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

RE: PMD discussion - parallel solutions




Hi

Parallel optics solution offers lowest cost for the transceiver ~1/2 other 10Gig 
equivalent.  The problem with the parallel optics is the cost of the cables as 
you exceed 50 meters and become absorbent over 100m.  In the Infiniband we only 
defined the parallel optics solution for up to 125m based on existing 50/125 um 
fiber, but relaxed the specifications for lower cost.  In my view if your are 
going over 100m then you need to switch to SMF with WDM or serial. 

I believe parallel optics is an attractive solution for the central office/ 
computer room.  The current PMD under consideration 802.3 are targeted for the 
backbone.  Some have suggested in the computer room /Centeral Office 
interoperablity will be established at XAUI, you just have to add similar VSR 
transceiver on both end of the link.

If your company is a member of the IB then you could look at the IB optical 
specifications for 10Gig and 30 Gig at http://www.infinibandta.org/home.html and 
look for volume 2B.

Thanks,

Ali Ghiasi
Sun Microsystems  


> From: mdonhowe@xxxxxxxxxx
> X-Lotus-FromDomain: WLGORE
> To: "'802.3Ae'" <Stds-802-3-Hssg@xxxxxxxx>
> Date: Wed, 31 May 2000 08:05:20 -0400
> Subject: RE: PMD discussion - parallel solutions
> Mime-Version: 1.0
> Content-Disposition: inline
> X-Resent-To: Multiple Recipients <stds-802-3-hssg@xxxxxxxxxxxxxxxxxx>
> X-Listname: stds-802-3-hssg
> X-Info: [Un]Subscribe requests to  majordomo@xxxxxxxxxxxxxxxxxx
> X-Moderator-Address: stds-802-3-hssg-approval@xxxxxxxxxxxxxxxxxx
> 
> 
> 
> 
> The viability of parallel solutions keeps coming up and I feel compelled to
> respond (sorry - usually I resist the temptation).
> 
> I believe parallel is the lowest cost transceiver for 10G---it's hard to beat
> one alignment for an array of VCSELs with one IC that sets the drive currents
> for the entire array.   Contrast that to (presumably) WDM's multiple 
alignments
> and multiple ICs or at least setpoints for each laser.  Or to the higher speed
> ICs required for serial.  The 'problem' with parallel is that at some run 
length
> it no longer is the lowest cost solution even if it is the lowest cost
> transceiver.
> 
> Although parallel is a great application for 10G (and higher) I did not feel 
it
> was a proper solution for 10GbE because of the distance objectives and market
> timing.  First, 12-wide parallel transmitters and receivers already exist from
> several vendors and by the time this standard is finished you would want to go
> at a higher data rate with fewer fibers.  A parallel 4-wide looks very 
promising
> but when I used the link model to analyze it at 3.125 it runs into the 3.6 dB
> ISI limit just short of 300m.  (Although technically one could argue 100m is 
all
> that is needed because you could 'conceivably' gang up installed
> fiber---practical implementations would install new ribbon cable).  So it 
failed
> to meet the distance objectives.  One could install a higher bandwidth fiber 
but
> why not go serial then since that meets the same distance objective and by the
> time the standard is finished will be a lower cost solution (providing IC cost
> trends stay the same).
> 
> I look at parallel and serial playing a leapfrog game where there is a market
> window for parallel that eventually closes when the higher data rate serial IC
> costs come down.  Now throw WDM in the mix and ask yourself if packaging costs
> are dominated by alignments (traditional?) or by components (specifically 
ICs).
> 
> More recently a 4-wide parallel scrambled approach has been proposed that 
would
> meet the 300m distance objective.  So now the question is by the time the
> standard is finished will it be a lower cost solution?
> 
> -Mark
> 
> Mark Donhowe
> W.L. Gore & Associates
> 750 Ott's Chapel Road
> Newark, DE  19713
> ph: (302) 368-2575
> fax: (302) 737-2819
> mdonhowe@xxxxxxxxxx
> 
>