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

RE: 850 nm solutions

I agree with Jim Tatum of Honeywell's summary of the economic and commercial
importance of 850nm VCSEL-based devices to support the 10Gb Ethernet LAN

Our research indicates that nearly 80% of the links will not exceed 300
meters, and almost 50% of the total will be made up of links of 100m or
less. Today’s Gigabit Ethernet market is still running at a large percentage
of SX ports (80-85% of the total ports) within the LAN infrastructure. Early
adopters of 10Gb Ethernet will likely mimic this deployment model.  The
primary concern of these customers are:  low cost, ease of integration( i.e.
low power, low EMI), and a desire to use multi-mode fiber.  Blaze proposed
an 850 nm VCSEL based CWDM solution, to address this short haul interconnect
low-cost marketplace.

Blaze will be presenting further technical data and analysis at the May,
Ottawa, IEEE interim meeting in support of the 850nm CWDM proposal.  The
original Blaze IEEE 802.3ae proposal can be found at:

Kirk Bovill
Director of Product Marketing
Blaze Network Products
Sierra Ct., Ste.G
Dublin, CA 94568
925-560-1629 tel, 408-497-3663 cell

-----Original Message-----
From: owner-stds-802-3-hssg@xxxxxxxx
[mailto:owner-stds-802-3-hssg@xxxxxxxx]On Behalf Of Tatum, Jim
Sent: Tuesday, April 18, 2000 12:45 PM
To: Jaime Kardontchik; stds-802-3-hssg
Subject: RE: 850 nm solutions

I guess I don't understand all of the problems here...

Why can't we have a serial solution at 850nm and 1310nm, 1550nm that
mirrors the 802.3Z approach, where we have a table that addresses each of
the fiber types. i.e. repeat the tables 38.2 and 38.5. Seems like an easy
thing to do, and straightforward. That is one PMD type. A second PMD type
would be the 4 wavelength WDM at 3.125GBd, with a similar table. To me
that is 2 PMD's, with options, with link lengths depending on what fiber
you have installed. The market has shown that 850nm technology is the low
cost approach.



Jim Tatum
(972) 470-4572

 -----Original Message-----
From: Jaime Kardontchik
Sent: Tuesday, April 18, 2000 2:07 PM
To: stds-802-3-hssg
Subject: 850 nm solutions

Edward and all,

Thank you Edward for your insight.

There are have been in Albuquerque  excellent presentations
using 850 nm VCSELs by J. Yorks, P. Kolesar et al, F. Peters,
R. Colla et al and J. Jewell. They were all targetted towards
the new 2200 MHz*km MMF using 10 Gbaud lasers. Unfortunately,
at this speed they cannot support the installed MMF (the
eye closes at about 25 meters).

However, there are two proposals using 850nm - 4WDM that
do support both the installed MMF and the new 2200 MHz*km
MMF. See the Albuquerque presentations:

B. Wiedemann et al: "Evaluating CWDM 10GBASE-SX"


J.E. Kardontchik et al: "850nm-4WDM-1.25Gbaud transceiver
   over MMF for 10 GbE"

Both provide ample support for the new MMF: the first
proposal, at 2.578 Gbaud, has a target of 550 meters, and
the second proposal, running at 1.25 Gbaud, has a target
of 1,000 meters. Hence, both meet with ample margin
the minimum objective of the HSSG regarding the new
2200 MHz*km MMF, 300 meters.

The 2.578 Gbaud proposal has a target of 100 meters on
installed MMF. The 1.25 Gbaud proposal has a target of
160 meters on installed MMF. Hence, both meet also the
minimum objective of the HSSG regarding the installed
MMF, 100 meters.

The 2.578 Gbaud proposal has the advantage of using
on-off optical modulation. Its PCS, using 64/66, is new
and needs still some work and debugging in the coming
months. Work is being presently done in both areas to
validate this proposal. It is a very good proposal.
I happen to co-sponsor it ... :-)

The 1.25 Gbaud proposal uses PAM-5 modulation, that
is new to the optical community. However, it only needs a
very small dynamic range (18-level ADC in the receiver),
corresponding to an SFDR (Spurious Free Dynamic Range)
of about

   SFDR = 20*log(18) + (2/3)*10*log(0.625e+9)
        = 84 dB*Hz^(2/3)

This SFDR is well within the present VCSEL technology
capabilities. (*) This proposal has the advantage that it reuses
the 1000BASE-T PCS, that has already been debugged and
standardized and some companies are already offering it
in commercial products. It uses the same symbol rate as
the 1 GbE optical transceivers, 1.25 Gbaud, simplifying
the coexistence of 1 and 10 GbE transceivers on the same
board and minimizing packaging and PCB costs and EMI.
It could be the lowest cost 10 GbE system solution on MMF,
both installed and new.

The only area that needs detailed work, in order to meet
the July 2000 schedule-milestone, is in providing actual
5-level pseudorandom stimulus to VCSELs at 1.25 Gbaud
and measuring the actual performance of the optical link.
My company has shown a lot of support towards developing
the whole proposal. However, it does not have the means
and optical expertise in this area and we will welcome
any optical companies interested in performing these
measurements and presenting them to the Task Force.

These are the 850 nm choices.

(*) See, for example:

   "Dynamic Range of VCSELs in Multimode Links"
   by: H.L.T. Lee, R.V. Dalal, R.J. Ram and K.D. Choquette
   IEEE Photonics Tech. Letters, vol 11, pp 1473-75, Nov 1999

Jaime E. Kardontchik

Edward Chang wrote on Mon, 17 Apr 2000 12:16:33 -0400:

> Brad:
> I appreciate your input, and we may have to speed up the proposal by
> responding to some questions.
> My main concern is that the sole technology, 850 nm VCSEL, which   created
> today's cost-effective gigabit rate LAN market is not included in the
> proposal.  The VCSEL technology will continue to play the major role in
> 10 GbE market to keep the cost affordable and to open up the market.
> Following the lead by the more affordable technology, the more   expensive
> technologies will be demanded by the market needs.   This was the
> happened in the Gigabit rate market.
> The mass market is always looking for the top performance at lowest
> It is true that July is approaching very quick, and we should speed up
> process.
> Nevertheless, our good common since taught us that do it "Right" is the
> to the success.
> I believe all of us had some regrettable experience: overly rigid to
> the rules set at the beginning, and not enough flexibility to
> the reality uncovered later.
> The mission of HSSG is to provide the right directions for 10 GbE
> therefore, we should continue to evaluate the over all progress and
> good balance.
> Regards,
> Edward S. Chang
> NetWorth Technologies, Inc.
> EChang@xxxxxxxxxxxxxxxx
> Tel: (610)292-2870
> Fax: (610)292-2872