Tutorial: New and Updated - Everything You Always Wanted to Know About Optical Networking

now in here right now we have reached Richards team Bergen who is walking up here who's going to talk to us but is going to give us a whole brand-new new and improved version of everything you always want to know about optical networking but we're afraid to ask if you're not afraid yet you will be already so you can see by the slide that this has updated June 6 because I managed to slip in some last-minute changes as per usual at the last second that's why the PC loves me so this talk is everything you always want to know about optical networking and if you've ever done any of my previous ones you'll know that this one is fairly revised I've probably changed out about 30-40 percent of the content I tried to modernize it I tried to really get into a lot more depth in some areas so hopefully there's some new contents and no one's ever seen at an A&R before so the purpose of this tutorial you know why am I here talking about optical networking at an emag you know the industry is largely based around fiber but there's a lot of people out there that just work with it at a packet level at a router level but really don't have any real technical understanding of what's happening inside that fibre so that leads to a lot of errors misconceptions all kinds of confusion and problems so really the purpose is presentation it's not going to make you an optical engineer or if you feel like it did just remember I left out all the math and the optical is a field full of math so really it's just to kind of help educate everyone about all the different pieces that are involved in running an optical system from the mundane to the advanced and unusual hopefully is interesting so I'm going to start with the very basics of fiber optic transmission so ultimately fiber is nothing more than a waveguide for light so all that means is light goes in one end comes out the other so it's commonly made of glass or silica but it can also be plastic but why do we use fiber in the first place well it's it's really low cost in terms of other materials so even compared to copper it's very low cost glass to cheap it's very light it's pretty flexible it carries a tremendous amount of information so today we have twenty terabit systems all over the place and that's only going up from there you can easily carry lots of independent signals with it so you can you can do things where you can add add capacity over time you can add channels over time you can add customers over time they don't interfere with each other it lets you carry these signals for thousands tens of thousands of kilometres without having to regenerate and ultimately technology keeps getting better and better so we have lots of fiber in the ground that was put there in the 80s and 90s that was designed for 100 Meg services that today is carrying you know terabytes so fortunately the technology of the glass doesn't really change we just get better and better with what we put on it so a quick flash back to high school physics when light propagates through something through a vacuum it's theoretically the fastest thing that can happen in the known universe right so that's that speed is what's called C and it's roughly 300 million meters per second so for shorthand math you just call it 300,000 kilometers per second well light passes through a material that's not a perfect vacuum it propagates slower so the speed that the light propagates relative to that that vacuum is what's called the refractive index so for example water has a refractive index of 1.3 3 so it means when light propagates through water is 1.33 times slower than if it was propagating through a vacuum so what happens when light tries to pass from one medium to another with a different index of refraction is that it you get a reflection instead so if you've ever looked under water you've looked up and you've seen a reflection down that's that's what you're seeing so it works like this it's a principle called total internal reflection so inside your fiber you actually have two different types of glass you have a core and a cladding and they have different refractive indexes so what happens is when the when the light comes in and it tries to go outside of the core when they tried to leave the core hit the cladding that different index of refraction causes it to be reflected back that's what keeps the light inside the fiber and keeps it propagating down instead of leaking out the sides and that happens as long as it comes in within a certain that's called an acceptance cone so there's a angles of light that can come in as long as it hits at just that right angle then it will continue to propagate forever so here's a demonstration from Wikipedia using a laser pointer and a little piece of acrylic and you can kind of see the behavior so what do we actually do with the fiber ruin - the vast majority of fiber that's out there is duplex systems right so we have fiber pairs we run two strands of fiber and one is used to transmit one direction wants to transmit the other direction so transmit/receive depending on which side you're looking at and it's not that that's the only configuration there's there's plenty of ways to do things with single strand fiber but this is kind of the cheapest easiest way if you just have the simplest components and so this kind of holds true whenever fiber is really cheap but like I said it's perfectly capable of doing more complex things it's just a cost-benefit trade-off as to how much you're willing to do that so the most basic distinction you can have when you talk about fiber is what's called multimode versus single mode so multimode fiber is specifically designed to be used with these cheap light sources so what's happening here is you've got an extremely wide core and the reason that you want that is lets you use these cheaper less precisely focused aimed at calibrated lasers or LEDs or whatever light source the downside of this is that it comes with expensive reach so you really you aren't able to get long distances with it but you're able to work with these cheaper optics but yet modal dispersion is the the name of the thing that limits your your reach here and I'll talk more about that in a second so there's a couple different types of multimode there's om one and om two so these are your your common orange fiber jackets that you're going to see when you're walking around a data center these are kind of known as fiddy gray this was originally fiber that was designed for hundred Meg's fitty services so there's there's a you know two versions one with a sixty two point five micron core one of the fifty and they're very similar they're designed for originally 100 Meg 1310 nanometer signals and it really starts to fall apart when you try to do ten gig right it becomes not scalable not tenable so then we came up with another type of fiber called om3 and om4 so there's different versions of it and this what's called laser optimized fiber and this is your aqua so your light blue colored jackets so these these fibers are specifically designed to work with these 850 nanometer short reach lasers they much more easily support speeds like 10 gigs so instead of doing only 26 meters on an I'm one you can go 300 to 500 meters with it with a 10 gig signal now and it's pretty much required for a forty or hundred gig signal which is actually ten and twenty five when you when you get into the insides of it so it's you see this more and more in modern high-speed networking but remember multimode fiber itself isn't actually cheaper multimode fiber is more expensive than single mode fiber so the reason you use it isn't to save money on the fiber it's to save money on the the transceiver so let you use these cheaper transceivers so single mode fiber is the other kind and that's really what you use for all your high bandwidth and all your long-distance stuff so it has a much smaller core it's going to be between eight to ten microns in size and it has no inherent distance limitations caused by the modal dispersions that limit you with multimode so typically you see distances of about 80 kilometers that's kind of the standard magic number between sites without amplification and with amplification you can transmit a signal thousands or tens of thousands of kilometers and and still get it through without having to regenerate but single mode fiber has a massive array of fiber types so it also has something very similar to that om one om two it's called os1 and os2 but it's not the same thing so instead of it being about the type of fiber or about the properties of the fiber it's actually about how it's how its buffered so OS one is what's called a tight buffered fiber your standard patch cables is what you would see for indoor use an OS two is loose and it's designed to be blown into duct so that when they when they have a duct and they'll a new fiber into it you blow it in with air so your classic SMS kind of the most common product that's out there is what's called SM f28 that's actually a corning brand name it's not a standard but there's there's lots of others there's low water peak dispersion shift that non zero dispersion ship it bent insensitive there's there's tons and tons of variants because this is the most prevalent type of fiber out there so here's kind of a diagram showing you the difference and you see how the core is smaller in the single mode fiber and you see here in the multimode fiber because it's so big you're able to get these modal reflections you're able to kind of have the light bounce around more inside the fiber and that's one of your your major limitations on distance so single mode is what you use for all of your high bandwidth long reach stuff so now I'm going to go into the basics of optical networking in terms and concepts just to kind of get everyone on the same page so the very first thing I'm going to talk about is what's called optical power and that's just the brightness of the light or the intensity so what happens when you shine light down a fiber is over time as that light propagates through the fiber some of that energy is going to be lost so that meat can be it being absorbed as it hits the glass particles and kind of gets converted to heat it can actually like not be perfectly reflect or reflected back into the fiber due to like microscopic imperfections and we call this loss of intensity attenuation and we measure that in decimal so it's the same decibel that you use when you're measuring sound and decibel is a logarithmic scale unit so a decibel doesn't actually tell you anything on its own it's what's called a dimensionless unit so all it does is express a ratio between two other things so the reason that we use a decibel is because it well talk about logarithmic values so minus 10 DB it's a logarithmic scale so minus 10 DB is one-tenth the signal and minus 20 DB is 1/100 the signal so you see how we we have that logarithmic scale another easy one is not entirely precise but it's really close is plus three DB is double and minus 3 DB is half so you can you can use that to kind of do your ballpark estimation but remember it doesn't tell you double of what so when you want to talk about an absolute value when you want to talk about like how bright is this relative to something you need a reference so an optical networking that's called a DBM so that's a decibel relative to one milliwatt of power so for example a 0 DBM signal is 1 milliwatt of power that's an absolute value that's a intensity of light so 3 DB is 2 milliwatts minus 3 DB is point 5 millivolts so a lot of times people ask what what actually should show up or what should happen when you have zero light in the signal and the correct answer is negative infinity DBM that's that's what that is but confusion between DB and DBM is probably one of the single biggest mistakes that you'll see in the real world just people don't know the difference between the two people will work they're light meters wrong they'll be confused about which one they're looking at and they'll just give you hideously wrong answers because you're these are two different ways to look at it one is when it's lost and the other is or change and the other is an absolute value so be clear about the difference between the two so why do we measure with a decimal what light just like sound follows what's called the inverse square law so think about it this soja signal is is traveling so a signal travels distance X and it loses half its intensity it travels another distance X it loses another half of its intensity so after 2 X you're down to 25% of your signal a 3x you're down to 12.5% of your signal so if you use a logarithmic scale to to explain this that it simplifies the calculations you can now do this with with elementary school arithmetic so you don't have to think about this analog of a logarithmic way you can just think oh an X plus an X plus an X I've gone from 3 DB to 6 DB to 9 DB very simple very easy to think about that's that's the only reason for using it so here's a little table that people could reference later that kind of shows some of the conversions but you know when you when you get down into even even 20 DB is only one percent of your signal left right so you very very quickly have a fall-off and then when you start to get to these very very low signals that's where you see you've got almost nothing left but technically 50 DB is still you've got a little tiny little bit signal still left in there so another concept in optical networking is called dispersion so dispersion just means to spread out so in optical networking this results in signal degradation so as the signals dispersed it's no longer distinguishable you took what was once a 0 and a 1 and something that was very clear that you could you could look at and you've now turn it into this blurred mush so that's that's all that means there's a couple different kinds of dispersion so when the most common ones you'll hear about is something called chromatic dispersion and what's happening here is in a non vacuum material different frequencies of light propagate at different speeds so this is actually how your optical prism works right so a higher frequency of light will propagate at a different speed and a lower frequency of light it causes it to spread out that's giving you that prism effect CMB called CMD for shorthand is one of the fundamental limitations and historic with a huge limitation in the the optical systems and what kind of speed we could transmit and it really becomes a limiting factor when you try to scale up with baud rate and I'll talk more about what that means later but this is just a fundamental concept of of networking the other major type of dispersion is something called PMD polarization mode dispersion what's happening here is you've got this this cylindrical fiber but nothing has ever manufactured exactly perfect so what happens is different polarizations of light will propagate faster so you see here you're looking at a an x and a y axis of this 3-dimensional fiber and at the end of some distance of fiber the horizontal polarization might have arrived at a different speed than the vertical polarization and that difference is called DGD differential group delay and that causes impairments to the fiber it makes it harder to to recover that signal so in fiber optics there's a couple of different transmission bands different windows that we talked about so there's there's really three big