Everything You Always Wanted to Know About Optical Networking [pdf] | Hacker News Reader

Everything You Always Wanted to Know About Optical Networking [pdf]

nanog.org

249 pointsandreasley34 comments

I just did my first municipal ISP design not too long ago and this document was an excellent introductory guide to the topic and understanding optics in general. The documentation for DWDM muxs (Ex: BTI and MRV) were also invaluable in learning how to apply the concepts outlined in the above document. This definitely expanded my horizons in the service provider networking field. Ex: http://www.juniper.net/support/downloads/?p=bti7200#docs

YouTube link for the most-recent talk: https://www.youtube.com/watch?v=__wn9zXFiy8

I've seen a couple versions of RAS' talk and learn something new every time.

yup, smart dude, learned plenty as a self-taught network grunt subscribed to nanog in the 2000-era and hanging out on irc, listening to guys like rs

Neat to see that optical is the new RF, in that it is using coherent modulation techniques that were once restricted to radio. In a similar vein, RF is also the new baseband (like audio and video before), in that the lower ends of RF (VHF and below) are transitioning to direct sampling at a reasonable cost.

The takeaway for me is that there are opportunities for people with experience in radio modulation and DSP to move into the optical game.

Is slide 6 (Fiber Works by "Total Internal Reflection") wrong? I thought that total internal reflection (TIR) happened when trying to cross from a higher refractive index (RI) medium to a lower one - not the other way around as depicted by the core-cladding boundary. Furthermore if the cladding has a higher RI, then the rays should become steeper upon transiting, not flatter like shown in the diagram.

I always thought fibers were single material (i.e. just the core) because the surrounding air (lower RI) lets TIR happen. It certainly seems to be the case in the next slide (Demonstration Using a Laser Pointer), TIR on Wikipedia [1], and pictures of fiber optic lamps [2].

[1] https://en.wikipedia.org/wiki/Total_internal_reflection [2] https://duckduckgo.com/?q=fiber+optic+lamp&t=canonical&iax=1...

Okay after reading a bit more [3] I'm pretty sure it's wrong, and the cladding RI's must be lower. So my follow up question is... what's the point of the cladding? Why not just rely on the surrounding air?

[3] https://en.wikipedia.org/wiki/Cladding_(fiber_optics)

The fiber optic strand isn't suspended in air, so you need the cladding for total internal reflection. Fiber optic strands are coated (protection and color indentification), bundled and installed inside cables with various fillings and strength elements.

You're correct about total internal reflection. The light in the core must move more slowly than in the cladding. The majority of the light remains in the core though some travels in the cladding as well. A higher index of refraction means lower speed of light so you would expect to see the core have a higher IoR than the cladding.

Yes, you're right the cladding is lower index, and it is a controlled index! It (along with other layers) protects the inner core. Most data transport fiber is single mode, which requires very tight control of the ~8um 'diameter' of the high index core, but allows for lower loss/km and you don't get mixing of the different modes (think of them as different reflective angles for simplicity). If air was used then any scratch or dirt contacting the core would couple light out.

In fact most multi-mode is graded index and it's much easier to get light into (8x larger core) without a specialized coupler. Note that even single mode allows DWDM- many different high-bandwidth wavelengths, but only 40nm spacing on 1560nm IR. The core, cladding, and outer protective layers are extruded and pulled to extremely precise dimensions, dispersions, and densities into multi-km rolls.

The games that are played to equalize loss and delay among the wavelengths over a range of power and temperature is crazy... and that's ignoring the EDFA optically pumped fiber amplifiers, which let under sea cables work. Amazing tech!

Has graded-index multimode caught on now? I last learned about fiber in the 90's when it was still something new.

Thanks this is the first time I have come across https://en.wikipedia.org/wiki/Yokogawa_Electric

Also it's refreshing to see that they mention "seeing" IR light from a remote control via your mobile phone's camera. This trick has been around for years (thanks to cameras tendency to colour shift) however it is rarely known about. The author is indeed a practitioner.

Modern phones don't see IR very well on account of improved filters. My iPhone can't see remotes at all. An IR source needs to be much more powerful to show up.

Back in the day I took the IR filter out of my digital camera and got some cool (dismal 3.2MP resolution) shots.

The rear camera has a very strong IR blocking filter. But the front camera doesn't. (We have a product that uses IR LEDs as a flash, and I use my front-facing camera to verify they're working.)

Thanks for this! Works on iphone 6

The effect is even more pronounced when you remove the IR filter.

