Hacker News Reader

Comment by msarrel | Hacker News Reader

Yup, always run extra everything. At least you can splice fiber now. And be careful of the minimum bend radius.

Once, in the 90s, we were having intermittent network failures in our data center. I kept trying to troubleshoot it with the fluke, but the problem kept moving. When I pulled up the raised floor, I discovered that rats were eating the exterior sheath of the network cables. That was some fun troubleshooting!

View on HN

"And be careful of the minimum bend radius."

The minimum bend radius of any good fiber is way less than cat6 at this point.

G.657A1/B1 = 10mm

G.657A2/B2 = 7.5mm

G.657A3/B3= 2.5mm

CAT6 = 25mm

Flexible cat6 = 15mm.

Fiber surpassed copper for flexibility almost 2 decades ago (G.657 was introduced in 2006, but let's call it 5 years before all forms became common).

Even G.652 was only 25mm and standardized in 1984.

It takes less force to bend, though.

could be solved by running reinforce wire along it

> could be solved by running reinforce wire along it

You could, but you might as well just use armoured fibre if you're doing that !

Some of the low core-count CSTA stuff out there these days could almost have you thinking its unarmored if you didn't know. Not at all bulky.

I accidentally put a knot in a fiber, and it still worked. I'm still using it today, 3 years later (after untying the knot, ofc).

for an example of how good modern g.657 bend loss insensitive fiber is, for ftth applications and drop/last hundred meters, here's an example of it being abused with a staple gun and the resulting loss in dB.

https://www.youtube.com/watch?v=h4wzJSInvww

the above is from 12 years ago

I've seen some g.657.b3 patch cords and work done for FTTH installs that are absolutely atrocious and yet continue to function properly.

This video is a cool demo. Trying to eyeball it, it seems all of the bend radii (except one) are over 10-20mm so this demo isn't that impressive. The only exception is the very last bend he makes with the weight where the radius is probably around 5mm. I find this ironic that when making this last bend is exactly when the video abruptly ends with some video corruption artifacts. As if they streamed the recording over this fiber :-)

This is proprietary technology and is definitely NOT representative of normal fibre cables. I know because I had had to deal with contractors doing exactly this, which crushed the cables we ran. There are many things which could be better if we were willing to pay higher prices for them. At the end of the day, cheap and good enough usually wins.

This is just G.657B3.

They are all proprietary in their coatings for sure, but it's made to a standard. The only thing is that it's not guaranteed compatible with old G.652 fiber if you try to fusion splice it.

You can get G.657A3 that is, if you want.

But this is representative of A3/B3 cables for sure, and easy to get.

AT&T fiber and Google Fiber actually use this particular stuff in their standard deployments.

Under the covers, most of the glass is usually corning clearcurve zbl (OFS actually makes their own i believe, but for others, less common).

The most common consumer cable these days is G.657A2/B2, which is easily capable of what you see in the video.

As for pricing, i'm not sure what you are paying, but this particular stuff is also easy to get (for a consumer) at 20 cents a foot or less. 1000ft is usually 200 bucks or better. Their are idiots on ebay trying to sell leftovers from fiber installs, but if you go to any really fiber cable seller, that's what you'll pay.

Here are totally custom cuts of 1.6mm outer for 0.25cents/foot: https://www.fiberinstrumentsales.com/tlc-simplex-sm-clearcur...

if you want 3mm outer, it's 0.35cents/foot: https://www.fiberinstrumentsales.com/tlc-simplex-sm-clearcur...

FIS is usually expensive, too. I didn't look hard since that seemed "cheap enough" already.

You can certainly get A1/B1 cable for like 18 cents/foot instead of 20, but if your contractors are crushing cables, seems a bit penny wise, pound foolish.

I also wouldn't run simplex cable, as it doesn't make sense anymore.

As expected, 6 fiber versions of the above cables are only marginally more expensive.

To be real frank if you're using FIS as a benchmark of price and market availability of product, you have already lost. Everything on there is 300% marked up from the price they buy it from the vendors in China and Taiwan. FIS is for enterprise end-user customers that don't know any better, not service providers who buy fiber stuff in quantity.

g.657.b3 is an industry standard and not proprietary, and after close to 14 years being on the market, is the default for ftth drop cables and inside patch cables now in the last mile.

As i'm sure you know - for invisible installs, they don't even bother to have an outer cable anymore, just the 250um fiber with no sheath running bare on the walls.

Kinda insane.

And clean the fiber connector ends with a special cloth or cleaning liquid. Once, I had an intermittent link due to dust particles.

Fun fact 1: When you don't have a proper OTDR fiber tester to check for breaks, you can shine a laser pointer down the fiber and see it projected out on a sheet of paper on the other end (you might have to call the person on the other end).