ones 850 nanometer is what's called the first window so this is really this is your area of highest attenuation it's really only used for short reach applications so you see it today with these these in data center in cabinet type very short reach applications but it doesn't doesn't do much more than that 1310 the second window is what's called the OBE and so this was historically the point of zero dispersion so dispersion as it goes through fiber changes depending on the frequency of light so this was the original latest spot that kind of had the least dispersion it was the easiest to work with so when people are starting to first deploy fiber optic networks this was this was the band that people used and you still see this used today for your kind of your medium reach your 10 kilometer type applications and you can you can do further with it you can do 30 40 kilometers on it if you really try but more likely these days is people have moved to the third window which is the 1550 nanometer window and this is called the C band let stands for a conventional band so it actually covers a range of 15 25 to 15 65 nanometers it's the point that has the lowest rate of attenuation so the least amount of loss over the fiber which has kind of become the the thing that we care about the most and it's really you're going to see this used for almost all of your long reach and DWDM applications and then there's a fourth window called the Elbe and the long band which is essentially a complete duplication of the c-band just a little bit higher up and if you get into places where you're really constrained on fiber and you really want to double your capacity you effectively you kind of have to put in a second set of every piece of tool to work with it but it basically doubles your capacity on a on a pair of fiber by going into the Elbe and so here's a little diagram showing a difference between them and you can see the progression of Technology so you see this dashed red line up at the top here this is kind of showing the behavior of the fiber in the early 80s so you see these different Peaks so that the peaks are showing you the optical loss the the rate of loss per per kilometer and you can see that there's these certain Peaks that were completely unusable because they were very high lossy this was called water peak so I'll talk more about that later but this was a section of the the band that was unusable and you see how the windows were kind of adapted around those those properties of the fiber itself and you see the modern fiber got better and better and better until these days it's not nearly as big a problem but that's kind of a lot of the reasoning for these original bands to have kind of been used another major concept something that that's really critical when you start talking about especially modern fiber systems is what's called the O SNR the optical signal to noise ratio so really the the difference here the thing here is you're looking at the difference between the noise floor so you see here we've got we've got some amount of optical noise in the system and the actual signal and the ratio between the two is your snr this is really one of the most determining factors on a modern system as to what you can do with the network as long as you keep your OS and are high you can you've got a good clean signal you can continue to amp it you can continue to do more things with it once that starts to fall down that's where you you hit problems so it's a key concept so now I'm going to talk a bit about Wave division multiplexing so WDM we all know that that light comes in many different colors so when we look look at light and we see white where we're perceiving it as white but it's not it's a mix of kind of all the other colors so what happens when a WDM system is you can actually mix different colors of lights and on the same fibre and they will propagate next to each other ideally not interfere with each other kind of traveling as ships on the night and it lets you increase your capacity so instead of only being able to carry one signal you might be able to carry 20 or 40 or 80 or more that's that's a concept that's what we're going for and you see here a little diagram showing the multiple colors of light being carried so there's so looking at the the WDM channels there's there's two main structures actually it's probably not a terribly good diagram I'll discuss regions so CWDM is the there's two types of WD map there's what's called coarse and dense so CWDM is pretty much you loose lea used to mean anything that's not DWDM so it's not a really well hard to find standard it kind of covers a lot of stuff one popular meaning especially historically was what something called eight channels with 20 nanometers in spacing between them so it's very wide channels and then if you were to add a low-water peak fiber you got access to another ten channels so you tend to see this a lot like this was very popular back in the one gig days when it was much cheaper to make one gig CWDM that was DWDM but it can also mean a lot of those things right it can be used to refer to 1310 1550 MUX and kind of any other combination notes of light DWDM is something that's actually well defined so there's an itu specification that defines a grid of channels and there's many different grids depending on the different channel sizes so the common ones that you'll see out there in the world are 200 150 and 25 gigahertz and the smaller the gigahertz the smaller the channel so your 200 gigahertz systems this is kind of typically your 2000 era technology you rarely rarely rarely see this in production anymore it's kind of old 100 gigahertz stuff that gets you your 42 48 channels kind of depending on on exactly how you do your channel allocation this is this is something that you see very commonly so you'll tend to see a lot of 100 gigahertz channels in Metro deployments largely especially where you're able to kind of bypass it muxes and tune DWDM channels the 50 gigahertz spacing is a lot more prevalent in commercial systems so this gets you between 80 to 96 channels so it's a point for nanometers between the channel so it's much smaller and then for awhile there was a there was a systems that use 25 gigahertz you could actually get between 162 192 channels what you saw happen in technology was we move smaller and smaller and smaller we moved all the way down to 25 and then we move it back to 50 because what we were doing with 25 was cramming high-density 10-gig systems in and then as we started to move to better technology we were able to work better with wider channels but these days you end up with these modern systems that are very flexible so you typically see systems that can be done in twelve point five or even six point seven five gigahertz increments so what are the advantages CWDM lets you use these cheaper less precisely focused lasers or more they need to be cooled as actively some what tends to happen is a you know a laser will transmit on a certain signal and then as it warms up that will will waver over time and so if you if you have a cheaper laser you don't need as good a cooling system or you don't need to actively cool it it just makes a lot easier to produce so if you've got this very wide window you the signal itself isn't inherently any wider but it gives it more range that it can it can move around in with DWDM you get a lot more channels that are possible in the fiber so all the way up to 160 to 190 to versus 8 in a massively much larger chunk of spectrum when you think about kind of a classic CWDM the other big advantage of DWDM is that it can stay entirely in the c-band and that's a band that gets a lot of other properties so one of the biggest ones is what's called ed says erbium doped fiber amplifiers this is the range of frequencies where these work to amplify the light make it brighter over time I talk more a lot more about this later but that property lets it let you work with that in a much easier way and this can all be duplicated again like I said in the elbe and effectively it's just a straight duplication so here's a little diagram showing the kind of relative differences in size between the between the channels and you can see here how wide a kind of a classic CWDM channel is so there's like I said there's there's a bunch of other definitions of it too so some other things that you'll see kind of used for WDM there's a classic 1310 1550 MUX so let you take your two different bands and put them together over the same fiber so it might let you use a an LR and ER optic together on the same fiber these days we see a lot of work on what's called four-lane ray optics so what happens here is whenever there's a new technology whenever there's leapin speed it tends to be much easier to implement it in parallel first and then it becomes serial later so you saw this when we move to 210 gig so some of the earliest standards that were much easier to implement was one called LX 4 so what it was was for 2.5 gig Cannell's inside rather than one native 10 gig channel and over time that it flipped right it came to a point where all the extra work of dealing with the four channels became more expensive and more difficult to work with than that native 10 gig but still in the world of 40 and hundred we're still in a world where largely especially on the client side you need parallel optics to make it work so 40 gig for the most part is really 4 by 10 when you actually look at it under the hood that's all it is and so it has these different you know 20 nanometer spacing grids a hundred gig when you look at it is is largely 4 by 25 under the hood and it has a different set of grids and they're of course different just to be a pain but you also see some other other things out there single strand optics there's a thing called BX bi-directional and what this does is it lets you without having to have any other knowledge or expertise take your pair of fiber and operate over one strand so you could just instantly double your capacity or you could work with a single strand fiber network and it's it's just transmitting and receiving it at different frequencies of light and having a little tiny MUX integrated into your into your pluggable transceiver so a quick word about DWDM channel terminology you see here is a diagram showing these different channels so the width of the channel itself the amount of signal that you can actually pass through it is what's called the channel pass band in this matter is because when you when you look at 100 gigahertz MUX and it says hundred gigahertz that doesn't mean you actually get 100 gigahertz that passes through that channel that means that there's a hundred gigahertz spacing between channels the amount of signal that you actually get through is the pass band and that can differ depending on different technologies and different types of optics require wider or narrower pass bands or you know Kim can be packed tighter if there's a wider pass man the like I said the the when you see hundred gigahertz or 50 gigahertz that's your channel spacing that's the ultra distance between Center of both channels the the adjacent channel isolation is how much how much this this filter is cutting off one channel from another so there's a little bit of interference that can can happen between channels even when you've got an active filter trying to filter that out even inside of the channel itself you can have variations so within within a channel you can have portions of it that go up and down that's called the ripple and then the the total channel ISIL isolation really the the signal from the noise floor is what's called the total channel ISIL ISIL isolation so now I'm going to talk about WDM networking components so probably one of the most common prevalent things that you'll see out there is what's called the passive mocks this is pretty much the first thing that you need to do any type of WDM it's a simple passive unpowered device that combines or splits multiple colors of light so you'll hear this call to a multiplexer sometimes as a filter or prism filters historically how it actually worked and actually these days it's kind of kind of gone back to being a prism but it's you know either way right you're splitting or combining different colors of light so in a complete system you actually have both components you have a MUX and Adeem ox historically these these were separate components and they still are but they kind of work in either direction so the technology works it doesn't care whether it's receiving or splitting or combining or splitting it works both ways but you need two of them and typically just see this sold in a single package so here's an example of a simple little 1u component an OE DM is a system that lets you selectively add or drop channels without paths so you can pass a channel through undisturbed while adding or dropping others so this is really useful when you're building these large rings when you when you want to have multiple multiple systems connected together you want to be able to have an A and a Z talk to each other but still have a BSC and a D in the middle that kind of drop signals and when you build a kind of a well-constructed o EDM ring you can reuse the the nodes you can reuse the channels between different segments you can do all kinds of nice things but an o EDM a classic om or what's called a photon these days a fixed OM is an entirely passive on power device it's it's very similar to a MUX and just let you add or drop certain channels so a little bit about the actual technology that's being used here so there's a couple different ways to to make a MUX or no idiom the three basic ones are the thin film filter the the fiber Bragg grating and what's called an a WG and a raved arrayed waveguide grading thin film filters are kind of your older technology they're they were really popular for for your 200 gigahertz arrow type stuff and they're actually still really useful they're still good when you've got these low channel counts right so if you've got an 8 channel MUX or something like that that's typically what you're what you're going to see an f BG is is kind of similar it's what's happening is you're itching inside the core of the fiber you're creating this pattern that causes certain frequencies of light to be reflected and when you pass it through this it splits your your fiber out and AWG is really all the modern technology it's kind of how all of your high channel count systems are built it's actually really interesting how it works so what happens is you've got these these two what's called free space propagation areas so it's basically this very wide prism that splits out all the light and then you have two of these right you have one on the on one side one on the other and you have these different these many many many strands of fibers that connect the two and they all have very precisely measured links that are that are slightly longer longer and longer as they progress what happens is as the the light spreads out and propagates through these many many many different fibers and then comes back together it constructively interferes or destructively interferes with each other and that resulting reemergence pattern breaks the signal out into your forty or a year or higher channels so this is this is really the highest and the nicest system to get you your low insertion loss and this is what almost all of your commercial muxes are going to be these days a DMS kind of moved beyond their their