Fiber Coupling Query

[Jamman960]

I'm hoping some one can help me, I've aquired a few 10m, 25m and 50m fiber patch cables which are all LC-LC along with a few switches and GBICS/SFP's. I'm looking at running the 50m patch between two buildings which are close togeather and currently linked via Cat5e.

 

The 2nd building has a fiber patch panel with 8 cores linked to our main building on the other side of the site via an overhead fiber link. The following is my current plan:

 

50M LC-LC Fiber - 1st building to 2nd building via underground pipe

LC-LC Coupler - Joining the 50m patch to the LC-ST Patch

LC-ST Patch - Connecting between the coupler and the patch panel(2nd building)

- 120M

5M ST-LC - Connectiing the coupler and patch panel again(main building)

LC-LC Coupler - Joining the 5M Patch to the final 10M patch

10M LC-LC Fiber - Final connection to the Switch/Server room

 

Total Distance ~190M

 

Another option is to just go from the 1st building and connect straight to the switch in there via an LC-LC coupler and an LC-SC cable, a direct connection to the main switch would be prefered though... I'm just not sure how much extra latency will be introduced by the above plan.

[CyberNerd]

distance is fine, what you need to watch for is the amount of decibel loss that you introduce from the extra connections. Each connector adds a loss of around 0.5db.

maybe you could bypass the patch panel and use less couplers?

Ideally you splice the connectors together rather than couple. but as the distance is short you might get away with it.

[Geoff]

Also check the fibre in your patch cables are the same grade.

[TNE]

Are all the cables the same spec ie 50/125 or 62.5/125 as i dont think you can mix and match

[Michael]

I'm looking at running the 50m patch between two buildings which are close togeather and currently linked via Cat5e.

 

If they're already linked and it works at 1000Mbps I don't see the point changing it for fibre. I'd only install it where there are lengths 90 metres + which is the getting towards the workable limit for CAT5e.

[CyberNerd]

Are all the cables the same spec ie 50/125 or 62.5/125 as i dont think you can mix and match

 

You can mix n match - it's just not advisable because it increases the decibel loss between the connections. you can go from 62.5 (Receiving) to 50 micron (transmission) with no loss but 50 microm (receiving) to 62.5 (transmission) incurs a 1-1.5 db loss. but yes, get the same grade if possible.

 

 

If they're already linked and it works at 1000Mbps I don't see the point changing it for fibre. I'd only install it where there are lengths 90 metres + which is the getting towards the workable limit for CAT5e.

 

Ideally, interconnected buildings should be linked by fibre.

[Michael]

Ideally, interconnected buildings should be linked by fibre.

 

Why? Unless the length is greater than 100 metres you'll get zero benefit. It's also much, much cheaper. You also need fibre compatible switches or require to buy additional modules which you wouldn't need with CAT5e.

 

I'd say the only exception would be if the area had very high electrical interference (for whatever reason) would I install fibre at shorter distances.

Edited July 29, 2009 by Michael

[bossman]

@Michael:

 

Totally agree with this advice, I would only look at going fibre if I was upgrading switch gear to run on 10Gb backbone and that wouldn't be too soon as the cost is still astronomical at the moment. I have OM3 fibre backbone and switch links already but the cost of the switches are too pricey at the moment. :)

[CyberNerd]

Why? Unless the length is greater than 100 metres you'll get zero benefit. It's also much, much cheaper. You also need fibre compatible switches or require to buy additional modules which you wouldn't need with CAT5e.

 

I'd say the only exception would be if the area had very high electrical interference (for whatever reason) would I install fibre at shorter differences.

 

I was trying to think of the reason as I typed. It was definitely recommended on the CISCO CCNA course that I did, and I'm pretty sure the technical specifications for institutional infrastructure recommend the same thing, and probably a bunch of other wiring standards that I cannot remember. Perhaps it is for electrical interference, IIRC there is also something about voltage differentials between external buildings that needs to be considered.

[bossman]

@Jamman960:

 

You could also trunk cat5E cables to give you larger throughput which would be far less expensive and give you better throughput.

[steveg]

I think the main reason for not using copper between buildings even over short distances is if each building has a different earth connection, there is a potential voltage difference between the 2 earths, so the network cable could carry this voltage. (pretty unlikely though)

 

Steve

[Michael]

Perhaps it is for electrical interference

 

Well other than 10GbE, exceeding 100 metres or electrical interference I can't think of any other reason. In saying that even CAT6 can do 10GbE. To add to this, going from building to building you need external rated cable which is going to be more expensive.

[bossman]

@CyberNerd:

 

As I have said all my switches are Cisco and I have a OM3 fibre backbone and switch links, my servers due to the school insisting I move them have cat6E copper links back to the core switch and I have had no degradation whatsoever and the cost was minuscule compared to fibre costs and also with no gain from installing fibre it was a no brainer.

 

I too have done the CCNA and some of what it teaches can be taken with a pinch of salt.

[srochford]

I think the main reason for not using copper between buildings even over short distances is if each building has a different earth connection, there is a potential voltage difference between the 2 earths, so the network cable could carry this voltage. (pretty unlikely though)

 

Steve

 

It's actually not that unlikely - it really can happen. There's also the risk that you're likely to be on different electrical phases in different buildings which can also lead to fun and games.

 

It's not a huge risk; as ever, there's a need to balance the cost of putting in the fibre against the risk of not doing so.

[Geoff]

It's also illegal. See section B of the Building Regulations.

