Anoka Bell Labs.
She received her PhD degree onoptical fiber communications in
2012 from Universal Melbourne inAustralia and from 2013 to 2015
(04:20):
, she was a research fellow atthe Universal Melbourne
conducting research on opticalfiber transmission and since
2015, she's now she has been inBell Labs since then with
current research interestsincluding advanced digital
signal processing and high-speedfiber transmission.
She was a awardee of the YoungInvestigators Award in 2020 for
Atropical Electronics Societyand she's a senior member at
(04:42):
Tripoli, a fellow of Optica, andshe published more than 150
papers in the era of high-speedoptical fiber communications
well over 60,000 citations.
So it's super exciting to haveyou here, Vivian.
Welcome to this podcast.
Speaker 1 (04:57):
Thank you, thank you Fatima, thank you Teri.
It's great to be here, sohopefully today we talk about
what we're doing in terms offiber communication and
high-speed optical transceiversand share some of our thoughts
and discussions, and hopefullythat would be interesting.
Speaker 3 (05:16):
Yeah, that's great, and then it's very interesting,
I think, about a couple of yearsago at the UFC.
We talk about 50 years of usingoptical fibers and so on, but
it took a long time for peopleto move from copper to optical
fiber.
So what is it that you're doingin terms of what is it that?
Optical fiber?
Sorry, let me start again.
(05:36):
Sorry, let's start again.
I said something else and Ijust shouldn't have said so.
We heard a couple of years agoback in the UFC the 50 years of
optical fibers.
It was super exciting to seethat happen.
It took a long time for peopleto really install fiber on the
ground and people have beenmoving from copper to fiber,
(05:58):
maybe started with submarinesystems and long-haul systems,
but now even coming closer topeople on access and to their
homes.
So where do we see opticalfiber installations and what is
it?
Optical fiber transmission?
What are we talking about?
Speaker 1 (06:17):
Okay, that's a good question.
I don't think you necessarilysee optical fibers in daily life
, although everyone of us mostof us using them all the time.
So they are basically buried inthe ground.
If we're talking aboutterrestrial links, that's the
technical term we use.
(06:37):
So they are buried a few metersdown to the ground or they are
literally sunk into the bottomof the ocean.
It's not like wirelesscommunication, like your phone.
You will see the base, towers,a tennis around you.
You don't really get to seethose, but I would say 99% of
(06:57):
the distance your signal istransmitting through the media
is not really air, it's fiber.
And so, for example, if youmake a phone call from the US to
Europe, china, india, any othercountry, it's only the first
maybe 10 centimeters that yoursignal transmitted from your
(07:17):
phone.
That is through air, which iswireless, and then, starting
from the point that antenna gotyour signal, everything most
likely is only fiber.
And then it gets to the otherend and another antenna is
transmitting your signal toanother phone and that's another
very little part that is notfiber, and everything in between
is most likely all fiber andthey are in the ground or at the
(07:42):
bottom of the ocean.
Speaker 3 (07:46):
I see, and why do we need research?
Then?
Optical fibers, what is it?
The challenges associated withinstalling the fiber on the
ground or making thosetransmissions and converting
between wireless and optical, Imean what are the challenges?
Here, yeah.
Speaker 1 (08:01):
So the challenges, as we can probably imagine, the
challenge is we need more speed,we need more data and we need
more bandwidth.
And if you think about it 20years ago, what kind of phones
we have and what kind ofapplications we're able to do
(08:22):
with our computers and phones,and that amount of bandwidth
compared to the bandwidth we'reconsuming today, that's the
factor of depending on theapplication it could be a few
hundred times or a few thousandtimes increase.
So that's a huge change interms of what we demand in terms
of bandwidth.
And so if we take thatperspective and think about the
(08:45):
future, in the next 10, 15, 20years, what we would need, if
you don't think about the pastand just standing now and try to
imagine what we might need andyou tell people, oh, I think we
need five thousand more times ofbandwidth for the next 15 years
, people will probably see thismight be crazy because how can
(09:08):
we need so much more?
We're pretty good in terms ofbandwidth and speed, but if you
think about the past and askpeople, if you imagine you ask
people on the street 15 yearsago, you think you need this
much bandwidth today and theyprobably think you're crazy.
So I always thought, even forpeople who work in technology,
(09:34):
even in the field oftelecommunication, they might
underestimate how much bandwidthwe may need in the future, and
the data sets so too.
If you look at the data,there's solid data saying every
year we are demanding about 50to 60% more data.
So that's the annual growth of50 to 60%, and that's a lot.
(09:58):
If you take that number and doa 10 years projection, you are
looking at about 100 times morebandwidth.
So that's the challenge we needto scale our electronics and
optics fast enough to satisfythe demand, and so what we want
(10:19):
to do is higher bandwidth, morethroughput at a reasonable price
, and we don't use all theenergy in this world, so power
consumption is also veryimportant.
So that's the main challenges.
Speaker 3 (10:37):
So how can be done?
So your research says thatyou're working with transceivers
and you're talking aboutincreasing bandwidth, make
higher speeds using less energy.
So what is it that is consumingthis energy and how does this
transmission work?
Yeah, okay.
Speaker 1 (10:53):
So let's start with.
What do we mean by bandwidth?
There are two bandwidths thatare relevant here.
One is the bandwidth of thefiber itself.
So if we take a step back thefiber transmission system on the
high level, you only have threethings.
You have a transmitter that issending the light, that carries
(11:16):
information, and then you have afiber link that's connecting
your two points and then youhave a receiver.
So when we talk about thebandwidth of the fiber, that's
about four terahors bandwidth,or roughly 35 nanometers
bandwidth.
That's a lot of bandwidth, andthis whole fiber typically
(11:38):
nowadays depends on theapplication and also distance.
This fiber can carry aboutanywhere between 20 to 40
terabits per second.
So one terabit is 1000 times of.
Say, if you have one gigabitper second for your home, this
is one terabit is 1000 times ofthat, and 20 to 40 is 20,000 to
(12:03):
40,000 times that.
So this is what a typical fibercan carry.
And then you have yourtransmitter and the receiver and
you have the bandwidth or thethroughput, or sometimes we call
interface range, of yourtransmitter and the receiver,
and that's a much smallerbandwidth.
So typically you have anywherebetween 50 to 100 gigahertz
(12:27):
bandwidth, depending on whetheryou're talking about product or
research, and that gives youabout one terabit or two terabit
per second data rate.
So you typically need anywherefrom 30 to 40 of them to fulfill
the whole bandwidth of thefiber.
And so when we talk aboutincreasing the bandwidth, for
(12:53):
the part I am working on, Imostly work on increasing the
data read or the bandwidth ofthe transmitter and receiver,
and after you increase that, youstill need multiple of them to
fill the bandwidth of the fiber.
So part of our research is onhow to develop better
transmitters and receivers thatcarries more information per
(13:14):
transmitter.
And then some other researchgroups or my colleagues or other
people, they are working on howto develop better fibers, a
fiber that can carry morebandwidth.
So those are the two main partsand for my research, for the
transmitter part, back to thequestion how do we make
(13:37):
transmitters that carry moreinformation or more data right?
Then we need to dig a littlebit deeper on how transmitter
works.
The transmitter on a high levelonly have three things.
It starts with a laser.
So the laser we are using fortelecommunication is a laser.
(13:57):
We call CW laser, continuouswave laser.
So that means the laser issending out light constantly.
It's a constant power and thelight that is constant or any
electromagnetic wave that isconstant doesn't carry
information.
So, despite this, for peoplewho are not necessarily familiar
(14:18):
with telecommunication, imagineyour city light or you have a
flashlight at torch and you turnit on.
That light is always on and itdoesn't carry information.
However, if you start to dimthe light or flash the light
according to, for example, amask code, then the light starts
(14:41):
to carry information and theprocess you use your hand to
change the on and off status ofthe light.
This process is calledmodulation.
So the same thing happens in anoptical transmitter.
You start with a laser and thatlight doesn't carry information
(15:01):
, and you send the laser into adevice that is called a
modulator.
So this modulator takes toinput.
It takes the laser, it alsotakes the electrical signal and
this electrical signal is yourdata or your information, or you
can say that's your zeroes andones.
So the modulator starts tomanipulate the light properties,
(15:24):
for example power of the light,according to the electrical
signal that is added to themodulator.
And then the output of themodulator is a.
You can see that as a lightthat is being manipulated or
dimmed or flashed to some extent, and that light starts to carry
(15:46):
information.
So that's how the transmitterworks.
And in order to make atransmitter that is fast, you
need basically two things.
That is fast it's not reallythe laser.
The laser is sending outconstant light and the laser
we're making today, especiallythe laser we make for high end
optical transmitters they aregood enough in terms of the
(16:10):
performance.
We are more working on how toreduce the power consumption,
how to reduce the form factorand all this, but not the
performance itself.
In order to make a fastertransmitter you don't
necessarily need to do much onthe laser itself.
You need two other things thatworks fast, that is, your
electrical signal generation andthe modulator.
(16:33):
So a lot of research andinnovation are done on this two
part to push the speed of thistwo.
For example, for the electricalsignal generation, for decades
we were relying on the CMOStechnologies, and when the CMOS
node gets smaller we get fastergates and we have faster
(16:54):
electrical signal generation.
That's how we have fasterelectrical signal to drive the
modulator.
But the speed that CMOS isscaling is clearly not quite
enough for the speed we need forthe transmitter itself to scale
.
So some of the researchersstart to do something outside
(17:16):
the CMOS.
They take other materials, forexample, in a phosphate or
silicon germanium, they combinemultiple CMOS signals to form a
higher speed one so that you canyou will not only be relying on
the CMOS scaling speed and youcan do double or triple speed
(17:42):
compared to the CMOS speed, andthat's one of the ways to do
higher speed.
Electrical signal generation.
And then, in terms of modulator,we are mostly relying on
materials.
For a long time, starting from,we widely deploy optical
(18:03):
transmission links, especiallycoherent transmission links.
We were using a material calledbulky lismai bit material for
modulator and that's anexcellent material.
The only problem is is not fastenough for today's modulator.
So nowadays, in the past maybefive to seven years, there has
(18:25):
been a lot of innovations onmaking new materials.
That is faster and also muchsmaller informed factor to make
modulators.
And at the moment what productis using very widely is either
silicon, photonics or in aphosphate and they are faster
(18:46):
speed, so they make higherbandwidth modulator.