classic fixed model into what's called a Rotom a reconfigurable om and what's happening here is it's literally just like an om but you can log into a box and you can tell it to instead of it being fixed only dropping these four channels or these eight channels you can tell it which channels drop so when you talk about a Rotom you usually talk about it in what's called degrees so the simplest example of an East and a West here this is a diagram showing a two degree Rotem so in this in this 2d Rotem you have basically two options you can either pass it through or you can drop it and you see there's both sides both degrees from both ends are doing that same operation so there's a WSS a wavelength selectable switch is a device that lets you steer the wavelength and this configuration you like I said you have two options you can either pass it to a or B if you pass it to a you send it all the way through and if you pass it to be you send it down to your MUX and you drop it but you can build these four degree a degree 20 degree rotom's are now fairly common so lets you build these very complex topologies you can you can have a central fiber hub and you can selectively route which pair of fiber which paths you're going to take so you can you can build protected systems you can reconfigure you can do all kinds of more rapid provisioning so rotom's have evolved quite a bit over the years so if you you look at how a classic basic Rotem mark like a first generation Rotem really what it was was like i said that WSS that selectable ability to steer a wavelength out port a or b but once you've steered it you sent it to a MUX you send it to the same passive box that you use historically for all of your other applications so the downside of this is that it Maps specific frequencies to specific ports so when you say oh I need to go turn up a new channel I need to turn it up on channel 31 you actually have to send a person into the field to plug a fiber into a port labeled channel 31 and that's the only channel that will pass through that so that's 32 evolved people said is there ways that we can eliminate the need to do that and make it easier to to software reconfigure this sort of having to send people around to do everything so the next evolution you saw was the color of this Rotem so what this does is it effectively eliminates the need to map to a specific port we are still very limited to to muxing or D MUX in one direction at a time so think of it like a you know replication right instead of it's sending the port the color to one specific port it sends to all the ports but it still only lets you send to this one one MUX so you get one direction one thing and if you ever want to change that you have to is still send someone out to the field to move cables the CDC Rotom is is really the modern Rotom so it stands for colorless directionless and contentious so what's happening here is you get that colorless property where you can use any port that you want you get the directionless property where you can actually see your instead of it only being able to be passed through or dropped you can say I want this to be sent to my north degree or my east degree or my West agree you get you get that flexibility and contingent list means that the channel itself can be reused within the MUX and there's no there's no inherent contention inside of it so you might have channel 41 being used going east-west you might also have it being used going north-south and it the MUX itself isn't going to care about that so why do this well the goal here is to be able to move to a system where this can operate entirely in software so the transponders one of the nice things you could do is you could reallocate the the transponders which physical path they go on depending on things like time of day fiber cuts any type of operational need that you you want so the big advantage here is when you think about the cost of a transponder train the transponder itself is a very expensive component and if you can have fewer of those out there and and dynamically change it so that the same transponder can you know fill a link going east during its peak traffic time and going west during its peak traffic time and you don't need to buy - that's a significant savings to the network so what's happening here is is people are moving to a system where this can all be automated so there's there's a system called peak app path computation element protocol where the routers are kind of able to do real-time traffic measurements and say hey I know that I need to pass 17 gigs from this device to this device it can take that information and send it offline to a control system that can then dynamically reallocate the optical waves so basically the network is calling upon the optical layer to say I need more or less capacity and it can tune up and tune down it can change the layout of these different waves over time so potentially opens up a lot of flexibility but it's also a lot of Interop issues and is historically always very difficult for the optical guys so just to give you guys an idea of kind of what goes into a modern CDC Rotem just to show you some of the complexity so the very first thing that you have to do when you come into a Rotom is amp it because you're going to lose a ton of power inside of a Rotom so you go through let's call it a modern one by 20 WSS so you've just lost a ton of power going through this so the very first thing you have to do is send it through another amp so you have this array of amps basically this WSS lets you let you send this signal to any one of these degrees so you basically wire them all up and then after you you select which degree you send it to which other degrees send it to dand you amp it again so there's these arrays these very large arrays of amps inside the system and then it goes to what's called a multicast switch so multicast switch is just literally an array of 1 by 16 so here's an example of an 8 by 16 multicast switch all that is is 8 different 1 by 16 splitters so you take the light you split it into you have 8 different ways to split it into 16 different things you just you spread this light out like crazy of course you got to amp it again because it's now very dim after that but at the end you've taken this and you've replicated this signal to all of these different ports so any one of them is capable of receiving the signal without having to be locked down to a specific thing but you see there's a lot of complexity there's a lot of loss there's a lot of amps that go into this system so you pay for it too so another concepts in d2 BDM is something called super channel so imagine what happens if you've got a bandwidth requirement that can't be delivered in a single carrier so what you can do with super channels is pack multiple carriers together in a single channel so let's you get more efficient in your bandwidth it lets you deliver a single data rate channel even if your optical technology doesn't scale to support it and all kinds of other nice things so here's an example showing what would happen if I needed to build a one terabit carrier a single single carrier I want a one terabyte Ethernet port because Randy Bush is in the audience and he says we need Terra vis Ethernet how would I do that so you have a couple options right if I was to do that with a single carrier today I would need this massive I needed like a 375 gigahertz wide channel I would need 320 gigabyte modulation I would need something that just doesn't exist today and probably won't exist for the next 10 years but what you could do to solve the same problem is you use 10 different subcarriers you can take 10 different components that we do have the technology to build today pack them in together in the same amount of space and deliver that same care a bit Ethernet type type client handoff if you want to do so that's what's happening at super channels is you're justyou're you're using sub carriers to to do different things optically so DWDM channels themselves have actually evolved quite a bit over the years so historically they they win one direction right they went smaller she started at 201 to 101 250 and with 25 and then like I said after 25 you actually started to see it moving backwards you moved back to 50 and now we're in a world where we want to go higher than that and so the reason for that is you know if someone says oh I need a 400 gig e or I need a 1 terabyte Ethernet solution how am I going to do that you can't pack that into a single small channel so what we've moved to is a system called flex grid what this does is it lets you in software reconfigure your MUX so that you have different channel sizes there's no longer one uniform single channel size that propagates through the network so if you've still got a you know a legacy 10 gig system on there and it only takes a 25 gigahertz worth of the channel you can give it a smaller piece of that spectrum meanwhile if you've come up with something that is really high bandwidth and needs a much wider Channel you can give it that dynamic Channel and this can all be tunable like I said down into the 12 to 6 gigahertz increments so this is this is what you see modern carriers deploying and any new system and gives you that next-gen flexibility and protection so now I'm going to talk about optical amplification so optical amplifiers are a way to increase the intensity of the signal so this happens purely optical purely at a physics level you don't need to regenerate the light you don't need to do what's called an OE oh you don't need to convert back to electric and then back to optical so there's different types there's there's different things there's different ways to do this optimum amplification so the three main things that you'll hear talked about the three main applications or what's called the booster the preamp and the in line so the booster amplifier is something that's designed to be used right at your transmitted it's it's designed to work with really high power levels and to get the signal as powerful as it can before it goes out into the fibre the in line amplifier is something that's that's in the middle so it's designed to it might not ever be going to a receiver it's just a hut somewhere in the middle and it's trying to make the signal brighter and a preamplifier is something that's actually designed to to be used on the receive side so what you're going for here is you want something that works with really low input powers you want something that doesn't minimize noise you need to make it bright enough that your receiver can see it comfortably and everything can work and like in line amplifiers kind of strike a balance between the two so these are the three types that you see so the most common amplifier that you'll see in though in the world is what's called an ED VII and erbium doped fiber amplifier so what's happening here is you've got a piece of fiber that's been doped with erbium ions and it could be literally as few as you know one per per millions and millions of atoms there's somewhere in the silicon there's some erbium that's been mixed in there and what happens is you you have a pump laser you have a laser of a different frequency that's injected into this piece of doped fiber via coupler when that pump energy comes in it excites these erbium electrons they move into a higher energy state and then when you have a 1550 nanometer when you have the c-band colored photon come along and hit one of these excited erbium ions that electron jumps to a lower state and it simultaneously emits a clone of the photon at the same frequency in the same direction the same properties so you basically you've made the signal brighter so there's there's two there's two states there's what's called the the unstable state in the quasi stable state so there's two different colors of optics that will do that and your best ones will actually use both but so the two the two frequencies are in 980 and 1480 so why can't you use this to kind of clone a signal forever well in addition to making the original signal brighter you also generate something called amplified spontaneous emission or just noise so the way this works in physics is whenever you've got this excited erbium ion you've got it now at a higher state and about ten milliseconds is the number once ten milliseconds have gone by and this excited electron hasn't encountered a good Photon it spontaneously decay and emits a junk photon and this is now noise it's not any any part of the signal it doesn't do anything good and once it's generated you can never pull it back out it's now a part of the signal it's indistinguishable from the original and after enough hops you ruin the signal so here's a diagram showing kind of what that that game profile looks like and you see there's two channels here that have been amplified but everything else that you see here that's noise that's that's junk that's been injected into the into the system so why does this really matter well what I'm trying to show here is one of the properties one of OSN are like I said o SN R is one of the most important things when you think about how do I get a long reach system and really what you're fighting here when you're fighting noise is what powerlevel did you receive the signal and so if you let the signal get really low before you amplify it you create a ton of noise if you amplify it before the signal has gotten really low you create very little so you see here this top line we start with a 24 DB o SNR it's a very very good very strong signal and you see that we amplify it 10 hops but we never let the signal get below negative 5 DBM before we amplify it you see that we have almost no loss in L SNR at all and you see as we move to 10 DB and 15 DB and 20 DB you see how quickly this falls off and then if you if you let the signal get down to negative 25 DBM before you amplify it you see here that at the end of 10 hops we've gone from 24 a very strong signal down to 17 a relatively weak signal that wouldn't support some of them the modern technology so when you're designing a system you're always thinking about how do i amplify before i let the power levels get too low that's why all of this matters another type of amplification that's different from Ed VII is what's called raman so it works on a principle called stimulated Raman scattering which I'll talk more about later but what's happening here is you've got instead of a piece of fiber that's been doped you're actually using the align side itself it's something called distributed so you're actually using the transmission fiber as your gain medium so instead of you and acting a pump laser into into a you know might only be as one meter of fiber that's sitting inside of your MUX you're injecting injecting a pump laser out into your long-haul transmission system and as it goes out into the system it encounters the the real signal and it causes a transfer of energy so the the pump signal is depleted and the original signal that you're trying to amplify is made brighter you see here that the three stages as that happens and the most common application that you see when you do this in Raman is something called counter-propagating so what's happening is you're actually shooting the laser the opposite direction of the propagation of the signal so here's why this matters in a net the only amplification here's here's an example where you start with a very high power level and you it declines over time as you propagate through the fiber and you