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

Re: [802.3BA] Discussion on 40G for => 10 km SMF



Chris-san

Thank you for your reply, but I found no answers for my questions about 
4x10G CWDM module in your comment.
  - Cost down technology
  - Module size (double XENPAK is correct?)

And I'd like to ask below question;
  - How to push 4-TOSAs, 4-ROSAs and 2-TFFs into the module?

Thanks,
Kazuyuki Mori


----- Original Message ----- 
From: "Chris Cole" <chris.cole@xxxxxxxxxxx>
To: <STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx>
Sent: Thursday, August 14, 2008 9:02 AM
Subject: Re: [802.3BA] Discussion on 40G for => 10 km SMF


Mori-san,

Thank you for clarifying that direct connection means the use of a GPPO
adaptor instead of GPPO cable between the SerDes IC PCB and the optics.

With respect to your specific comments and questions:

1) Cost

You are correct that 802.3ba TF consensus is that 40GE-CWDM cost will be
between 4x and 8x of 10GE-LR.

HOWEVER this is relative to 10GE-LR cost over time. So a constant cost
multiplier ratio like 4x, means that the absolute cost is decreasing by
following the decrease in the cost of 10GE-LR.

Your statement that no CWDM solution will realize less then 4x cost
reduction is misleading, as already pointed out by Steve Trowbridge in
his 8/4/08 reflector email.

You are comparing a falling cost ratio of 40GE-Serial against a constant
cost reference (10GE LR in 2010,) versus a constant cost ratio of
40GE-CWDM (4x) against a falling cost reference (10GE-LR over time.)
This is not apples-to-apples.

So let's do a fair apples-to-apples comparison. On page 15 of
traverso_02_0708, 40GE-Serial with GPPO IF cost (your reflector email
proposal) in 2010 is 5.9x 2010 10GE LR cost, and in 2012 is 3.7x 2010
10GE LR cost. This represents a huge 8x cost reduction from the cost of
40GE-Serial in 2008, and then very steep cost reduction from 2010 to
2012 (1.6x.)

On the same page, 40GE-CWDM cost in 2010 is 5.9x 2010 10GE LR cost, and
in 2012 is 4.9x 2010 10GE LR cost. Since just the cost of 10GE LR will
drop from 1x in 2010 to 0.73x in 2012, this means you are claiming that
in 2012 40GE-CWDM cost is 6.7x 2012 10GE LR cost, i.e. the cost
multiplier is increasing! This will NOT be the case, and would only
happen if there was NO cost reduction investment in 40GE-CWDM.

A much more accurate cost reduction prediction is that if 40GE-CWDM cost
in 2010 is 5.9x 2010 10GE LR cost, then in 2012 it is 4.3x 2010 10GE LR
cost, i.e. following 10GE LR cost decrease.

If the same type of aggressiveness is used to predict CWDM cost as is
being used for Serial cost predictions, then 40GE-CWDM cost in 2010 is
5x 2010 10GE LR cost, and in 2012 is 3.7x 2010 10GE LR cost (less then
4x!)

40GE-Serial has no cost advantage over 40GE-CWDM in the foreseeable
future. Contrary to many statements, 40GE-CWDM cost will fall below 4x
10GE LR cost if the cost reference is 10GE LR in a given year, like 2010
which is what is used for 40GE-Serial cost predictions.

2) Size, form factor

You are making an unfair comparison of what is possible for 40GE-Serial
advanced technology with a huge investment, to what is possible with
40GE-CWDM off the shelf technology with modest investment.

To package today's 40GE-Serial technology (GPPO IF ICs and Optics) in a
pluggable form factor, something like double XENPAK is required. A
XENPAK or X2 size is not feasible.

As I commented in my 7/31/08 email to Takai-san, if we are to make a
fair comparison of advanced technology with huge investment, CWDM has a
size advantage over Serial. CWDM CDRs can be separated from the optics,
and an un-retimed form factor like QSFP becomes possible with optics
integration. For Serial, the SerDes always have to be packaged with the
optics leading to a size disadvantage for both space and thermal
reasons.

