Re: XAUI AC coupling
Your argument is exactly why I favor NOT specifying the XAUI signal
coupling method. Once again, I'm in favor of allowing either AC or
DC-coupling for XAUI as an implementation detail. I strongly support not
specifying or even recommending any DC bias (common-mode) voltages for
XAUI is defined as an XGMII extender. Therefore, XAUI IS a backplane
application. The first thing your customers are going to ask you when
you tell them that the purpose of your XAUI chips is to extend the XGMII
is: Can XAUI go across backplanes, and connectors and leap tall
buildings in a single bound?
How the heck is a backplane application or any other short or long
chip-to-chip interconnect a non-standard usage of XAUI, as XAUI is
Joel Dedrick wrote:
> Boy - you look away for a minute and this thing goes off the rails.
> First, let me say that the proposed DC bias point of 0.75V is in my view
> decidedly nonoptimum and wastes power. Even if an optimum point for some
> particular technology *were* chosen, it would be nonoptimum in other
> technologies or technology generations.
> While I don't find this in the email thread, I seem to remember from New
> Orleans that the impetus for supporting DC coupling comes from backplane
> applications. I think it's a mistake to create a standard where
> implementations of the standard are permanently penalized in order to
> support some reuse of the silicon in non-standards based applications. If
> I've misinterpreted the reason for this initiative, my appologies.
> Now to the point: Any IC manufacturer, faced with a "recommended" common
> mode specification, will be obliged to meet it in order to not be frozen out
> of applications where DC coupling was chosen as a system implementation.
> All silicon guys try hard to superset known specifications whenever
> possible. So, while making DC coupling "recommended" or "optional" may make
> it more palatable to some, it is really just a ruse. It will have the same
> impact as a mandatory specification in terms of the behavior of suppliers.
> So, we will be mandating a solution that wastes power in at least some of
> the technologies people will use to build these things, even for cases where
> AC coupling *is* used, since we will have to assume it is not. There's no
> way we'll build two output buffers, or have some kind of mode switch, or
> have two different versions of the chip. Everyone will bear the power
> and/or margin penalty, whether they AC couple or not.
> I vote that we require AC coupling, and that no DC voltages are specified or
> Joel H. Dedrick
> Director, Marketing
> Optical Networking Division
> PMC-Sierra, Inc.
> (408) 626-2070
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