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

RE: XAUI AC coupling




Hi Pat

Regardless of the nature of each standard or the forum size.  I was just
proposing a solution which can coexist with AC and DC coupled link. 
Obviously you need to choose a common mode volatage and it would be nice to
be the same as the other standards.  But looks like this might start a
religious war consume lots of CPU cycle.

Every standard leverages other standards, but regardless I already said lets
stick with AC coupling for 10 Gig Ethernet and skip the religious war.

Thanks,

Ali



------Original Message------
From: pat_thaler@xxxxxxxxxxx
To: aghiasi@xxxxxxxxxxxxx, rtaborek@xxxxxxxxxxxxx
Sent: September 28, 2000 9:55:26 PM GMT
Subject: RE: XAUI AC coupling


Ali,

The other standard to which you refer is, I assume, the Infiniband
specification. Infiniband operates at a somewhat lower frequency
than we do and has different objectives. There is already divergence
between Infiniband and our specification of this interface. For
instance, they have chosen different idle signalling and alignment
mechanisms. Most of the Phy decisions in that body were made by votes
of ~7 voting member companies with simple majority voting.

I'm a member of the Infiniband Link Working Group. A lot of hard
work and contribution has gone into the specification and I believe
it will serve its application space well. However, I strongly disagree
with your assertion that decisions by that body provide a precident
for what 802.3ae "should choose."

The ~200 voters of 802.3 need to choose the direction that is right
for Ethernet's future.

Regards,
Pat

-----Original Message-----
From: Ali Ghiasi [mailto:aghiasi@xxxxxxxxxxxxx]
Sent: Wednesday, September 27, 2000 11:54 AM
To: rtaborek@xxxxxxxxxxxxx
Cc: HSSG
Subject: Re: XAUI AC coupling



Hi

There is already another Standard based on Hari, which specifies both AC
and DC coupling.
I agree majority of applications will be AC coupled, but there are some
applications which
require DC coupling.  We should spec AC coupling and as ED said in a
section describe how to
implement DC coupling.  The other standard already has chosen 0.75 volts
of common mode, so we
should choose 0.75 V.