3) CWDM optics connection

The proposal for quick time to market 40GE-CWDM product is to use
existing discrete 10G OSAs (with DFBs on the CWDM grid) connected with
separate fibers to discrete TFF, similar to what was presented in
paatzch_01_0708 but for CWDM.

Thank you

Chris

-----Original Message-----
From: Kazuyuki Mori [mailto:mori.kazuyuki-1@xxxxxxxxxxxxxx]
Sent: Wednesday, August 13, 2008 9:50 AM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] Fwd: Re: [802.3BA] Discussion on 40G for => 10 km
SMF

Chris-san,

Actually, a female-female adaptor is used for GPPO connection. I know
the
cost of every such GPPO component is relatively high when compared to
10GE
LR XFP component cost.
This kind of RF interconnection technologies enables us to reduce size
(XENPAK or X2) and cost of 40GE Serial modules.

Anyway, I don't know the details of 4x10G CWDM module.
1)Cost
    TF consensus are 4x-8x of 10GE LR and no solution to realize under
4x.
2)Size, form factor
    What kind of form factor do you use, double size of XENPAK?
3)Connection between optics
    Do you use discrete optics? Do you use fiber to connect between
optics?

Thanks,
Kazuyuki Mori

----- Original Message ----- 
From: "Chris Cole" <chris.cole@xxxxxxxxxxx>
To: <STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx>
Sent: Wednesday, August 13, 2008 4:01 AM
Subject: Re: [802.3BA] Fwd: Re: [802.3BA] Discussion on 40G for => 10 km
SMF


Mori-san,

The only commercially available PCB edge mount GPPO connectors I know
have male polarity, which is the same as XLMD GPPO connectors.

Are you developing commercial female polarity PCB GPPO connectors that
could be directly connected to the XLMD GPPO connectors?

Or by direct connection do you mean that instead of flexible
female-female polarity GPPO cable you use rigid female-female polarity
GPPO bullet?

As you know, the cost of every such GPPO component is significant when
compared to 10GE LR XFP component costs.

Thank you

Chris

-----Original Message-----
From: Kazuyuki Mori [mailto:mori.kazuyuki-1@xxxxxxxxxxxxxx]
Sent: Tuesday, August 12, 2008 12:25 AM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] Fwd: Re: [802.3BA] Discussion on 40G for => 10 km
SMF

Chris-san,

We connect the PCB edge mount GPPO connectors and GPPO connectors on the

XLMD optics directly. Therefore we can eliminate the GPPO cable and
reduce
module size.

Thanks,
Kazuyuki Mori


----- Original Message ----- 
From: "Chris Cole" <chris.cole@xxxxxxxxxxx>
To: <STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx>
Sent: Tuesday, August 12, 2008 3:50 PM
Subject: Re: [802.3BA] Fwd: Re: [802.3BA] Discussion on 40G for => 10 km
SMF


Hello Mori-san,

Thank you for continuing to fill in the details of the 40GE-Serial
baseline
proposal.

You correctly point out that edge mount type GPPO connectors are
commercially available. Various types of such edge connectors are used
on
all 40G GPPO IC and Optics packages, and ship in 10Ks volume into
telecom
applications.

Is is reasonable to assume that the interconnect between the PCB edge
mount
GPPO connectors and GPPO conectors on the XLMD optics are GPPO cables,
similar to cables used in 40G modules today, for example as shown in the

left hand photograph on page 8 of traverso_02_0708?

Chris

________________________________

From: Kazuyuki Mori [mailto:mori.kazuyuki-1@xxxxxxxxxxxxxx]
Sent: Mon 8/11/2008 10:37 PM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] Fwd: Re: [802.3BA] Discussion on 40G for => 10 km
SMF



Hello Chris-san,

Sorry for our insufficient information for RF interconnection. In case
of
XLMD, we use an edge mount type GPPO connector at the end of PCB, so we
can
apply the SMT type package for SerDes as shown on the right side of page
6
of traverso_02_0708. We already confirmed the characteristics of this RF
interconnection; the result shows 50GHz BW. As you know well, the edge
mount
type GPPO connector is already commercially available.

Thanks,
Kazuyuki Mori