# When to choose fiber instead of twisted pair (copper)

These days both fibre and twisted pair are supporting 1000BASE-T and 10GBASE-T, but when do you choose one over the other? The obvious one is maximum cable length, but what are other factors which might come to mind when making this choice.

• When chosing fiber, there are electromagnetic disturbances. Edit: Of course fiber doesn't support POE. – user217 May 8 '13 at 7:02
• While you can't do poe over fiber you can run fiber close to your device and then add a fiber media converter to add poe to a shorter run of cat5/6. POE is supposedly limited to 100m distance. – D. Kermott Aug 5 '16 at 22:24

There are a fairly large number of factors to consider and it is also important to accept that not all fiber is the same just like not all twisted pair is the same. Here are a few things that occur to me. Of course these may vary depending on environment.

Fiber: Much longer max distances, depending on specification and throughput. No electromagnetic interference. Note however that fiber comes in a large number of grades and planning for the future may be somewhat more complex.

Copper: Shorter max distances per link. Susceptible to electromagnetic interference. Fewer number of grades/variations (but still mind these!) meaning fewer accidents when the rollout team grabs the wrong spool (see below for a funny story). Supports power over ethernet.

Funny story: A particular county in a Western state runs a fiber network that began as a control system for their hydroelectric dams. As time went on they found they didn't need nearly as much fiber as was laid for this and so they separated out many of the strands and used it to build a county-wide network linking homes and businesses to ISP's and telco's. Of course single mode fiber was used for longer transmissions and multi-mode for short runs.

So when they were running the longer runs to connect towns and cities with this new network, the team that ran the fiber grabbed the wrong spool, and ran multi-mode fiber on all these long distance links. The team was quite unhappy when they found out they would have to re-run all the fiber again! Moral of the story: be very careful about ensuring that everyone knows which grades of fiber go where....

• Fiber cabling has a standard color scheme to depict the type of fiber to help prevent problems like these: - Yellow == 1 Gig SM (9 micron) - Orange == 1 Gig MM (50/62.5 micron) - Aqua == 10 Gig MM (50/62.5 micron) – Mat Wood May 8 '13 at 14:25
• Normally I see orange as 62.5 micron MM and aqua as 50 micron MM, whether this is 1G or 10G. – YLearn May 8 '13 at 19:22
• The colour code is only for single pairs, mult-pair is always coloured individually per-core to allow identification. Usually the type of fibre is printed on the jacket. – LapTop006 May 11 '13 at 13:24
• Yellow is generally single mode and supports 10G as well as several different 10G waves simultaneously (such as when using WDM.) saying it is 1G is misleading. – netdad May 20 '13 at 4:20
• Outside plant cable (the sort you'd be running on telephone poles and/or most outside conduit runs) is almost always black on the outside - holds up best in sunlight. So you always have to read the text that's imprinted if you have more than one kind of cable. There is no 62.5 10Gig cable. Singlemode is not only "able to do 10 gig" it's essentially unlimited - only the electronics driving it limit the speed. – Ecnerwal Aug 9 '13 at 16:26

While most other answers have talked about Length limits, Interference issues and are all correct. One important factor while choosing the cable is the speed of the network and the devices you plan to use.

The general specifications for currently available ethernet network speeds are as follows

