[c-nsp] 15km optics

Richard A Steenbergen ras at e-gerbil.net
Fri Mar 13 16:48:17 EDT 2009


On Fri, Mar 13, 2009 at 12:48:15PM -0300, Ray Burkholder wrote:
> I have a carrier which has about 13km to 15km of fibre to their CO.  They
> run 1310 nm alcatel equipment on their end.  I was wanting to use a Cisco
> GLC-LH-SM on my end but that is rated for 10km.  The GLC-ZX-SM is 1550 nm
> 70km device which is incompatible.  

I think this has already been answered to death by everyone else on the 
list, but I'd like to try and summarize a few points into something 
which can be pointed to on the list archives the next time someone asks 
this question.

1. All SMF optics have wideband receivers and will happily accept any
signal from any other source (typically 1270-1610nm), with absolutely no
requirement that it match what the other side is sending. Don't think of
a fiber circuit as a single link, think of it as two completely distinct
links running in opposite directions which just happen to be parallel to
each other for TX and RX.

2. Distances specified on optics mean absolutely nothing, they are there 
to give you a rough estimation of what you could expect under the worst 
possible conditions. Optical budget (and at higher speeds, things like 
dispersion penalties) are pretty much all you care about (ignoring goofy 
things like reflections which aren't relevent to this discussion).

3. Not all optics perform the same. Think of it like how processors are
made, they make a batch and some of them perform better than others. A
certain number perform completely outside of spec and have to be thrown
away, but even within the "good" batch there will be varying levels of
performance and quality.

For example, Cisco LX/LH optics have a TX power spec of between -3 to
-9.5 dBm, and a RX power spec of between -3 to -20 dBm. The max TX of -3
is so that incase you have two LX optics side by side, the signal isn't
so strong that it blinds the other side. The distance spec is calculated
assuming the worst possible conditions, a TX over -9.5 dBm, running over
old crappy SMF28 fiber. In reality, you're likely to find that most of 
your LX/LH optics operate much closer to the -3 side than the -9.5 side, 
and infact you may find some that are hotter than spec. Fixing something 
thats too hot is easy, you just add an attenuator. You'll find that you 
can almost always go at least 2dB below your RX budget too, before you 
actually start taking errors. Think of it like the max capacity sign on 
an elevator, they actually make the steel cables to hold many times more 
than that max capacity, "just incase".

Also optics tend to cool over time, so what shipped as a +1 might end up 
as a -1 after a few years, and they don't want your link randomly going 
out and you complaining. :)

4. Often times, links can be extended simply by changing out one side of 
the pair with a longer reach optic. When you upgrade one side, not only 
do you increase the transmitted signal strength (and often switch the 
signal from 1310 to 1550, which has a lower rate of loss per unit of 
distance), you're also increasing the receiver sensitivity. The far side 
can hear you because you're transmitting a stronger signal in a band 
that has less loss over distance, and you can hear the far side because 
you have a more sensitive receiver.

At the end of the day, almost all your questions (well at least all the 
simple ones :P) can be answered by a light meter or integrated optical 
monitoring on newer optics. Anyone who doesn't have a light meter would 
do well to get themselves over to ebay and spend a couple hundred bucks. 

<rant>
Also, please for the love of god read the instruction manual and learn 
the difference between "absolute power mode" (i.e. how much signal am I 
receiving?) and "relative loss mode" (i.e. how much loss does this cable 
have assuming that there is a known fixed-strength signal generator on 
the other side?). I'm seriously sick and tired of 95% of supposedly 
trained remote hands techs telling me I'm transmitting +65dBm. :)
</rant>

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)


More information about the cisco-nsp mailing list