FOA Guide

The Fiber Optic Association - Tech Topics

Copper or Fiber? What's the real story? (Here is a specific look at fiber vs copper in LANs)

Every time we read another article about copper and fiber cabling that was written by someone who was either ignorant or prejudiced - or both - it is full of pure, unadulterated garbage. What is the real comparison of fiber and copper? Let's take a more careful look.

Point of view: Well, everybody has a point of view. We're the professional society of fiber optics, aren't we? But we have an interest in both technologies and know they are complementary in many ways, so if we're prejudiced, it's toward making sure everybody uses lots of both of them, but gets trained so they install either properly!

And as part of the point of view, we are focusing on premises cabling for communications, although we'll look at some other applications like outside plant telephones.


Start With Phones

First of all, telephone wire is over a century old. The actual way a phone works is that old too!

The telephone wire used for POTS (plain old telephone service, like in your home or a small office) goes to a switch, where the analog voice signal is digitized and most probably sent off on a fiber optic cable. It it isn't now, it will be soon.

The copper phone wire has very limited bandwidth. It was designed to provide 3,000 Hz bandwidth, perfectly adequate for a voice signal. The modem has to play some pretty sophisticated games to get higher bit rates over the limited bandwidth of the POTS line. Two digital systems work on this old cable style: ISDN and xDSL. ISDN is a low bit rate digital signal that never caught on - since analog modems are just as fast and a lot cheaper and easier to set up. xDSL or digital subscriber loop is faster - as high as 1.5 Mb/s - but only if you are close to the phone switch - less than about 15,000 feet.

So the phone system is mostly fiber optics beyond the short subscriber link. Fiber links offer over 1,000 times as much bandwidth over distances over 100 times further. Specifically, you can have:

   Distance  Bandwidth  Voice Channels
 Copper  2.5 km 1.5 Mb/s  24
 Fiber  200 KM 2.5+ Gb/s  32,000 +

That tremendously higher capacity comes at a price. But even if it costs 1000 times as much for two fibers (one to transmit and receive), it is still 1/100th the cost of copper per voice channel. And it has other advantages, like a fiber optic cable being a fraction of the size and weight of a copper cable, a big point in underground conduits in crowded cities!

Depending on the application, fiber costs are typically 1-5% as much as copper in the backbone.

In the subscriber loop, where one connection only is used, the economics are quite different. A drop to the home is less than $100, while a fiber to the home would cost over $1000 and require an onsite way to power the transceivers. So fiber to the home is a non-player, except in rural areas where the line is long and would require a repeater. Then the longer distance capability of fiber makes it cheaper to run fiber than have copper with repeaters and providing power to the repeater.


What about the coax cable used for CATV? Well, it has lots of bandwidth (100 MHz to 1000 MHz depending on how old the installation is), but it even cheaper than telephone wire to install. CATV systems are using this coax for everything, (television signals, Internet connections, and even telephones) but it too is quickly converted to fiber, which provides the backbone connectivity due to lower loss (and subsequently longer runs betweem repeaters) and much greater reliability. Both those features translate into cost savings, of course. And the fiber backbones are bi-directional to allow for all the new services being offered.

In both telephone and CATV systems, fiber and copper coexist, with each being used where the economics dictate.

The Real Controversy:

When most people talk about copper vs. fiber, they are talking about LANs or premises cabling. Here there is a lot of controversy, a lot of "positioning" and a lot of misinformation.

The wire we use for LANs is a lot younger than fiber optics. Fiber use is over 20 years old, but computer networks on unshielded-twisted-pair cable (UTP) have only been around about 15 years. In that time, UTP has gone through at least 5 generations, each time to keep up with the increasing bandwidth requirements of LANs. Today, it's hardly the "telephone wire" that some people think it is.

In that time, LANs have grown in capacity via:

 LAN Bandwidth
Ethernet 10 Mb/s
FDDI 100 Mb/s
Fast Ethernet 100 Mb/s
ATM 55, 155 Mb/s
Gigabit Ethernet 1,000 Mb/s (1 gigabit/s)
10 Gigabit Ethernet  10 Gb/s

The copper cabling manufacturers should be praised for their technical efforts to expand the capacity of UTP cabling. And the hardware manufacturers deserve a pat on the back too! They have been able to keep up with networks by some really sophisticated product development (including the electronics that perform the miracles of getting signals on and off the cabling.)

But all their efforts have produced a product that is not easily installed if one needs the maximum performance offered. Recently, a number of magazine articles and even a representative of AMP was quoted as saying that as much as 80-90% of all Cat 5 cabling was improperly installed and would not provide the rated performance. Contractors have told us that 40% of their Cat 6 installations pass certification tests.

The performance of the Cat 5 cable is dependent on close control of the physical characteristics of the cable and the materials used in the insulation. Untwist the wires too much at a connection or remove too much jacket and the cable may fail crosstalk testing. Pull it too hard (only 25 pounds tension allowed!) or kink it and loss the performance you paid for.

Even if top performance is not necessary, getting all 8 of the wires connected correctly requires a lot of care. 1-5% of all connections will not be correct first time around, according to the installers I have talked with. And did we mention the problem of electromagnetic interference (EMI) from motors, flourescent light ballasts, etc.?