1. Twisted Pair Copper Cable
• Category 5e
• 10/100/1000BASE-T Networks (i.e Upto Gigabit speeds) can be run on cables up to 100M in length
• 10GBASE-T Networks cannot be run on these cables.
• Category 6
• 10/100/1000BASE-T Networks can be run on cables up to 100M in length
• 10GBASE-T Networks can be run on cables up to 37M to 50M. (This limit value is a range because it depends upon the alien-crosstalk environment)
• Category 6A
• 10/100/1000BASE-T Networks and 10GBase-T Networks can be run on cables up to 100M in length
2. Optic Fiber Cable
• The distance and speed limits of optical fiber cable are heavily depend on the kind of transceivers used on the ends of the cable. I'll summarize some of the "ideal" combinations here, you can take a look at all possible link specifications here
• Single Mode Fiber (SMF) Optic Cable
• Gigabit upto 5Km with 1000Base-LX transceivers
• 10G is supported up to 40Km with 10GBASE-E transceivers
• 40G is supproted upto 10Km with 40GBASE-LR4 transceivers
• 100G is supproted upto 10Km with 100GBASE-LR4 transceivers
• SMF is widely used for the long-haul where installed fiber is typically reused for more than 25 years. It is one of the most future-proof standards available.
• Multi Mode Fiber Cables
• OM1 & OM2 Grade cables
• Gigabit upto 550m with 1000Base-LX modules.
• 10G is supported up to 300m with 10GBASE-LX4 modules.
• 40G & 100G is not supported.
• Gigabit upto 550m with 1000Base-LX modules.
• 10G is supported up to 300m with 10GBASE-LX4 modules.
• 40G is supported up to 100m with 40GBASE-SR4 modules.
• 100G is supported up to 100m with 100GBASE-SR10 modules.
• Gigabit upto 550m with 1000Base-LX modules.
• 10G is supported up to 300m with 10GBASE-LX4 modules.
• 40G is supported up to 125m with 40GBASE-SR4 modules.
• 100G is supported up to 125m with 100GBASE-SR10 modules.

Now keeping these specifications in mind, I'd pick a cable that will support the current network requirement, while having sufficient scope for expansion. It is important to choose a medium that will support enough growth because usually replacing switches or modules on both ends of a link is just a one/two man job, while running new cabling for an upgrade is highly labor intensive and might cost a lot of money in comparison.

Let me also list down some of the scenarios where I'd pick each cable.

1. Category 6 would be my choice for last mile connectivity to the desk/wall outlet for any organization that either has
• A very low bandwidth requirement, like small cafes, retail stores etc
• Or where the expected life of the interior fitout and office space is less than 5 years
• It might be worth noting that Category 5e might seem cheaper, but its only by a small margin. So I'd just skip that for the better performance.
2. Category 6A will be my choice for both last mile connectivity for the following reasons
• Ability to run at 10G speeds later on
• Less susceptible to noise, interference and cross talk compared to Cat 6. (Since allmost all Cat6A cables are shielded)
• I'd also run some Cat6A within Racks in the same Aisle or Zone of a Datacenter.
• Cat6A would be my choice for the last mile haul in any industrial or factory environment if they cannot use fiber all the way to their equipment.
3. Single Mode Fiber would be my cable of choice for any kind of backbone cabling where the long haul is needed. It has stood the test of time and I dont think it'll get obsolete any time soon. I'd pick the OS1 grade for indoor applications and OS2 grade for outdoor applications.
4. Multi Mode Fiber is by far my most favorite of the lot. I'd use it in between my telecom rooms, between racks in the data center. I'd pick OM3/OM4 depending on the budget available.
5. If possible to convince decision-makers I'd use a combination of SMF and MMF across the entire network, even to the desk or wall outlet. Going for a completely fiber only networks saves up a lot of space and energy costs and provides a lot of management flexibility by drastically reducing the number of telecommunication rooms in large facility. Also such a network will have a large capacity of bandwidth and provided you've run enough number of cores, it will be future proof to a large extent. (Even the costs of ducting and pathways needed for an all-fiber network will be cheaper than an equivalent copper based solution). Read more about this here.
6. The only technology that fiber cant support today is analog telephones, at least not without a converter on both ends.

I know this has been a long answer, but there is no such thing as the perfect cable for everything. In the end you need to balance out between cost, capabilities and provision for growth based on the requirements of your particular application and business.

• "Read more about this here." It basically seems to come down to "we assumed ridicuously overspecced wiring closets to make copper look expensive". It also seems to count the cost of real-estate for switches in distributed closests but ignore that their centralised system will need nearly as much real-estate in the central closest. – Peter Green Aug 24 '16 at 13:36

If cost is a factor for you, I would look at it like this;

Scenario 1

If you are making a short run, say 50M, I would go for copper, simply because of cost (This 50M run is between two comms cabs in an office building for example).

Copper is cheaper and using switches would save on the cost of SFPs or GBICs and so on. We obviously aren't talking thousands here but some people are on strict budgets.

Scenario 2

If you are making a short run between two cabs in a DC between core routers, it would be fibre for me hands down.