Therefore you can, you can makehigher speed transmitted out of
it.
And there are also a lot ofvery promising material in
research domain.
For example, one of them wouldbe some families and I bet they
are kind of they're similar totraditional is not admit, but
(19:09):
much faster speed and muchsmaller in terms of form factor.
There are other very innovativematerials like organic
materials or post money and allthis.
They're all under research.
But for modulator itself mostlywe rely on the material
innovation.
That's the two main area peopleare doing very active research
(19:35):
on to try to bring up the speedof an optical transmitter.
Speaker 3 (19:40):
That's very interesting.
That's on the speed, but theother thing you're talking about
is throughput.
How can we increase the overallfiber throughput using those
new devices that you're talkingabout?
Speaker 1 (19:53):
So throughput of the fiber doesn't necessarily get
increased because thetransmitters are faster.
So, as I said, fiber has acertain bandwidth, so the
transmitter has a much slowerbandwidth, smaller bandwidth
than that.
So we need multiple.
The reason why we push eachtransmitter to higher speed is
(20:17):
we want to use less number oftransmitters.
That's better, because laser isvery expensive.
Every transmitter comes withone laser, so at the moment the
laser is not cheap enough.
It's a significant cost to eachtransmitter.
So we try to use less lasers,which means each laser has to
carry more information.
That's the main motivation whywe're pushing the per
(20:41):
transmitter speed to as high aspossible.
But when it comes to thethroughput of the fiber itself,
the most widely deployed fibernow is called standard single
mode fibers and we're usingabout 35 nanometer of the fiber.
(21:01):
The fiber support more thanthat but amplifiers doesn't.
So we're limited by the opticalamplifier bandwidth and that's
what we call the conventionalband, the C band.
It's about four carats or 35nanometer.
In order to have more bandwidthfrom the fiber or more
throughput from the fiber, thereare several things you can
(21:23):
potentially do.
You can extend the bandwidth ofthe fiber itself because the
fiber does have a very wide lowloss window and you can try to
develop other amplifiers thatsupport more bandwidth.
That's one way to do it.
You might get a factor of two,three or even more out of that
(21:47):
and the system quickly getspretty complicated because every
band you have to use differentamplifiers.
So when an optical signal, avery broad optical signal, gets
to a point that it needs to beamplified and by the way that's
dependent on the transmissionlink that's anywhere between 50
kilometers to 100 kilometers.
So if you have a link that isbetween, say, europe and here,
(22:15):
then you need many, manyamplifiers.
I mean the summary links isanywhere between 10,000 to 8,000
range, so you need a lot ofamplifiers.
So every time you need toamplify a signal you have to
split them into different bandsand amplify them through
different amplifiers and combinethem.
(22:36):
So it is a less scalable way ifyou try to increase the bands
or the useful wavelengths out offiber.
The other way to do it is tojust add more fibers.
So you can either add morefiber in the way that we call
fiber bundle, so simply havemore fibers there, or you can
(22:59):
use more noble fibers.
For example, you can use multicore fibers, which means single
mode fiber has one cladding withone core in it.
You can have one cladding withmultiple core in it, for example
three or five or seven or evenmore.
Or you can use something calleda few mode fiber.
(23:20):
So the single mode fiber.
As the name suggests, itsupports one mode.
You can make fiber that hasmore than one mode, for example
three or six or 15.
And in that way you have moremodes, which means you can do
multiplexing by effect of numbermodes.
(23:42):
Or another very interestingtype of fiber is the fiber we
call holocore fiber, so it's airin the middle.
It naturally supports much morebandwidth, much more wavelength
range than single mode fiber.
And it is possible to developone amplifier that supports
(24:05):
wider bandwidth than single modefibers C-band amplifier, that's
also one way to do it.
And all these are very activeresearch domain in fiber
fabrication and also systemtransmission link as a whole,
and those also affect our futureprojection on how an optical
(24:25):
transceiver will look like.
Speaker 3 (24:28):
I see, and Teri, do you want to ask a question?
Speaker 2 (24:36):
I don't want to step on what you were going to ask,
but I did have one relatedquestion to what Vivian was
discussing.
Speaker 3 (24:42):
Yeah, you can ask a related question.
And then I was going to go andask something else.
I wanted to go into digitalsignal processing.
So if it's related, just askthe question.
Speaker 2 (24:51):
Okay, perfect.
So I guess what I wanted to askalso is now you brought up
different architecture forfibers, right?
So multi-core, multi-mode,holocore, and you said you've
been looking into possibly goingin those directions.
How does that?
How challenging is it to adaptyour optical transceiver design
(25:12):
for different types of fibers?
Speaker 1 (25:16):
That's a great question and principle, does not
?
You can okay, let me put thisway.
You can use the currenttransceiver for some of the new
fibers, for example.
If we are talking about fiberbundles, then basically you
don't need to change anything.
You just put more transmittersand receivers there.
(25:37):
They're parallel, they areindependent.
Or for multi-core fibers, thenit depends.
If the core couples to eachother, meaning there are
crosstalk or on a system levelwe may call it information
exchange among the cores, thenyour receiver should design in a
(26:02):
way that it can deal with thecrosstalk, and that's a typical
process we call the MIMOprocessing.
So you need to process all thesignals from all the cores at
the same time.
To redo the crosstalkMulti-core fiber, then sorry, a
few more fiber, then you do needto have a new set of signal
(26:24):
processing.
The hardware wouldn'tnecessarily change much, but
it's the digital signalprocessing party to whether you
need to deal with the couplingbetween different dimensions
that you're using now.
Speaker 3 (26:39):
It's interesting because that lists the way that
I was thinking on the digitalsignal processing side.
So I suppose not all signalsare perfect and there's lots of
impairments and, as you said,crosstalk that might occur.
So explain to us a little biton the work that is needed on
digital signal processing,perhaps not only the receiver
but sometimes even at thetransmitter as well.
Speaker 1 (27:03):
There are a few important pieces in terms of
digital signal processing.
So all we talk about at themoment we are talking about
high-end optical transceivers.
That means the transceivers aredeveloped for long-haul
distance transmission and thatmeans we are using, we are using
(27:28):
, we're trying to use thebandwidth very efficiently, and
that means the transmitterusually have one important
digital signal processing thatis called pulse shaping.
The name sounds very confusing,but basically what it does is
it tries to squeeze theinformation into a bandwidth
(27:48):
that is theoretically as smallas possible and you try to shape
your spectrum in a way that ittheoretically cannot be smaller
enough.
That means you are using yourbandwidth the most efficient way
, and that's a processing.
It's a digital filter.
It's a digital linear filterthat we use as transmitter to
(28:12):
try to increase the spectralefficiency, basically, and both
the transmitter and receiveralso have error correction.
So that means you typically addabout 20 to 25% overhead in
terms of error correction,coding, and that's the part to
(28:39):
protect your signal from noiseand distortions and to recover
your bits without errors.
And then.
So error coding and pulseshaping are probably the two
very important part of thetransmitter processing.
When it comes to the receiver,it's a little bit more process
(29:01):
because the signal nowpropagates through a long fiber
channel and it has experienced alot of distortions and noise.
So the receiver, what it does,is basically try to undo
whatever happened in the channel, and there are a few important
pieces there.
For fiber there is chromaticdispersion and that means every
(29:25):
wavelength or frequency of yoursignal experience a different
delay and your digital signalprocessing needs to undo that.
Basically, it's a phase profilethat you add to your signal to
do the reverse of what thedigital and what the chromatic
(29:46):
dispersion has done, and thenafter that you need to
compensate the crosstalk.
Even for single mode fiber,there are two polarizations that
we're using.
They are orthogonal so but theway you receive it they are
mixed, because your projectionof the two polarizations cannot
(30:10):
be the same as how you send itat the beginning.
So it has been rotated.
It's a unitary matrix if youuse a two by two matrix to
describe that.
So your receiver has to figureout what kind of rotation has
been done and then rotate itback, basically to separate your
(30:31):
two orthogonal positions.
And it also needs to be able totrack, because your fiber is
not in the in a very staticenvironment.
There is mechanical vibration.
There are other things.
That is keep changing.
So your receiver needs to beable to track the rotation over
(30:52):
time and keep correcting that.
And after that, there is a laserdifference between the
transmitter and receiver.
They are the transmitter has alaser, the receiver has a laser.
These two lasers have slightlydifferent frequencies.
They cannot be aligned becausethey don't talk to each other.
So the receiver has to figureout what's the difference
(31:13):
between these two and compensatethat.
And the laser also have face,and face is a noise is random.
So the receiver also needs tobe able to estimate how much
face difference between the twoand being able to compensate and
track that.
And then, after it goes throughall those processing, the
(31:37):
signal, should go back to whatyou sent at the transmitter with
noise, and then you make adecision and do the error
correction.
So that's the importantbuilding blocks of transmitter
and receiver digital signalprocessing.
Speaker 3 (31:53):
Yeah, that's.
That's extremely fascinatingand how things have evolved.
But you also spoke about asworking with transceivers.
One of your.
One of the important boxes thatyou also need to take is the
energy efficiency.
So I suppose the fiber ispassive, nothing happens there,
so the energy consumption iscoming from your transceivers,
(32:14):
and maybe the optical amplifiersas well, obviously.
But you know what is the effortthat is required in order to
make that receiver increasingspeed, increasing data rate,
better digital signal processing, and yet you have to market on
the energy efficiency.
How is that possible?
Speaker 1 (32:32):
Yeah, it's hard.
It's hard so once in a while.
Yes, fiber is passive, but howdoes fiber relate it to power
consumption?
Fiber has loss.
So although it's passive, if ithas a lot of loss, then you
need a lot of energy to boostthe signal back.
(32:53):
So one of the importantprogress in fiber fabrication in
the past many decades is toreduce the loss.
So nowadays fiber is prettygood in terms of loss, but
that's also part of the energyreduction there.
(33:14):
If you have a lower loss fiber,you need less powerful
amplifier.
Then you save energy.
So now let's look at thetransmit and the receiver.
I wouldn't say there's onesingle thing you do to reduce
the power consumption.
It's a combination of prettymuch everything, from the
(33:37):
hardware design, from the laserdesign and from digital signal
processing.
When we talk about digitalsignal processing, a major
factor is battery, cmos, whichmeans smaller nodes and consumes
less energy, and also thealgorithm itself.
(33:57):
You need to be able to designthe efficient algorithm that
does a good job but doesn't havea high complexity.
That's important.