Fun fact 2: If the other end is plugged into a transceiver, don't look at the fiber end directly; instead, look at it through your phone camera. If all is working correctly, you should be able to see a faint purple light out of one of the strands.

P.S. always remember to roll it if no link initially :)

> If the other end is plugged into a transceiver, don't look at the fiber end directly

Let me fix that for you.

Don't look at the fiber end directly, always assume it is plugged in unless you can see the other end yourself with your own eyes.

Fibre light meters can be bought cheaply for a basic troubleshooting bag kit-out if you don't fancy splashing out on a fancy device.

The last one may not work anymore. A ton of phones nowadays have IR filters.

They all do and always have, they just aren't perfect.

Really? I thought they usually used to have no filters, especially on the front camera. Well of course the lenses always filtered some IR light, but I wouldn't call that a filter, they also "filter" non IR light. A filter to me is a dedicated element or coating that prevents the majority of IR to be transmitted.

works on my iphone 13

Your typical Rodent Accelerated Transmission Signal And System Suppression

> Yup, always run extra everything

Cables ain't free, duct capacity is finite and duct rental from the local incumbent is costly too... Please calculate the financial optimum of pay now vs. pay later - taking into account growth, various forms of attrition, cost of capital, opportunity costs and appetite for risk. Or everyone would be running 1152 strands cables everywhere.

But then I see that from a telco perspective and, now that I've read the article, it seems to be from a small-scale hosting perspective - entirely different economics.

Duct capacity is finite, but fiber cables are tiny and they are very nearly free compared to the cost of having them run.

Not quite tiny - a basic armored (rats !) 12 strands cable may be around 10-15 millimeters, whereas similarly specced cables in the 500 strands range might reach 20-30 mm. In crowded Paris downtown, it adds up fast. On the countryside, that is a lot more weight to hang on poles.

Interesting. I would use the ability to get 40x the strands for 4x the weight as evidence against extra strands adding up fast.

    > Paris downtown

Where is downtown Paris? 1/2/3 arrondissement? I never heard the term used for Paris. Mostly, it is used to describe the Wall Street area of Manhattan (NYC) as the island is relatively north-south and Wall Street is on the southern end.

It’s also come to mean the densest part of any city. Even the town I group up in would call the densest part “downtown”, even though the only buildings taller than two stories were city hall and the historic theater.

Today I learned!

I always thought "downtown" and "uptown" were more like "south from X" and "north from X" where X is some reference point in context (e.g. "where I'm standing"). Turns out, it's more similar to "downwell" and "upwell", i.e. towards or away from some central point (gravity wells being spherical-ish and all).

Uptown can mean anything out of: geographically north, upstream on a river, higher elevation, or higher income and wealth of the residents.

It just so happens that for Manhattan which is the most famous uptown, all four of those indicators point in the same direction. In general I think an area would more likely be called an uptown if it met more of those criteria relative to some downtown. (And there would be correlation, wealthier residents would buy higher-elevation land.)

I notice that non US English-speaking tend to use the term CBD instead of downtown, my assumption is that it is because of the NYC thing you mention

I don't think there's any association with New York. etymonline claims that "downtown" originally refers to the idea that the suburbs are at a higher elevation than the city proper. I don't know how much stock I'd put in that, but I trust it more than someone's random intuition about the geography of New York.

( https://www.etymonline.com/word/downtown : "The notion is originally literal, of suburbs built on heights around a city." )

It's the only available word for the concept in American English. I wouldn't recommend saying CBD; that would refer to cannabidiol or to marijuana in general.

Reading more about it, accounts vary. There are historians who posit it was in fact inspired by Manhattan, or maybe Boston, but it's not cut and dry.

Edit forgot link: https://downtownny.com/news/where-did-downtown-come-from/

I'm not especially impressed by the argument "if we assume it came from Manhattan, then we can know it came from Manhattan":

> Let’s suppose that Fosdick, incoherent from distress, forgot to include the words “to the” between “down” and “town” when describing the direction in which people were fleeing. With this in mind, we can assume that Lower Manhattan is in fact the original “downtown,” as posited by Schwarzer, and that the possible Bostonian origin of the word is simply a misunderstanding.

I'm also frustrated by the lack of sourcing here but the wiki article for downtown [1] also points to Manhattan, citing a book I don't have a copy of. Allegedly it was in use in Manhattan in the 1830s, which aligns with the date given by the etymology entry linked previously.

[1] https://en.m.wikipedia.org/wiki/Downtown

https://www.etymonline.com/word/downtown The word was in want of a closer look in fresh archives. On a closer search, it's older (at least 1780s), American, but not apparently a New York reference at first.

British people would say city centre or town centre.

Australians are the ones who say CBD.

New Orleans seems to like giving directions relative to it's CBD (just south of the French Quarter). It's otherwise not common here in the US.

In New Zealand we refer to it as the CBD, but you could also say city centre or town centre (or just “town”).