In scenario 1, you could in fact drop in two or three 50M lengths of copper for almost the same price as one run. This would cover any likely future expansion. In scenario 2, fibre would better cover future expansion and give a greater ROI because of technologies like WDM.

As others have mentioned, fibre will go futher. If you want greater than 100M, you have one option - fibre. If you need a shorter run, it simply becomes a cost trade off.

If you ever feel that you are going to run FCoE over the medium then it is better to run fibre than copper as the BER for copper is significantly higher and most of the time will be outside of the tolerance for the 'lossless' nature of FCoe.

• What is FCoE acronym? – Trevor Boyd Smith Feb 16 at 17:21
• FCoE stands for fiber channel over Ethernet and is usually used when you have a converged storage network. It is very intolerant to errors and so you are best with fiber in this case as it has a much lower BER. – David Rothera Feb 19 at 9:05

I will add another factor that hasn't been mentioned.

I can quickly and easily make copper cables, measured and cut to the exact length I need for a run...resulting in neater cable plants.

While you can cut to length and put ends on fiber optic cable, I find it considerably more difficult and time consuming...and my experience doing this is with multi-mode fiber, I can only imagine that single-mode is that much worse.

• Not if you're using gigabit ethernet (and especially not 10g). Hand crimping simply can't reliably match the quality needed. It will probably work, but makes little sense. These days it also costs more to have someone crimping cable then stocking in half meter increments. – LapTop006 May 11 '13 at 13:26
• I hand crimp for gig with very high success rate. We don't have any 10g at this point and I probably will not try to hand crimp for that, but its not hard to get good quality runs and crimps for gig. – Jeff McAdams May 11 '13 at 13:56
• But do you actually test with a qualification tester? Unless you actually monitor error counters on every link that's the only way you can be fairly sure it's actually a good cable. – LapTop006 May 11 '13 at 14:05
• Yes, I have relatively recently gotten a qualification tester. Since getting it, I haven't yet made one that failed qualification. – Jeff McAdams May 11 '13 at 15:12
• I agree with Jeff, I've never had any problems with the cables I've used for gigabit Ethernet, and I use cheap connectors from China with a 10-year-old crimping tool. Also, like Jeff, I cannot yet speak for 10-gigabit, though. – WaxTrax May 24 '13 at 0:40

Some things that I haven't seen mentioned:

1) 10GBASE-T uses a lot of power compared to fiber to DAC.

2) 10GBASE-T transceivers have much higher latency than other options. This could be important in a compute cluster or other low latency environment like automated financial trading.

Lots of people have answered already; and I agree: Fiber beats copper as soon as distance becomes a factor.

But beware of multimode fiber; it also has serious distance issues. You can only go 550 metres on GE (1000BASE-SX), and some of the 10GE modes manage 330 metres on really good multimode fiber (OM3 or OM4). In a larger colo or CO or in a campus setting, this is severely limiting. IMNSHO multimode is a "rack row" technique only for 10G.

Combining this with the many available grades of multimode and you've got a recipe for confusion and investment one-way alleys.

Singlemode, OTOH, is simpler, and the price delta down to multimode keeps getting smaller. I am not recommending anything except singlemode for new structural installs, limiting multimode new installs to backwards compatibility (ie. new computer room needs to interconnect with old FC infrastructure on 50µm or to closet switches on 1000BASE-SX).

• "In a larger colo or CO or in a campus setting, this is severely limiting. IMNSHO multimode is a "rack row" technique only for 10G. " and it gets worse for 40G/100G. The multimode standards require multiple fiber pairs and seem to have even shorter distance limits. – Peter Green Aug 24 '16 at 13:39

Fiber over Copper...

1. High electrical noise environment.
• The first place I used fiber was on the production floor of a textile mill. twisted pair was a nightmare there.
2. Long spans -- in excess of 100m.
3. Anywhere future in-place upgrades are desirable
• I know of a few places where SM fiber from the 80's is still being used today for 10Gbe.
4. Anywhere that maximum speed is desired.

Yes, there are many different grades/qualities of fiber, but unless we're talking 10km+ spans, the differences aren't that major for 1/10/40/100Gbe. Higher speeds reduce overall distance, but that's an unavoidable reality.

