[f-nsp] 10G-XFP-LR optics used on short links

[Richard A Steenbergen]

On Sat, Nov 22, 2008 at 05:28:16PM +0100, Niels Bakker wrote:
> * helmwork at ruraltel.net (Eric Helm) [Sat 22 Nov 2008, 05:24 CET]:
> >Hello, I'm working some XMR routers that only have 10G-XFP-LR optics 
> >available. There is one short link connecting to XMR routers of around 
> >3 meters. Should an attenuator be used on these links? I've looked at 
> >the output of 'show optics' and the XFPs are operating within a normal 
> >range. I'm concerned that some damage may occur to the XFPs in this 
> >case.
> 
> LR was designed to work without attenuators.

This is true, but I will mention that I have come across "unusually
strong" LR optics in my time, particularly early generation ones (XENPAK
mostly). I have a friend who is lighting a 22km fiber span with LR using
optics that put out +3dBm (way above spec, and would probably require
attenutation if used in a 3 meter config), though this will probably
cool as the optic ages. If in doubt, try an air gap (1 inch free-space
optics, aka gods attenuator :P).

> Please don't fearmonger.  We're talking run-off-the-mill standard optics 
> here so the saturation point for the receiver is far removed from the 
> damage point.  The only damage that can occur is if you drop them or 
> cough into them.

The potential for damage is a function of the sensitivity of the
receiver, not the power of the transmitter. For example, the difference
in transmit power between a 40km and 80km 1550 optic is negliable, it is
the difference in receive sensitivity that makes an 80km link work. If
you shine a very bright (say +3dBm) signal directly into a 80km optic
for a long period of time you risk damage, but LR is pretty much
impossible to hurt with another router optic. Routers just don't have
particularly interesting lasers. Nothing you can plug into a router will
ever transmit more than ~2.5mW (with most optics actually putting out
less than 1mW, even the 80km ones), whereas you can go buy a 50mW laser
pointer for $50 and shine it at planes all night long. :)

You can of course temporarily blind the optic by transmitting too strong
a signal, which you will see as bit errors just as if you had too low a
signal. If in doubt, a $75 light meter (and 30 seconds spent RTFMing so 
you know the difference between absolute power mode and fixed light 
source metering mode, an investment 95% of colo techs refuse to make) is 
your friend.

-- 
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)