And you need to find modulationformat that is more efficient
for the limited amount of poweryou're using.
(34:18):
And when it comes to hardware,laser is getting more efficient
these days and the modulator nowis much higher bandwidth.
Or the other way to look at itis actually much higher
bandwidth.
It actually means it needs lessdriving voltage.
(34:41):
So the progress we made ondesigning a better modulator is
also a relatively big factor onwhy we are having lower power
(35:01):
consumption devices.
And on top of all this thereare a lot of progress made on
electrical amplifiers andcircuits that are making things
much smaller than before.
So that means the electricaltraces on the device is much
shorter and that means less lossand higher power efficiency.
(35:25):
So it's a lot of hard work frompretty much everywhere to try
to reduce the power consumption.
Speaker 3 (35:32):
Oh, I can see that that's something that we're
going to continue to work for afew years.
Speaker 2 (35:37):
Yes, Come, Thierry you have a question.
So the group I'm in right nowwork on OAM beams, right?
So I guess we mentioned themulticore fibers, multimode
fibers, and one of thechallenges for energy efficiency
right now, if we switch to theschemes, is the use of the
(35:57):
algorithms.
The digital signal processinglike MIMO is very effective, but
it's also very Not inefficient.
I think it is efficient but itwill consume some power.
So I guess the question wouldbe are there schemes such as OAM
beams that eliminate, at leastin part, the need for algorithms
(36:18):
like MIMO, that you take outpotential in the future?
Speaker 1 (36:22):
Yeah, that's a great question.
So one way people tend to go tois try to eliminate crosstalk
between modes and cores as muchas possible.
My personal view, first of all,if we are talking about
high-end long-distanceapplications, the transceivers
(36:44):
for those MIMO is not a partthat is very heavy in terms of
power consumption.
So even if we don't do any MIMOI don't really have numbers in
my head but we're notnecessarily saving a lot A lot
of power is consumed by numberone, chromatic dispersion
(37:04):
compensation and number two,error correction.
That's a very big part of thewhole power consumption in terms
of digital signal processing.
But it can be a problem if yourMIMO size gets too big.
I wouldn't say it's a problem ifyou have three modes or five
modes, but if you start to have15, 20, or even more, number one
(37:29):
needs to look into how toimplement that in a efficient
way.
And if you try to eliminate allthe crosstalk, then I guess
it's a good idea to look atoverall are we saving in terms
of performance Some of thefibers?
It works better if you havecrosstalk because when you start
(37:52):
to randomize the signal indifferent chords or modes, it
averaged some of the non-lineareffects and that gives you some
sort of advantage.
So it depends on theapplication If you're doing
long-haul, short-haul, how muchDSP you want, or do you want DSP
at all?
And we should always start withthe application and what we're
(38:17):
sensitive about for thatapplication.
Speaker 2 (38:21):
Thank you, that's very interesting.
Speaker 3 (38:23):
Yeah, it is very interesting.
And the final question that Ihave on the current devices and
systems is on energy.
Is there any role for safephotonic integration there?
Does photonic integration helpswith energy consumption or does
it help with the footprint or,as you said, a combination of
(38:46):
things?
Speaker 1 (38:47):
Yeah, integration helps with everything.
I would say.
It definitely reduces the formfactor, which is very good, and
it most likely helps a lot interms of power consumption
because when things are smaller,transmission distance are
shorter and you save loss andalso you can design your
(39:10):
impedance in a different way.
It opens a lot of possibilitiesand, honestly, for the future,
if we think about we are goingto need, if we just project the
50 or 60% annual growth, we aregoing to have probably 100 times
(39:30):
actually more than 100 timesmore interface rate in 10 years
range.
And if we do look at how fastwe're scaling the interface of a
single transceiver, it is about20%.
So you see the big discrepancybetween the two.
So very, very likely at somepoint we will need to put
(39:52):
multiple transmitters in onepackage and nowadays we are
seeing signals from onetransmitter as an entity and
then if you have a terabitsignal here, you treat this as a
single, independent piece ofspectrum.
You add and job and send it todifferent receivers, but maybe
(40:14):
in 10 years range you alwaysneed to treat 10 of them as a
group because no one wants oneof this, one terabit or 10
terabits.
I need much more than that, sothere's no need to separate them
at any point of your network.
Then you will put all thehardware together, and then,
when you have 10 or 20transmitters in one package, you
(40:37):
need some innovations on how tointegrate them and you need to
make them small enough so thatthey fit whatever form factor
you need to plug this thing toand you also need to Well, you
have a lot of them in onepackage.
(40:58):
You probably cannot tolerate alot of electrical interface and
then convert signal to a certainsocket and then use another
socket to connect it.
It's just not practical.
So integration is the way to go, more integration.
We're doing a lot ofintegration these days already,
compared to 10 years ago.
(41:19):
It's fascinating how smallthings get to and how many
layers you see in terms of oneInside one, one small little
package.
But we're going to do more, I'mpretty sure.
Speaker 3 (41:33):
And I think I have a final question here in terms of
research and your work.
So what does the future holdand what is it that you're
working at the moment inresearch that we are going to
see implemented in the next, say, three, five, 10 years time?
Speaker 1 (41:51):
I think that's something we touched upon a
little bit already.
I think we are going to do moreintegration, and that opens up
possibility of design thingsstructurally in a very different
way.
So integration doesn'tnecessarily mean we just put 10
of them or five of them in onepackage and make them as small
(42:12):
as possible, and it's verylikely we're going to see a
different ways of putting themtogether so that we have a jump
in terms of either energyefficiency or performance and
maybe we design our signalprocessing in a very different
way, because now you have 10channels in one package and you
(42:32):
may want to see how can I comeup with a brand new digital
signal processing method thatcan compensate some hardware
defect that was not able to beaddressed before?
And how do you design yoursignal processing together with
(42:53):
hardware design?
Will that help and make theoverall global performance much
better than many of theindependent one just simply put
together?
So that's something we arethinking about for now and
besides that, I guess it's justwe're trying to see how to have
(43:15):
higher electrical signalgeneration and at the same time
keep in mind that is the powerconsumption still makes sense?
Is the cost still makes sense?
Do we want to think things verymuch outside the box and doing
something very different fromnow.
That's things we're activelythinking about.
Speaker 3 (43:36):
Thank you, vivian.
It's fascinating to see that,when we had to look at those
problems at a more systemsperspective, network perspective
, rather than the justindividual components alone,
because it's the sum of them allthat is making a difference, I
suppose, for, as I said,efficiency, energy efficiency or
even cost, is that true?
(43:56):
Yeah, exactly.
Speaker 1 (43:58):
So I think the more developed this field it's like
maybe many other fields, youneed people who understand
pretty much everything.
So in 10 years, 10 years ago,15 years ago, people are making
transceivers.
People are very specialized.
If you are a device person,then you design a laser, you
(44:19):
design a modulator, a photo diet, and you are very good at that,
and you don't necessarily knowmuch about error correction or
digital signal processing.
And people who do DSP, theydon't necessarily have to
understand the device.
They know on a high level howthey work and that's all.
But nowadays it's getting moreand more obvious that we need
(44:40):
people who understand more thanthat small thing that used to be
a standalone thing.
And if you know the wholepicture, if you are doing signal
processing and you understandwhat are the limitations of your
devices and why those arelimitations and how hard it is
(45:02):
to overcome that limitation, youmight start to think, oh, if
this is so hard and hard, welldesign, can I do something in
the signal processing?
Is that easy overall?
And so this is getting veryinteresting.
We will see people who canhandle the same studies outside
(45:26):
a smaller topic probably cancome up with very clever ideas
and we might be surprised abouttheir solutions.
We'll see.
Speaker 3 (45:34):
Yeah, thank you, thank you.
Thanks a lot, vivian.
Thank you, I don't want to bedeleted from this, from to be
added out, but the comments arethe following I made this
question on purpose because,depending on the audience, they
might be working, say, ondevices, or they might be
working as a signal processing,and your answer was perfect.
It's exactly what I wanted tohear.
So for them to see that you hadto understand a lot of stuff.
(45:58):
Excellent, I love that.
Thank you.
Sorry, teri, I'll pass it on toyou now.
Speaker 2 (46:04):
Oh, thank you so much .
It was super interesting and Iguess it ties into the fact that
since it's ever relevant field,right for 100 years and more
now telecommunications has beena field where a lot of
interesting research has beendone, and I guess now some young
professionals might beinterested in getting into that
(46:24):
field and they couldn't bebetter served than by asking
questions to someone like you.
So I guess the next part wouldbe to ask you a little bit of
career advice for anyone whowould be interested in getting
into that field.
I guess the first advice you.
Speaker 1 (46:40):
That is pretty interesting, so ask.
Speaker 2 (46:44):
I guess the first piece of advice already that
you've given is to be good atmultiple things within that
field.
So study your DSP and studyfibers, study sources,
electronics.
So the first question I hadwasso you've talked about a lot
of the development of opticaltransceivers, fibers, all those
(47:06):
things.
So, and from my understandingalso, you do both work in the
lab also as probably somesimulation, theoretical work.
But one thing that might beinteresting is to share your
thoughts on how do you run agood experiment?
How do youbecause that's a bigpart of the development process,
right?
Speaker 1 (47:27):
Yeah, Okay, Thank you .
I guess in order to run a goodexperiment, you need to like
doing experiments.
I guess that's number one.
If you have fun, you do abetter job.
I don't think people have to doexperiments to be successful or
(47:50):
doing good things.
In telecommunication A lot ofpeople are great in doing theory
and also information theorythat's a huge part of this and
design modulation format and runreally complicated simulations.
So you don't really have to begood at experiments to make
(48:10):
something good.
And if you don't likeexperiment, I think that's fine.
You can do simulations oftheory and you will still be
really good and produce veryuseful things for the community.
But if you want to do experiment, I think for our experiment I
(48:32):
would say in my experience Iwould say pay attention to
things.
That doesn't quite make sense.
What I mean by that is when youget into the lab you probably
have a goal.
This is a system I'm going tobuild and then I'm seeing the
(48:53):
result from the computer or apiece of equipment and maybe it
already looks like what you'regetting.
You're getting what you wantthe signal, the data rate and
you hit the required the bit toerror ratio and all this.
So you can probably alreadyconclude the experiment at that
point, but then you might haveobserved something that doesn't
(49:18):
really make sense, somethingthat you were not expecting.
Maybe you observe something alittle bit strange from your
spectrum or during your signalprocessing.