hit an amp and you amplify ting and you make it brighter the two areas that you want to avoid are two higher power and two lower power if you go too high you get into what's called nonlinear effects this is where the that you start introducing all kinds of weird properties that aren't easy to predict and aren't easy to deal with it start to calls impairments in your signal but if you let the signal get too low that's where you start to introduce the high noise so this is this is showing an advanced application you see the penalties that you're going to pay here you had to launch at a very high power to get it out and you had to receive at a very low power but before you amplified it so you've lost a lot of OSN are in the process in a hybrid EDA Raman amplification what happens is you add a Raman pump and you send it the opposite direction so you're sending it back the opposite direction that the the signal is coming in what happens is it actually makes the signal get brighter within the fiber so you see here in this red diagram that you needed to launch at a lower power so now you've stayed out of the zone of nonlinear effects and then as it started to get closer and closer and closer to the end it encounters the Raman pump and it starts to get brighter and brighter and whiter so by the time you receive the signal you've now received it at a much higher strength you've now avoided that noise that was being introduced so this is very important this lets you get much longer reaches and keep much higher o SNR and lets you do more data over over longer distances so does it matter we'll look at this example let's assume that we've got a 21 DB OS and our signal and we've got 23 DB of spacing between these things so it's kind of your 100 kilometer between huts very very common in at the only system we were only able to do seven hops so we were only able to do 700 kilometres as soon as we add ramen to this we're now able to go to 2,000 kilometers we're now able to get 220 hops and in fact you can probably do better because this is assuming a very very high OS and our and technology has actually gotten quite a bit better but you see here because you you were able to amplify with the ramen the signal you no longer need is much gain on the FO when you when you send it and you've kept that higher OSR so another big thing that comes into play and you think about amplifiers is power balance so it's unbalanced channel powers calls all kinds of problems you have a system that that's not stable one of the problems is that amplifiers the amount of gain that they introduce varies across the frequencies so the very first thing that you see happen to compensate for this is something called a glint I gain flattening filter so this is really a static attenuator that's designed kind of match the game profile via and it's really engineered it's very specific to the SSL so when a manufacturer makes the edge' they look at the gain profile that they get out of it and then they design that gain flattening filter to be the inverse of it so what you're trying to do here is you know you might have had this one channel be much brighter in this other channel be much lower you want them all to be consistent flat so you kind of you flatten them all out so they're all about the same frequency but even these very small power variations that are going to sneak through this add up over time so eventually you need something called DGE dynamic game Equalization and it tends to be you know every seven or eight hops or something like that but as these over time as these different amps start to incur these little little bits of variation you end up with with a couple of DB of difference between your channels and that causes problems so typically what you see in most works as you use a Rotom to fill this role but you can use other things as well but every so often you need something that's actually aware of the content that takes a look at it and and rebalances the channels as well so amplifiers also have limits in their total system power both and what they can output and what they can input so this matters because the total input power changes over time as you add a channel so imagine you've got a single channel and you're sending a plus 0 DBM so you've got 1 milliwatt of input power now you go to a DWDM system where you've got 40 channels and each one of them is outputting the same amount of power so you've now got 40 milliwatts of power that's plus 16 DBM of power that's actually incredibly bright strong light so if your amplifier has a maximum input power of minus 6 and you run all 40 channels that means before you put the channel into the amp you needed to attenuate it all the way down to negative 22 so you need to think about these different types of amplifiers and the different power balance requirements for the channels to make the system work another thing that happens is this can change over time as you add or lose channel so imagine you're in a situation where you've got these 40 channels going through and now all of a sudden your data center loses power and you lose half your channels suddenly the power levels across this amp system have changed dramatically so you need a system that is is able to react to that quickly and a good fo will have a constant monitoring loop that will look at the input signal look at the output signal and will make adjustments dynamically to keep the power levels right and the best edge cells will actually communicate with each other they'll have a communication channel band they'll talk to each other so it'll know what the what the SS upstream are doing it'll know what it's doing and it can communicate and adapt to changing fibre conditions and spices and things like that without you ever having to go in and configure things so some other optical networking concepts here tells just a lot of material so one thing you'll see out there is something called an optical switch what's happening here is it's letting you move the light around between ports without doing that Oh EO conversion without going from optical to electrical back to optical so the most common design that you'll see as something called 3d MEMS and it's basically an array of these tiny little mirrors that are controlled electrostatically so you can very click very quickly but then within milliseconds or less steer these mirrors so you can take a take a mirror and take a channel of light and say i want it to go to port 1 and 50 milliseconds later i wanted to go to port 10 so you can use this for optical cross connects you can use this for fiber protection probably the the most common area that you see it where it makes the most economic sense as people who do lab work where you've got a system where you're constantly moving and reconfiguring cables it could very quickly pay for itself i talked earlier in the in the road i'm about the WSS that's a component called the wavelength selectable switch and it lets you route these individual wavelengths across ports so the first generation WSS is used these these 3d men's technologies and then as we move to more modern ones they actually use what's called al-kahf liquid crystal on silicon you're kind of able to to dynamically steer these beams of light so there's a lot of very interesting physics that are going into these these high-end Rotem systems to make all this work there's a component out there called a circulator that's typically not seen by the end-user but you you use it quite a bit in manufacturing these components so the way a circulator works is it has three ports so when light comes in port one it goes out port two and the light comes in port two it goes out port three so here's an example of using a Bragg grating filter combined with two circulators to create an O ADM so you see when the light comes in from the left side here it goes out to the right and when it hits the the Bragg grating filter which only reflect reflect certain frequencies of light it sends it back and it sends it to the drop port so here you get this behavior I want all these frequencies to propagate except for this one frequency which matches the filter that sends it back and now the drop board and you see these views in other systems I use them quite a lot when you're building single strand systems you'll you'll see them inside muxes you'll see them inside dispersion compensation spool so you get to reuse the same fiber twice you'll see them inside amplifiers for the same there's another component out there called an optical splitter and it does exactly what it sounds like it splits the signal so some really common examples of things you're going to encounter in the real world the 5050 splitter is typically what's used for your very simple optical protection so the philosophy here is you take your signal and you split it in half and you send it down to different fiber paths and then on the other side you've got an optical switch and all it's doing is looking at power levels and it's using whichever side has the strongest so it might be using side a because that's your your short path that's where your strong a signal is now let's say you go out and get a fiber cut and that line goes down within milliseconds it can detect the loss of power and it can switch to using the other strand and you did all of this for a 3 DB hit or a 3 DB to split the signal in half so it's a very cheap way to protect me imagine you had you had terabits of bandwidth you had dozens of transponders are hundreds of transponders and millions of dollars of components rather than having to duplicate both of them to light both systems you could throw this very simple component in and use both sides of the fiber to get that active protection and deal with fiber cuts another thing that you'll see out there is the 99 one or the 98 two splitter and this is often used for what's called OPM optical performance monitoring so what's happening here is you're tapping a very small portion of the signal and you send it over to a little spectrum analyzer and that spectrum analyzer looks at what's happening it can tell you what your OS and our levels are what light channels are in use what power levels are it gets you insight to your network without you having to send a person out there so very very helpful when you're troubleshooting things another major concept has becoming increasingly important is something called feck and that stands for forward error correction so what's happening here is you're adding extra redundant information to your transmission so that when the receiver gets it it can actually computationally recover from any errors that it sees so think of it like raid 5 for wavelengths right you've added a little bit of extra information but in the event of data loss you can computationally recover it so in practice what happens here is you use spec 2 to lower the required Oh snr so you take a signal that otherwise would have been come pletely unusable because it's gone past the point of being able to be decoded and by adding a little bit more overhead you can now make it go much further so a really common example is you can take a 10 gig signal and Pat it by 7% and make it go from from 80 kilometers of reach to 120 kilometers of reach not by doing anything other than making it possible to kind of recover that signal better so this where this really matters is when you're kind of upgrading old DWDM systems because the problem is it's usually not practical to go out and move your huts closer moving cuts is a huge construction operations cost lots of money you really don't get much flexibility when you're doing that so when you want to move to a new generation technology an easy way to solve that problem is to make the feck systems better so what you've seen here there's there's three generations effect the first generation typically something called RS factory Solomon's Beck so you tended to see like a 6% overhead and you got about 60 60 Bo what's called net coding ganger 60 be more signal that you were able to work with versus what the base signal would have been for OS and requirements when you went to second gen you added a little bit more you went to 7% overhead that you got eight to nine DB of game then when you went to third gen you really you moved up quite a bit you you move into something called s defect soft decision effect so you're adding a significant amount of overhead you might add 20 to 25 percent more data to the encoded data but that's getting you a 10 to 11 DB gain of OS and R that can actually be huge that can make the difference between a system working or not working that can make a difference between a system that can do 100 gig per channel or they can do 200 gig per channel quite easily so it's worth it to kind of put all that extra data in and you see feck has actually become very critical to even even kind of commodity pluggable components as well so things like hundred gig SR for technology pretty much require effect to work even though you're only going a few meters over a local patch cable you you've kind of got that built in and you'll see that more and more and more most systems have to have feck built in just to work with kind of these commodity optics so here's a diagram showing the benefits that you get from these different levels of forward error correction so you see here an unencoded signal what this curve is showing is as the Oh snr as your x-axis gets gets further to the left your OS and ours going down you're moving up this curve of the bit error rate so you're getting more and more and more bit errors and this happens in an exponential curve so in a completely uncoated signal you might need in this example an 18 DB Oh SNR to get this this low rate of bit errors that you want by moving to different generations effect you see how far back we pushed it to now maybe I only need 9 DB of Oh SNR to get that same performance level that lets me get much much much better performance out of the system another thing that you'll see referenced when you when you talk about optical technology something called OTN digital wrapper technology or G 709 so this stands for optical transport network and think of it as a replacement for SONET and SDH so what it's giving you here is the ability to MUX and D mocks at a TDM level you're able to take for example a 100 gig wave and turn it into 10 by 10 by by taking these individual components and putting it inside a larger channel so historically saw it in SDH were completely wavelength unaware they didn't know or think anything about WDM so OTN technology makes it so that they are aware of this you can like I said take it 200 gig service break it up into 10 by 10 and for the most part when you buy a 10 gig wave you almost never get it as a 10 gig wave you get it as 1/10 of 100 gig wave that kind of thing so it lets you mix multiple protocols inside the same wavelength it helps with troubleshooting and just think of it like a like a sonet/sdh extension so now I'm going to talk about all the many many many different types of single mode fiber out there so I already discussed that single mode fiber is is really essential for all of your long reach applications but there's many different types of it so the most common types that you'll see out there these are the basic standards so you're very you're very common basic standard single-mode fiber as what's called itu-t g65 - you might hear that called SM f28 like I said that's a corning brand-name but that's a very common name because it was one of the most popular