Joel:
Your comment “Opens up
the door for someone to develop a channel on several different kinds of media
that would be infomative or normative compliant but not actually work.” Is very important. If we are working
on a “Right by Design”, that will happen all too often. We
need a specification which gives reliable design information.
Ed
Dr. Edward P. Sayre, P. E.
North East Systems Asscoiates, Inc.
9 Maple Lane
PO Box 807
Marshfield, MA 02050
USA
[T]: +1-781-837-9088
[F]: +1-781-837-9083
[C]: +1-978-314-4940
[E]: esayre@xxxxxxxx
From: Joel Goergen
[mailto:jgoergen@xxxxxxxxx]
Sent: Thursday, February 16, 2012
1:58 PM
To:
STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GCU]
Objective changes for the Back Plane Initiative
Rich
You just made my point. Without the length, the loss number is
meaningless. Further to my requests for the edits to the proposed changes
... Without at least referencing copper traces, we will have to make sure the
modeling works on printed wiring. I can live without a materials
reference, though for the life of me, I do not understand why people think we
should leave it out ... Opens up the door for someone to develop a channel on
several different kinds of media that would be infomative or normative
compliant but not actually work.
Joel
On 2/15/12 2:50 PM, "Mellitz, Richard" <richard.mellitz@xxxxxxxxx> wrote:
But I would expect
serdes mfgs to support that -35 dB across all lengths. The channels are
different. The test details should take care of this. The biggest problem
I’ve had with ap is the 1 meter objective because this gets set in
concrete before the detail designers ever get involved. I have seen this
many times since ap. The architects always think they have good material.
On the average the mechanicals are get set for the 1 meter even though
the design is subbed to an ODM. This happens more than I’d care to admit.
I keep hearing from the architects that if you told me loss first I
would be thinking I need to get into the details more before I do the
mechanicals. So the 1 meter objective is only useful for the experts.
… Rich
From: Joel Goergen [mailto:jgoergen@xxxxxxxxx]
Sent: Tuesday, February 14, 2012
11:09 PM
To: STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GCU]
Objective changes for the Back Plane Initiative
Hi
Pravin
I did not realize that the length had been removed from the objective.
If you remove the length from the objective, you are taking choices away from
the system vendors and you are giving complete freedom to the SERDES vendors.
The proof is in the channel. If defined as loss only, then you are
assuming that a –35dB channel at 40in is the same as a channel
–35dB at 30in is the same as a channel –35dB at 20in is the same as
–35dB at 10in and is the same as –35dB at 1in. And that
simply is not true. The methods used in design, debug, and manufacturing
of a system may be in peril because we address loss as the same across length.
There now is no contract with which to make sure a serdes can work across
lengths.
Joel
On 2/14/12 3:21 PM, "Pravin Patel" <pravinp@xxxxxxxxxx>
wrote:
Rich,
Let's give system vendors the freedom to create a channel that meets loss
requirements. I see no need to add another variable, which may lead to more
confusion and a burden being placed on System Vendors. Howard's January
presentation indicates that Joel has supported the same objective change. His
presentation does not address a specific length.
Pravin Patel
Senior
Technical Staff Member; IBM Master Inventor IBM SystemX
High
Speed Interconnect Architecture Bldg:205 Room:EE161
E-mail: pravinp@xxxxxxxxxx <mailto:pravinp@xxxxxxxxxx> 3039 Cornwallis Road
Phone: 919-543-3837 RTP, NC 27709
Let's Build A Smarter Planet
From:
"Mellitz, Richard" <richard.mellitz@xxxxxxxxx>
To:
STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx
Date:
02/14/2012 05:11 PM
Subject:
Re: [802.3_100GCU] Objective
changes for the Back Plane Initiative
How’s this for a compromise?
Define a 4 lane PHY for operation over backplanes with a insertion loss of
<= 35 dB at 12.9 GHz with printed circuit board routed lengths which may
reach 1 meter.
Define a 4 lane PHY for operation over backplanes with a insertion loss of
<= 33 dB at 7.0 GHz with printed circuit board routed lengths which may
reach 1 meter.
From: Mike
Dudek [mailto:mike.dudek@xxxxxxxxxx <mailto:mike.dudek@xxxxxxxxxx> ]
Sent: Tuesday, February 14, 2012
4:23 PM
To: STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_100GCU]
Objective changes for the Back Plane Initiative
I think the channels defined by IEEE
std.802.3ap with lengths up to at least 1m is too restrictive. To me that
would imply all backplanes allowed by ap. (ie no opportunity to tighten
ILD etc.)
Mike Dudek
QLogic Corporation
Senior Manager Signal
Integrity
26650 Aliso Viejo Parkway
Aliso Viejo CA 92656
949 389 6269 - office.