There are some response of afilter that looks a little not
so normal.
According to your pastexperience, Although you have
(49:38):
done the experiment and you getall your data and you could
write a paper about it already,I would say pay attention to
those things and don't let itslip away, and make sure you
understand as much as possible.
And that always, sometimessurprisingly, leads to something
that you never understoodbefore and it's something maybe
(50:00):
you realize no one has everunderstood.
So pay attention to smallthings, especially those that
look like quite make sense.
Speaker 2 (50:10):
That's awesome, thank you.
It's very interesting becausethat's something I've been told
as well that science happens inthe year.
Sometimes the mundane detailsis where you find the very
interesting things.
So that's definitely like yousee a little discrepancy and
that's where you can go from.
Speaker 1 (50:25):
Right, right.
Sometimes it's nothing, butsometimes you get a lot of joy
understanding things that youdidn't expect at all.
Speaker 2 (50:35):
That's fantastic advice, and I guess that ties
into the next question, whichwould be is there a piece of
advice, a specific piece ofadvice that you've received
early in your career that hasbeen very influential on your
path?
Speaker 1 (50:53):
I wouldn't say it's a specific sad piece of advice,
but what I've seen from mycolleagues and people I worked
with when I was studying andalso later I'm working what I've
seen that influenced me a lotare people, their passion, I
(51:18):
would say the researchers aroundme.
I see that when people reallylike what they are doing and
they had a lot of fun and theygenuinely like what they are
doing they do a much better job,consistently and sustainably
and also they are happier andthey are producing better things
(51:41):
and life is better that way.
So I guess that's what Iobserved.
And then I start to maybeconstantly check with myself do
I still like this?
Am I having fun?
And just make sure I still likewhat I'm doing and so far I
(52:05):
like it very much and I thinkit's good.
Yeah, I guess it's a passion.
I see people around me I reallylike that and that has an
impact on me for sure.
Speaker 2 (52:19):
And so being passionate is probably the most
important thing.
If you're happy at doing whatyou do, like you said, you're
probably going to be much betterat it and also you're going to
enjoy the time you spend,because we do spend a lot of
time working, after all.
Speaker 1 (52:34):
Yeah, you have eight plus hours every day and that's
a big chunk of your life, so ifyou don't like it.
That's too bad.
Speaker 2 (52:42):
And is there any other advice that you would like
to give to someone who'sstarting out in the field, or
you would like to give yourself?
If you could talk to yourselfwhen you were first starting out
.
Speaker 1 (52:55):
Okay, that's a good question.
I will say stay open-minded andcurious.
And that probably reflected onwhen I say pay attention to
small things that doesn't checkout in your experiment, and
open-minded it's not.
(53:17):
And curious those two.
Sometimes, when people mentionthis too, maybe they were more
referring to learning new things, things you don't know about,
but I'm more thinking about thepart that, since that you have
opinion on.
For example, you think thisalgorithm or this field or this
(53:38):
topic is not promising, it's notgood and it's not useful.
Or you conclude something fromyour past experiment and you are
very likely making thatconclusion because you had
enough data point and you did itin a logical way.
But still be open-minded thatyou might be wrong because you
(54:02):
are limited from what you see.
Maybe you are very logical, butthe fact you see maybe it's not
a whole choice, you might notunderstand everything.
So when people have an oppositeopinion than you, I would say
try your best.
To this is also advice tomyself.
(54:24):
When you ask that question.
I try to do this as much aspossible.
Maybe I probably want to doeven better.
Spend time, listen to people whodisagree with you and not only
be implied and just listen butunderstand why they have their
opinion and then you can make adecision from there.
(54:47):
Are you convinced?
Or maybe most likely you don't,because there's a reason why
you believe what you believe,and the researchers are usually
pretty good, they're logical,they're good at thinking.
So the opinion you have mightbe a solid and more reasonable
(55:08):
opinion.
But still listen to people whodisagree and understand why they
hold that opinion and decide ifthey can convince you and my
experience most often they don'tconvince me.
But it is those moments like oneout of 10 times.
There's one time that you areconvinced and that is actually
(55:33):
the moment you realize you havegained a lot of understanding of
something.
So it's a small chance that yourealize, oh, I was wrong.
But in those moments you reallystart to have another level of
understanding what you weredoing.
So I would say it's almostalways worth the time.
(55:57):
Take some time, listen andunderstand why some people
disagree with you.
I think that I should do moreand I think if someone asks me
what's important, this isprobably what I would say.
Speaker 2 (56:16):
That's great.
Thank you so much, Fadima.
You wanted to add something.
Speaker 3 (56:20):
I would say I like your advice saying to listen to
people who disagree with you.
As I said, they might be right.
There's a small percentage ofthem that might be right, but
sometimes in my frame, thequestion is slightly different
way and maybe communicating isslightly different way and then
we learn.
As I said, we learn from thatprocess in trying to communicate
in a different way.
(56:41):
So I love that advice.
It's so true.
Thank you for that.
Speaker 2 (56:45):
Yeah, and I think, as you mentioned also, like, okay,
maybe you only become convincedone time out of 10, but the
other nine times you're stillhave to refine your
understanding.
If you want to at leastconvince yourself of your
initial belief, you need to makesure.
So even then, it's valuable toexchange with other people of
(57:05):
differing opinions.
I guess the last question thatwe were curious about was the
most important leadership lessonyou've learned, and how has it
proven valuable?
Speaker 1 (57:18):
This is an interesting one.
Okay, so I'm not in a At themoment.
I'm leading a very small team,but I'm not in this position for
long, so I am owning a peoplemanager for about two years now.
I wouldn't say I learned a biglesson or something, but since I
(57:43):
start to have this job, I wasexposed to materials and courses
that are for how to be a bettermanager, how to be a better
leader, and all this.
There are a lot of advices toteach you how to show that you
respect other people and how you, I would say, create a better
(58:11):
manager image.
But what I start to realize is,in order to be a good leader I
wouldn't say I am already, but Iaim to be for the future the
thing is, maybe we should havethink a lot on how to show
(58:35):
respect to other people and allthis the real question to ask is
do you actually respect them?
Do you care?
Do you want good things forthem?
Of course you want success foryour team, you want success for
the company, but if you don'tgenuinely care and just try to
(58:59):
behave like you care, it'sprobably not a good starting
point.
So I try to ask myself wheneverI approach a people problem or
something I need to supportother people, I think I try to
(59:20):
tell myself or ask myself do Icare?
Do I genuinely care about them,their career and starting from
there, it's probably a good idea, and that's also what I
observed from good leaders thatI get to see around me.
(59:40):
The better ones are doing this,so I'm trying to do the same
thing and I hope I will do good.
Speaker 2 (59:53):
That's great.
I think that's an importantpoint is, if you actually care
about the people around you, Iguess leadership becomes more
natural as well.
And that's a great challenge,because sometimes I guess you
don't initially always care, butyou have to find ways to get
there.
I don't know if that's beenyour experience.
Speaker 1 (01:00:16):
You can always care about someone.
I think it's more when you addan actual moment, there's a
crisis or there's something youneed to quickly fix, and then
you may be taken over by thegoal that you need to achieve,
and then you start to think alot about the consequences If
(01:00:38):
this does not get done or wedon't achieve this, and then
something will happen.
So we need to quickly fix this,and then you might be carried
away by that emotion or thaturgency of trying to fix
something, and then you start tocare less about your people.
You might fix that thing for ashort term, but a long term I
(01:01:04):
don't think it's good and youmight have more problems in the
future.
So it's just not.
I don't think that's a success.
If you only fix one thing andthen people around you are not
happy or they are not gettingthe benefit of fixing that thing
.
I guess it's just a loss.
Speaker 2 (01:01:29):
Yeah, it's a difficult challenge.
To make everything cometogether cohesively, yeah,
Vivian.
So what are your plans for theyear?
What are your hobbies,interests, anything you'd like
to share with people about whatyou're looking forward to?
Speaker 1 (01:01:49):
Yeah, I mean I'm quite excited to see I do think
we're at a point that, in termsof high speed optical
transceivers, we're going to seesome changes.
How we design them, they willlook different than what they
traditionally look like.
So I'm quite excited to see allthe innovations coming together
(01:02:12):
to make better communicationsystems.
And, yeah, and I do hope, moreyounger people join us and who
are excited about this field andwant to make contributions.
Speaker 2 (01:02:28):
Thank you so much, and thank you, fatima, as well.
It was a pleasure to do this,but, yes, with both of you.
Speaker 3 (01:02:36):
Thank you, thank you, thank you, cherie, thank you,
vivian, that was great.
Illuminated by IEEE.
Photonics is a podcast seriesthat shines light on the hot
topics in Photonics and thesubject matter experts advancing
technology forward.
Speaker 2 (00:14):
Hello everyone, welcome to today's episode of
Illuminated, an IEEE Photonicspodcast.
My name is Thierry LaPointe-Leclère and, as the IEEE
Photonics Society EmergingTechnology Task Force
representative for the YoungProfessionals Program, it's my
pleasure to be your host today,I guess.
Before getting any further, Iwould like to introduce myself,
(00:34):
because all of you will beunfamiliar with who I am not
that I blame you so I basicallyam a first year PhD student at
Boston University, working inProfessor Siddharth's or
Mashantran group, who specializeat the moment in optical
orbital angular momentum beams.
I have a master's degree inphotonics from Laval University
(00:58):
and a bachelor's degree inengineering physics from the
same university.
I think it's fair at this pointto say that I enjoy the field
of optics in Photonics quite abit, because I've spent already
at least like seven yearsworking in various groups and on
various projects in Photonicsand I'm absolutely delighted to
be hosting this episode ofIlluminated.
(01:18):
So on to today's podcast.
This one is particularlyinteresting because we're
looking at how we keep findingnew ways of sending more
information and at faster ratesin optical systems.
In particular, we're going tobe discussing high speed optical
system and high throughputfibers, and our expert yesterday
is Viviane Cheng from Bell Labsand our moderator is Dr Fatima
(01:43):
Gunning from Tyndall NationalInstitute.
We will go into their technicalwork, career journeys and
advice for young professionals.
Well, thank you for listeningto this episode.
I think it's going to be a goodone.
First of all, I'd like tointroduce Fatima.
I'll give a brief bio.
So Fatima received a PhD inOpto-electronics in 2000 from
(02:07):
Pontifica University at Catalicade Rio de Janeiro I hope I said
that correctly Investigatingfundamental optical properties
of glasses, chemical fibers andplanar waveguides transformed by
a technique called thermalpolling.