Ivor's of its type you saw an evolution to what's called full spectrum or low-water peak fibre so this was just that same type of fiber but now we remove those water Peaks we started to make those certain frequencies that were unusable before usable again and you see it's actually the same standard it's just a newer version of it so as we move to two new versions of g65 2 they got better and better and better about removing that water peak and that's mostly what you see out there then there was a technology called dispersion shifted fiber g65 3 and that turned out not to be a terribly good idea so then there was a technology called non zero dispersion shifted fiber that was G 6 v 5 there was a low loss fiber so G 6 5 4 is typically what you see for for oceanic systems undersea cables we were very specifically concerned about low loss because you really don't have a way to to go in and kind of swap out cards you get access to the the fiber once it's in the water and you've got those amps out there and then you've got another standard called G 6 v 7 which is your bend insensitive fiber so your standard single-mode fiber this is the fiber that you saw that was kind of out there and these original fiber cables so it's if fibrin was kind of deployed in the 90s you can almost guarantee this is what it was like I said sometimes called SMS 28 you'll also see it called NDS F non dispersion shifted fiber so just kind of basic basic fiber this was originally designed to be used by the 1310 band as I said so this all happened before WDM was ever was ever a thing all right this in fact this fiber was deployed before amplification was even discovered or available or commercialized so ironically what's happened here is we've actually gone full circle we've gone to creating all these different fiber types to try to optimize the network to these days this standard fibers actually the preferred type of fiber to do high bandwidth systems so ironic the low water peak fiber so what I talked about earlier those those water Peaks what happens here is you get these hydroxyl molecules they get into the glass during the manufacture of the fiber and they absorb certain frequencies of light and the manufacturing techniques just got better and better and better to kind of remove those and you see the removal of of that water peak or turning into something that's very low that's all it is so dispersion shifted fiber was an attempt to improve the situation so what happened here was they engineered the fiber in such a way that it moved the point of zero dispersion over into the 1550 nanometer ban so they thought this will be great will have now fiber that I can send low dispersion and low low loss and everything will work great and it turned out to be a really bad idea so this worked well for single channel systems before the advent of WDM and the problem when you start doing any type of WDM on the system is it causes a lot of nonlinear interaction I'll talk a lot more about that later on but basically this type of fiber was a giant dead end and you almost never see it in the real world it's pretty much used for pull string and nothing else so then they said well alright how can we how can we do this but how can we fix this so they move to something called non zero dispersion shifted fiber which is very similar in concept but they move that point of the zero dispersion so that it's just outside of the the range that you're you're actually sending your signal out so you still get a little bit of dispersion but it's it's not you know is not nearly as high as it was but it's not zero and so then to manage this when to compensate for dispersion over time what you have are two different types of fiber so you had your plus or minus what happens here is these have different slopes so your transmission fiber will will cause the signal to spread out over time and then at the end of the transmission will come to somewhere where you compensate for it and you basically run it through a spool of fiber that goes the opposite direction and it recompress is it so with physics purely purely at a fiber level you've kind of compensated for that dispersion that happens over time that was how a lot of 10-gig systems were built and operated so other single other common fiber types that are out there like I said g6 5-4 is your ultra low attenuation really what it's designed for is to be able to send really high powers so I'll talk more about this later but it has a very large core it has a very large effective area so let's you transmitted really high powers without nonlinear issues and you see this use for for undersea cables and G six five seven is your Bend insensitive fiber so this is the stuff that you'll see a data centers this is the stuff you'll see any time you have humans around it and if you just have regular fiber it's actually very sensitive to being bent if you bend it just a little bit too far the light will actually leak out the sides so you really want your patch cables to be made from from G six five seven it's pretty much the only way to do kind of patch cable type stuff and remember in modern fibers actually is much better than some of these original specs but also remember that lots and lots and lots of fiber that's out in the ground is old old old old old fiber so here's a diagram comparing kind of the different rates of dispersion for some major brands of commercial fiber here so you see the the top one the six five two that's your classic SMF you see that it's got this very high rate of dispersion so the magic number for kind of the center of the c-band tends to be 18 the unit of measure here is called picoseconds per nanometer so you see about 18 as very common and then there's different types of dispersion shifted fiber so tera light true wave leaf they each have different slopes but you see how they're trying to reduce that rate of dispersion so now I'm going to talk a bit about nonlinear impairments so this is a fun little piece of physics so what do I mean when I say nonlinear impairments really a better description here might be high power problems if you don't transmit at high powers you will never have these problems it why do you care you want to transmit at high powers because if you transmit at a high power that means you receive at a higher power that means you introduce less noise that means you can continue to amplify for longer so you do want to push it as far as you can but so what do I mean by high power well usually the magic number is somewhere in the +4 DBM per channel so you actually have to be pushing and it's not something that you will you will see in your regular life if you are using kind of even a hot optic that's a single output you really have to put it through through another amp first so there's there's two main types of nonlinear impairments out there the first type is called stimulated scattering of which there's two types the stimulated Bri Wan scattering called SBS and stimulated Raman scattering called SRS and the other type is called the kerf echt so what's happening in the Qura fact is the intensity of the light if a light gets bright enough inside the fiber it actually dynamically changes the refractive index of the fiber as the the bright signal comes out so it causes a couple different things but the three big ones are what's called for wave mixing self phase modulation and cross phase modulation so SBS is actually really interesting what's happening here is when you when you start to fill this fiber with lots and lots of optical power you actually start to make the atoms vibrate and they start to vibrate in a pattern right and as they as they start to line up and as they start to kind of cause these acoustic waves it actually creates a dynamic Bragg grating inside the fiber so it actually causes your optical signal instead of continuing to propagate through the fiber it actually starts to be reflected back so past a certain point instead of any more power getting into the fiber all you're doing is causing that power to be spit right back out at your transmitter and that can cause anything from errors to actual damage on the transmitter so this bad thing but the SBS is really dependent on the power density of the fiber so what's happening here is the wider the signal the wider the line width of the signal the more you spread that optical power out so as we move to modern systems kind of hundred gig 200 gig and higher systems we've moved to a system that spreads that power out quite a bit so what this used to be a huge area of concern for 10-gig signals it's not nearly as much anymore because you've gotta you've got the power dispersed over over a wider signal so another reason that that matters is when you think about dispersion and why dispersion causes some of these effects dispersion is causing that power to be spread out so very quickly you have this high power in this one specific frequency and over time it spreads out so it reduces the impact of this issue but if the power doesn't spread out if the signal is not dispersed and that's where you start to see these so SBS really kind of requires long fibre distances to see so if you put a really high power system but you do it throw one meter patch cable you probably won't notice but if you do it through a 20 meter or 20 meters or 20 kilometer fiber so 20 kilometer is kind of the magic number once you go past 20 kilometers the signal has fallen off below the point where it's causing that impact but for that first 20 kilometers you're starting to get those reflections all the way through so SRS stimulated Raman scattering is a very similar phenomenon but what's happening here is when you put too much power into the system it causes power to be transferred from one frequency to another when you do this intentionally this is how Raman amplification works when you do this unintentionally you cause noise so ironically actually making the channels tighter reduces the SRS effects so here's some example launch power showing different so non zero dispersion shifted fiber being a worst case scenario for this showing the different power level so if you you're trying to send 80 channels and you're sending them down a 50 gigahertz spacing system you can only transmit 0.5 DBM per channel and pass that you're going to start to see impact due to the 2 to SRS 4 wave mixing is a really cool issue so what happens here is when you have channels that are regularly spaced they actually interact with each other and they create harmonics so the closer they're spaced the worse this effect yet so you see here we have 3 channels that are that are regularly spaced and what happened is the interaction the nonlinear interaction because of the high power levels and because of the low dispersion caused it to create noise harmonics both in band and these three channels and on the side bands even though those side bands aren't being used so this is this is a completely transmission rate independent behavior it's just a property of physics an easy way or one way to kind of deal with this is to have uneven channel spacing so rather than have things lined up precisely and in rows you can kind of move them around a little bit but remember this is really prevalent on low dispersion fibers so here's an example showing why that is so if you look at this this top line this is an example of a zero dispersion fiber and you see that you go all the way out to so point eight nanometer spacing is 100 gigahertz channel and you see that at point eight nanometer spacing we're basically causing huge amounts of mixing this is the mixing efficiency so how easy it is for these signals to mix even adding a single picosecond per nanometer of dispersion causes this to be greatly reduced so you see that we went from massive efficiency and mixing down to a negative 30 DB of efficiency but then you see that if we had standard single-mode fiber with a high dispersion rate that it would be even lower it would be down into the negative 50 and effectively not be an issue so this is an area where dispersion is actually a good thing so I have some examples here of different different things that you can look at over time so the next the next one is the curve X R what's called the inner channel effect so cross phase modulation is what happens when one one wavelength of light affects the phase of another and it causes inter channel cross talk so probably the most common area you'll see this in real life is when you try to mix like a 10 gig and 100 gig technology on the same fiber you're mixing what's called an energy system I'll talk more about this later you're mixing you're mixing an amplitude modulated only system with a phase modulated system and the the 10 gig signal causes noise on the hundred gig signal so it actually it doesn't hurt the tanguay signal at all but having 10 gig signals on your fiber will cause impact to your hundred gig signals and here having a high chromatic dispersion actually helps prevent this self phase modulation is what happens when when the the change in power level between a 0 & 1 is so so extreme that it causes the curve fact it causes the change in refractive index to happen dynamically in the fiber and what's really bad here is you end up with a property called self focusing so imagine that inside of this waveguide a fiber you create a smaller waveguide of fiber and you focus the power because of these different refractive indexes and in this example having lower chromatic dispersion helps with SPM so all of these things are balanced but ultimately all of these nonlinear effects are all about power and power density so one of the really easy ways to defeat the entire thing besides just transmit at a lower power is to have have wider fiber more more fiber that you can spread the power into so this is this is something in a fiber called the mode field diameter the MFD or you'll see it translated into the fibers effective area and if you don't see the effective area it's literally the mode field diameter PI R square so if you get a wider effective area fiber so here's an example kind of using your very standard non zero dispersion shifted fiber you might have 50 microns squared worth of effective area and leaf and some of the the better dispersion shifted fibers made that a little bit better but your your your classic sm f28 is better still right it's a wider fiber so you're able to put higher power levels into the fiber and not cause these issues and then your submarine fiber is the best of all this is a system that's pushing right up to them the maximum boundary of what can still be a single mode fiber but by having this very wide core you're able to inject much much larger amounts of power one trade-off here is that the larger the effective area the less Raman gain you get so if you depend on a Raman system you need to take that into account I'm using this large effective area I can transmit at a higher power but now I get less gain out of my Raman system as well so a quick word about what we actually transmit over the fiber so very first thing we're going to talk about is modulation so remember at the end of the day everything that we do is an analog world we think we've got these beautiful digital signals but we still have to send it over an analogue system and light is really just another type of electromagnetic wave and it propagates in a wave-like fashion so the very simplest form of modulation the system that you see that it's in use for effectively all of your 10 gig and lower systems or