Mike.Dudek@xxxxxxxxxx
<mailto:Mike.Dudek@xxxxxxxxxx>
From: Brad
Booth [mailto:Brad_Booth@xxxxxxxx] <mailto:[mailto:Brad_Booth@xxxxxxxx]>
Sent: Tuesday,
February 14, 2012 12:31 PM
To: STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx <mailto:STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx>
Subject: Re:
[802.3_100GCU] Objective changes for the Back Plane Initiative
Joel,
I understand where you’re coming from on the insertion loss. That sounds
like an objective with a specification built in.
How about the following:
1) Define a 4 lane PHY for operation up to 12.9 GHz over backplane channels
defined by the IEEE P802.3bj Task Force with lengths up to at least 1m.
2) Define a 4 lane PHY for operation over backplanes channels defined by IEEE
Std. 802.3ap with lengths up to at least 1m.
My hope is that #1 represents the new backplane channels that can be defined by
the task force and supports the NRZ position while #2 represents supporting
existing channels being developed for 10GBASE-KR and 40GBASE-KR4. By removing
insertion loss from #1, the task force is also permitted to have some wiggle
room relative to what the final loss number needs to be.
One note is that the word “defined” in both #1 and #2 may be too
literal for some. In such a case “referenced” may be the better
word.
Thanks,
Brad
From: Joel
Goergen [mailto:jgoergen@xxxxxxxxx] <mailto:[mailto:jgoergen@xxxxxxxxx]>
Sent: Tuesday, February
14, 2012 2:14 PM
To: STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx <mailto:STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx>
Subject: Re:
[802.3_100GCU] Objective changes for the Back Plane Initiative
Hi Dave
My goal was to make sure loss and reach are defined. I added the material
because we want to make sure we are talking about an fr-4 rated kind of design
... Not define a specific set of materials. Rather, make sure we are not
trying to do this on the lowest fr-4 out there ... Hence improved fr-4
Take care
Joel
On 2/14/12 11:57 AM, "Chalupsky, David" <david.chalupsky@xxxxxxxxx
<david.chalupsky@xxxxxxxxx> >
wrote:
Hi Joel,
Is there a way to differentiate the wording regarding materials for the two
objectives? I think your intent is that “improved FR-4”
has a different definition for each objective, but with the same terminology on
materials it is not clear why the two objectives are distinct.
I am in favor of two objectives, just anticipating that a good deal of
word-smithing will ensue.
Would a compromise be to keep the loss and reach, but drop the material?
Thx,
Dave
From: Joel Goergen [mailto:jgoergen@xxxxxxxxx <mailto:jgoergen@xxxxxxxxx> ]
Sent: Tuesday, February 14, 2012
11:12 AM
To: STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx <STDS-802-3-100GCU@xxxxxxxxxxxxxxxxx>
Subject: [802.3_100GCU]
Objective changes for the Back Plane Initiative
All
There have been some discussions around changing the following objective:
Define a 4-lane 100 Gb/s backplane PHY for operation over links
consistent with copper traces on “improved FR-4” (as defined by
IEEE P802.3ap or better materials to be defined by the Task Force) with lengths
up to at least 1m.
To something like:
•Define a 4 lane PHY for operation over backplanes
with a insertion loss of <= 35 dB at 12.9 GHz
•Define a 4 lane PHY for operation over backplanes
with a insertion loss of <= 33 dB at 7.0 GHz
I don’t support this because it leaves to much variation, along with
additional specifications, in the definition of the transmitter and receiver.
Systems implementation of this type of specification will see 1) an
increased design cost attributed to tools, modeling, and re-spins, 2)
incomplete and inconsistent models providing false positives to a successful
implementation, and 3) a complex set of additional metrics that will make the
finished standard too complicated to follow.
The following is a much better set of objectives that allow systems vendors a
path forward using current design processes and implementations.
1) Define a 4 lane PHY for operation over backplanes with a insertion loss of
<= 35 dB at 12.9 GHz over links consistent with copper traces
on “improved FR-4” (as defined by IEEE P802.3ap or better materials
to be defined by the Task Force) with lengths up to at least 1m.
2) Define a 4 lane PHY for operation over backplanes with
a insertion loss of <= 33 dB at 7.0 GHz over links
consistent with copper traces on “improved FR-4” (as defined by
IEEE P802.3ap or better materials to be defined by the Task Force) with lengths
up to at least 1m.
I heard reference that by defining loss, material, and length, we can not meet
that. I disagree. I think we are defining a loss up to 1m on
improved fr-4. If it makes people feel better, the following is also
acceptable from a system designers point of view”
1) Define a 4 lane PHY for operation over backplanes with a insertion loss of
<= 35 dB at 12.9 GHz over links consistent with copper
traces on “improved FR-4” (materials definition to be defined by
the Task Force) with lengths up to at least 1m.
2) Define a 4 lane PHY for operation over backplanes with
a insertion loss of <= 33 dB at 7.0 GHz over links
consistent with copper traces on “improved FR-4” ( materials
definition to be defined by the Task Force) with lengths up to at least 1m.
I do not support a loss only model. It is not economically feasible.
Take care
Joel Goergen