As a postgraduate researcherstudent, or as a postgraduate
research student, she visitedBritish Telecom Research
(02:29):
Laboratories in the UK andjoined Corning Research Centre
at Adelstral Park, also in theUK, in 2001 to work on advanced
electroabsorption based opticalmodulators for applications in
high speed optical transmissionnetworks.
She co-founded the PhotonicsSystem Group at Tyndall National
Institute in Cork, ireland, in2003.
(02:50):
There are research focuses onspectrally efficient techniques
for high speed opticaltransmission, including all
optical OFDM, coremwdm andfrequency comb generation,
expanding communications to newwavelength windows, namely to
micron, and with fibers such asOOCOR fibers.
She actively works closely withPhotonics Devices integration
(03:15):
and packaging teams to enablenext generation components for
communications.
She was at IEEE, so her rolewithin IEEE Photonics Society
over the years have involvedbeing a VP of Multicultural
Outreach from 2018 to 2019 andVP of Membership Outreach for
the IEEE Photonics Society from2020 to 2022.
(03:38):
And even though right now Idon't see a position listed, an
official position, she's stillvery active within the society,
being a volunteer throughmentorship, conference
committees and co-hosting events, so I think she's probably
quite busy.
Speaker 3 (03:56):
So thanks, Therese, for the introduction.
It's a delight to be here andit's in delight to introduce our
guest speaker today, which isVivian Cheng.
Vivian Cheng is a distinguishedmember of technical staff and
department head ofOptoelectronic subsystems group
Anoka Bell Labs.
She received her PhD degree onoptical fiber communications in
2012 from Universal Melbourne inAustralia and from 2013 to 2015
(04:20):
, she was a research fellow atthe Universal Melbourne
conducting research on opticalfiber transmission and since
2015, she's now she has been inBell Labs since then with
current research interestsincluding advanced digital
signal processing and high-speedfiber transmission.
She was a awardee of the YoungInvestigators Award in 2020 for
Atropical Electronics Societyand she's a senior member at
(04:42):
Tripoli, a fellow of Optica, andshe published more than 150
papers in the era of high-speedoptical fiber communications
well over 60,000 citations.
So it's super exciting to haveyou here, Vivian.
Welcome to this podcast.
Speaker 1 (04:57):
Thank you, thank you Fatima, thank you Teri.
It's great to be here, sohopefully today we talk about
what we're doing in terms offiber communication and
high-speed optical transceiversand share some of our thoughts
and discussions, and hopefullythat would be interesting.
Speaker 3 (05:16):
Yeah, that's great, and then it's very interesting,
I think, about a couple of yearsago at the UFC.
We talk about 50 years of usingoptical fibers and so on, but
it took a long time for peopleto move from copper to optical
fiber.
So what is it that you're doingin terms of what is it that?
Optical fiber?
Sorry, let me start again.
(05:36):
Sorry, let's start again.
I said something else and Ijust shouldn't have said so.
We heard a couple of years agoback in the UFC the 50 years of
optical fibers.
It was super exciting to seethat happen.
It took a long time for peopleto really install fiber on the
ground and people have beenmoving from copper to fiber,
(05:58):
maybe started with submarinesystems and long-haul systems,
but now even coming closer topeople on access and to their
homes.
So where do we see opticalfiber installations and what is
it?
Optical fiber transmission?
What are we talking about?
Speaker 1 (06:17):
Okay, that's a good question.
I don't think you necessarilysee optical fibers in daily life
, although everyone of us mostof us using them all the time.
So they are basically buried inthe ground.
If we're talking aboutterrestrial links, that's the
technical term we use.
(06:37):
So they are buried a few metersdown to the ground or they are
literally sunk into the bottomof the ocean.
It's not like wirelesscommunication, like your phone.
You will see the base, towers,a tennis around you.
You don't really get to seethose, but I would say 99% of
(06:57):
the distance your signal istransmitting through the media
is not really air, it's fiber.
And so, for example, if youmake a phone call from the US to
Europe, china, india, any othercountry, it's only the first
maybe 10 centimeters that yoursignal transmitted from your
(07:17):
phone.
That is through air, which iswireless, and then, starting
from the point that antenna gotyour signal, everything most
likely is only fiber.
And then it gets to the otherend and another antenna is
transmitting your signal toanother phone and that's another
very little part that is notfiber, and everything in between
is most likely all fiber andthey are in the ground or at the
(07:42):
bottom of the ocean.
Speaker 3 (07:46):
I see, and why do we need research?
Then?
Optical fibers, what is it?
The challenges associated withinstalling the fiber on the
ground or making thosetransmissions and converting
between wireless and optical, Imean what are the challenges?
Here, yeah.
Speaker 1 (08:01):
So the challenges, as we can probably imagine, the
challenge is we need more speed,we need more data and we need
more bandwidth.
And if you think about it 20years ago, what kind of phones
we have and what kind ofapplications we're able to do
(08:22):
with our computers and phones,and that amount of bandwidth
compared to the bandwidth we'reconsuming today, that's the
factor of depending on theapplication it could be a few
hundred times or a few thousandtimes increase.
So that's a huge change interms of what we demand in terms
of bandwidth.
And so if we take thatperspective and think about the
(08:45):
future, in the next 10, 15, 20years, what we would need, if
you don't think about the pastand just standing now and try to
imagine what we might need andyou tell people, oh, I think we
need five thousand more times ofbandwidth for the next 15 years
, people will probably see thismight be crazy because how can
(09:08):
we need so much more?
We're pretty good in terms ofbandwidth and speed, but if you
think about the past and askpeople, if you imagine you ask
people on the street 15 yearsago, you think you need this
much bandwidth today and theyprobably think you're crazy.
So I always thought, even forpeople who work in technology,
(09:34):
even in the field oftelecommunication, they might
underestimate how much bandwidthwe may need in the future, and
the data sets so too.
If you look at the data,there's solid data saying every
year we are demanding about 50to 60% more data.
So that's the annual growth of50 to 60%, and that's a lot.
(09:58):
If you take that number and doa 10 years projection, you are
looking at about 100 times morebandwidth.
So that's the challenge we needto scale our electronics and
optics fast enough to satisfythe demand, and so what we want
(10:19):
to do is higher bandwidth, morethroughput at a reasonable price
, and we don't use all theenergy in this world, so power
consumption is also veryimportant.
So that's the main challenges.
Speaker 3 (10:37):
So how can be done?
So your research says thatyou're working with transceivers
and you're talking aboutincreasing bandwidth, make
higher speeds using less energy.
So what is it that is consumingthis energy and how does this
transmission work?
Yeah, okay.
Speaker 1 (10:53):
So let's start with.
What do we mean by bandwidth?
There are two bandwidths thatare relevant here.
One is the bandwidth of thefiber itself.
So if we take a step back thefiber transmission system on the
high level, you only have threethings.
You have a transmitter that issending the light, that carries
(11:16):
information, and then you have afiber link that's connecting
your two points and then youhave a receiver.
So when we talk about thebandwidth of the fiber, that's
about four terahors bandwidth,or roughly 35 nanometers
bandwidth.
That's a lot of bandwidth, andthis whole fiber typically
(11:38):
nowadays depends on theapplication and also distance.
This fiber can carry aboutanywhere between 20 to 40
terabits per second.
So one terabit is 1000 times of.
Say, if you have one gigabitper second for your home, this
is one terabit is 1000 times ofthat, and 20 to 40 is 20,000 to
(12:03):
40,000 times that.
So this is what a typical fibercan carry.
And then you have yourtransmitter and the receiver and
you have the bandwidth or thethroughput, or sometimes we call
interface range, of yourtransmitter and the receiver,
and that's a much smallerbandwidth.
So typically you have anywherebetween 50 to 100 gigahertz
(12:27):
bandwidth, depending on whetheryou're talking about product or
research, and that gives youabout one terabit or two terabit
per second data rate.
So you typically need anywherefrom 30 to 40 of them to fulfill
the whole bandwidth of thefiber.
And so when we talk aboutincreasing the bandwidth, for
(12:53):
the part I am working on, Imostly work on increasing the
data read or the bandwidth ofthe transmitter and receiver,
and after you increase that, youstill need multiple of them to
fill the bandwidth of the fiber.
So part of our research is onhow to develop better
transmitters and receivers thatcarries more information per
(13:14):
transmitter.
And then some other researchgroups or my colleagues or other
people, they are working on howto develop better fibers, a
fiber that can carry morebandwidth.
So those are the two main partsand for my research, for the
transmitter part, back to thequestion how do we make
(13:37):
transmitters that carry moreinformation or more data right?
Then we need to dig a littlebit deeper on how transmitter
works.
The transmitter on a high levelonly have three things.
It starts with a laser.
So the laser we are using fortelecommunication is a laser.
(13:57):
We call CW laser, continuouswave laser.
So that means the laser issending out light constantly.
It's a constant power and thelight that is constant or any
electromagnetic wave that isconstant doesn't carry
information.
So, despite this, for peoplewho are not necessarily familiar
(14:18):
with telecommunication, imagineyour city light or you have a
flashlight at torch and you turnit on.
That light is always on and itdoesn't carry information.
However, if you start to dimthe light or flash the light
according to, for example, amask code, then the light starts
(14:41):
to carry information and theprocess you use your hand to
change the on and off status ofthe light.
This process is calledmodulation.
So the same thing happens in anoptical transmitter.
You start with a laser and thatlight doesn't carry information
(15:01):
, and you send the laser into adevice that is called a
modulator.
So this modulator takes toinput.
It takes the laser, it alsotakes the electrical signal and
this electrical signal is yourdata or your information, or you
can say that's your zeroes andones.
So the modulator starts tomanipulate the light properties,
(15:24):
for example power of the light,according to the electrical
signal that is added to themodulator.
And then the output of themodulator is a.
You can see that as a lightthat is being manipulated or
dimmed or flashed to some extent, and that light starts to carry
(15:46):
information.
So that's how the transmitterworks.
And in order to make atransmitter that is fast, you
need basically two things.
That is fast it's not reallythe laser.
The laser is sending outconstant light and the laser
we're making today, especiallythe laser we make for high end
optical transmitters they aregood enough in terms of the
(16:10):
performance.
We are more working on how toreduce the power consumption,
how to reduce the form factorand all this, but not the
performance itself.