But most networks only run at Ethernet at 10 Mb/s to the desktop. Even if you use 10/100 Ethernet cards, they will fall back to 10 Mb/s if the cabling won't support the faster speeds. And I'm told quite a few networks do. It's hard to tell unless you have a sophisticated network management system.

So what about fiber ? Fiber is not that easy to install either. Pulling the cable is easy - in fact it can be pulled at 8 times the pulling tension of Cat 5 and the typical cables used include strength members and stiffeners that make it hard to kink and damage. Fiber, by the way, is a lot stronger than steel - remember they don't reinforce fiber glass boats, airplanes, or even tires now with steel - they use glass fiber or aramid fibers, the strength members used in fiber optic cable.

Terminating fiber optic cable is not as simple as copper. While manufacturers have developed crimp-on connectors, they are expensive, high loss and have not been very reliable. Fiber optic connectors need adhesives for reliability and low cost. And most installation involves stripping fibers, injecting adhesives and polishing the ends. No IDC (insulation displacement connectors) here. Any good installer can learn how to terminate fiber in less than 2 hours.

Fiber does not have infinite bandwidth either! At least not the multimode fiber used in most premises networks. It's a lot higher than copper, but as you approach gigabit speeds, you are limiting the distances available for links to 500 meters or so.

Singlemode fiber, as used in telco and CATV networks, practically has infinite bandwidth. But it uses higher cost components and can be pricey for shorter links. It's not necessary for today's networks but may be for the next generation. (You've heard they are working on 10 GB Ethernet, haven't you?)

How about cost? Isn't that the bottom line for copper vs. fiber? Well, fiber prices continue to fall while copper prices (and the more sophisticated hardware needed to support high bandwidth transmission) rise. The cable plant is a wash - comparing Cat 5 and multimode fiber - both are about the same price by the time you consider everything (including fiber testers at $995 and copper testers at $4,000-6,000 and going up for Cat 6, 7 etc.)

It's the electronics that make the difference. A 10/100 Ethernet card for Cat 5 is $10-100, while a 10/100 fiber card (the new 100BASE-SX style) is $100-200 or so. And double the difference because you have to have electronics at each end of the link. Media converters from copper to fiber sell for as little as $100 and solve the problem of getting a fiber connection where a copper one now exists.

The additional cost of fiber is usually offset by the additional cost imposed by copper hubs (with the limited distance of copper - 90m - you need local telecom closets) and the cost of conditioned, uninterruptable power supplies (UPS), data-quality grounds and HVAC for every closet! These costs made one job estimate only $9 per desktop more with fiber - yet the customer still chose copper! Here's another viewpoint.

So why are 99% of all desktops connected with copper? The comfort factor. Installers and customers are both more comfortable with that old familiar copper wire (even if what they install has little in common with their simplistic perception!) and they like the fact that they save $100 or so on every desktop with copper.

The majority of backbones in large companies are fiber. They want the bandwidth and reliability of fiber, and the networking equipment vendors recommend fiber for the backbone. They expect to upgrade to higher bandwidth in the future, and only fiber offers upgradeability.

So what do we think you should do?

If you are a typical user, Cat 5e to the desk is fine. You probably use Fast Ethernet and might use Gigabit Ethernet in the future, and that's what Cat 5e is designed for. However, I suggest you make sure it is installed properly - or you are wasting your money.

Power users should go fiber to the desk today, running a cable with 2 multimode 62.5/125 fibers of the new high bandwidth 50/125 laser-optimized type (>500 MH-km at 850 and 1300 nm) for future 10 GBE applications. Will you need 10 GBE at the desk? Not unless you are a heavy graphics user, like CAD or prepress? Not designing airliners or cars on the desktop? Go back to Cat 5e.

What about Cat 5E or Cat 6?

Right now, Cat 6 is in limbo - no network calls for it and it needs a boost in performance for 10GbE. You get more performance, but no network is yet designed to operate on it. If you want a higher performance UTP cabling system - buy a complete solution from one vendor, require documentation on performance, and hope it has a purpose someday!

Can't I future-proof my network?

At the desktop -no. Flatly, no. You have no guarantee that anything you install for cabling today, Category "whatever" copper or multimode fiber, will be useful in another few years. That's why manufacturers offer 15-25 year warranties - they know you will not be using that cabling more than another few years! A lifetime warranty is the only one that makes sense, since its lifetime is very short! Maybe you could install multimode and singlemode fiber, but you would be the first!

In the backbone, maybe, if you install a large fiber count optical fiber cable with lots of spare singlemode fibers, you probably have a good chance of supporting your network for ten or twenty years. The backbone is easier to deal with since it changes a lot less than the desktop connection.

Silly Move Dept.

For some reason, the LAN migrated away from coax cable to UTP. Granted old "thicknet" Ethernet was a pain to deal with and expensive, but RG-58 ("thinnet" or "cheapernet") was cheap and reliable. GBE would run like a charm over CATV RG-6, and it's cheaper than everything but string!

Oh well, nobody said it was gonna make sense!


The Fiber Optic Association - Tech Topics


(C) 2002-5, The Fiber Optic Association, Inc.

Return To The FOA Home Page

Return To FOA Tech Topics