Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Going the distance

At 9:16 AM 99/7/1, Larry Miller wrote:
>In our case the huge majority of increased link length came from the fact
>that the LX receivers are 6dB or more quieter than SX (or the standard), and
>on SM fiber this translates into a much larger link budget. One of the
>reasons we standardized on this is that transceivers with lower noise seem
>to be more reliable as well.
>It was not until we got into some custom 50 km links that we had to go to
>"death ray" lasers of increased power.
>This was a case of getting a 2X improvement over the standard "for free".
>Obviously, this may not happen in the 10GbE world.
>I suspect that if the Committee, in its wisdom, legislates under-achieving
>link lengths to accommodate the simultaneous worst-case everything as was
>done in 802.3z, then we will have a similar phenomenon as we have today: if
>you cannot easily double the link lengths (at least) then you probably
>bought some pretty cheesy equipment. Sales people, I think, would rather
>have the standard reflect more closely what the technology will reliably do,
>not some pruned-back number that they have to talk around.

We in TIA FO-2.2 are in the process of researching and standardizing such a
doubling of the GbE minimum guaranteed range, based precisely on such
observations.  In my own systems, I regularly get 750-meter actual ranges
at 1.20 Gbaud (using Fibre-Channel FC-0 and FC-1 coding and chips) over
standard-issue fddi-grade 62.5-micron cabling at 850 nm.  This is 3.4 times
what the standard guarantees, so lots of link length was left on the table;
it would have taken too long to do the work necessary to increase the GbE
link lengths.

If you recall from the GbE Modal Bandwidth Investigation effort of a few
years ago, the key issue in setting those guaranteed minimum link lengths
was the wide variability in fiber characteristics (with uncontrolled laser
launches).  Because the then objective was to ensure that at least 99% of
conforming links would work, the 1% tail of the fiber bandwidth
distribution wagged the dog, resulting in grossly overdesigned (~4:1) links
in the typical case.   A good bit of this tail was due to ancient fibers
already installed in walls.

So, if we can find a way to reduce the variability, even if the average
bandwidth is unchanged, we may be able to significantly increase the
guaranteed lengths.  Better fiber is one way, better launch optics is
another, and one can do both.   FO-2.2 is presently pursuing better launch
optics, and a simple field test for delivered bandwidth of new and existing

Joe Gwinn