In order to make a fastertransmitter you don't
necessarily need to do much onthe laser itself.
You need two other things thatworks fast, that is, your
electrical signal generation andthe modulator.
(16:33):
So a lot of research andinnovation are done on this two
part to push the speed of thistwo.
For example, for the electricalsignal generation, for decades
we were relying on the CMOStechnologies, and when the CMOS
node gets smaller we get fastergates and we have faster
(16:54):
electrical signal generation.
That's how we have fasterelectrical signal to drive the
modulator.
But the speed that CMOS isscaling is clearly not quite
enough for the speed we need forthe transmitter itself to scale
.
So some of the researchersstart to do something outside
(17:16):
the CMOS.
They take other materials, forexample, in a phosphate or
silicon germanium, they combinemultiple CMOS signals to form a
higher speed one so that you canyou will not only be relying on
the CMOS scaling speed and youcan do double or triple speed
(17:42):
compared to the CMOS speed, andthat's one of the ways to do
higher speed.
Electrical signal generation.
And then, in terms of modulator,we are mostly relying on
materials.
For a long time, starting from,we widely deploy optical
(18:03):
transmission links, especiallycoherent transmission links.
We were using a material calledbulky lismai bit material for
modulator and that's anexcellent material.
The only problem is is not fastenough for today's modulator.
So nowadays, in the past maybefive to seven years, there has
(18:25):
been a lot of innovations onmaking new materials.
That is faster and also muchsmaller informed factor to make
modulators.
And at the moment what productis using very widely is either
silicon, photonics or in aphosphate and they are faster
(18:46):
speed, so they make higherbandwidth modulator.
Therefore you can, you can makehigher speed transmitted out of
it.
And there are also a lot ofvery promising material in
research domain.
For example, one of them wouldbe some families and I bet they
are kind of they're similar totraditional is not admit, but
(19:09):
much faster speed and muchsmaller in terms of form factor.
There are other very innovativematerials like organic
materials or post money and allthis.
They're all under research.
But for modulator itself mostlywe rely on the material
innovation.
That's the two main area peopleare doing very active research
(19:35):
on to try to bring up the speedof an optical transmitter.
Speaker 3 (19:40):
That's very interesting.
That's on the speed, but theother thing you're talking about
is throughput.
How can we increase the overallfiber throughput using those
new devices that you're talkingabout?
Speaker 1 (19:53):
So throughput of the fiber doesn't necessarily get
increased because thetransmitters are faster.
So, as I said, fiber has acertain bandwidth, so the
transmitter has a much slowerbandwidth, smaller bandwidth
than that.
So we need multiple.
The reason why we push eachtransmitter to higher speed is
(20:17):
we want to use less number oftransmitters.
That's better, because laser isvery expensive.
Every transmitter comes withone laser, so at the moment the
laser is not cheap enough.
It's a significant cost to eachtransmitter.
So we try to use less lasers,which means each laser has to
carry more information.
That's the main motivation whywe're pushing the per
(20:41):
transmitter speed to as high aspossible.
But when it comes to thethroughput of the fiber itself,
the most widely deployed fibernow is called standard single
mode fibers and we're usingabout 35 nanometer of the fiber.
(21:01):
The fiber support more thanthat but amplifiers doesn't.
So we're limited by the opticalamplifier bandwidth and that's
what we call the conventionalband, the C band.
It's about four carats or 35nanometer.
In order to have more bandwidthfrom the fiber or more
throughput from the fiber, thereare several things you can
(21:23):
potentially do.
You can extend the bandwidth ofthe fiber itself because the
fiber does have a very wide lowloss window and you can try to
develop other amplifiers thatsupport more bandwidth.
That's one way to do it.
You might get a factor of two,three or even more out of that
(21:47):
and the system quickly getspretty complicated because every
band you have to use differentamplifiers.
So when an optical signal, avery broad optical signal, gets
to a point that it needs to beamplified and by the way that's
dependent on the transmissionlink that's anywhere between 50
kilometers to 100 kilometers.
So if you have a link that isbetween, say, europe and here,
(22:15):
then you need many, manyamplifiers.
I mean the summary links isanywhere between 10,000 to 8,000
range, so you need a lot ofamplifiers.
So every time you need toamplify a signal you have to
split them into different bandsand amplify them through
different amplifiers and combinethem.
(22:36):
So it is a less scalable way ifyou try to increase the bands
or the useful wavelengths out offiber.
The other way to do it is tojust add more fibers.
So you can either add morefiber in the way that we call
fiber bundle, so simply havemore fibers there, or you can
(22:59):
use more noble fibers.
For example, you can use multicore fibers, which means single
mode fiber has one cladding withone core in it.
You can have one cladding withmultiple core in it, for example
three or five or seven or evenmore.
Or you can use something calleda few mode fiber.
(23:20):
So the single mode fiber.
As the name suggests, itsupports one mode.
You can make fiber that hasmore than one mode, for example
three or six or 15.
And in that way you have moremodes, which means you can do
multiplexing by effect of numbermodes.
(23:42):
Or another very interestingtype of fiber is the fiber we
call holocore fiber, so it's airin the middle.
It naturally supports much morebandwidth, much more wavelength
range than single mode fiber.
And it is possible to developone amplifier that supports
(24:05):
wider bandwidth than single modefibers C-band amplifier, that's
also one way to do it.
And all these are very activeresearch domain in fiber
fabrication and also systemtransmission link as a whole,
and those also affect our futureprojection on how an optical
(24:25):
transceiver will look like.
Speaker 3 (24:28):
I see, and Teri, do you want to ask a question?
Speaker 2 (24:36):
I don't want to step on what you were going to ask,
but I did have one relatedquestion to what Vivian was
discussing.
Speaker 3 (24:42):
Yeah, you can ask a related question.
And then I was going to go andask something else.
I wanted to go into digitalsignal processing.
So if it's related, just askthe question.
Speaker 2 (24:51):
Okay, perfect.
So I guess what I wanted to askalso is now you brought up
different architecture forfibers, right?
So multi-core, multi-mode,holocore, and you said you've
been looking into possibly goingin those directions.
How does that?
How challenging is it to adaptyour optical transceiver design
(25:12):
for different types of fibers?
Speaker 1 (25:16):
That's a great question and principle, does not
?
You can okay, let me put thisway.
You can use the currenttransceiver for some of the new
fibers, for example.
If we are talking about fiberbundles, then basically you
don't need to change anything.
You just put more transmittersand receivers there.
(25:37):
They're parallel, they areindependent.
Or for multi-core fibers, thenit depends.
If the core couples to eachother, meaning there are
crosstalk or on a system levelwe may call it information
exchange among the cores, thenyour receiver should design in a
(26:02):
way that it can deal with thecrosstalk, and that's a typical
process we call the MIMOprocessing.
So you need to process all thesignals from all the cores at
the same time.
To redo the crosstalkMulti-core fiber, then sorry, a
few more fiber, then you do needto have a new set of signal
(26:24):
processing.
The hardware wouldn'tnecessarily change much, but
it's the digital signalprocessing party to whether you
need to deal with the couplingbetween different dimensions
that you're using now.
Speaker 3 (26:39):
It's interesting because that lists the way that
I was thinking on the digitalsignal processing side.
So I suppose not all signalsare perfect and there's lots of
impairments and, as you said,crosstalk that might occur.
So explain to us a little biton the work that is needed on
digital signal processing,perhaps not only the receiver
but sometimes even at thetransmitter as well.
Speaker 1 (27:03):
There are a few important pieces in terms of
digital signal processing.
So all we talk about at themoment we are talking about
high-end optical transceivers.
That means the transceivers aredeveloped for long-haul
distance transmission and thatmeans we are using, we are using
(27:28):
, we're trying to use thebandwidth very efficiently, and
that means the transmitterusually have one important
digital signal processing thatis called pulse shaping.
The name sounds very confusing,but basically what it does is
it tries to squeeze theinformation into a bandwidth
(27:48):
that is theoretically as smallas possible and you try to shape
your spectrum in a way that ittheoretically cannot be smaller
enough.
That means you are using yourbandwidth the most efficient way
, and that's a processing.
It's a digital filter.
It's a digital linear filterthat we use as transmitter to
(28:12):
try to increase the spectralefficiency, basically, and both
the transmitter and receiveralso have error correction.
So that means you typically addabout 20 to 25% overhead in
terms of error correction,coding, and that's the part to
(28:39):
protect your signal from noiseand distortions and to recover
your bits without errors.
And then.
So error coding and pulseshaping are probably the two
very important part of thetransmitter processing.
When it comes to the receiver,it's a little bit more process
(29:01):
because the signal nowpropagates through a long fiber
channel and it has experienced alot of distortions and noise.
So the receiver, what it does,is basically try to undo
whatever happened in the channel, and there are a few important
pieces there.
For fiber there is chromaticdispersion and that means every
(29:25):
wavelength or frequency of yoursignal experience a different
delay and your digital signalprocessing needs to undo that.
Basically, it's a phase profilethat you add to your signal to
do the reverse of what thedigital and what the chromatic
(29:46):
dispersion has done, and thenafter that you need to
compensate the crosstalk.
Even for single mode fiber,there are two polarizations that
we're using.
They are orthogonal so but theway you receive it they are
mixed, because your projectionof the two polarizations cannot
(30:10):
be the same as how you send itat the beginning.
So it has been rotated.
It's a unitary matrix if youuse a two by two matrix to
describe that.
So your receiver has to figureout what kind of rotation has
been done and then rotate itback, basically to separate your
(30:31):
two orthogonal positions.
And it also needs to be able totrack, because your fiber is
not in the in a very staticenvironment.
There is mechanical vibration.
There are other things.
That is keep changing.
So your receiver needs to beable to track the rotation over
(30:52):
time and keep correcting that.
And after that, there is a laserdifference between the
transmitter and receiver.
They are the transmitter has alaser, the receiver has a laser.
These two lasers have slightlydifferent frequencies.
They cannot be aligned becausethey don't talk to each other.
So the receiver has to figureout what's the difference
(31:13):
between these two and compensatethat.
And the laser also have face,and face is a noise is random.
So the receiver also needs tobe able to estimate how much
face difference between the twoand being able to compensate and
track that.
And then, after it goes throughall those processing, the
(31:37):
signal, should go back to whatyou sent at the transmitter with
noise, and then you make adecision and do the error
correction.
So that's the importantbuilding blocks of transmitter
and receiver digital signalprocessing.
Speaker 3 (31:53):
Yeah, that's.