signals is what's called IMD d stands for intensity modulation with direct detection and the most common version of that is something called nrz which stands for non-return-to-zero you'll also see this whole type of system called a SK amplitude shift keying or amplitude modulation it's it's like your AM radio your modulating signal by increasing and decreasing the power level and all of this is just a really fancy way of saying bright for a1 and dim for zero that's all this is and direct detect just means that all you need to detect this is a photodiode you look at the light and you see I celebrate I saw at them you're you turn that back into a one or zero so historically like I said all these these fiber systems up until very recently have all been based on this technology and all we've done is signal faster and faster and faster so the the signal rate the rate at which we actually change a signal is what's called the baud so technically it's defined as the symbol rate per second so when you see a 10 gig signal so you go out and buy a 10 gig Ethernet port what's happening is you're actually flashing from a bright to a dim 10 billion times per second that translates into Giga bod so in this case it's a direct mapping we we encode one bit per baud we have a brighter atom so we know a one or a zero and we just change the speed at which we flash the light really quickly and that worked well to a point we build a lot of very successful networks with it but it really only worked up until about 10 gig of technology and then it kind of started to peter out but you know obviously we we needed more bandwidth than that so what's this had internet to do when we can't keep up with our cat pictures well we start looking for ways to increase the amount of information that we can encode per symbol change so there's a couple different ways you can do that in the world of ASX in the world of amplitude keying the easy way to do that is introduce more than two states so instead of having a bright and atom we now have a bright and a really bright so you can have you can encode more information with different levels of state and the other method is to move to something called phase shift keying so that's that's modulating on a property of the analog signal called the phase so what's happening here think about an analog signal represented as a sine wave and it goes from from you know high to a low and propagates what you can do is modulate on the difference in phase the offset between between where that peak and that on peak is so what you do is you have a receiver that is actually specifically looking for it's locked on to the phase that it's expecting to see and it can calculate the difference between what it thinks it should be receiving and what you're sending and you're able to encode data that way so you can actually start to pack a lot more more data into a system without having to just rely on brightness and dimness so that introduces something called coherent optical technology so coherent comes from from the the concept called phase coherence and what coherence means is the ability to measure phase change over over signals so like I said what happens here is you actually you you have a known reference signal so on your on your receive side you actually have a laser and you configure that laser to be exactly what the transmit side should be and you compare the difference between your local oscillator and what you receive you're able to calculate phase information and you're able to encode data in an entirely new and different kind of way this is what helped us break the 10-gig barrier kind of in a 2011 11 area timeframe so tying all this together doing phase shift keying is not actually as easy as it sounds because with amplitude keying all you needed was a photodiode right all you needed was to see your brighter them and you could make a decision of a 0 or 1 but doing phase shift keying actually kind of combines a lot of different things so to do it you you need a purpose-built DSP a digital signal processor so this is just a micro processor that is specifically designed to do real-time processing of signals and they get they tie all this information together they make it usable and coalesce it into some some kind of useful data so when we introduce coherent technology you basically saw a jump from kind of the the previous limit though the wall we were hitting our heads against 160 10 gig channels to immediately moving 200 gig panels and 29.6 terabits per per fiber so a huge jump in capacity like I said they delivered 200 gig optical signals for the first time it eliminated the need for physical dispersion compensation because the the property of of being able to modulate on phase kind of lets you calculate what that dispersion is unless you compensate for it and enable these these really high bandwidth so ver these really massive distances so then what what do you do and you need to go from there well then you start tying the two together so you move to a system called quadrature amplitude modulation or qualm and really all this is is using amplitude modulation and phase modulation at the same time so take to amplitude modulated carrier send them the same time and do it without women offset deal with assigning the cosign offset and then rely on your D your DSP to kind of computationally recover all this data so we create more and more complex versions this over time you can add different possible states you can add more bits per symbol so you know you see in the real world today eight and sixteen qualm very soon you're going to start to see thirty-two quantum systems and then there's already 64 common and higher systems and development that are going to be out very soon so another thing that you see in these modern commercial systems is something called polarization multiplexing so remember that light and the many things that we don't really understand about what it is it's an electromagnetic wave propagating through space it can actually do that in 3-dimensional space in two different polarizations so you can have a horizontal and a vertical polarization of light the two will pass each other and not interfere with each other and propagate differently along the X and the y axis of the fiber so in a moderate system what happens is you actually take two different two different lasers and send them at two different polarizations and MUX them together so it kind of gives you that double the capacity for the for the same pair of fiber right off the bat so when you look at a moderate system and you think about what is the what is the bit rate was the amount of data that I'm able to get basically think of it as a multiplication of polarization times laud times modulation so your polarization like I said today we have dual pole or dual polarization and you'll double capacity and we're already looking at things like optical angular orbital momentum and all kinds of different ways to multiplex additional properties of light but today we kind of get to polarization so you get a doubling of bandwidth the baud rate the higher the baud rate goes the wider the channel size that you need and the the better the DAC needs to be so today a very very calm systems that you'll see are kind of in that 32 gigabyte range and we're starting to see technology come out to move into the 50 and 60 gigawatts but as you do that you need wider channel sizes modulation is when you're encoding more data per per bought change so for the same signaling rate and the same everything else you can you can encode instead of being four bits per symbol you could be five bits or per symbol but it's harder to detect so basically with modulation you need better OS in our levels to make all this work so when you're when you're building a system you're trying to strike a balance you're trying to say let's say for example I want to do a hundred gig signal I could very easily do that with DP QPSK and I can do that tens of thousands of kilometers so 10 DB o SNR is very low so I can I can do that for very long distances and I only need thirty seven point five gigahertz worth of channel to do it but I only get two bits per symbol per Hertz of efficiency so I'm transmitting 128 gig signal with Beck I'm getting down to 100 that's what I'm able to get out of it so now you say well what if I what if I want to move to 200 gigs well there's a couple ways you could do that one is you do a 32 gig about 16 qualm system so this is this is very common in a lot of the DCI boxes that you'll see from folks like infinera and Corian tand and Nokia and folks like that they're all very similar technology and it requires a 19 DB o SNR so that can either mean anything from a few hundred kilometers to a couple thousand kilometers depending on your line system but it's a much higher OS in our requirement the other way that you could do it would be to have a lower modulation move back to a qualm but to do it at a higher baud rate so for the same effective data rate in the end you could do it at a 56 giga bob rate this now needs a 60 2.5 gigahertz channel but maybe you don't care about packing the maximum maximum number of channels in so you're able to use this wider channel to support it but now you're able to get the same 200 gig signal and take it down to to a 17 DB SNR so this might let you get you know hundreds of thousands of kilometers more distance out of it by doing it in a different way and you see the technology that we'll probably be doing by the end of the year or certainly into next year as people shift new systems is the 64 gigabyte the 64 qualm so this is purely a metro application for the most part but you'll be able to get 600 gigs per wave so this is this is what's going to support 70 6.8 terabit systems expect that to be out within certainly under a year widely so a couple more things about coherent like I said it eliminates the need for that physical dispersion compensation because it's able to to computationally recover for it so it you went from a world where you might have been limited to a couple thousand picoseconds per nanometer to one where you can now do hundreds of thousands and not care and in fact coherent systems actually work better with dispersion and as well the nonlinear effects are better with dispersion so this is why we've kind of moved back to going back to classic single mode instead of any of these these other things that were created for 10-gig technology so another cool thing with coherent is that because of the property of having a local oscillator and knowing what you're looking for you can actually lock on to one specific frequency and ignore everything else so this is what makes your CDC rotom's work all you're doing is taking the signal and sending it out every port and then your coherent receiver is able to lock on to the frequency that it cares about and ignore all the rest so it lets you do a lot of network building in software but there's a lot of downsides to this - it's very power hungry it's very expensive these are these are all many proprietary DSPs they're they're very difficult decree it's a whole field of science to go you'll make these and they today don't really integrate into pluggable technology in fact they probably never will because even as the DSP technology gets better and better the pluggable technology also tends to want to shrink - so as much as we all hope for a coherent in the QSF p28 by the time it happens QSP 28 will be dead and we'll all be moved on to something else so it's a lot more difficult to work with you tend to need external transponders to work with it so the other way that you could do this is back in the in the world of ASX of amplitude shifting is to introduce more more state levels so when I talk about nrz you'll typically never hear it call this but this is also called Pam to let stands for pulse amplitude modulation with two levels so again all it is is a bright or dim a one or zero so there's a lot of work that's happening in the space of Pam for right now so introducing four levels of brightness using this same direct detect technology so you can have lower power systems that are cheaper that integrate into pluggable in fact this is going to be one of the basis of 400 gig technology is very likely to need this type of technology and today there's there's vendors that are shipping hundred gig DWDM tune pluggable x' they use this technology so this is what's happening here you're introducing these four four levels of state and you're able to encode two bits of information per symbol change so a little bit about how you actually engineer your optical network so when you go out and connect fibers even the best connectors and splices aren't perfect so every time you take two fibers and you put them together you're going to get some kind of loss so the typical budgetary figure is 0.5 DB per connector and it might be less than that it might be worse than that depends on on the quality of the connector and how well it's cleaned and particular data center but word insertion loss is also used to describe the loss that you see from a mock so basically it's the penalty that you pay just for inserting the fiber so some real-life examples in kind of a common 40-channel DWDM AWG design MUX you're going to see a insertion loss of about 3.5 DB meanwhile if you move to an ad channel system of 50 gigahertz much you're going to see a much higher insertion loss you're going to see 9.5 DB all 50 gigahertz MUX is is 200 gigahertz muxes of an offset plus an inter leaver so you see that that add up of cost and insertion loss another easy way to kind of cause problems is when you have mismatch core sizes so here you see that we have a wider core and a narrower core come together and in fact when you see that on an OTDR you'll actually see all kinds of weird behaviors if you have a misaligned core if you have fibers that aren't perfectly mated you'll tend to see some some loss and if you have any kind of an air gap between the fibers the connector isn't perfectly inserted as it clicked in all the way air really doesn't transmit this very well so you'll see a ton of loss that way the signal will still work you'll just see loss so when you start doing an optical network you need to build a budget you need to to actually look at how this is going to work so when an optics is 10 kilometers that is a complete guideline I've taken 10 kilometer optics and I've made them do 25 kilometers all day and night long but you have to know what you're doing so but that 10 kilometer is purely a guideline it's purely a standard stated thing the actual distance is going to depend on the quality of your fiber the quality of your connectors what what you're doing with them but you want to leave some margin in this too right you need to deal with what happens over time if someone comes along and cuts your fiber and then it repairs it and now you've introduced more loss because of the splice or what happens if some data center tech comes along and their grubby little hands and starts working in your cable tray and start spending your fiber you're going to if you're going to see some impacts and then the actual transmitters themselves cool over time so when you buy an optic out of the box and you plug it in you come back 10 years later so you're still running it it's not going to be transmitting at the same power level that it was when it was brand-new so you need to add some budget in there to make sure that the link still works and doesn't just die randomly over