Not sure I have ever heard downtown being used to refer to part of a New Zealand city although I understand its meaning.

I've seen it in Canada, NZ and India as well. Def not AU specific.

And while some people preach hard-wired everything I’d probably increasingly not bother at home. . I’ll have to see how much networking and audio stuff I even do given a kitchen fire with s

Maybe in an apartment. In a house, you want wired backhaul on access points, and wired streaming devices.

Why? Is this an actual problem? I have a bunch of wired connections (or at least had) but I'm not sure I ever saw a difference and I'm certainly not going to spend time or money retrofitting them for some Platonic ideal--ADDED, at least in recent years or outside of specific situations.

Latency. WiFi latency can spike which ruins video calls. Simple speed (given lots of time) on the other hand is rarely a problem.

There’s also the problem of sometimes, video calls just stutter and seize up and it’s the internet’s fault, not the local network. But with a wired connection, I am immediately confident the problem is not in my network. With WiFi, who knows, maybe it’s my fault. I used to continuously ping the router so I could verify this, and indeed ping spikes would happen. Now with a wired connection I don’t worry about this.

But yeah if you run no latency-sensitive applications, WiFi might be fine.

WiFi is convenient and (usually) cheaper (esp if it's a retrofit), but not deterministic. And less 'cluttered' I guess, if that's your pet peeve. All sorts of things can muck around with Wifi speed, reliability and latency and most people have little or no ability to troubleshoot it effectively.

I retrofit wired my house[1] about 20 years ago; I only have had a small handful of issues with it in that time (all basically mechanical fatigue of jacks). If I have speed, latency or connectivity issues, I can be reasonably sure the last thing I have to debug is the physical network.

[1] Copper, not fiber, but no runs are over 50m. Works to 10G so far. I used a Leviton residential structured wiring system (network, coax and phone at the time; the last 2 long since obsoleted). I did a lot myself since I'd done enough quality time in datacenters to be adequately proficient with low voltage Cat5. Not cheap but not extravagant.

It's potentially a problem in a typical brick house in Europe.

I'm sure it's less of a problem in houses with other materials.

The answer is brick and concrete. You’d be amazed how well a load bearing wall in a high rise will block a signal.

Which are, of course, also difficult to retrofit with a lot of new wiring. Old plaster lath is something of an issue too but not nearly as much and is easier to retrofit--or just replace.

For me it's the metal studs in my walls. WiFi reception is truly awful. Fortunately it was prewired with Cat5e. But that will need to be replaced some day.

I just had luck and have my ISP modem close to my desktop PC and desk so I run cable to desktop, laptop dock, TV and consoles all on the same wall.

So worth it as in apartment everyone and everything is blasting wifi. I see like 20+ SSIDs.

Or, take the money you would spend calculating that, and that's the amount you spend for your margin.

Haha, I'll tell that to the outside plant capacity planning team !

But seriously, I'm telco-biased, and it seems that the article has a small scale hosting... Different worlds. At small scale, sure - the bigger cable is a rounding error.

Have fusion splicers come down in price? Can you recommend any?

You don't need them unless you are running a mile or two of fiber - mechanical fast connectors (IE no polish, etc) are less than 0.3db loss at this point.

Fusion splicing will still be 0.01db or better, but you will be well within the power budget of any transceiver you find with mechanical connectors, if we are talking home networking - even a 2km 10gb transceiver that is 20 bucks has a 6db power budget.

Heck, the 10km are cheaper than 2km at this point - 17bucks on amazon, power budget of 16db.

This is VERY true - especially for home use. You could have many db of loss and still have a perfectly functional 100g network.

AliExpress sells some basic ones from $300-700 CAD ($200-500 USD) with reasonable reviews.

Whatever you do, don't try using the razor-style "hand splicers" and adhesive splice kits. Without a splicer that has a scope, you're just making bets each time that are difficult to test.

I've learned the hard way that it's just not worth it unless you can _see_ what you're working on.

I have a $1000 fusion splicer from Amazon, and it is fantastic. It is very straightforward to use, and for my home use it is perfect. If I did spicing as a job this wouldn't be the tool to use, but for occasional splices in my homelab it works great.

https://www.amazon.com/dp/B0CW2HGVT2?ref_=ppx_hzsearch_conn_...

For Multimode, they now make splices where you jam the two cleaved ends into the splice. It costs about $5 per splice.

You can have them second hand for ~500€ on ebay

Any advice on making sure you get a good splicer?

I know a little about fiber connectors, and the different connector modules for speed, but I am not really sure what I need for a splicer for fiber.

_minimum_ bend?

minimum bend _radius_

A straight cable has an infinite radius, the more bend the smaller the radius

Though if it is on or under the surface of the earth, “straight” will be a bend radius of around 6,370km. We don’t make a lot of buildings that deal with this but transcontinental or transoceanic cables certainly do. If someone designed a fiber that required absolutely no bend in order to work you’d have to use it in buildings or dig much deeper holes.

