Range using single-mode SFPs across multi-mode fiber - was Re: NANOG Digest, Vol 47, Issue 56

Keegan Holley keegan.holley at sungard.com
Wed Dec 14 21:04:51 UTC 2011

2011/12/14 Justin M. Streiner <streiner at cluebyfour.org>

> On Wed, 14 Dec 2011, Keegan Holley wrote:
>  inappropriate. We are attempting to use Juniper single-mode SFPs (LX
>>>> variety) across multi-mode fiber. Standard listed distance is always
>>>> for SFPs using the appropriate type of fiber. Does anyone out there
>>>> know how much distance we are likely to get? Thanks for your help in
>>>> advance.
>>> Single mode just has a smaller core size for the smaller "beam" emitted
>> by
>> laser vs. LED.  it works although I've never done it outside of a lab (MM
>> is cheaper). As for the distance it theory that should come down to the
>> optics and your transmit power.  Hopefully this is just a cable connecting
>> the router to a long line.  I've never heard of a 10K MM fiber run since
>> SX
>> optics can't shoot that far.  You should be able to get through the 500m
>> or
>> so that MM optics are rated for, but YMMV (errors, light levels, bounces,
>> etc etc)
> In a nutshell, don't do it if at all possible.  This issue gets
> significantly
> worse at 10G.  If there's any way to get SMF in place for this link, do it.
> +1 probably should have added that.  I guess I just assumed.

> In practice, you will likely get something less than the rated distance,
> particularly if the MM fiber in question is an older type, such as OM1. If
> you're using OM1, mode-conditioning jumpers at both ends are pretty much a
> must.
> The problems with shooting an LX/LH beam over MMF are threefold:
> 1. Attenuation on some flavors of MMF can be significantly higher than an
> equivalent run of SMF.
+1 Assumed again..

> 2. Modal dispersion on MMF will scatter and distort the LX beam, likely
> resulting in link errors because the receiver can't recover the data
> correctly.

Not that I'm advocating this, but it's fine over short distances.  I did
this for a few lab routers where I wasn't concerned with link quality, but
I was able to fill a 10G pipe with no errors/retransmit over about 10M.

> 3. Shooting a 9 micron beam into a 50 (or worse, 62.5) micron core, and
> getting enough of the beam to reach the 9 micron target at the other end to
> result in a recoverable signal is problematic.

Again for short distances it's doable.  I agree not to even try over 62.5

More information about the NANOG mailing list