[CyberNerd]

@CyberNerd:

 

As I have said all my switches are Cisco and I have a OM3 fibre backbone and switch links, my servers due to the school insisting I move them have cat6E copper links back to the core switch and I have had no degradation whatsoever and the cost was minuscule compared to fibre costs and also with no gain from installing fibre it was a no brainer.

 

I too have done the CCNA and some of what it teaches can be taken with a pinch of salt.

 

yes. i know what you mean. I was talking from an 'ideal' perspective, and I am pretty sure that in an ideal situation that external buildings should be linked by armoured fibre.

I think srochfords explanation of electrical phases is the reason.

 

Cat5e or Cat6 will work, I am sure of that, but where do you draw the line?

[bossman]

@CyberNerd:

 

Agree with you armoured fibre should be the norm but in this case I would advise the cat5E or cat6 for cost and outcome purposes. :)

[webman]

Agree with you armoured fibre should be the norm but in this case I would advise the cat5E or cat6 for cost and outcome purposes. :)

 

But if, as Geoff says, it is illegal due to Building Regulations; then copper will be (or should be) out of the question.

[Jamman960]

Just checked all the patch cables, they're 50/125 and so is the 8core cable(luckily... was installed before I joined)

 

At the moment the switch in the 2nd building has 2 Gbic/GigE ports and is linking to 3 buildings - 1 via Fiber, 2 via Cat5E so one is currently only connected at 100mbps. In theory if I could move the closer building over to Fiber & bypass the switch in that building I could then move the other building over to the GigE port and be closer to a star topology.

 

I've worked out that I'd only have to spend around £50 on additional patch cables & couplers so it may still be worth going ahead with. The couplers I've seen are £2.99 each(ebay), based on losing 0.5db per connection I'd probably lose around 3db - is that too much?

[CyberNerd]

What Loss Should You Get When Testing Cables?

While it is difficult to generalise, here are some guidelines:

 

-For each connector, figure 0.5 dB loss (0.7 max)

-For each splice, figure 0.2 dB

-For multimode Fibre, the loss is about 3 dB per km for 850 nm sources, 1 dB per km for 1300 nm. This roughly translates into a loss of 0.1 dB per 100 feet for 850 nm, 0.1 dB per 300 feet for 1300 nm.

-For singlemode Fibre, the loss is about 0.5 dB per km for 1300 nm sources, 0.4 dB per km for 1550 nm. This roughly translates into a loss of 0.1 dB per 600 feet for 1300 nm, 0.1 dB per 750 feet for 1300 nm.

So for the loss of a cable plant, calculate the approximate loss as:

 

(0.5 dB x number of connectors) + (0.2 dB x number of splices) + Fibre loss on the total length of cable

 

Fiber Optics - Testing

 

 

3db should be ok - just like running a 1km cable

TBH I'm not sure what the 'actual' figures should be - I guess it depends on the fibre modules.

[SYNACK]

That many couplers are going to cause trouble, you have a very limited power budget with fibre and connectors are the worst as they drain a huge amount of that power and can add jitter. You also have the issue of all of the additional places for dust to get in. At most I would suggest a patch - fibre - patch - fibre - patch where the two fibres are prepperly installed and speced fibre.

 

You may be able to manage it depending on the quality of the fibre and the connectors on each bit but it will always be a vunrable point where little things like vibration jaring the conectors or dust may compromise it.

 

If it is not to much of an investment then it may be worth a go though as a temporary solution until better cable is avalible. The idea would be to connect it all up then go onto the switch managment pages for the switches at each end. Clear the port counters then send several really large files over the link. Once you have done that check the switch pages for errors on the ports in question. There should be none, if there are look for a different solution, if not then it may well work properly for some time.

 

The issues with running copper between building are real especially for above ground cable which can be affected with lightning and fry two switches rather than one let alone the nasty potential voltage that can be generated between two differing ground points and/or phases.

 

EDIT: Can't find the spec for multimode but the low end single mode stuff - which is more powerful - seems to have a power budget of around 10.5dB ftp://files.dlink.com.au/products/DEM-311GT/Datasheet/DEM-31xxx_datasheet_04.pdf

Edited July 29, 2009 by SYNACK

[bossman]

But if, as Geoff says, it is illegal due to Building Regulations; then copper will be (or should be) out of the question.

 

I agree also with what Geoff states but it is only advice based on costs and outcome not legalities, if it has to conform to wiring regulations then there is only one option and that is armoured fibre.

 

Case closed

[webman]

I agree also with what Geoff states but it is only advice based on costs and outcome not legalities, if it has to conform to wiring regulations then there is only one option and that is armoured fibre.

 

So we can ignore the legalities as long as we do something for cost reasons (and 'outcome', whatever that means exactly )?

[Oops_my_bad]

We have a science block that was originally connected (blatantly illegally) via overhead cat5e I was replacing the switches either end after about every 6 months due to each port sequentially getting blown.

 

Needless to say they now have a nice overhead OM3.

[bossman]

So we can ignore the legalities as long as we do something for cost reasons (and 'outcome', whatever that means exactly )?

 

Not at all that is not what I am saying far from it, if you look at the scenario it states that 50m is underground and this is what I was referring to, as for the rest it is a logical decision based on his requirements and that is around 140m of fibre albeit patched.

 

In my honest opinion I would go for copper underground 50m as stated and the rest get someone in to fibre overhead in one link and thus cut down on the need to join everywhere.