RE: (SSIG) Taking the winning route
All,
Let's clarify the difficult task of comparing different PHY costs by looking
at the different cost components for each PHY. In the context of this
fascinating discussion, it is worth pointing out that 802.3z also did not
"need" the SX solution. It was included as a cost-saving solution, and it
currently represents roughly 80% of the volume sales of GbE
opto-Transceivers. Somehow that MUST be relevant. In the table below I've
added the GbE SX solution to the fray. Due to space constraints in email,
we'll have to put up with abbreviations. More explanation below.
GbE-SX GbE-LX Ser.-850 WWDM Ser.-1300
Laser Wavelength and Tolerance
1 VCSEL 1 F-P 1 VCSEL 4 DFBs 1 VCSEL/DFB
850+-5% 1310+-3% 850+-0.9% 1310+-0.2% 1310+-1.1%
Detector (# required)
1 1 1 4 1
Fiber Coupling
Tx 1 MM 1 SM 1 MM 4 SM 1 SM
Rx 1 MM 1 MM 1 MM 4 MM 1 SM
Mux/Demux
N/A N/A Elect.4:1 Opt.4:1 Elect.4:1
Tx + Rx IC's
1+1 @1G 1+1 @1G 1+1 @10G 4+4 @2.5G 1+1 @10G
Fiber
Inst MM Inst SM InstMM to 55m Inst SM/MM Inst SM
COST 0.5X 1.0X 1.25X 4X 2X
The laser wavelength tolerance is included since it can become a significant
cost issue in the 10GbE lasers, particularly for the WWDM lasers. In all
cases, about 0.25% was deleted from the nominal wavelength tolerance due to
temperature variations. For WWDM, of course the wavelengths are different;
the 1310nm represents the wavelength region. Detector comparisons are
straightforward. For fiber coupling it should be pointed out that
single-mode (SM) coupling is more difficult on the transmit end and MM
coupling is more difficult on the receive end. I took the WWDM Mux/Demux
out of the optical comparison and put it into a general Mux/Demux comparison
row. This made sense because only the WWDM needs Optical Mux/Demux and only
the 10GbE serial solutions need Electrical Mux/Demux. For the fiber
comparisons, "Inst" means "Installed." This should add a dash of realism
for the final row which is the cost comparison which assumes 1.0X for the
1GbE-LX to keep it consistent with Rich's email.
One topic of cost discussions is that of the 10Gig electronic IC's. While
some vendors are currently charging ludicrously high prices, this is a sign
of current lack of availability. The cost of the actual processed/packaged
silicon is not high. For long term prediction of IC costs, put it this way.
Who has ever bet against the scalability of Silicon - and won?
Jack Jewell
-----Original Message-----
From: Mike Bennett [mailto:mjbennett@xxxxxxx]
Sent: Saturday, April 08, 2000 5:50 PM
To: rtaborek@xxxxxxxxxxx
Cc: HSSG
Subject: Re: (SSIG) Taking the winning route
Rich,
I should have read more carefully. Somehow I got the idea you were talking
about High
Bandwidth MMF. The relative cost estimate is very informative. Thanks,
Mike
Rich Taborek wrote:
> Hi Mike,
>
> My comment on Serial solutions was with respect to the installed based of
MMF,
> which is "FDDI grade" 62.5 um MMF as I understand it. The bandwidth of
this
> fiber @ 850 nm is typically no more than 200 MHz*km. At 10.3125 GBaud,
this
> translates to roughly:
>
> 200 MHz*km
> --------------- = ~39 meters (the fastest signaling rate is 1/2 the
Baud.
> This number is inflated since it does
> 10.3125 GBaud/2 not consider laser rise/fall time which
> corresponds to a higher effective signaling BW)
>
> Serial 850 nm solutions can easily achieve 100 m, but require new
"enhanced"
> MMF. This fiber does not correspond to the installed fiber base.