There was an encoding mechanism proposed about 10-15 years ago that used spirally polarized light to carry more channels, but it required the surface of the fiber to be polished to a much higher degree than existing cables in order for the light to go around bends properly.

If you’re using the planet as your “flat surface” then sure. If, however, you’re willing to deal with exiting the atmosphere at each end, you can use Real Straightness. But I don’t know anyone running a single segment for that distance.

I’m sure there are some microwave antennas still out there doing the Lord’s Work. At least in the Plains states where hills are low and putting antennas on two of them gives you some extra distance. How far do microwaves bend over the horizon?

Don't they generally do the opposite of bend over the horizon? Two towers that are observed (visible wavelength = tiny Fresnel zone) to have line-of-site can easily be obstructed (microwave = huge Fresnel zone).

https://en.wikipedia.org/wiki/Fresnel_zone

> How far do microwaves bend over the horizon?

that depends on the weather and frequency among other things

https://en.wikipedia.org/wiki/Line-of-sight_propagation

If you're going out of the atmosphere may as well skip the fiber and just point lasers through the vacuum of space directly and reap the benefits of the faster speed of light through a vacuum vs glass!

works great until a group of space bats get in the way.

Fibers can be bent to a radius of an inch or so, that's the tolerated minimum bend radius.

I guess you could call it "the maximum sharpness tangency" or something like that, but that's not the standard verbiage.

G.657A1, which is basically the crappiest you would ever buy at this point, is not even half an inch bend radius.

Minimum bend radius.

It’s usually 20x the outer diameter of the cable for fiber, a cable with a .3 inch OD has a 6 inch minimum bend radius.

What?

This is 100% wrong.

This is well standardized. As I posted elsewhere:

G.657A1/B1 = 10mm (half an inch)

G.657A2/B2 = 7.5mm (a little over a quarter inch)

G.657A3/B3 = 2.5mm (less than 1/8th of an inch)

The only cable you will find with a 6 inch bend radius is going to be armored interlocking fiber cable or something.

You can take 2mm G657.A3 fiber cable, wrap it around a pencil, and it will work 100% fine.

Even old G652 cable only had a 1 inch bend radius.

If I am wrong, then explain why this 24-strand plenum rated fiber optic cable has a minimum bend radius of 5.9”: https://leviton.com/content/dam/leviton/network-solutions/pr...

Part number: LTP12B024

I am running a project where the low voltage contractor is pulling this fiber cable into a conduit my electricians are installing.

I don’t doubt your bend radius claims on the fiber optic cables you listed, but there are far more types of fiber cable than the ones you listed.

There are not far more types of fiber cable, there are far more types of jackets :)

You are confusing the cable jacket with the fiber cable inside of it. This is the spec on the jacket they use and they didn't design this particular jacket to be flexible enough to survive flexing at the level the fiber cable inside of it can. Sure - not all jackets can be bent as much as the fiber. Some because they are not made to be as flexible (as here), some because it's basically impossible (armored interlocking jackets), etc.

I'm not sure what this changes?

The part number with suffix AB0403 (the SMF A1 fiber), if you cut the jacket, you should be able to bend the fiber cable inside at a radius of 10mm and have it be fine. It appears to be standard G.657A1 fiber cable inside.

There are really not far more types of fiber cable than i listed, and this spec sheet definitely does nothing to support that claim, since as i said, it has A1 fiber cable inside the cable jacket that will happily support a 10mm bend radius.

BTW, as an aside, I wouldn't use this stuff - just looking at this sheet and pricing at my normal fiber distributors - it is both overpriced and underspec'ed for its price. Which seems pretty typical for leviton :)

Most 24 standard I/O plenum cable is going to be 7.8mm (this is 9.9, so 30% thicker), and have less than half that bend radius even with non-flexible OFNP rated jackets. If you are being charged more than 2 bucks a foot for it, you should consider other options.

> There are not far more types of fiber cable, there are far more types of jackets :)

Ahhhh, thanks for clarifying! Now I understand what you’re saying, the jacket and fibers have different bend radiuses (radii). I appreciate you taking the time to explain :) also, noted about there only being a few types of fiber, with a multitude of jackets.

I’ll let my low voltage subcontractor know about the underwhelming Leviton cable, it’s a plan and spec public job so it’s possible the Leviton rep is friends with the engineer and got their cable flat specced, wouldn’t be the first time!

I’m only responsible for the raceway, it’s a 2500’ run between two college campuses separated by a highway that runs outside for a good chunk of the distance so it’s possible the jacket is beefy and inflexible for temp resistance. I’m much more familiar with electrical conductor jackets, THHN and XHHW-2, which are PVC and XLPE, respectively.

Reply on news.ycombinator.com