That's extremely fascinatingand how things have evolved.
But you also spoke about asworking with transceivers.
One of your.
One of the important boxes thatyou also need to take is the
energy efficiency.
So I suppose the fiber ispassive, nothing happens there,
so the energy consumption iscoming from your transceivers,
(32:14):
and maybe the optical amplifiersas well, obviously.
But you know what is the effortthat is required in order to
make that receiver increasingspeed, increasing data rate,
better digital signal processing, and yet you have to market on
the energy efficiency.
How is that possible?
Speaker 1 (32:32):
Yeah, it's hard.
It's hard so once in a while.
Yes, fiber is passive, but howdoes fiber relate it to power
consumption?
Fiber has loss.
So although it's passive, if ithas a lot of loss, then you
need a lot of energy to boostthe signal back.
(32:53):
So one of the importantprogress in fiber fabrication in
the past many decades is toreduce the loss.
So nowadays fiber is prettygood in terms of loss, but
that's also part of the energyreduction there.
(33:14):
If you have a lower loss fiber,you need less powerful
amplifier.
Then you save energy.
So now let's look at thetransmit and the receiver.
I wouldn't say there's onesingle thing you do to reduce
the power consumption.
It's a combination of prettymuch everything, from the
(33:37):
hardware design, from the laserdesign and from digital signal
processing.
When we talk about digitalsignal processing, a major
factor is battery, cmos, whichmeans smaller nodes and consumes
less energy, and also thealgorithm itself.
(33:57):
You need to be able to designthe efficient algorithm that
does a good job but doesn't havea high complexity.
That's important.
And you need to find modulationformat that is more efficient
for the limited amount of poweryou're using.
(34:18):
And when it comes to hardware,laser is getting more efficient
these days and the modulator nowis much higher bandwidth.
Or the other way to look at itis actually much higher
bandwidth.
It actually means it needs lessdriving voltage.
(34:41):
So the progress we made ondesigning a better modulator is
also a relatively big factor onwhy we are having lower power
(35:01):
consumption devices.
And on top of all this thereare a lot of progress made on
electrical amplifiers andcircuits that are making things
much smaller than before.
So that means the electricaltraces on the device is much
shorter and that means less lossand higher power efficiency.
(35:25):
So it's a lot of hard work frompretty much everywhere to try
to reduce the power consumption.
Speaker 3 (35:32):
Oh, I can see that that's something that we're
going to continue to work for afew years.
Speaker 2 (35:37):
Yes, Come, Thierry you have a question.
So the group I'm in right nowwork on OAM beams, right?
So I guess we mentioned themulticore fibers, multimode
fibers, and one of thechallenges for energy efficiency
right now, if we switch to theschemes, is the use of the
(35:57):
algorithms.
The digital signal processinglike MIMO is very effective, but
it's also very Not inefficient.
I think it is efficient but itwill consume some power.
So I guess the question wouldbe are there schemes such as OAM
beams that eliminate, at leastin part, the need for algorithms
(36:18):
like MIMO, that you take outpotential in the future?
Speaker 1 (36:22):
Yeah, that's a great question.
So one way people tend to go tois try to eliminate crosstalk
between modes and cores as muchas possible.
My personal view, first of all,if we are talking about
high-end long-distanceapplications, the transceivers
(36:44):
for those MIMO is not a partthat is very heavy in terms of
power consumption.
So even if we don't do any MIMOI don't really have numbers in
my head but we're notnecessarily saving a lot A lot
of power is consumed by numberone, chromatic dispersion
(37:04):
compensation and number two,error correction.
That's a very big part of thewhole power consumption in terms
of digital signal processing.
But it can be a problem if yourMIMO size gets too big.
I wouldn't say it's a problem ifyou have three modes or five
modes, but if you start to have15, 20, or even more, number one
(37:29):
needs to look into how toimplement that in a efficient
way.
And if you try to eliminate allthe crosstalk, then I guess
it's a good idea to look atoverall are we saving in terms
of performance Some of thefibers?
It works better if you havecrosstalk because when you start
(37:52):
to randomize the signal indifferent chords or modes, it
averaged some of the non-lineareffects and that gives you some
sort of advantage.
So it depends on theapplication If you're doing
long-haul, short-haul, how muchDSP you want, or do you want DSP
at all?
And we should always start withthe application and what we're
(38:17):
sensitive about for thatapplication.
Speaker 2 (38:21):
Thank you, that's very interesting.
Speaker 3 (38:23):
Yeah, it is very interesting.
And the final question that Ihave on the current devices and
systems is on energy.
Is there any role for safephotonic integration there?
Does photonic integration helpswith energy consumption or does
it help with the footprint or,as you said, a combination of
(38:46):
things?
Speaker 1 (38:47):
Yeah, integration helps with everything.
I would say.
It definitely reduces the formfactor, which is very good, and
it most likely helps a lot interms of power consumption
because when things are smaller,transmission distance are
shorter and you save loss andalso you can design your
(39:10):
impedance in a different way.
It opens a lot of possibilitiesand, honestly, for the future,
if we think about we are goingto need, if we just project the
50 or 60% annual growth, we aregoing to have probably 100 times
(39:30):
actually more than 100 timesmore interface rate in 10 years
range.
And if we do look at how fastwe're scaling the interface of a
single transceiver, it is about20%.
So you see the big discrepancybetween the two.
So very, very likely at somepoint we will need to put
(39:52):
multiple transmitters in onepackage and nowadays we are
seeing signals from onetransmitter as an entity and
then if you have a terabitsignal here, you treat this as a
single, independent piece ofspectrum.
You add and job and send it todifferent receivers, but maybe
(40:14):
in 10 years range you alwaysneed to treat 10 of them as a
group because no one wants oneof this, one terabit or 10
terabits.
I need much more than that, sothere's no need to separate them
at any point of your network.
Then you will put all thehardware together, and then,
when you have 10 or 20transmitters in one package, you
(40:37):
need some innovations on how tointegrate them and you need to
make them small enough so thatthey fit whatever form factor
you need to plug this thing toand you also need to Well, you
have a lot of them in onepackage.
(40:58):
You probably cannot tolerate alot of electrical interface and
then convert signal to a certainsocket and then use another
socket to connect it.
It's just not practical.
So integration is the way to go, more integration.
We're doing a lot ofintegration these days already,
compared to 10 years ago.
(41:19):
It's fascinating how smallthings get to and how many
layers you see in terms of oneInside one, one small little
package.
But we're going to do more, I'mpretty sure.
Speaker 3 (41:33):
And I think I have a final question here in terms of
research and your work.
So what does the future holdand what is it that you're
working at the moment inresearch that we are going to
see implemented in the next, say, three, five, 10 years time?
Speaker 1 (41:51):
I think that's something we touched upon a
little bit already.
I think we are going to do moreintegration, and that opens up
possibility of design thingsstructurally in a very different
way.
So integration doesn'tnecessarily mean we just put 10
of them or five of them in onepackage and make them as small
(42:12):
as possible, and it's verylikely we're going to see a
different ways of putting themtogether so that we have a jump
in terms of either energyefficiency or performance and
maybe we design our signalprocessing in a very different
way, because now you have 10channels in one package and you
(42:32):
may want to see how can I comeup with a brand new digital
signal processing method thatcan compensate some hardware
defect that was not able to beaddressed before?
And how do you design yoursignal processing together with
(42:53):
hardware design?
Will that help and make theoverall global performance much
better than many of theindependent one just simply put
together?
So that's something we arethinking about for now and
besides that, I guess it's justwe're trying to see how to have
(43:15):
higher electrical signalgeneration and at the same time
keep in mind that is the powerconsumption still makes sense?
Is the cost still makes sense?
Do we want to think things verymuch outside the box and doing
something very different fromnow.
That's things we're activelythinking about.
Speaker 3 (43:36):
Thank you, vivian.
It's fascinating to see that,when we had to look at those
problems at a more systemsperspective, network perspective
, rather than the justindividual components alone,
because it's the sum of them allthat is making a difference, I
suppose, for, as I said,efficiency, energy efficiency or
even cost, is that true?
(43:56):
Yeah, exactly.
Speaker 1 (43:58):
So I think the more developed this field it's like
maybe many other fields, youneed people who understand
pretty much everything.
So in 10 years, 10 years ago,15 years ago, people are making
transceivers.
People are very specialized.
If you are a device person,then you design a laser, you
(44:19):
design a modulator, a photo diet, and you are very good at that,
and you don't necessarily knowmuch about error correction or
digital signal processing.
And people who do DSP, theydon't necessarily have to
understand the device.
They know on a high level howthey work and that's all.
But nowadays it's getting moreand more obvious that we need
(44:40):
people who understand more thanthat small thing that used to be
a standalone thing.
And if you know the wholepicture, if you are doing signal
processing and you understandwhat are the limitations of your
devices and why those arelimitations and how hard it is
(45:02):
to overcome that limitation, youmight start to think, oh, if
this is so hard and hard, welldesign, can I do something in
the signal processing?
Is that easy overall?
And so this is getting veryinteresting.
We will see people who canhandle the same studies outside
(45:26):
a smaller topic probably cancome up with very clever ideas
and we might be surprised abouttheir solutions.
We'll see.
Speaker 3 (45:34):
Yeah, thank you, thank you.
Thanks a lot, vivian.
Thank you, I don't want to bedeleted from this, from to be
added out, but the comments arethe following I made this
question on purpose because,depending on the audience, they
might be working, say, ondevices, or they might be
working as a signal processing,and your answer was perfect.
It's exactly what I wanted tohear.
So for them to see that you hadto understand a lot of stuff.
(45:58):
Excellent, I love that.
Thank you.
Sorry, teri, I'll pass it on toyou now.
Speaker 2 (46:04):
Oh, thank you so much .
It was super interesting and Iguess it ties into the fact that
since it's ever relevant field,right for 100 years and more
now telecommunications has beena field where a lot of
interesting research has beendone, and I guess now some young
professionals might beinterested in getting into that
(46:24):
field and they couldn't bebetter served than by asking
questions to someone like you.
So I guess the next part wouldbe to ask you a little bit of
career advice for anyone whowould be interested in getting
into that field.
I guess the first advice you.
Speaker 1 (46:40):
That is pretty interesting, so ask.
Speaker 2 (46:44):
I guess the first piece of advice already that
you've given is to be good atmultiple things within that
field.
So study your DSP and studyfibers, study sources,
electronics.