>
> Assuming that by VCSEL, you mean "serial VCSEL at 850nm" for purposes of
cost
> comparison, I estimate the cost differences between the various PHYs as
(very
> roughly):
>
> 1 GbE LX PHY current cost = 1
> 10 GbE Serial VCSEL @ 850 nm in (2002) = 2-3
> 10 GbE WWDM @ 1300 nm in (2002) = 3-4
> 10 GbE Serial 1300 nm in (2002) = 2-4
>
> Best Regards,
> Rich
>
> --
>
> Mike Bennett wrote:
> >
> > Hi Rich,
> >
> > I have a few questions for you: What distance would you say the 850nm
serial VCSEL
> > solution would reach? If 100m is not possible for serial VCSEL at
850nm, please
> > explain why not. What would you estimate the relative cost difference
between the VCSEL
> > and 1300nm WWDM PHYs will be, say normalized to current LX PHYs?
> >
> > Respectfully,
> >
> > Mike Bennett
> > Lawrence Berkely Lab
> >
> > Rich Taborek wrote:
> >
> > > Martin,
> > >
> > > I apologize, I should have included a bit more detail with my last
response. I'm
> > > currently way backed-up with reflector traffic for the week and now
see that my
> > > response left out some details.
> > >
> > > WWDM 1300nm is the only strongly supported PHY proposal which
addresses the
> > > intalled base of MMF.
> > >
> > > No Serial solutions, including the 850nm serial VCSEL solution,
address the
> > > installed base of MMF at reasonable distance. I peg this disatance at
100m as
> > > does the corresponding distance/cable plant objective.
> > >
> > > The point of my previous note was that the 850nm serial VCSEL solution
over
> > > enhanced MMF meets only one distance/cable plant objective, whereas
WWDM 1300nm
> > > meets the same one plus three others, rendering it as a much more
flexible and
> > > encompasing solution. In simply counting objective checkmarks, it's
4-to-1 in
> > > favor of WWDM.
> > >
> > > We all agreed to HSSG objectives. It's time to select PHY's according
to these
> > > objectives. I agree with Vipul's picks.
> > >
> > > --
> > >
> > > Best Regards,
> > > Rich
> > >
> > >
> > > Martin Nuss wrote:
> > > >
> > > > Of all the MM-fiber PMD proposals, only the 850nm serial VCSEL
solution over the
> > > > new high-performance fibers has so far been shown to work under
stressed systems
> > > > conditions, with Bit Error Rate measurements and careful analysis of
the systems
> > > > impairments to support that, and working with VCSELs from many
vendors. The
> > > > suggestions below are highly puzzling to me...
> > > >
> > > > Martin
> > > >
> > > > Rich Taborek wrote:
> > > > >
> > > > > Vipul, Rob,
> > > > >
> > > > > It should be pointed out that a Serial 850nm solution only
partially meets one
> > > > > HSSG distance/cable plant objective: 300m on MMF. However, this
MMF must be the
> > > > > new, enhanced MMF. The Serial 850nmsolution addresses no SMF
objectives
> > > > >
> > > > > WWDM meets all HSSG MMF objectives as well as SMF objectives to
10km.
> > > > >
> > > > > I agree with Vipul's choice of 3 PMDs as the best possible PMD set
to address
> > > > > HSSG HSSG distance/cable plant objectives.
> > > > > Best Regards,
> > > > > Rich
> > > > >
> > > > > --
> > > > >
> > > > > Vipul Bhatt wrote:
> > > > > >
> > > > > > Rob,
> > > > > >
> > > > > > Rob Marsland wrote:
> > > > > > >
> > > > > > > Finally, I hate to be annoying, but this is the SERIAL sig.
Since when is
> > > > > > > WWDM a serial solution?
> > > > > > >
> > > > > >
> > > > > > There is nothing annoying about your question. I should answer.
I believe it
> > > > > > is in our (the Serial SIG's) best interest to rise above our
Serial focus and
> > > > > > recognize that an "all Serial" set of solutions that meets all
the distance
> > > > > > objectives is not something our customers are willing to sign up
for. By
> > > > > > proposing a set of three solutions - two of which are Serial - I
am proposing
> > > > > > a set that has the highest chance of being accepted by our
customers and the
> > > > > > majority of 802.3ae members. I understand you disagree, and I
respect your
> > > > > > opinion.
> > > > > >
> > > > > > Vipul
>
> -------------------------------------------------------
> Richard Taborek Sr. Phone: 408-845-6102
> Chief Technology Officer Cell: 408-832-3957
> nSerial Corporation Fax: 408-845-6114
> 2500-5 Augustine Dr. mailto:rtaborek@xxxxxxxxxxx
> Santa Clara, CA 95054 http://www.nSerial.com