time one thing that you'll encounter in the world is is different types of connectors the PC and the UPC versus the a PC so these are different types of ferrule connectors the the physical contact or the ultra physical contacts you'll sometimes hear called Polish contact but it's actually physical are the blue tipped fibers so what's happening here is this is just a straight the fibers on line perfectly flat against each other so in a PC standard you're guaranteed a minus 30 DBM Rd be back reflection and an ultra ultra physical Eustace angles were polished but in a UPC you're kind of guaranteed minus 55 back reflection so less less reflection that happens and this the green tip fibers that you'll encounter out there or what's called a PC angled physical contact so what's happening here is you've got an 8 degree angle at the end of the fair and it's getting you much lower reflection at the end of the fiber so why do you care about that well when you disconnect that fiber even a UPC cable once it's not plugged into something and once it's not propagating in that light it's going to cause huge amounts of reflection it's going to cause all this power to come back at you so if you're running a high power system and someone goes and unplug the cable and now all that power where's it going to go it's going to come right back to the transmitter it can actually damage the transmitter to have all that power come back to it so in a high-power system you tend to see these APC systems out there other than that there's there's really no no need for it in a kind of low power system dispersion compensation units so this was remember I talked about dispersion and the opposite slope so originally this was a big full of fiber in a box with an opposite dispersion slope and so at the end you run through 100 kilometers of fiber and then you might run it through another 10 kilometers of dispersion compensating fiber and riess Cui's the signal back together and then eventually it moved to being done with bragg gratings and so you don't need an actual spool of fiber because it spools of fiber introduced latency and people don't like those but historically it added like 20% overhead so when you when you have these old spool based systems is added a significant amount of latency to the system but you you tended to see these deployed at amp sites and the best place to put it is in the middle of a two stage amp with what's called mid stage access so you want to have your first amp take a pass at it make the signal brighter then you can incur some more loss going through this but you you haven't let the signal get low you made it brighter before you dropped it and then you make it brighter again when it comes out of it and you use circulators to reduce the amount of fiber that's needed and all of that so this is this was very prevalent in 10-gig technology and you'll see it become more prevalent for things like pam-4 actually for things like Pam floor you need interactive tunable dispersion compensation to make it work so a couple different ways to deal with this version largely we like to move to a role where we don't have to care so if you can deal with it electronically so much the better so for example when we when we created standards like 10 base 10 G base L R M we were able to do 300 meters of 10 gig over over multimode fibre that was all done by introducing electron version compensation but dispersion is is the worst for direct detect systems and like I said Pam fors is very finicky so today with the you know like dual 56 gig Avadh channel systems that are out there you need a tolerance of plus or minus 100 picoseconds per nanometer so you need very very tight you need interactive tunable dispersion compensation and all kinds of other properties that go into making this work well this coherent systems really don't care they eat dispersion for lunch and like I said it actually makes the system perform better to have it so you really you when you build a 800 gig only system or above you don't care another concept that you'll see out there is what's called the the three RS so three RS are reeling and retiming and so you you move from doing one to doing both to doing all three in reamp Liffe eyeing all you're doing is making the signal brighter you're completely unaware of what the signal was or spacing or brightness or anything you're purely you're creating another photon right you're making the signal be brighter in a two are you're actually restoring the shape of the pulse so you're you're aware that this was electronic pulse and you've done some level of decoding but you haven't fully fully fully D converted it and in a three hour you're actually retiming it so you know either you restore the shape of the pulp but you actually restore the the distance between the pulses and basically this is what's called a no Josi you literally you go from optical to electronic topical so this would be the equivalent of any router taking optics turning it into a packet turning it back into an optics you've completely regenerated the signal from scratch so when you work with fiber systems and you encounter bit error rates what happens when you when you encounter these impairments the links don't just die like that would be nice if they just died but they don't what happens is they just start taking errors they get worse and worse and worse over time so the probability that this will happen is what's called the bit error rate so here's some some time showing you what what you need to be 99% confident that a bit error rate is of a particular so if you want to guarantee for example that you're only going to have one bit every ten to the negative thirteen which is decent at 100 Gig you need to test this system for three and a half hours at one gig you need to test this for 14 days you have to send a signal for 14 days you have to transmit a binary for 14 days to look to see if you're going to encounter that bit error once so that's that so OSN are what's happening here as the signal gets worse and worse and worse the bit error rate goes up so here's a here's a diagram showing kind of some example comparisons of a QPSK signal and 8n8 qualm system and a 16 comm system you see that the more the more complexity you pack into the modulation the higher the o snr that's required for the bit error rate to still be be good so now a quick word about the tools of the trade and working with fiber systems so the most common thing that you'll see in the real world is something called an optical power meter or just a light meter and all this is is a little tool that measures the brightness of a signal and it displays results but remember that most of these light meters out there have what's called a relative loss mode so this is what I talked about the distinction between a DB and a DBM if you use it in the mode the DB mode what what's happening is you're expecting to have a known transmitter on the other side and you put it in this mode because you want to measure the loss over a piece of fiber when you put it in DBM mode you want to know how much light am I actually seeing so if you tell your field text to go out to your router and plug this in and get it reading how much light is coming out of my optics you need to be in DBM mode you're seeing how much light do I actually have not measuring loss over a Ferro fiber and literally if I had a nickel for every time someone told me they measured something ridiculous on my fiber I'd be a very rich man so another thing that you see in the in the real world is something called an OTDR an optical time-domain reflectometer so what's happening here is you're launching light and you're measuring the reflections that come back and you're using that to characterize the fiber so you're able to detect with an OTDR you're able to figure out where the splice points are where any breaks are and use it to kind of characterize the fiber is is good and clean and working as expected so here's an example of an OTDR output and in fact when you see when you see these little sections where it jumps up and then jumps back down that's a that's a core size mismatch so that's an example of a bad OTDR actually I think goes above net in 1999 to those so questions people ask me all the time is can I really blind myself by looking at the the fiber or you caution beware of big scary lasers so here's some words about laser safety guidelines so lasers are broken up into four classes for safety so your class 1 lasers are completely harmless either because they're so low powered that they can't do anything to you or because there's some way that you can't access them so for example your DVD player if you still have one probably has a powerful laser but it's inside of a metal case that you can't get out while it's operating so it's a class 1 a class 2 is something that's only harmful if you intentionally stare at it so if you really wanted to hurt yourself with a laser pointer you could do it and you could hold your eye open you could shine it in your eye but you have to like actually want to do it because if you were doing this by accident you would have a blink reflex and you'd be protected so it's it's generally considered safe unless unless you're trying to hurt yourself class 3 is something that you can't view directly because it would cause damage to the eye and you wouldn't be protected by the blink reflex so there's two sub classifications in there and there was an old system and a new system so you might see it called 3r it got renamed 3a and 3d so the distinction here is your class 3a zoar typically your 1 to 5 milliwatts your high-powered internet laser pointers your your things that you get from China that shouldn't really be legal those those tend to fall in there anything that's in a class 3b bylaw should require a key and a safety interlock system and if you have people to work around it they all have to have goggles and protection systems and all this stuff in class for things that burn meld or destroy Alderaan so when we talk about laser safety in the eye networking lasers everything that we're thinking about is actually occurring in the infrared spectrum so infrared of course is not visible to human eye but there's there's two classifications of infrared so there's what's called IRA or near-infrared and that's between 700 to 1400 nanometer and then there's IRB or far infrared or shortwave infrared and that's the higher frequencies so the laser safety levels are actually based around what can actually enter the eye and cause damage because the human eye didn't evolve to see infrared your cornea actually does a really good job of filtering out IRB so you might think that oh it's a high-powered laser and it's a really long reach system so I'm probably going to be more risk from that and you're actually at less risk from a medium reach application so the maximum power they continuous amount of power that can be transmitted without some kind of a auto shutdown feature for a 1550 nanometer IR IR B type signal if you want to be in a class one laser you've got to be below 10 DBM so typically you're your single channel 10 gig 10 kilometer type optics are going to be in the 0 to 1 DBM so every one of those is perfectly safe to be below a 3 are there to be in a 3 are you have to be below 17 DBM and this tends to be is a maximum amount of power like if you were to load up all 40 channels on a WDM system or if you were to load up a single stage esse this is kind of the maximum power that you get out of there to be below a class 3 B you got to be below 27 DBM which is a really really really powerful signal but you know that's your your raman pumps your ultra long-haul type applications and then class 4 is something that's really above that and actually causes a lot of physical damage so in routers essentially everything that you ever look at if you go to a router even if it's a 200 kilometer reach optic even if it's anything crazy like that it's going to be classified ironically probably the closest thing that comes that comes close to doing actually meeting the harm of the category of being outside of a class 1 and potentially able to cause damage is like a 40 gig lr4 so this is something that's centered around 1300 nanometer it actually has four parallel lanes so in a four lane optic centered around here you can actually be really close to eight milliwatts so a ten kilometer optic something that you wouldn't think would be especially dangerous is actually the worst thing that you're able to encounter on a router inside of an OPS of an amplifier they can very easily have output powers that go into 3 R + 3 B and raman amplifiers are almost always class 4 but they they legally have to have auto power shutdown or reduction systems and if they don't they can't legally be sold so the the DWDM equipment like the probably the most dangerous thing you could ever encounter in an in a real data center is a is a signal coming out of a passive MUX so like you've taken 80 channels you put them all into one fiber and you don't have any active components you don't have an amplifier or something that actually has a power shut off mechanism if it's a textile break so now you've started to get enough power from all these individual things put into a you know piece of component that's concentrated all this light energy into one piece of fiber that can actually start to get you into into 3 B territory and doesn't have an auto shutdown feature so should I be wearing goggles to the Colo probably not if you're playing with your your standard client optics you're in no danger whatsoever even on your amplified WDM systems light really disperses quickly once it's inside air so unless you take the fiber and hold it right up to your eye you're probably fine and remember that everything about 1,400 nanometers is actually mostly blocked by the human eye so don't worry about it too much but don't look at it what a microscope is so why would you even look into a fiber anyways right if this is infrared like you can't see any light at all so the human eye is really designed to see between 390 to 750 so technically no fiber even the 850 nanometer signal that you see you could see but if you've ever looked into a you know an SR and you've seen red there's a reason for that like the signal is not perfect the generation is not perfect so it's generating some sidebands that are falling down into the into the human eye visibility one trick you can use depending on the camera that's out there is digital cameras can typically see infrared so if you take your camera and hold it up to like your remote control you can see if you can see CD infrared signal that way but unfortunately that's tend to go away because then the nicer cameras are putting optical filters on there to try and block this out for better photo quality so sorry iPhone users as of like 4s and iPhone 5 they started putting iron filters on the camera so people ask all the time can optical transceivers be damaged by overpowered transmitters and well the answer is yes and now so remember that most optics transmitted at kind of the same power no matter what so there's not really much variation between a 10-kilometer optic and an ad common optic when you when you make an 80 kilometer optic you're not making it stronger what you're doing is making the receiver better so you're able to work with a lower lower signal you might have 10 DB more of sensitivity out of it and the sensitive receiver is What's in damaged danger of burning out so there's three