So the first question I hadwasso you've talked about a lot
of the development of opticaltransceivers, fibers, all those
(47:06):
things.
So, and from my understandingalso, you do both work in the
lab also as probably somesimulation, theoretical work.
But one thing that might beinteresting is to share your
thoughts on how do you run agood experiment?
How do youbecause that's a bigpart of the development process,
right?
Speaker 1 (47:27):
Yeah, Okay, Thank you .
I guess in order to run a goodexperiment, you need to like
doing experiments.
I guess that's number one.
If you have fun, you do abetter job.
I don't think people have to doexperiments to be successful or
(47:50):
doing good things.
In telecommunication A lot ofpeople are great in doing theory
and also information theorythat's a huge part of this and
design modulation format and runreally complicated simulations.
So you don't really have to begood at experiments to make
(48:10):
something good.
And if you don't likeexperiment, I think that's fine.
You can do simulations oftheory and you will still be
really good and produce veryuseful things for the community.
But if you want to do experiment, I think for our experiment I
(48:32):
would say in my experience Iwould say pay attention to
things.
That doesn't quite make sense.
What I mean by that is when youget into the lab you probably
have a goal.
This is a system I'm going tobuild and then I'm seeing the
(48:53):
result from the computer or apiece of equipment and maybe it
already looks like what you'regetting.
You're getting what you wantthe signal, the data rate and
you hit the required the bit toerror ratio and all this.
So you can probably alreadyconclude the experiment at that
point, but then you might haveobserved something that doesn't
(49:18):
really make sense, somethingthat you were not expecting.
Maybe you observe something alittle bit strange from your
spectrum or during your signalprocessing.
There are some response of afilter that looks a little not
so normal.
According to your pastexperience, Although you have
(49:38):
done the experiment and you getall your data and you could
write a paper about it already,I would say pay attention to
those things and don't let itslip away, and make sure you
understand as much as possible.
And that always, sometimessurprisingly, leads to something
that you never understoodbefore and it's something maybe
(50:00):
you realize no one has everunderstood.
So pay attention to smallthings, especially those that
look like quite make sense.
Speaker 2 (50:10):
That's awesome, thank you.
It's very interesting becausethat's something I've been told
as well that science happens inthe year.
Sometimes the mundane detailsis where you find the very
interesting things.
So that's definitely like yousee a little discrepancy and
that's where you can go from.
Speaker 1 (50:25):
Right, right.
Sometimes it's nothing, butsometimes you get a lot of joy
understanding things that youdidn't expect at all.
Speaker 2 (50:35):
That's fantastic advice, and I guess that ties
into the next question, whichwould be is there a piece of
advice, a specific piece ofadvice that you've received
early in your career that hasbeen very influential on your
path?
Speaker 1 (50:53):
I wouldn't say it's a specific sad piece of advice,
but what I've seen from mycolleagues and people I worked
with when I was studying andalso later I'm working what I've
seen that influenced me a lotare people, their passion, I
(51:18):
would say the researchers aroundme.
I see that when people reallylike what they are doing and
they had a lot of fun and theygenuinely like what they are
doing they do a much better job,consistently and sustainably
and also they are happier andthey are producing better things
(51:41):
and life is better that way.
So I guess that's what Iobserved.
And then I start to maybeconstantly check with myself do
I still like this?
Am I having fun?
And just make sure I still likewhat I'm doing and so far I
(52:05):
like it very much and I thinkit's good.
Yeah, I guess it's a passion.
I see people around me I reallylike that and that has an
impact on me for sure.
Speaker 2 (52:19):
And so being passionate is probably the most
important thing.
If you're happy at doing whatyou do, like you said, you're
probably going to be much betterat it and also you're going to
enjoy the time you spend,because we do spend a lot of
time working, after all.
Speaker 1 (52:34):
Yeah, you have eight plus hours every day and that's
a big chunk of your life, so ifyou don't like it.
That's too bad.
Speaker 2 (52:42):
And is there any other advice that you would like
to give to someone who'sstarting out in the field, or
you would like to give yourself?
If you could talk to yourselfwhen you were first starting out
.
Speaker 1 (52:55):
Okay, that's a good question.
I will say stay open-minded andcurious.
And that probably reflected onwhen I say pay attention to
small things that doesn't checkout in your experiment, and
open-minded it's not.
(53:17):
And curious those two.
Sometimes, when people mentionthis too, maybe they were more
referring to learning new things, things you don't know about,
but I'm more thinking about thepart that, since that you have
opinion on.
For example, you think thisalgorithm or this field or this
(53:38):
topic is not promising, it's notgood and it's not useful.
Or you conclude something fromyour past experiment and you are
very likely making thatconclusion because you had
enough data point and you did itin a logical way.
But still be open-minded thatyou might be wrong because you
(54:02):
are limited from what you see.
Maybe you are very logical, butthe fact you see maybe it's not
a whole choice, you might notunderstand everything.
So when people have an oppositeopinion than you, I would say
try your best.
To this is also advice tomyself.
(54:24):
When you ask that question.
I try to do this as much aspossible.
Maybe I probably want to doeven better.
Spend time, listen to people whodisagree with you and not only
be implied and just listen butunderstand why they have their
opinion and then you can make adecision from there.
(54:47):
Are you convinced?
Or maybe most likely you don't,because there's a reason why
you believe what you believe,and the researchers are usually
pretty good, they're logical,they're good at thinking.
So the opinion you have mightbe a solid and more reasonable
(55:08):
opinion.
But still listen to people whodisagree and understand why they
hold that opinion and decide ifthey can convince you and my
experience most often they don'tconvince me.
But it is those moments like oneout of 10 times.
There's one time that you areconvinced and that is actually
(55:33):
the moment you realize you havegained a lot of understanding of
something.
So it's a small chance that yourealize, oh, I was wrong.
But in those moments you reallystart to have another level of
understanding what you weredoing.
So I would say it's almostalways worth the time.
(55:57):
Take some time, listen andunderstand why some people
disagree with you.
I think that I should do moreand I think if someone asks me
what's important, this isprobably what I would say.
Speaker 2 (56:16):
That's great.
Thank you so much, Fadima.
You wanted to add something.
Speaker 3 (56:20):
I would say I like your advice saying to listen to
people who disagree with you.
As I said, they might be right.
There's a small percentage ofthem that might be right, but
sometimes in my frame, thequestion is slightly different
way and maybe communicating isslightly different way and then
we learn.
As I said, we learn from thatprocess in trying to communicate
in a different way.
(56:41):
So I love that advice.
It's so true.
Thank you for that.
Speaker 2 (56:45):
Yeah, and I think, as you mentioned also, like, okay,
maybe you only become convincedone time out of 10, but the
other nine times you're stillhave to refine your
understanding.
If you want to at leastconvince yourself of your
initial belief, you need to makesure.
So even then, it's valuable toexchange with other people of
(57:05):
differing opinions.
I guess the last question thatwe were curious about was the
most important leadership lessonyou've learned, and how has it
proven valuable?
Speaker 1 (57:18):
This is an interesting one.
Okay, so I'm not in a At themoment.
I'm leading a very small team,but I'm not in this position for
long, so I am owning a peoplemanager for about two years now.
I wouldn't say I learned a biglesson or something, but since I
(57:43):
start to have this job, I wasexposed to materials and courses
that are for how to be a bettermanager, how to be a better
leader, and all this.
There are a lot of advices toteach you how to show that you
respect other people and how you, I would say, create a better
(58:11):
manager image.
But what I start to realize is,in order to be a good leader I
wouldn't say I am already, but Iaim to be for the future the
thing is, maybe we should havethink a lot on how to show
(58:35):
respect to other people and allthis the real question to ask is
do you actually respect them?
Do you care?
Do you want good things forthem?
Of course you want success foryour team, you want success for
the company, but if you don'tgenuinely care and just try to
(58:59):
behave like you care, it'sprobably not a good starting
point.
So I try to ask myself wheneverI approach a people problem or
something I need to supportother people, I think I try to
(59:20):
tell myself or ask myself do Icare?
Do I genuinely care about them,their career and starting from
there, it's probably a good idea, and that's also what I
observed from good leaders thatI get to see around me.
(59:40):
The better ones are doing this,so I'm trying to do the same
thing and I hope I will do good.
Speaker 2 (59:53):
That's great.
I think that's an importantpoint is, if you actually care
about the people around you, Iguess leadership becomes more
natural as well.
And that's a great challenge,because sometimes I guess you
don't initially always care, butyou have to find ways to get
there.
I don't know if that's beenyour experience.
Speaker 1 (01:00:16):
You can always care about someone.
I think it's more when you addan actual moment, there's a
crisis or there's something youneed to quickly fix, and then
you may be taken over by thegoal that you need to achieve,
and then you start to think alot about the consequences If
(01:00:38):
this does not get done or wedon't achieve this, and then
something will happen.
So we need to quickly fix this,and then you might be carried
away by that emotion or thaturgency of trying to fix
something, and then you start tocare less about your people.
You might fix that thing for ashort term, but a long term I
(01:01:04):
don't think it's good and youmight have more problems in the
future.
So it's just not.
I don't think that's a success.
If you only fix one thing andthen people around you are not
happy or they are not gettingthe benefit of fixing that thing
.
I guess it's just a loss.
Speaker 2 (01:01:29):
Yeah, it's a difficult challenge.
To make everything cometogether cohesively, yeah,
Vivian.
So what are your plans for theyear?
What are your hobbies,interests, anything you'd like
to share with people about whatyou're looking forward to?
Speaker 1 (01:01:49):
Yeah, I mean I'm quite excited to see I do think
we're at a point that, in termsof high speed optical
transceivers, we're going to seesome changes.
How we design them, they willlook different than what they
traditionally look like.
So I'm quite excited to see allthe innovations coming together
(01:02:12):
to make better communicationsystems.
And, yeah, and I do hope, moreyounger people join us and who
are excited about this field andwant to make contributions.
Speaker 2 (01:02:28):
Thank you so much, and thank you, fatima, as well.
It was a pleasure to do this,but, yes, with both of you.
Speaker 3 (01:02:36):
Thank you, thank you, thank you, cherie, thank you,
vivian, that was great.
Popular Podcasts
40s and Free Agents: NFL Draft Season
Daniel Jeremiah of Move the Sticks and Gregg Rosenthal of NFL Daily join forces to break down every team's needs this offseason.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Bobby Bones Show
Listen to 'The Bobby Bones Show' by downloading the daily full replay.