Ali Ghiasi
Newport Comunication


Rich Taborek wrote:
>
> Tom, Ed,
>
> I agree with this direction and side with Ed on the issue of spec'ing
> DC-coupling for the reasons that Ed provides. Thanks for your support on
> this issue.
>
> Best Regards,
> Rich
>
> --
>
> Ed Grivna wrote:
> >
> > Hi Tom,
> >
> > I agree with the consideration, that IF we are to standardize a
> > DC coupled interface, then it is required to also spec the levels.
> > I think my main concern is that as soon as you do this, you will
> > open a Pandora's box of technologies that do not interoperate
> > at a DC level, but work fine when AC-coupled.  For my own part,
> > would love to specify DC lvels, just so long as they match up
> > with the DC-levels _I_ want.  If anything else goes in, escpecially
> > this late in the game, I'll be one of the first to scream restraint
> > of trade.
> >
> > As stated earlier, I have no problem with allowing DC-coupling, but
> > I don't want it to be mandatory.  However, once you spec the levels
> > in the standard, you now have requirements that must be met.
> >
> > -Ed Grivna
> >
> > > All,
> > >       I concur that it would be advantageous to support both AC and DC
> > > coupling. Maybe we have a section with AC coupling specs and a
separate
> > > table for DC coupling specs. This would allow a vendor to be just AC
> > > compliant or just DC compliant or both. If we leave out the DC specs
all
> > > together (respectfully disagreeing with my fellow Minnesotan) than I
> > > don't think we have standardized anything for the DC coupled
interfaces
> > > and we will not have interoperability.
> > >
> > > Tom
> > >
> > >
> > > Ed Grivna wrote:
> > > >
> > > > Hi Rich,
> > > >
> > > > I agree with the Dawson implementation; i.e., where capability for
> > > > AC-coupling is mandatory, and you CAN DC-couple if both ends
> > > > will support it.
> > > >
> > > > But in that regard, I don't believe it apporpriate to add ANY
> > > > DC-level numbers to the standard for XAUI signaling, not even as
> > > > a guideline.  This places the responsibility on the implementer
> > > > (where it always belongs) to ensure that the devices they choose
> > > > will support DC-coupling (if they wish to implement it).
> > > >
> > > > The other place I find a minor dissagrement is with your statement
> > > > that if both ends of the link are made by the same manufacturer
> > > > that a DC-coupled link will generally be possible.  This is not
> > > > necessarily the case, especially when CML drivers are used.  These
> > > > switch very close to the power rail, and are not generally
positioned
> > > > in the middle of the common mode operating range of the receiver.
> > > > While a DC-coupled connection MAY work, it is not optimal.
> > > >
> > > > Regards,
> > > >
> > > > Ed Grivna
> > > > Cypress Semiconductor
> > > >
> > > > >
> > > > > All,
> > > > >
> > > > > By my count, I have 4 votes for allowing XAUI DC-coupling against
0
> > > > > votes for requiring only AC-coupling. The 4 votes are:
> > > > >
> > > > > Ed Grivna - Cypress
> > > > > Dawson Kesling - Intel
> > > > > Jeff Porter - Motorola
> > > > > Rich Taborek - nSerial
> > > > >
> > > > > Any other opinions out there?
> > > > >
> > > > > Best Regards,
> > > > > Rich
> > > > >
> > > > > --
> > > > >
> > > > > "Jeff Porter (rgbn10)" wrote:
> > > > > >
> > > > > > Folks,
> > > > > >
> > > > > > I feel consensus emerging here.
> > > > > >
> > > > > > Rich writes
> > > > > >
> > > > > > > a) A XAUI implementer can always get away with AC-coupling and
> > > > > > >    AC-coupling details for XAUI are readily available;"
> > > > > >
> > > > > > and
> > > > > >
> > > > > > > That said, I'd be happy to go with (1) or (2).
> > > > > >
> > > > > > Dawson writes
> > > > > >
> > > > > > > An alternative is to mandate CAPABILITY for AC coupling. This
allows
> > DC
> > > > > > > coupling where compatible implementations permit, but ensures
that ALL
> > > > > > > implemenations will interoperate via AC coupling.
> > > > > >
> > > > > > I agree.  Specify the differential signal.  Require the receiver
> > > > > > to function *when* driven by ac coupled signals to provide a
method
> > > > > > that insures interoperability.  After all, we've increased baud
rate,
> > among
> > > > > > other reasons, to permit ac coupling as an approach to
interoperability.
> > > > > > Do not require ac coupling since dc coupling will often work,
and we've
> > > > > > left a way to interoperate.
> > > > > >
> > > > > > The remaining technical work is to include in an (informative)
XAUI link
> > > > > > budget (if we choose to explain how this could work) the
attenuation,
> > > > > > skew, and jitter, etc. budgeted for ac coupling.
> > > > > >
> > > > > > Proposals and justification for this budget item?
> > > > > >
> > > > > > Jeff
> > > > > >
> > > > > > Rich Taborek wrote:
> > > > > > >
> > > > > > > Dawson,
> > > > > > >
> > > > > > > In terms of specsmanship, I believe that we have two
alternatives with
> > > > > > > regard to coupling for XAUI:
> > > > > > >
> > > > > > > 1) Leave coupling out altogether as an implementation detail;
> > > > > > > 2) Specify detail for both AC-coupling and DC-coupling.
> > > > > > >
> > > > > > > It sound like you're leaning towards (2) where I'm leaning
towards
> > (1).
> > > > > > > My argument is that (2) is a whole heck of a lot more work
than (1)
> > and
> > > > > > > may be more costly since compliance verification has some non
zero
> > cost.
> > > > > > > I believe that (1) works and is interoperable because:
> > > > > > >
> > > > > > > a) A XAUI implementer can always get away with AC-coupling and
> > > > > > > AC-coupling details for XAUI are readily available;
> > > > > > > b) A savvy XAUI implementer may save $$$, increase reliability
(fewer
> > > > > > > components), increase signal fidelity (fewer vias), etc. by
going with
> > > > > > > DC-coupling if possible given their component selection.
> > > > > > >
> > > > > > > The only other possibilities are not palatable to me:
> > > > > > >
> > > > > > > 3) Mandate AC-coupling;
> > > > > > > 4) Mandate DC-coupling.
> > > > > > >
> > > > > > > That said, I'd be happy to go with (1) or (2).
> > > > > > >
> > > > > > > Best Regards,
> > > > > > > Rich
> > > > > > >
> > > > > > > --
> > > > > > >
> > > > > > > "Kesling, Dawson W" wrote:
> > > > > > > >
> > > > > > > > Rich and all,
> > > > > > > >
> > > > > > > > I agree that it would be nice to avoid AC coupling if we can
still
> > > > ensure
> > > > > > > > interoperability.
> > > > > > > >
> > > > > > > > If we remove reference to coupling altogether, we must add a
common
> > mode
> > > > > > > > specification or definite logic levels; we cannot only
specify
> > > > peak-to-peak
> > > > > > > > swing as we are now doing and expect interoperability. (All
> > chip-to-chip
> > > > > > > > interconnect spec's I know of specify either DC-referenced
logic
> > levels
> > > > or
> > > > > > > > common mode and differential mode levels. Is there an
exception? We
> > have
> > > > > > > > avoided this by mandating AC coupling up to this time.)
> > > > > > > >
> > > > > > > > An alternative is to mandate CAPABILITY for AC coupling.
This allows
> > DC
> > > > > > > > coupling where compatible implementations permit, but
ensures that
> > ALL
> > > > > > > > implemenations will interoperate via AC coupling.
> > > > > > > >
> > > > > > > > -Dawson Kesling
> > > > > > > >  Intel Corporation, NCD
> > > > > > > >  916 855-5000 ext. 1273
>
> -------------------------------------------------------
> 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