points they come into play here this what's called the saturation point and the damage point or sorry two points so basically speaking the only thing that's in in damage is your your AV Commodore obstacle your optics with very sensitive receivers if you plug them directly into each other without any fiber in between you can actually cross the line and cause damage for the most part if you're in the 40 kilometer space what happens if you put in back-to-back because you just blind the laser and it doesn't work but do you don't cause any damage and then the ten kilometer optics don't care at all so do I really need to be concerned about Bend radius yeah so you can actually they sell tools where you want to troubleshoot and figure out what fiber you're looking at you can stick a red light generator on the end and you can go find the fiber and you can bend it you can see the light leaking out besides and this is what's happening here so even bent insensitive fibers are really just been less sensitive so don't go crazy like clipping them together but for the most part you you really want to use Bend IN sensitive fiber anytime you let your humans and your fiber come near each other so here's kind of an example showing the different practical Bend radiuses so when you got kind of a standard g65 to cable something that you might see designed for being in the ground it's it really expects a very slow bend radius compared to much much smaller Bend radiuses for the the cable that you can use in data centers where you can just take it wrap it around a pencil and it'll still work so a lot of time people ask can contain savers of different wavelengths talk to each other absolutely for the most part so all optical receivers have a wideband photo detector so when you've got a laser receiver it can basically see anything between 1260 and 1620 you can see everything that you would ever transmit out of a laser you won't be able to see in 850 those are those are separate systems but like anything that's designed to work over single mode is for the most part going to be going to be completely the wideband and coherent receivers like I said can even lock on to one specific signal ignore all the rest so lots of DWDM systems are built around this premise so you can have a system with wavelengths color one going one direction wavelength color two going the opposite direction and you can run this over your single strand of fiber and everything works just fine so the only real gotcha to this is that your optical power meter needs to know what frequency you're receiving or else it will give you a wrong power so if you set your meter and you tell it I'm looking at a 1310 signal and then you send it to 1550 your power readings just going to be off by a few DB but other than that it works completely so you can also mismatch frequencies to get extra reach out of it so consider a situation where you've got a ten kilometer optic and a 40 kilometer optic and an ER and an LR pair you can probably get nearly as much distance in this matching a 1310 and a 1550 as you can with a 15 15 15 tickets because the ER is going to transmit at 1550 which has a lower rate of attenuation so it's going to lose less signal as it gets to the other side and the ER is also going to have a more sensitive receiver so it's going to be able to detect that lr signal further so you probably very easily get 30 kilometers out of a mismatch 10.com in 40 kilometer optic just by by using these properties and so you might only need to have the optic on one side to make it work so a lot of times people ask do I need to clean my fiber yes so an interesting thing about fiber is you've actually got a lot of mating force you've got 45,000 pounds per square inch of pressure coming together on this tiny tiny little ferrule and if you've got a piece of dirt in there it will actually embed itself in the glass it will chips the end of your cable and it will damage and destroy it so buy a cheap cleaning kit go online get them for a few bucks it never hurts like if there's a piece of dust in there it will it will damage a whole bunch of stuff so you really don't want that to happen so a bunch of other miscellaneous information about fiber people ask me all the time how fast is light travel and fiber so the math here is this so light travels it you know almost 300,000 kilometers per second so your your standard single-mode fiber has a refractive index of one point four six seven nine so you do the math and you see that light propagates through fiber at about 200,000 kilometers per second and conveniently enough that turns out into 204 kilometers per second or 125 isch miles per millisecond and then if you want to compare that to what you see in a ping or traceroute you have to cut that in half again to account for round-trip time so more or less you see about one millisecond of latency for every hundred kilometers or sixty-two point five miles of fiber in the real world so if you ever go on Google Maps and do the distance and wonder why am I seeing such a higher value remember that fiber never gets light in a straight line it usually zigzags between population centers it gets late in rings there might be dispersion compensation spools there might be all kinds of other mechanisms that are happening here so the fiber itself is actually a much longer route than than the mileage might indicate and that's pretty much it questions Jake semantics first of all great talk and I've seen the previous versions and yeah it's a lot of great extra material put into this I have a question about your experiences with going out and doing a deployment and you've introduced uh before you talked about a lot of different technologies but compared to what you can go to a carrier and what you can grab from them in terms of dark fiber whatever what kind of technologies did you settle upon to use based upon availability of what's out there sure so you know a lot of the practical realities of how you design your fiber system depend on things like what type of technology are you trying to support so if you're if you're a carrier and you need to service everything from someone who wants a one gig or 100 Meg or you know everything like that and you've been in this business for 20 years you've got a collection of customers all across the map you probably have some legacy systems and you probably have a system that is designed a certain way and if you were to build a system that is truly new and optimized for nothing but high bandwidth and optimized for for creating very efficient systems or very but in my case I care about creating cheap massive bandwidth to feed into my packet network so I don't sell waves I don't need to do anything like that so I get to go out and design things in a different way I I focus on things like how do I get more reach out of a higher modulation system because effectively it's no cheaper to make a hundred gig component transceiver than to make a 200 gig they're basically the same thing there's no way to say I'm going to make this only do 100 it's going to be half the price you're going to pay or you're going to pay for the transceiver and the cleaner your line system you could just double your data rate for the same amount of money so that's what I tend to focus on is how do you how do you design your system for that so lots of hybrid for Raman lots of just low noise focused design but if you're a traditional carrier and you're now trying to support some of these older services you might not have network looks anything like that and you might still have a bunch of 10-gig waves running around preventing you from getting those types of reaches so Thomas McCarthy from Transit Wireless has two questions one in reference to CWDM and DWDM is there one that's more robust than another in terms of you know bad fibre and whatnot not really so basically the signal itself is the same the difference between the CWDM and DWDM is how tightly you pack them together so you you can increase robustness by changing the signal you can you can send a different modulation or you can add more effect or you can do all kinds of things like that but the difference in channel spacing other than those nonlinear effects I talked about it tends to be unrelated to that so the smaller window you get and you might need more precisely engineered you might need better cooling you may need the ability to control it make sure that it doesn't waver outside of that range but effectively it's the same it's the same signal the same laser okay second one is there a training path or certification path you recommend for I have high level engineers know a lot about optical networking however I've got a lot of field techs out there I'd like to learn you know quite a few things that you went over here good question no I actually have no idea so maybe someone else does but yeah I've never seen that yeah seems specific training like you know how to splice fiber and whatnot but I haven't seen a broad-based one like this yeah I've never seen something like that so that's kind of why I give this talk alright thank you David Holub called data center services great talk Richard I want to go back to the road immerses the photonics which in the MEMS technology and assume the application is something about a lot of people around here do which is sort of a metro network for rapid provisioning of you know ten and hopefully now hundred gig plus waves and circuits and so forth and I know there's different feature sets but when you focus on the sort of key feature sets of protection debugging scalability some of these feature sets are certainly duplicated between the photonic switch and the and the Rotom when you're making some of these equipment choices what are the considerations we should be thinking about and also do you see gmpls as they're part of the future and how would that potentially impact those equipment choices so the difference between a photonic switch and a Rotom indigo in adds drop multiplexer in that kind of stuff yeah yeah so really what you wish you get with a Rotom is the ability to build more complex topologies and to be able to kind of steer your wavelengths around so where it really pays off if you're selling waves you're selling a particular piece of light that runs through a particular piece of glass and you define it to go exactly this and you sell that to a customer and you want that to be provisioned quickly and you want that to work and you potentially want the ability to reroute that if all your building is bulk capacity to yourself it might actually be cheaper for you to just throw more bandwidth at the problem instead of putting in a very expensive Rotom with very expensive amps and on all the latest stuff it might depending on the cost of the transceivers and what you're doing it might be cheaper just to have more channels and so if you send it to your router and you don't really care which port it lands on or something like that it can completely change the dynamic as for 4G MPLS so having looked at the optical industry for a bit my personal opinion is that these people are some of the were software writers I've ever seen in my life I mean there's there's platforms out there that still run on Java and leave windows and yeah look like they were written by someone in DOS just just hideous stuff so I'm not a fan or believer in any type of vendor Interop being scalable and successful a I don't think they have the the chops for it and B I think they're highly motivated not to have it work because the optical vendor would much rather have you locked into their line system buying their transponders and they'll make a pass at saying that they care but they don't really care so I personally not putting much faith into into gmpls I mean gmpls has been around for a long long time it hasn't gone anywhere and I'm not putting much faith in to any other evolutions of it actually being delivered in a truly open line system in a truly open way it's great for press release material but it's not really good for practical implementation and result CrowdStrike I'm I wanted to go back to the for error correction part around hundred gig waves and we during our turn off of a couple of hundred gateways we came to realize that the 100 gig wave effects standard is not really standard and it's kind of like a bunch of different vendors have their different flavors yep so when you went through that slide I was kind of reminded that maybe something too to add into that deck about the different flavors and maybe we'll get a standard someday so the question I guess is do you see them collecting around the standard or you think it's going to be this mismatch on compatible things at this point I think it's going to be a mismatch so you there are standards they do exist and this question of whether you use the standard or not right so there's people to do and there's people that use different standards and there's people that go oh but we made our own proprietary version it's just a little bit better and of course again they want you to buy more of their transponders so for the most part I personally would never really try to like take two different brands of transceivers and make them work and talk to each other you get 98% of the benefit that you want by being able to take any vendors transponder and put it on your line system that way you can just oh I knew vendor came out with a new technology I can go buy that transponder but you still kind of want same system talking the same system or for the most part they've they've all got some way to have some proprietary version of their effect and probably other components as well so it so not only that but also kind of the way the DSPs work they tend to be tuned so that they only work against themselves and it you introduced a whole different can of worms when you try to make mismatched optics in coherent world or in high-performance tough talk to each other Nick guy on all communications thank you for such a brilliant succinct description this in very complex things and the reason why I tell people they need to clean jumpers my question is around dispersion shifted fibre do you think that it's still necessary to remove legacy dispersion shift of fiber out of kind of longer whole networks or maybe leave it in place given it a turn yeah it's definitely not necessary to remove it but you you change the characteristics of what you can get so if you were to go put in a nice new modern SMS 28a plus from Corning you might be able to get three thousand comet or DP sixteen comm system and if you have some old leaf stuff in there you might be able to only get two thousand kilometres and so if you need to sell something that's twenty five hundred kilometres you've doubled the cost if you didn't replace that system so it really depends what you're trying to do with it but no it's certainly supportable it's just not as easy it's not as good as if you were to put in kind of modern fiber and personally I'm kind of surprised more people haven't gone to putting in Submarine fiber and really trying to get some high power systems out of it but you know largely I think a lot of people are kind of in the field depending on what their vendors engineer for them so it might be a factor of that as much as it is a factor of what could be done with the technology cool that's it

Channel: NANOG

Views: 47,440

Rating: 4.8163671 out of 5

Keywords: NANOG 70, Kaskadian

Id: __wn9zXFiy8

Length: 119min 29sec (7169 seconds)

Published: Wed Jun 07 2017