I just tried this with my phone and mildly blew my mind, and then realized I could actually see the IR from the remote with my eyes, and am now confused.

He mentioned it in the presentation. You're not actually seeing the IR, you're seeing a sideband. Just like with a heat lamp, the red light isn't the IR, it's produced alongside the IR.

If its a 808nm or other really close wavelength, then you could be seeing the primary wavelength, but the apparent intensity will be very low as the eye has very little sensitivity.

As someone who used to work with optical equipment this is really awesome. My team was responsible for the SDN layer that sat on top of the nodes and configured them so we were taught enough to be dangerous but I always wondered what was really going on under the hood.

Why are we still stuck with Copper RJ45 Ethernet twisted Cable @ 1Gbps for Local Networking / last Mile? While there are up coming NBase-T which offer up to 2.5/5Gbps Ethernet, they are expensive and no where near as cheap as 1Gbps Port. Not to mention most Vendor will likely put out 2.5 Gbps instead of 5Gbps due to cost reason.

Why is there no common, simple, cheap Fibre Cable and Port to replace our age old RJ45 and Copper Cable.

P.S - I actually rather have NBase-T, but at the moment no one is offering it. At least not at the price bracket most would pay for it.

I don't know if it's the actual reason, but they break really easily and they cost a lot more.

Cat-6 can take 10Gbps over ~50m runs, and the cards are under $100 now

can you point out such a card? the closest i've seen is https://www.newegg.com/Product/Product.aspx?Item=9SIAD5G5M36...

You can usually find refurbished ones, e.g.: https://www.newegg.com/Product/Product.aspx?Item=9SIA9AX5D88...

And B&H has a new card on preorder for just barely under $100: https://www.bhphotovideo.com/c/product/1344847-REG/asus_90ig...

thanks!

it looks like asus has also introduced lowish-priced switch with some 10G ports, too. https://www.amazon.com/XG-U2008-Unmanaged-2-Port-8-Port-Giga...

Reason for existence of 2.5/5G does not have that much to do with the cost of PHYs themselves (it.s more or less the same circuitry as for 10G) but with power consumption.

Even 1000-base-T is surprisingly power hungry due to the fact that the used signaling is essentially analog and not some simple binary/ternary line code, which necessitates inefficient analog final Tx amplifiers and such things.

Probably all current 1G PHYs contain some TDR-ish magic that measures the length of attached cable and determines how much can be the Tx power scaled back in order for the link to still work (~10 years ago this feature was often the main point of manufacturers' advertising, today it is essentially must-have).

10G PHY is even more power hungry and 2.5/5G is way to not only scale Tx power back but to get additional power headroom by reducing the symbol rate.

This feels like a great resource to me, it's 125 slides chock full of stuff.

>> Light, like sound, follows the inverse square law. • The signal is inversely proportional to the distance squared. • A signal travels distance X and loses half of its intensity.

I thought light (at least photons) don't lose intensity? Is this through the medium that fiber optics are made of?

Yes, no material is 100% reflective, and some of those photons will either be absorbed by the material or reflected in the wrong direction.

It's explained two pages prior on Page 14.

I think the slide on page 16 is wrong; the inverse square law comes into play when light is unconstrained and spreads out in all directions. Losses that come from impurities in the glass and so forth attenuate differently.

For instance, if you're looking at light intensity from a lightbulb, every time you double the distance you get 1/4th the intensity, or that comes out to about a 6 decibel loss.

If you're shining a light through an optical cable, it might lose half it's intensity every time it goes, say, 10km (I don't know what the number is for proper high-quality single-mode fiber, so we'll just go with 10km). So, your loss is about 3 decibels per 10km.

Over short distances, the small fixed loss of the cable isn't significant, but over very long distances, the 3 decibels per 10 kilometer loss adds up a lot faster than 6 decibels every time you double the distance.

I'd say it's unclear rather than wrong, and what you say is perfectly correct.

Decibels are useful when dealing with exponential decay, meaning that the output of a process is always proportional to its input. An inverse square law is an example of this more general case, so I think the author is just holding it up as an example, rather than saying it applies to fibres. Other examples are inverse (which holds for fibres as you point out), or any other choice of exponent.

Photons will scatter off the atoms of the medium, yes. Ones that end up with momentum pointing out the side of the core will exit the fiber and be absorbed in the jacket.

More like physics than networking. I was hoping for more about what goes on between light and IP, but all I learned is that SONET/SDH is considered old school now.

This is a fantastic guide. Thanks for putting it up!

Super practical and concise guide.

Reply on news.ycombinator.com