There are technologies like wavelength division multiplexing (WDM), that you can't use with copper. I would say that two best reasons to use fiber over copper is max cable length and WDM.

When running data cabling between different electrical bonding zones (eg buildings with different electrical earths). If this is done with copper cable, different ground potentials might try to equalise over the cable. Fun to watch, not so funny for the owner of the equipment. Better done with fibre.

One non-techncial reason to install fiber over copper for switch uplinks and the like, ignoring costs, etc: People are more afraid to touch a fiber cable compared to copper, so you'll get less of the "Oops, I didn't realize that was the uplink".

• I love this answer. I'd like to point out that in addition to fear (for non-technical people) it also provides a differentiation for technical people. It's a lot easier for me to guess that the fiber is probably the uplink and I shouldn't be touching it unless I really mean to. – CrazyCasta Apr 7 '14 at 23:25

One thing that none of the other answers have mentioned yet is size. Fiber is physically smaller and you can fit more runs of fiber in a given space than copper, especially cat6. Depending on your environment, this can be a consideration.

• Panduit (and probably others) are starting to offer "small diameter" copper patch cables, down to 28 AWG for Cat5e and Cat6. They're still not quite as small as 50/125 multimode, but they're a good bit smaller than typical Cat5/6 cabling. – James Sneeringer May 31 '13 at 17:42
• However bend radius control is more important for fibre. – Peter Green Dec 4 '15 at 2:09

I have the feeling copper is a bit more robust then fiber. In a datacenter this is not an issue because the cables stay where they are once the links are up.

However, in harsher conditions (temporary setups in open air, equipment hauled around the country and networks rebuilt often, ...) sturdiness is a factor. I haven't kept score, but I think fibers fail me a bit more often than copper, cables getting damaged or connectors getting dirty.

YMMV of course, and as Lucas may remember, those little clip thingies on RJ45 connectors aren't the most student-proof invention either...

Fiber is also useful to protect the equipment when the placement of the cable has the risk of getting hit by lightnings. A stretch of fiber at both ends of a copper link can also serve this same purpose.

• It's not so much the cable itself that causes an issue but the potential difference between the two buildings should one be hit, electricity from such strike can conduct down copper but not fibre. – David Rothera May 26 '13 at 23:08

No one seems to have mentioned that with Fibre - you have the future option for DWDM.

However, I guess if you're debating between fibre & copper on a run - DWDM would probably be your last thought as if you really needed additional bandwidth you'd just run another cable (since the run would be short enough)

But over LONG runs & links - DWDM allows you to future-proof your fibre investment if you have the correct hardware.

Things to keep in mind regarding Fiber vs Copper:

• Potential for unidirectional link failure (Which is not a problem with copper)
• A lot depends on the Transceiver e.g 1000 Base LX/LH is only capable of ~ 5KM so for longer distances plan for higher end SFP.
• More expensive to install fiber (Splicing, pulling etc.)
• Fiber is not as rugged (patch cables) as Copper (Watch those bends)

A bit off-topic: if you are planning to go with fiber, in my experience Cisco branded SFP are ten times the cost of other not as well known brands, e.g. startech. This was a few years ago but it was something like $650 vs.$70. I had both types installed and working without a hitch.

Unless you report to a manager who really believes "No one was ever fired for buying Cisco"-

I use a mix of both in central offices (usually multi-mode), and the primary reason I choose fiber over copper is that fiber patches are pre-built and it's much harder for a technician to make mistakes on the install. When you have techs building cables, you are adding a lot of potential for human error. Minimizing install mistakes is a huge priority when you are on a time budget.

Cost factor i think is everything when trying to choose.

If speed and long distance coverage is essential in your network design and implementation Optical Fiber cable is your best suite

Direct point-to-point fibre provides fast failure detection The default debounce timer on GigE and 10GigE fibre linecards is 10 msec The minimum debounce for copper is 300 msec

By and large, copper is preferred because the interfaces are less expensive and copper has the ability to deliver power concurrent with the transmission of data. Always run copper to endpoints unless it is not possible. As far as the links between networking devices it could be copper or fiber and like you said link distance is the primary determinant.